{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 100, "global_step": 8000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.5e-06, "loss": 28.3115, "step": 1 }, { "epoch": 0.0, "learning_rate": 5e-06, "loss": 28.3019, "step": 2 }, { "epoch": 0.0, "learning_rate": 7.5e-06, "loss": 28.1625, "step": 3 }, { "epoch": 0.0, "learning_rate": 1e-05, "loss": 28.1932, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.25e-05, "loss": 27.7263, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.5e-05, "loss": 27.1889, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.7500000000000002e-05, "loss": 26.3331, "step": 7 }, { "epoch": 0.0, "learning_rate": 2e-05, "loss": 25.4267, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.2499999999999998e-05, "loss": 23.3616, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.5e-05, "loss": 21.8006, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.75e-05, "loss": 21.284, "step": 11 }, { "epoch": 0.0, "learning_rate": 3e-05, "loss": 21.2637, "step": 12 }, { "epoch": 0.0, "learning_rate": 3.2500000000000004e-05, "loss": 21.1415, "step": 13 }, { "epoch": 0.0, "learning_rate": 3.5000000000000004e-05, "loss": 20.9429, "step": 14 }, { "epoch": 0.0, "learning_rate": 3.75e-05, "loss": 20.4776, "step": 15 }, { "epoch": 0.0, "learning_rate": 4e-05, "loss": 19.892, "step": 16 }, { "epoch": 0.0, "learning_rate": 4.25e-05, "loss": 19.233, "step": 17 }, { "epoch": 0.0, "learning_rate": 4.4999999999999996e-05, "loss": 18.6345, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.75e-05, "loss": 18.2748, "step": 19 }, { "epoch": 0.0, "learning_rate": 5e-05, "loss": 18.1278, "step": 20 }, { "epoch": 0.0, "learning_rate": 5.25e-05, "loss": 17.7551, "step": 21 }, { "epoch": 0.0, "learning_rate": 5.5e-05, "loss": 17.5592, "step": 22 }, { "epoch": 0.0, "learning_rate": 5.75e-05, "loss": 16.7518, "step": 23 }, { "epoch": 0.0, "learning_rate": 6e-05, "loss": 16.5106, "step": 24 }, { "epoch": 0.0, "learning_rate": 6.25e-05, "loss": 16.3567, "step": 25 }, { "epoch": 0.0, "learning_rate": 6.500000000000001e-05, "loss": 15.9238, "step": 26 }, { "epoch": 0.0, "learning_rate": 6.75e-05, "loss": 15.4527, "step": 27 }, { "epoch": 0.0, "learning_rate": 7.000000000000001e-05, "loss": 15.3176, "step": 28 }, { "epoch": 0.0, "learning_rate": 7.25e-05, "loss": 14.9926, "step": 29 }, { "epoch": 0.0, "learning_rate": 7.5e-05, "loss": 14.4565, "step": 30 }, { "epoch": 0.0, "learning_rate": 7.75e-05, "loss": 14.2002, "step": 31 }, { "epoch": 0.0, "learning_rate": 8e-05, "loss": 13.7769, "step": 32 }, { "epoch": 0.0, "learning_rate": 8.25e-05, "loss": 13.726, "step": 33 }, { "epoch": 0.0, "learning_rate": 8.5e-05, "loss": 12.6503, "step": 34 }, { "epoch": 0.0, "learning_rate": 8.75e-05, "loss": 12.6933, "step": 35 }, { "epoch": 0.0, "learning_rate": 8.999999999999999e-05, "loss": 12.1913, "step": 36 }, { "epoch": 0.0, "learning_rate": 9.25e-05, "loss": 11.834, "step": 37 }, { "epoch": 0.0, "learning_rate": 9.5e-05, "loss": 11.5715, "step": 38 }, { "epoch": 0.0, "learning_rate": 9.750000000000001e-05, "loss": 11.3092, "step": 39 }, { "epoch": 0.01, "learning_rate": 0.0001, "loss": 10.8852, "step": 40 }, { "epoch": 0.01, "learning_rate": 0.0001025, "loss": 10.7544, "step": 41 }, { "epoch": 0.01, "learning_rate": 0.000105, "loss": 10.1788, "step": 42 }, { "epoch": 0.01, "learning_rate": 0.0001075, "loss": 10.3947, "step": 43 }, { "epoch": 0.01, "learning_rate": 0.00011, "loss": 9.8711, "step": 44 }, { "epoch": 0.01, "learning_rate": 0.00011250000000000001, "loss": 9.5748, "step": 45 }, { "epoch": 0.01, "learning_rate": 0.000115, "loss": 9.754, "step": 46 }, { "epoch": 0.01, "learning_rate": 0.0001175, "loss": 9.2442, "step": 47 }, { "epoch": 0.01, "learning_rate": 0.00012, "loss": 9.2299, "step": 48 }, { "epoch": 0.01, "learning_rate": 0.0001225, "loss": 9.0341, "step": 49 }, { "epoch": 0.01, "learning_rate": 0.000125, "loss": 8.9441, "step": 50 }, { "epoch": 0.01, "learning_rate": 0.0001275, "loss": 8.7826, "step": 51 }, { "epoch": 0.01, "learning_rate": 0.00013000000000000002, "loss": 8.8958, "step": 52 }, { "epoch": 0.01, "learning_rate": 0.00013250000000000002, "loss": 8.6397, "step": 53 }, { "epoch": 0.01, "learning_rate": 0.000135, "loss": 8.493, "step": 54 }, { "epoch": 0.01, "learning_rate": 0.0001375, "loss": 8.5221, "step": 55 }, { "epoch": 0.01, "learning_rate": 0.00014000000000000001, "loss": 8.1613, "step": 56 }, { "epoch": 0.01, "learning_rate": 0.0001425, "loss": 8.7778, "step": 57 }, { "epoch": 0.01, "learning_rate": 0.000145, "loss": 8.4581, "step": 58 }, { "epoch": 0.01, "learning_rate": 0.0001475, "loss": 8.9698, "step": 59 }, { "epoch": 0.01, "learning_rate": 0.00015, "loss": 8.6242, "step": 60 }, { "epoch": 0.01, "learning_rate": 0.0001525, "loss": 8.661, "step": 61 }, { "epoch": 0.01, "learning_rate": 0.000155, "loss": 8.2074, "step": 62 }, { "epoch": 0.01, "learning_rate": 0.0001575, "loss": 9.0654, "step": 63 }, { "epoch": 0.01, "learning_rate": 0.00016, "loss": 8.4233, "step": 64 }, { "epoch": 0.01, "learning_rate": 0.00016250000000000002, "loss": 9.5135, "step": 65 }, { "epoch": 0.01, "learning_rate": 0.000165, "loss": 8.4864, "step": 66 }, { "epoch": 0.01, "learning_rate": 0.0001675, "loss": 9.1542, "step": 67 }, { "epoch": 0.01, "learning_rate": 0.00017, "loss": 8.8938, "step": 68 }, { "epoch": 0.01, "learning_rate": 0.0001725, "loss": 8.6653, "step": 69 }, { "epoch": 0.01, "learning_rate": 0.000175, "loss": 8.9115, "step": 70 }, { "epoch": 0.01, "learning_rate": 0.0001775, "loss": 8.3977, "step": 71 }, { "epoch": 0.01, "learning_rate": 0.00017999999999999998, "loss": 8.0126, "step": 72 }, { "epoch": 0.01, "learning_rate": 0.0001825, "loss": 8.0421, "step": 73 }, { "epoch": 0.01, "learning_rate": 0.000185, "loss": 8.1183, "step": 74 }, { "epoch": 0.01, "learning_rate": 0.0001875, "loss": 7.7236, "step": 75 }, { "epoch": 0.01, "learning_rate": 0.00019, "loss": 7.5251, "step": 76 }, { "epoch": 0.01, "learning_rate": 0.00019250000000000002, "loss": 8.5536, "step": 77 }, { "epoch": 0.01, "learning_rate": 0.00019500000000000002, "loss": 7.8101, "step": 78 }, { "epoch": 0.01, "learning_rate": 0.0001975, "loss": 8.3544, "step": 79 }, { "epoch": 0.01, "learning_rate": 0.0002, "loss": 8.4304, "step": 80 }, { "epoch": 0.01, "learning_rate": 0.00020250000000000002, "loss": 7.4948, "step": 81 }, { "epoch": 0.01, "learning_rate": 0.000205, "loss": 7.4432, "step": 82 }, { "epoch": 0.01, "learning_rate": 0.0002075, "loss": 8.522, "step": 83 }, { "epoch": 0.01, "learning_rate": 0.00021, "loss": 7.6184, "step": 84 }, { "epoch": 0.01, "learning_rate": 0.0002125, "loss": 8.3915, "step": 85 }, { "epoch": 0.01, "learning_rate": 0.000215, "loss": 9.3057, "step": 86 }, { "epoch": 0.01, "learning_rate": 0.0002175, "loss": 7.9554, "step": 87 }, { "epoch": 0.01, "learning_rate": 0.00022, "loss": 9.2703, "step": 88 }, { "epoch": 0.01, "learning_rate": 0.00022250000000000001, "loss": 8.307, "step": 89 }, { "epoch": 0.01, "learning_rate": 0.00022500000000000002, "loss": 8.0812, "step": 90 }, { "epoch": 0.01, "learning_rate": 0.0002275, "loss": 7.9097, "step": 91 }, { "epoch": 0.01, "learning_rate": 0.00023, "loss": 7.8502, "step": 92 }, { "epoch": 0.01, "learning_rate": 0.0002325, "loss": 10.5609, "step": 93 }, { "epoch": 0.01, "learning_rate": 0.000235, "loss": 8.2943, "step": 94 }, { "epoch": 0.01, "learning_rate": 0.0002375, "loss": 9.2673, "step": 95 }, { "epoch": 0.01, "learning_rate": 0.00024, "loss": 10.059, "step": 96 }, { "epoch": 0.01, "learning_rate": 0.00024249999999999999, "loss": 7.9417, "step": 97 }, { "epoch": 0.01, "learning_rate": 0.000245, "loss": 7.8239, "step": 98 }, { "epoch": 0.01, "learning_rate": 0.0002475, "loss": 10.081, "step": 99 }, { "epoch": 0.01, "learning_rate": 0.00025, "loss": 7.9449, "step": 100 }, { "epoch": 0.01, "eval_acceptance_rate": 0.08749189972877502, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 0.8767000516375396, "eval_expected_tokens": 1.0958750645469244, "eval_loss": 9.101815223693848, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.02396876613673113, "eval_runtime": 50.3694, "eval_samples_per_second": 9.669, "eval_steps_per_second": 0.318, "step": 100 }, { "epoch": 0.01, "learning_rate": 0.0002525, "loss": 9.2231, "step": 101 }, { "epoch": 0.01, "learning_rate": 0.000255, "loss": 10.0614, "step": 102 }, { "epoch": 0.01, "learning_rate": 0.0002575, "loss": 8.4643, "step": 103 }, { "epoch": 0.01, "learning_rate": 0.00026000000000000003, "loss": 9.3793, "step": 104 }, { "epoch": 0.01, "learning_rate": 0.00026250000000000004, "loss": 10.0504, "step": 105 }, { "epoch": 0.01, "learning_rate": 0.00026500000000000004, "loss": 8.283, "step": 106 }, { "epoch": 0.01, "learning_rate": 0.0002675, "loss": 8.3914, "step": 107 }, { "epoch": 0.01, "learning_rate": 0.00027, "loss": 8.8283, "step": 108 }, { "epoch": 0.01, "learning_rate": 0.0002725, "loss": 8.7213, "step": 109 }, { "epoch": 0.01, "learning_rate": 0.000275, "loss": 8.6531, "step": 110 }, { "epoch": 0.01, "learning_rate": 0.0002775, "loss": 8.5586, "step": 111 }, { "epoch": 0.01, "learning_rate": 0.00028000000000000003, "loss": 9.9497, "step": 112 }, { "epoch": 0.01, "learning_rate": 0.0002825, "loss": 9.0021, "step": 113 }, { "epoch": 0.01, "learning_rate": 0.000285, "loss": 10.9348, "step": 114 }, { "epoch": 0.01, "learning_rate": 0.0002875, "loss": 8.9401, "step": 115 }, { "epoch": 0.01, "learning_rate": 0.00029, "loss": 8.6853, "step": 116 }, { "epoch": 0.01, "learning_rate": 0.0002925, "loss": 13.409, "step": 117 }, { "epoch": 0.01, "learning_rate": 0.000295, "loss": 9.0949, "step": 118 }, { "epoch": 0.01, "learning_rate": 0.00029749999999999997, "loss": 15.4847, "step": 119 }, { "epoch": 0.01, "learning_rate": 0.0003, "loss": 14.8908, "step": 120 }, { "epoch": 0.02, "learning_rate": 0.0003025, "loss": 13.2725, "step": 121 }, { "epoch": 0.02, "learning_rate": 0.000305, "loss": 12.6308, "step": 122 }, { "epoch": 0.02, "learning_rate": 0.0003075, "loss": 27.4928, "step": 123 }, { "epoch": 0.02, "learning_rate": 0.00031, "loss": 21.0197, "step": 124 }, { "epoch": 0.02, "learning_rate": 0.0003125, "loss": 22.7951, "step": 125 }, { "epoch": 0.02, "learning_rate": 0.000315, "loss": 20.3566, "step": 126 }, { "epoch": 0.02, "learning_rate": 0.0003175, "loss": 26.2958, "step": 127 }, { "epoch": 0.02, "learning_rate": 0.00032, "loss": 26.9744, "step": 128 }, { "epoch": 0.02, "learning_rate": 0.00032250000000000003, "loss": 44.1375, "step": 129 }, { "epoch": 0.02, "learning_rate": 0.00032500000000000004, "loss": 34.9571, "step": 130 }, { "epoch": 0.02, "learning_rate": 0.00032750000000000005, "loss": 38.8785, "step": 131 }, { "epoch": 0.02, "learning_rate": 0.00033, "loss": 85.0352, "step": 132 }, { "epoch": 0.02, "learning_rate": 0.0003325, "loss": 51.5333, "step": 133 }, { "epoch": 0.02, "learning_rate": 0.000335, "loss": 40.2932, "step": 134 }, { "epoch": 0.02, "learning_rate": 0.0003375, "loss": 36.5462, "step": 135 }, { "epoch": 0.02, "learning_rate": 0.00034, "loss": 75.8937, "step": 136 }, { "epoch": 0.02, "learning_rate": 0.00034250000000000003, "loss": 86.1037, "step": 137 }, { "epoch": 0.02, "learning_rate": 0.000345, "loss": 53.7456, "step": 138 }, { "epoch": 0.02, "learning_rate": 0.0003475, "loss": 83.0468, "step": 139 }, { "epoch": 0.02, "learning_rate": 0.00035, "loss": 110.7143, "step": 140 }, { "epoch": 0.02, "learning_rate": 0.0003525, "loss": 52.4007, "step": 141 }, { "epoch": 0.02, "learning_rate": 0.000355, "loss": 36.0668, "step": 142 }, { "epoch": 0.02, "learning_rate": 0.0003575, "loss": 22.3388, "step": 143 }, { "epoch": 0.02, "learning_rate": 0.00035999999999999997, "loss": 28.0379, "step": 144 }, { "epoch": 0.02, "learning_rate": 0.0003625, "loss": 43.6223, "step": 145 }, { "epoch": 0.02, "learning_rate": 0.000365, "loss": 55.8699, "step": 146 }, { "epoch": 0.02, "learning_rate": 0.0003675, "loss": 33.7208, "step": 147 }, { "epoch": 0.02, "learning_rate": 0.00037, "loss": 31.072, "step": 148 }, { "epoch": 0.02, "learning_rate": 0.0003725, "loss": 30.4092, "step": 149 }, { "epoch": 0.02, "learning_rate": 0.000375, "loss": 21.5141, "step": 150 }, { "epoch": 0.02, "learning_rate": 0.0003775, "loss": 30.8876, "step": 151 }, { "epoch": 0.02, "learning_rate": 0.00038, "loss": 83.116, "step": 152 }, { "epoch": 0.02, "learning_rate": 0.00038250000000000003, "loss": 51.4657, "step": 153 }, { "epoch": 0.02, "learning_rate": 0.00038500000000000003, "loss": 24.4031, "step": 154 }, { "epoch": 0.02, "learning_rate": 0.00038750000000000004, "loss": 26.7915, "step": 155 }, { "epoch": 0.02, "learning_rate": 0.00039000000000000005, "loss": 32.5022, "step": 156 }, { "epoch": 0.02, "learning_rate": 0.0003925, "loss": 42.6491, "step": 157 }, { "epoch": 0.02, "learning_rate": 0.000395, "loss": 20.6333, "step": 158 }, { "epoch": 0.02, "learning_rate": 0.0003975, "loss": 23.2271, "step": 159 }, { "epoch": 0.02, "learning_rate": 0.0004, "loss": 34.7075, "step": 160 }, { "epoch": 0.02, "learning_rate": 0.0004025, "loss": 30.369, "step": 161 }, { "epoch": 0.02, "learning_rate": 0.00040500000000000003, "loss": 71.6912, "step": 162 }, { "epoch": 0.02, "learning_rate": 0.0004075, "loss": 43.949, "step": 163 }, { "epoch": 0.02, "learning_rate": 0.00041, "loss": 61.2892, "step": 164 }, { "epoch": 0.02, "learning_rate": 0.0004125, "loss": 42.7592, "step": 165 }, { "epoch": 0.02, "learning_rate": 0.000415, "loss": 40.9293, "step": 166 }, { "epoch": 0.02, "learning_rate": 0.0004175, "loss": 44.9396, "step": 167 }, { "epoch": 0.02, "learning_rate": 0.00042, "loss": 56.4062, "step": 168 }, { "epoch": 0.02, "learning_rate": 0.00042249999999999997, "loss": 46.9524, "step": 169 }, { "epoch": 0.02, "learning_rate": 0.000425, "loss": 39.3676, "step": 170 }, { "epoch": 0.02, "learning_rate": 0.0004275, "loss": 41.0684, "step": 171 }, { "epoch": 0.02, "learning_rate": 0.00043, "loss": 54.7731, "step": 172 }, { "epoch": 0.02, "learning_rate": 0.0004325, "loss": 33.7425, "step": 173 }, { "epoch": 0.02, "learning_rate": 0.000435, "loss": 98.6551, "step": 174 }, { "epoch": 0.02, "learning_rate": 0.0004375, "loss": 144.1547, "step": 175 }, { "epoch": 0.02, "learning_rate": 0.00044, "loss": 84.7202, "step": 176 }, { "epoch": 0.02, "learning_rate": 0.0004425, "loss": 63.0181, "step": 177 }, { "epoch": 0.02, "learning_rate": 0.00044500000000000003, "loss": 43.0474, "step": 178 }, { "epoch": 0.02, "learning_rate": 0.00044750000000000004, "loss": 49.2202, "step": 179 }, { "epoch": 0.02, "learning_rate": 0.00045000000000000004, "loss": 38.8897, "step": 180 }, { "epoch": 0.02, "learning_rate": 0.00045250000000000005, "loss": 44.342, "step": 181 }, { "epoch": 0.02, "learning_rate": 0.000455, "loss": 51.8579, "step": 182 }, { "epoch": 0.02, "learning_rate": 0.0004575, "loss": 38.3193, "step": 183 }, { "epoch": 0.02, "learning_rate": 0.00046, "loss": 36.7328, "step": 184 }, { "epoch": 0.02, "learning_rate": 0.0004625, "loss": 41.7626, "step": 185 }, { "epoch": 0.02, "learning_rate": 0.000465, "loss": 55.2153, "step": 186 }, { "epoch": 0.02, "learning_rate": 0.00046750000000000003, "loss": 33.0427, "step": 187 }, { "epoch": 0.02, "learning_rate": 0.00047, "loss": 30.9342, "step": 188 }, { "epoch": 0.02, "learning_rate": 0.0004725, "loss": 32.2735, "step": 189 }, { "epoch": 0.02, "learning_rate": 0.000475, "loss": 28.561, "step": 190 }, { "epoch": 0.02, "learning_rate": 0.0004775, "loss": 40.5324, "step": 191 }, { "epoch": 0.02, "learning_rate": 0.00048, "loss": 29.7317, "step": 192 }, { "epoch": 0.02, "learning_rate": 0.0004825, "loss": 46.3316, "step": 193 }, { "epoch": 0.02, "learning_rate": 0.00048499999999999997, "loss": 27.4074, "step": 194 }, { "epoch": 0.02, "learning_rate": 0.0004875, "loss": 26.0797, "step": 195 }, { "epoch": 0.02, "learning_rate": 0.00049, "loss": 24.7339, "step": 196 }, { "epoch": 0.02, "learning_rate": 0.0004925, "loss": 36.57, "step": 197 }, { "epoch": 0.02, "learning_rate": 0.000495, "loss": 27.9487, "step": 198 }, { "epoch": 0.02, "learning_rate": 0.0004975, "loss": 27.6223, "step": 199 }, { "epoch": 0.03, "learning_rate": 0.0005, "loss": 26.5779, "step": 200 }, { "epoch": 0.03, "eval_acceptance_rate": 0.06711464375257492, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 0.8575532944856898, "eval_expected_tokens": 1.0719416181071124, "eval_loss": 23.701913833618164, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.017985404526778086, "eval_runtime": 50.2725, "eval_samples_per_second": 9.687, "eval_steps_per_second": 0.318, "step": 200 }, { "epoch": 0.03, "learning_rate": 0.0004999999797222134, "loss": 23.8953, "step": 201 }, { "epoch": 0.03, "learning_rate": 0.0004999999188888571, "loss": 22.5139, "step": 202 }, { "epoch": 0.03, "learning_rate": 0.0004999998174999408, "loss": 25.8961, "step": 203 }, { "epoch": 0.03, "learning_rate": 0.0004999996755554811, "loss": 19.1636, "step": 204 }, { "epoch": 0.03, "learning_rate": 0.0004999994930555008, "loss": 17.5353, "step": 205 }, { "epoch": 0.03, "learning_rate": 0.0004999992700000297, "loss": 25.6684, "step": 206 }, { "epoch": 0.03, "learning_rate": 0.000499999006389104, "loss": 18.4557, "step": 207 }, { "epoch": 0.03, "learning_rate": 0.0004999987022227664, "loss": 17.3078, "step": 208 }, { "epoch": 0.03, "learning_rate": 0.0004999983575010662, "loss": 17.8456, "step": 209 }, { "epoch": 0.03, "learning_rate": 0.0004999979722240593, "loss": 22.6902, "step": 210 }, { "epoch": 0.03, "learning_rate": 0.0004999975463918083, "loss": 29.1765, "step": 211 }, { "epoch": 0.03, "learning_rate": 0.0004999970800043822, "loss": 26.3337, "step": 212 }, { "epoch": 0.03, "learning_rate": 0.0004999965730618567, "loss": 19.5495, "step": 213 }, { "epoch": 0.03, "learning_rate": 0.0004999960255643141, "loss": 18.4896, "step": 214 }, { "epoch": 0.03, "learning_rate": 0.0004999954375118431, "loss": 25.6386, "step": 215 }, { "epoch": 0.03, "learning_rate": 0.0004999948089045392, "loss": 19.0871, "step": 216 }, { "epoch": 0.03, "learning_rate": 0.0004999941397425042, "loss": 26.4831, "step": 217 }, { "epoch": 0.03, "learning_rate": 0.0004999934300258469, "loss": 16.7648, "step": 218 }, { "epoch": 0.03, "learning_rate": 0.0004999926797546822, "loss": 15.9966, "step": 219 }, { "epoch": 0.03, "learning_rate": 0.000499991888929132, "loss": 17.5632, "step": 220 }, { "epoch": 0.03, "learning_rate": 0.0004999910575493245, "loss": 19.0823, "step": 221 }, { "epoch": 0.03, "learning_rate": 0.0004999901856153945, "loss": 27.0615, "step": 222 }, { "epoch": 0.03, "learning_rate": 0.0004999892731274836, "loss": 33.3205, "step": 223 }, { "epoch": 0.03, "learning_rate": 0.0004999883200857396, "loss": 35.3689, "step": 224 }, { "epoch": 0.03, "learning_rate": 0.0004999873264903174, "loss": 20.7511, "step": 225 }, { "epoch": 0.03, "learning_rate": 0.000499986292341378, "loss": 22.3797, "step": 226 }, { "epoch": 0.03, "learning_rate": 0.0004999852176390893, "loss": 30.7062, "step": 227 }, { "epoch": 0.03, "learning_rate": 0.0004999841023836253, "loss": 31.6417, "step": 228 }, { "epoch": 0.03, "learning_rate": 0.0004999829465751674, "loss": 27.0171, "step": 229 }, { "epoch": 0.03, "learning_rate": 0.0004999817502139026, "loss": 22.0755, "step": 230 }, { "epoch": 0.03, "learning_rate": 0.0004999805133000254, "loss": 29.2434, "step": 231 }, { "epoch": 0.03, "learning_rate": 0.0004999792358337362, "loss": 20.3132, "step": 232 }, { "epoch": 0.03, "learning_rate": 0.0004999779178152425, "loss": 18.2305, "step": 233 }, { "epoch": 0.03, "learning_rate": 0.0004999765592447578, "loss": 21.9615, "step": 234 }, { "epoch": 0.03, "learning_rate": 0.0004999751601225027, "loss": 16.277, "step": 235 }, { "epoch": 0.03, "learning_rate": 0.000499973720448704, "loss": 23.2164, "step": 236 }, { "epoch": 0.03, "learning_rate": 0.0004999722402235953, "loss": 20.3794, "step": 237 }, { "epoch": 0.03, "learning_rate": 0.0004999707194474168, "loss": 19.7031, "step": 238 }, { "epoch": 0.03, "learning_rate": 0.0004999691581204152, "loss": 15.925, "step": 239 }, { "epoch": 0.03, "learning_rate": 0.0004999675562428437, "loss": 13.9604, "step": 240 }, { "epoch": 0.03, "learning_rate": 0.0004999659138149623, "loss": 15.5548, "step": 241 }, { "epoch": 0.03, "learning_rate": 0.0004999642308370372, "loss": 15.5763, "step": 242 }, { "epoch": 0.03, "learning_rate": 0.0004999625073093416, "loss": 20.7866, "step": 243 }, { "epoch": 0.03, "learning_rate": 0.0004999607432321551, "loss": 17.608, "step": 244 }, { "epoch": 0.03, "learning_rate": 0.000499958938605764, "loss": 16.5295, "step": 245 }, { "epoch": 0.03, "learning_rate": 0.0004999570934304606, "loss": 17.1243, "step": 246 }, { "epoch": 0.03, "learning_rate": 0.0004999552077065447, "loss": 18.822, "step": 247 }, { "epoch": 0.03, "learning_rate": 0.0004999532814343218, "loss": 14.7749, "step": 248 }, { "epoch": 0.03, "learning_rate": 0.0004999513146141048, "loss": 13.8481, "step": 249 }, { "epoch": 0.03, "learning_rate": 0.0004999493072462126, "loss": 16.4543, "step": 250 }, { "epoch": 0.03, "learning_rate": 0.0004999472593309706, "loss": 12.4118, "step": 251 }, { "epoch": 0.03, "learning_rate": 0.0004999451708687113, "loss": 13.0785, "step": 252 }, { "epoch": 0.03, "learning_rate": 0.0004999430418597735, "loss": 12.9176, "step": 253 }, { "epoch": 0.03, "learning_rate": 0.0004999408723045024, "loss": 11.5831, "step": 254 }, { "epoch": 0.03, "learning_rate": 0.0004999386622032501, "loss": 11.2252, "step": 255 }, { "epoch": 0.03, "learning_rate": 0.000499936411556375, "loss": 11.2441, "step": 256 }, { "epoch": 0.03, "learning_rate": 0.0004999341203642424, "loss": 12.2222, "step": 257 }, { "epoch": 0.03, "learning_rate": 0.0004999317886272238, "loss": 11.2427, "step": 258 }, { "epoch": 0.03, "learning_rate": 0.0004999294163456975, "loss": 10.8432, "step": 259 }, { "epoch": 0.03, "learning_rate": 0.0004999270035200482, "loss": 13.0501, "step": 260 }, { "epoch": 0.03, "learning_rate": 0.0004999245501506676, "loss": 10.902, "step": 261 }, { "epoch": 0.03, "learning_rate": 0.0004999220562379536, "loss": 10.6306, "step": 262 }, { "epoch": 0.03, "learning_rate": 0.0004999195217823106, "loss": 10.6132, "step": 263 }, { "epoch": 0.03, "learning_rate": 0.00049991694678415, "loss": 11.1806, "step": 264 }, { "epoch": 0.03, "learning_rate": 0.0004999143312438893, "loss": 14.698, "step": 265 }, { "epoch": 0.03, "learning_rate": 0.000499911675161953, "loss": 17.9601, "step": 266 }, { "epoch": 0.03, "learning_rate": 0.0004999089785387717, "loss": 11.6929, "step": 267 }, { "epoch": 0.03, "learning_rate": 0.0004999062413747832, "loss": 11.6784, "step": 268 }, { "epoch": 0.03, "learning_rate": 0.0004999034636704312, "loss": 13.5016, "step": 269 }, { "epoch": 0.03, "learning_rate": 0.0004999006454261666, "loss": 11.8914, "step": 270 }, { "epoch": 0.03, "learning_rate": 0.0004998977866424462, "loss": 13.722, "step": 271 }, { "epoch": 0.03, "learning_rate": 0.0004998948873197343, "loss": 12.489, "step": 272 }, { "epoch": 0.03, "learning_rate": 0.0004998919474585007, "loss": 11.3088, "step": 273 }, { "epoch": 0.03, "learning_rate": 0.0004998889670592225, "loss": 11.2748, "step": 274 }, { "epoch": 0.03, "learning_rate": 0.0004998859461223834, "loss": 10.0905, "step": 275 }, { "epoch": 0.03, "learning_rate": 0.0004998828846484732, "loss": 9.7439, "step": 276 }, { "epoch": 0.03, "learning_rate": 0.0004998797826379887, "loss": 9.8809, "step": 277 }, { "epoch": 0.03, "learning_rate": 0.0004998766400914329, "loss": 9.9973, "step": 278 }, { "epoch": 0.03, "learning_rate": 0.0004998734570093159, "loss": 11.0598, "step": 279 }, { "epoch": 0.04, "learning_rate": 0.0004998702333921537, "loss": 9.4643, "step": 280 }, { "epoch": 0.04, "learning_rate": 0.0004998669692404696, "loss": 9.3287, "step": 281 }, { "epoch": 0.04, "learning_rate": 0.0004998636645547929, "loss": 9.5712, "step": 282 }, { "epoch": 0.04, "learning_rate": 0.0004998603193356599, "loss": 8.9837, "step": 283 }, { "epoch": 0.04, "learning_rate": 0.000499856933583613, "loss": 8.9764, "step": 284 }, { "epoch": 0.04, "learning_rate": 0.0004998535072992015, "loss": 8.5344, "step": 285 }, { "epoch": 0.04, "learning_rate": 0.0004998500404829815, "loss": 8.2698, "step": 286 }, { "epoch": 0.04, "learning_rate": 0.000499846533135515, "loss": 8.4141, "step": 287 }, { "epoch": 0.04, "learning_rate": 0.0004998429852573711, "loss": 8.1562, "step": 288 }, { "epoch": 0.04, "learning_rate": 0.0004998393968491256, "loss": 7.8572, "step": 289 }, { "epoch": 0.04, "learning_rate": 0.0004998357679113603, "loss": 7.9894, "step": 290 }, { "epoch": 0.04, "learning_rate": 0.0004998320984446641, "loss": 8.0055, "step": 291 }, { "epoch": 0.04, "learning_rate": 0.0004998283884496321, "loss": 7.6254, "step": 292 }, { "epoch": 0.04, "learning_rate": 0.0004998246379268662, "loss": 7.3884, "step": 293 }, { "epoch": 0.04, "learning_rate": 0.0004998208468769749, "loss": 7.4248, "step": 294 }, { "epoch": 0.04, "learning_rate": 0.0004998170153005733, "loss": 7.1405, "step": 295 }, { "epoch": 0.04, "learning_rate": 0.0004998131431982825, "loss": 7.752, "step": 296 }, { "epoch": 0.04, "learning_rate": 0.0004998092305707312, "loss": 7.4073, "step": 297 }, { "epoch": 0.04, "learning_rate": 0.0004998052774185537, "loss": 7.1519, "step": 298 }, { "epoch": 0.04, "learning_rate": 0.0004998012837423916, "loss": 6.997, "step": 299 }, { "epoch": 0.04, "learning_rate": 0.0004997972495428924, "loss": 8.5223, "step": 300 }, { "epoch": 0.04, "eval_acceptance_rate": 0.20126701891422272, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.0012554966708689, "eval_expected_tokens": 1.2515693708385862, "eval_loss": 6.866159915924072, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.06289234270964655, "eval_runtime": 50.2774, "eval_samples_per_second": 9.686, "eval_steps_per_second": 0.318, "step": 300 }, { "epoch": 0.04, "learning_rate": 0.0004997931748207109, "loss": 6.9558, "step": 301 }, { "epoch": 0.04, "learning_rate": 0.0004997890595765078, "loss": 6.6138, "step": 302 }, { "epoch": 0.04, "learning_rate": 0.0004997849038109511, "loss": 8.0751, "step": 303 }, { "epoch": 0.04, "learning_rate": 0.0004997807075247146, "loss": 7.8057, "step": 304 }, { "epoch": 0.04, "learning_rate": 0.0004997764707184791, "loss": 7.6458, "step": 305 }, { "epoch": 0.04, "learning_rate": 0.000499772193392932, "loss": 8.1545, "step": 306 }, { "epoch": 0.04, "learning_rate": 0.0004997678755487671, "loss": 8.357, "step": 307 }, { "epoch": 0.04, "learning_rate": 0.0004997635171866848, "loss": 7.0666, "step": 308 }, { "epoch": 0.04, "learning_rate": 0.0004997591183073923, "loss": 6.9085, "step": 309 }, { "epoch": 0.04, "learning_rate": 0.000499754678911603, "loss": 7.4728, "step": 310 }, { "epoch": 0.04, "learning_rate": 0.0004997501990000371, "loss": 6.6544, "step": 311 }, { "epoch": 0.04, "learning_rate": 0.0004997456785734215, "loss": 7.0804, "step": 312 }, { "epoch": 0.04, "learning_rate": 0.0004997411176324895, "loss": 7.4997, "step": 313 }, { "epoch": 0.04, "learning_rate": 0.0004997365161779808, "loss": 7.1559, "step": 314 }, { "epoch": 0.04, "learning_rate": 0.0004997318742106419, "loss": 6.892, "step": 315 }, { "epoch": 0.04, "learning_rate": 0.0004997271917312259, "loss": 8.1523, "step": 316 }, { "epoch": 0.04, "learning_rate": 0.0004997224687404926, "loss": 6.8491, "step": 317 }, { "epoch": 0.04, "learning_rate": 0.0004997177052392076, "loss": 7.4137, "step": 318 }, { "epoch": 0.04, "learning_rate": 0.0004997129012281442, "loss": 6.5662, "step": 319 }, { "epoch": 0.04, "learning_rate": 0.0004997080567080817, "loss": 6.7059, "step": 320 }, { "epoch": 0.04, "learning_rate": 0.0004997031716798056, "loss": 6.656, "step": 321 }, { "epoch": 0.04, "learning_rate": 0.0004996982461441086, "loss": 7.3175, "step": 322 }, { "epoch": 0.04, "learning_rate": 0.0004996932801017898, "loss": 7.1065, "step": 323 }, { "epoch": 0.04, "learning_rate": 0.0004996882735536546, "loss": 6.473, "step": 324 }, { "epoch": 0.04, "learning_rate": 0.0004996832265005154, "loss": 6.7444, "step": 325 }, { "epoch": 0.04, "learning_rate": 0.0004996781389431908, "loss": 6.6422, "step": 326 }, { "epoch": 0.04, "learning_rate": 0.0004996730108825061, "loss": 6.4991, "step": 327 }, { "epoch": 0.04, "learning_rate": 0.0004996678423192933, "loss": 6.4407, "step": 328 }, { "epoch": 0.04, "learning_rate": 0.0004996626332543907, "loss": 6.3121, "step": 329 }, { "epoch": 0.04, "learning_rate": 0.0004996573836886434, "loss": 6.3483, "step": 330 }, { "epoch": 0.04, "learning_rate": 0.0004996520936229031, "loss": 6.0121, "step": 331 }, { "epoch": 0.04, "learning_rate": 0.000499646763058028, "loss": 6.2077, "step": 332 }, { "epoch": 0.04, "learning_rate": 0.0004996413919948824, "loss": 6.2116, "step": 333 }, { "epoch": 0.04, "learning_rate": 0.0004996359804343381, "loss": 6.0766, "step": 334 }, { "epoch": 0.04, "learning_rate": 0.0004996305283772729, "loss": 6.2542, "step": 335 }, { "epoch": 0.04, "learning_rate": 0.000499625035824571, "loss": 5.9732, "step": 336 }, { "epoch": 0.04, "learning_rate": 0.0004996195027771235, "loss": 5.8911, "step": 337 }, { "epoch": 0.04, "learning_rate": 0.0004996139292358281, "loss": 5.9175, "step": 338 }, { "epoch": 0.04, "learning_rate": 0.000499608315201589, "loss": 5.8257, "step": 339 }, { "epoch": 0.04, "learning_rate": 0.0004996026606753167, "loss": 5.7552, "step": 340 }, { "epoch": 0.04, "learning_rate": 0.0004995969656579287, "loss": 5.5939, "step": 341 }, { "epoch": 0.04, "learning_rate": 0.0004995912301503486, "loss": 5.8897, "step": 342 }, { "epoch": 0.04, "learning_rate": 0.0004995854541535071, "loss": 5.83, "step": 343 }, { "epoch": 0.04, "learning_rate": 0.000499579637668341, "loss": 5.6405, "step": 344 }, { "epoch": 0.04, "learning_rate": 0.000499573780695794, "loss": 5.8075, "step": 345 }, { "epoch": 0.04, "learning_rate": 0.0004995678832368163, "loss": 5.3335, "step": 346 }, { "epoch": 0.04, "learning_rate": 0.0004995619452923644, "loss": 5.4934, "step": 347 }, { "epoch": 0.04, "learning_rate": 0.0004995559668634016, "loss": 5.5131, "step": 348 }, { "epoch": 0.04, "learning_rate": 0.0004995499479508979, "loss": 5.2661, "step": 349 }, { "epoch": 0.04, "learning_rate": 0.0004995438885558294, "loss": 5.4474, "step": 350 }, { "epoch": 0.04, "learning_rate": 0.0004995377886791793, "loss": 5.4994, "step": 351 }, { "epoch": 0.04, "learning_rate": 0.0004995316483219371, "loss": 5.4208, "step": 352 }, { "epoch": 0.04, "learning_rate": 0.0004995254674850989, "loss": 5.4833, "step": 353 }, { "epoch": 0.04, "learning_rate": 0.0004995192461696675, "loss": 5.2662, "step": 354 }, { "epoch": 0.04, "learning_rate": 0.0004995129843766518, "loss": 5.5907, "step": 355 }, { "epoch": 0.04, "learning_rate": 0.0004995066821070679, "loss": 5.5164, "step": 356 }, { "epoch": 0.04, "learning_rate": 0.000499500339361938, "loss": 5.2688, "step": 357 }, { "epoch": 0.04, "learning_rate": 0.0004994939561422912, "loss": 5.201, "step": 358 }, { "epoch": 0.04, "learning_rate": 0.0004994875324491628, "loss": 5.3675, "step": 359 }, { "epoch": 0.04, "learning_rate": 0.0004994810682835951, "loss": 5.2587, "step": 360 }, { "epoch": 0.05, "learning_rate": 0.0004994745636466365, "loss": 5.1196, "step": 361 }, { "epoch": 0.05, "learning_rate": 0.0004994680185393424, "loss": 4.8852, "step": 362 }, { "epoch": 0.05, "learning_rate": 0.0004994614329627742, "loss": 5.0357, "step": 363 }, { "epoch": 0.05, "learning_rate": 0.0004994548069180007, "loss": 4.9316, "step": 364 }, { "epoch": 0.05, "learning_rate": 0.0004994481404060966, "loss": 4.8047, "step": 365 }, { "epoch": 0.05, "learning_rate": 0.0004994414334281432, "loss": 4.8066, "step": 366 }, { "epoch": 0.05, "learning_rate": 0.0004994346859852288, "loss": 4.754, "step": 367 }, { "epoch": 0.05, "learning_rate": 0.0004994278980784478, "loss": 5.0793, "step": 368 }, { "epoch": 0.05, "learning_rate": 0.0004994210697089013, "loss": 4.9239, "step": 369 }, { "epoch": 0.05, "learning_rate": 0.0004994142008776972, "loss": 5.0001, "step": 370 }, { "epoch": 0.05, "learning_rate": 0.0004994072915859496, "loss": 4.7425, "step": 371 }, { "epoch": 0.05, "learning_rate": 0.0004994003418347796, "loss": 4.7094, "step": 372 }, { "epoch": 0.05, "learning_rate": 0.0004993933516253143, "loss": 4.6378, "step": 373 }, { "epoch": 0.05, "learning_rate": 0.0004993863209586878, "loss": 4.7203, "step": 374 }, { "epoch": 0.05, "learning_rate": 0.0004993792498360406, "loss": 4.9021, "step": 375 }, { "epoch": 0.05, "learning_rate": 0.0004993721382585199, "loss": 4.8228, "step": 376 }, { "epoch": 0.05, "learning_rate": 0.0004993649862272793, "loss": 4.7177, "step": 377 }, { "epoch": 0.05, "learning_rate": 0.0004993577937434789, "loss": 4.5112, "step": 378 }, { "epoch": 0.05, "learning_rate": 0.0004993505608082856, "loss": 4.5726, "step": 379 }, { "epoch": 0.05, "learning_rate": 0.0004993432874228728, "loss": 4.6325, "step": 380 }, { "epoch": 0.05, "learning_rate": 0.0004993359735884202, "loss": 4.6898, "step": 381 }, { "epoch": 0.05, "learning_rate": 0.0004993286193061145, "loss": 4.5298, "step": 382 }, { "epoch": 0.05, "learning_rate": 0.0004993212245771486, "loss": 4.5357, "step": 383 }, { "epoch": 0.05, "learning_rate": 0.000499313789402722, "loss": 4.4383, "step": 384 }, { "epoch": 0.05, "learning_rate": 0.0004993063137840412, "loss": 4.4729, "step": 385 }, { "epoch": 0.05, "learning_rate": 0.0004992987977223186, "loss": 4.515, "step": 386 }, { "epoch": 0.05, "learning_rate": 0.0004992912412187735, "loss": 4.4767, "step": 387 }, { "epoch": 0.05, "learning_rate": 0.0004992836442746317, "loss": 4.4115, "step": 388 }, { "epoch": 0.05, "learning_rate": 0.0004992760068911258, "loss": 4.3689, "step": 389 }, { "epoch": 0.05, "learning_rate": 0.0004992683290694946, "loss": 4.3043, "step": 390 }, { "epoch": 0.05, "learning_rate": 0.0004992606108109836, "loss": 4.3722, "step": 391 }, { "epoch": 0.05, "learning_rate": 0.0004992528521168449, "loss": 4.4702, "step": 392 }, { "epoch": 0.05, "learning_rate": 0.0004992450529883372, "loss": 4.3436, "step": 393 }, { "epoch": 0.05, "learning_rate": 0.0004992372134267257, "loss": 4.3791, "step": 394 }, { "epoch": 0.05, "learning_rate": 0.000499229333433282, "loss": 4.3022, "step": 395 }, { "epoch": 0.05, "learning_rate": 0.0004992214130092845, "loss": 4.2811, "step": 396 }, { "epoch": 0.05, "learning_rate": 0.0004992134521560181, "loss": 4.4499, "step": 397 }, { "epoch": 0.05, "learning_rate": 0.0004992054508747742, "loss": 4.2457, "step": 398 }, { "epoch": 0.05, "learning_rate": 0.0004991974091668508, "loss": 4.2514, "step": 399 }, { "epoch": 0.05, "learning_rate": 0.0004991893270335525, "loss": 4.1986, "step": 400 }, { "epoch": 0.05, "eval_acceptance_rate": 0.30389800667762756, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.1462779697710423, "eval_expected_tokens": 1.432847462213803, "eval_loss": 4.307589054107666, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.10821186555345075, "eval_runtime": 50.3354, "eval_samples_per_second": 9.675, "eval_steps_per_second": 0.318, "step": 400 }, { "epoch": 0.05, "learning_rate": 0.0004991812044761903, "loss": 4.3158, "step": 401 }, { "epoch": 0.05, "learning_rate": 0.0004991730414960819, "loss": 4.2263, "step": 402 }, { "epoch": 0.05, "learning_rate": 0.0004991648380945514, "loss": 4.4214, "step": 403 }, { "epoch": 0.05, "learning_rate": 0.0004991565942729299, "loss": 4.3698, "step": 404 }, { "epoch": 0.05, "learning_rate": 0.0004991483100325542, "loss": 4.2459, "step": 405 }, { "epoch": 0.05, "learning_rate": 0.0004991399853747688, "loss": 4.3668, "step": 406 }, { "epoch": 0.05, "learning_rate": 0.0004991316203009238, "loss": 4.2012, "step": 407 }, { "epoch": 0.05, "learning_rate": 0.0004991232148123761, "loss": 4.272, "step": 408 }, { "epoch": 0.05, "learning_rate": 0.0004991147689104895, "loss": 4.2226, "step": 409 }, { "epoch": 0.05, "learning_rate": 0.0004991062825966341, "loss": 4.0848, "step": 410 }, { "epoch": 0.05, "learning_rate": 0.0004990977558721863, "loss": 4.2749, "step": 411 }, { "epoch": 0.05, "learning_rate": 0.0004990891887385297, "loss": 4.0998, "step": 412 }, { "epoch": 0.05, "learning_rate": 0.0004990805811970539, "loss": 4.1286, "step": 413 }, { "epoch": 0.05, "learning_rate": 0.0004990719332491552, "loss": 4.1383, "step": 414 }, { "epoch": 0.05, "learning_rate": 0.0004990632448962364, "loss": 4.3261, "step": 415 }, { "epoch": 0.05, "learning_rate": 0.0004990545161397072, "loss": 4.1462, "step": 416 }, { "epoch": 0.05, "learning_rate": 0.0004990457469809834, "loss": 3.8712, "step": 417 }, { "epoch": 0.05, "learning_rate": 0.0004990369374214877, "loss": 4.1626, "step": 418 }, { "epoch": 0.05, "learning_rate": 0.0004990280874626491, "loss": 3.9395, "step": 419 }, { "epoch": 0.05, "learning_rate": 0.0004990191971059033, "loss": 4.0896, "step": 420 }, { "epoch": 0.05, "learning_rate": 0.0004990102663526924, "loss": 3.947, "step": 421 }, { "epoch": 0.05, "learning_rate": 0.0004990012952044654, "loss": 4.0189, "step": 422 }, { "epoch": 0.05, "learning_rate": 0.0004989922836626773, "loss": 4.1789, "step": 423 }, { "epoch": 0.05, "learning_rate": 0.0004989832317287904, "loss": 3.9781, "step": 424 }, { "epoch": 0.05, "learning_rate": 0.0004989741394042727, "loss": 3.8718, "step": 425 }, { "epoch": 0.05, "learning_rate": 0.0004989650066905995, "loss": 4.0818, "step": 426 }, { "epoch": 0.05, "learning_rate": 0.0004989558335892521, "loss": 3.869, "step": 427 }, { "epoch": 0.05, "learning_rate": 0.0004989466201017188, "loss": 3.8314, "step": 428 }, { "epoch": 0.05, "learning_rate": 0.000498937366229494, "loss": 3.8664, "step": 429 }, { "epoch": 0.05, "learning_rate": 0.000498928071974079, "loss": 3.7923, "step": 430 }, { "epoch": 0.05, "learning_rate": 0.0004989187373369816, "loss": 3.9447, "step": 431 }, { "epoch": 0.05, "learning_rate": 0.000498909362319716, "loss": 3.7237, "step": 432 }, { "epoch": 0.05, "learning_rate": 0.000498899946923803, "loss": 3.7731, "step": 433 }, { "epoch": 0.05, "learning_rate": 0.00049889049115077, "loss": 3.9019, "step": 434 }, { "epoch": 0.05, "learning_rate": 0.000498880995002151, "loss": 3.9955, "step": 435 }, { "epoch": 0.05, "learning_rate": 0.0004988714584794866, "loss": 3.6386, "step": 436 }, { "epoch": 0.05, "learning_rate": 0.0004988618815843235, "loss": 3.7553, "step": 437 }, { "epoch": 0.05, "learning_rate": 0.0004988522643182156, "loss": 3.6825, "step": 438 }, { "epoch": 0.05, "learning_rate": 0.0004988426066827229, "loss": 3.7576, "step": 439 }, { "epoch": 0.06, "learning_rate": 0.0004988329086794121, "loss": 3.7945, "step": 440 }, { "epoch": 0.06, "learning_rate": 0.0004988231703098565, "loss": 3.8234, "step": 441 }, { "epoch": 0.06, "learning_rate": 0.0004988133915756358, "loss": 3.7815, "step": 442 }, { "epoch": 0.06, "learning_rate": 0.0004988035724783362, "loss": 3.8596, "step": 443 }, { "epoch": 0.06, "learning_rate": 0.0004987937130195509, "loss": 3.9209, "step": 444 }, { "epoch": 0.06, "learning_rate": 0.0004987838132008791, "loss": 4.159, "step": 445 }, { "epoch": 0.06, "learning_rate": 0.0004987738730239269, "loss": 3.8071, "step": 446 }, { "epoch": 0.06, "learning_rate": 0.0004987638924903066, "loss": 3.7721, "step": 447 }, { "epoch": 0.06, "learning_rate": 0.0004987538716016374, "loss": 3.6595, "step": 448 }, { "epoch": 0.06, "learning_rate": 0.0004987438103595449, "loss": 3.8838, "step": 449 }, { "epoch": 0.06, "learning_rate": 0.0004987337087656614, "loss": 3.7296, "step": 450 }, { "epoch": 0.06, "learning_rate": 0.0004987235668216254, "loss": 3.6496, "step": 451 }, { "epoch": 0.06, "learning_rate": 0.0004987133845290822, "loss": 3.6354, "step": 452 }, { "epoch": 0.06, "learning_rate": 0.0004987031618896837, "loss": 3.7368, "step": 453 }, { "epoch": 0.06, "learning_rate": 0.0004986928989050881, "loss": 3.5884, "step": 454 }, { "epoch": 0.06, "learning_rate": 0.0004986825955769604, "loss": 3.868, "step": 455 }, { "epoch": 0.06, "learning_rate": 0.0004986722519069719, "loss": 3.6925, "step": 456 }, { "epoch": 0.06, "learning_rate": 0.0004986618678968007, "loss": 3.5753, "step": 457 }, { "epoch": 0.06, "learning_rate": 0.0004986514435481312, "loss": 3.5262, "step": 458 }, { "epoch": 0.06, "learning_rate": 0.0004986409788626546, "loss": 3.563, "step": 459 }, { "epoch": 0.06, "learning_rate": 0.0004986304738420684, "loss": 3.7193, "step": 460 }, { "epoch": 0.06, "learning_rate": 0.0004986199284880767, "loss": 3.7125, "step": 461 }, { "epoch": 0.06, "learning_rate": 0.0004986093428023903, "loss": 3.5574, "step": 462 }, { "epoch": 0.06, "learning_rate": 0.0004985987167867265, "loss": 3.5798, "step": 463 }, { "epoch": 0.06, "learning_rate": 0.000498588050442809, "loss": 3.5445, "step": 464 }, { "epoch": 0.06, "learning_rate": 0.0004985773437723679, "loss": 3.6113, "step": 465 }, { "epoch": 0.06, "learning_rate": 0.0004985665967771405, "loss": 3.8043, "step": 466 }, { "epoch": 0.06, "learning_rate": 0.0004985558094588699, "loss": 3.6768, "step": 467 }, { "epoch": 0.06, "learning_rate": 0.000498544981819306, "loss": 3.4812, "step": 468 }, { "epoch": 0.06, "learning_rate": 0.0004985341138602056, "loss": 3.6944, "step": 469 }, { "epoch": 0.06, "learning_rate": 0.0004985232055833313, "loss": 3.5295, "step": 470 }, { "epoch": 0.06, "learning_rate": 0.000498512256990453, "loss": 3.8204, "step": 471 }, { "epoch": 0.06, "learning_rate": 0.0004985012680833466, "loss": 3.5764, "step": 472 }, { "epoch": 0.06, "learning_rate": 0.0004984902388637949, "loss": 3.3971, "step": 473 }, { "epoch": 0.06, "learning_rate": 0.0004984791693335871, "loss": 3.546, "step": 474 }, { "epoch": 0.06, "learning_rate": 0.0004984680594945186, "loss": 3.5567, "step": 475 }, { "epoch": 0.06, "learning_rate": 0.0004984569093483921, "loss": 3.5542, "step": 476 }, { "epoch": 0.06, "learning_rate": 0.0004984457188970163, "loss": 3.5067, "step": 477 }, { "epoch": 0.06, "learning_rate": 0.0004984344881422063, "loss": 3.6667, "step": 478 }, { "epoch": 0.06, "learning_rate": 0.0004984232170857841, "loss": 3.398, "step": 479 }, { "epoch": 0.06, "learning_rate": 0.0004984119057295783, "loss": 3.3549, "step": 480 }, { "epoch": 0.06, "learning_rate": 0.0004984005540754235, "loss": 3.3134, "step": 481 }, { "epoch": 0.06, "learning_rate": 0.0004983891621251616, "loss": 3.4443, "step": 482 }, { "epoch": 0.06, "learning_rate": 0.0004983777298806402, "loss": 3.4267, "step": 483 }, { "epoch": 0.06, "learning_rate": 0.0004983662573437144, "loss": 3.4911, "step": 484 }, { "epoch": 0.06, "learning_rate": 0.0004983547445162447, "loss": 3.4347, "step": 485 }, { "epoch": 0.06, "learning_rate": 0.000498343191400099, "loss": 3.3699, "step": 486 }, { "epoch": 0.06, "learning_rate": 0.0004983315979971518, "loss": 3.4394, "step": 487 }, { "epoch": 0.06, "learning_rate": 0.0004983199643092832, "loss": 3.3633, "step": 488 }, { "epoch": 0.06, "learning_rate": 0.000498308290338381, "loss": 3.3491, "step": 489 }, { "epoch": 0.06, "learning_rate": 0.0004982965760863385, "loss": 3.2727, "step": 490 }, { "epoch": 0.06, "learning_rate": 0.0004982848215550564, "loss": 3.3203, "step": 491 }, { "epoch": 0.06, "learning_rate": 0.0004982730267464412, "loss": 3.3926, "step": 492 }, { "epoch": 0.06, "learning_rate": 0.0004982611916624067, "loss": 3.3463, "step": 493 }, { "epoch": 0.06, "learning_rate": 0.0004982493163048725, "loss": 3.4808, "step": 494 }, { "epoch": 0.06, "learning_rate": 0.000498237400675765, "loss": 3.4102, "step": 495 }, { "epoch": 0.06, "learning_rate": 0.0004982254447770174, "loss": 3.5078, "step": 496 }, { "epoch": 0.06, "learning_rate": 0.0004982134486105692, "loss": 3.423, "step": 497 }, { "epoch": 0.06, "learning_rate": 0.0004982014121783663, "loss": 3.2627, "step": 498 }, { "epoch": 0.06, "learning_rate": 0.0004981893354823614, "loss": 3.354, "step": 499 }, { "epoch": 0.06, "learning_rate": 0.0004981772185245135, "loss": 3.2188, "step": 500 }, { "epoch": 0.06, "eval_acceptance_rate": 0.41847267746925354, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.3580332269673243, "eval_expected_tokens": 1.6975415337091553, "eval_loss": 3.2558422088623047, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.17438538342728882, "eval_runtime": 50.3025, "eval_samples_per_second": 9.681, "eval_steps_per_second": 0.318, "step": 500 }, { "epoch": 0.06, "learning_rate": 0.0004981650613067883, "loss": 3.2548, "step": 501 }, { "epoch": 0.06, "learning_rate": 0.0004981528638311579, "loss": 3.4537, "step": 502 }, { "epoch": 0.06, "learning_rate": 0.0004981406260996013, "loss": 3.449, "step": 503 }, { "epoch": 0.06, "learning_rate": 0.0004981283481141033, "loss": 3.3523, "step": 504 }, { "epoch": 0.06, "learning_rate": 0.0004981160298766559, "loss": 3.2196, "step": 505 }, { "epoch": 0.06, "learning_rate": 0.0004981036713892574, "loss": 3.3392, "step": 506 }, { "epoch": 0.06, "learning_rate": 0.0004980912726539127, "loss": 3.159, "step": 507 }, { "epoch": 0.06, "learning_rate": 0.0004980788336726328, "loss": 3.3043, "step": 508 }, { "epoch": 0.06, "learning_rate": 0.0004980663544474359, "loss": 3.3585, "step": 509 }, { "epoch": 0.06, "learning_rate": 0.0004980538349803463, "loss": 3.3334, "step": 510 }, { "epoch": 0.06, "learning_rate": 0.0004980412752733951, "loss": 3.2397, "step": 511 }, { "epoch": 0.06, "learning_rate": 0.0004980286753286195, "loss": 3.1741, "step": 512 }, { "epoch": 0.06, "learning_rate": 0.0004980160351480636, "loss": 3.1555, "step": 513 }, { "epoch": 0.06, "learning_rate": 0.000498003354733778, "loss": 3.1663, "step": 514 }, { "epoch": 0.06, "learning_rate": 0.0004979906340878196, "loss": 3.2551, "step": 515 }, { "epoch": 0.06, "learning_rate": 0.0004979778732122521, "loss": 3.2959, "step": 516 }, { "epoch": 0.06, "learning_rate": 0.0004979650721091455, "loss": 3.3525, "step": 517 }, { "epoch": 0.06, "learning_rate": 0.0004979522307805766, "loss": 3.2146, "step": 518 }, { "epoch": 0.06, "learning_rate": 0.0004979393492286284, "loss": 3.1586, "step": 519 }, { "epoch": 0.07, "learning_rate": 0.0004979264274553905, "loss": 3.1547, "step": 520 }, { "epoch": 0.07, "learning_rate": 0.0004979134654629592, "loss": 3.2544, "step": 521 }, { "epoch": 0.07, "learning_rate": 0.0004979004632534374, "loss": 3.2021, "step": 522 }, { "epoch": 0.07, "learning_rate": 0.000497887420828934, "loss": 3.1441, "step": 523 }, { "epoch": 0.07, "learning_rate": 0.000497874338191565, "loss": 3.1916, "step": 524 }, { "epoch": 0.07, "learning_rate": 0.0004978612153434526, "loss": 3.1986, "step": 525 }, { "epoch": 0.07, "learning_rate": 0.0004978480522867257, "loss": 3.207, "step": 526 }, { "epoch": 0.07, "learning_rate": 0.0004978348490235196, "loss": 3.0928, "step": 527 }, { "epoch": 0.07, "learning_rate": 0.0004978216055559761, "loss": 3.0325, "step": 528 }, { "epoch": 0.07, "learning_rate": 0.0004978083218862438, "loss": 3.1976, "step": 529 }, { "epoch": 0.07, "learning_rate": 0.0004977949980164773, "loss": 3.229, "step": 530 }, { "epoch": 0.07, "learning_rate": 0.0004977816339488383, "loss": 3.1409, "step": 531 }, { "epoch": 0.07, "learning_rate": 0.0004977682296854945, "loss": 3.184, "step": 532 }, { "epoch": 0.07, "learning_rate": 0.0004977547852286206, "loss": 3.332, "step": 533 }, { "epoch": 0.07, "learning_rate": 0.0004977413005803975, "loss": 3.0433, "step": 534 }, { "epoch": 0.07, "learning_rate": 0.0004977277757430127, "loss": 3.1675, "step": 535 }, { "epoch": 0.07, "learning_rate": 0.0004977142107186602, "loss": 3.0438, "step": 536 }, { "epoch": 0.07, "learning_rate": 0.0004977006055095405, "loss": 3.1236, "step": 537 }, { "epoch": 0.07, "learning_rate": 0.0004976869601178609, "loss": 3.2369, "step": 538 }, { "epoch": 0.07, "learning_rate": 0.0004976732745458348, "loss": 3.1288, "step": 539 }, { "epoch": 0.07, "learning_rate": 0.0004976595487956823, "loss": 3.1917, "step": 540 }, { "epoch": 0.07, "learning_rate": 0.0004976457828696301, "loss": 3.1879, "step": 541 }, { "epoch": 0.07, "learning_rate": 0.0004976319767699114, "loss": 3.13, "step": 542 }, { "epoch": 0.07, "learning_rate": 0.0004976181304987657, "loss": 3.0169, "step": 543 }, { "epoch": 0.07, "learning_rate": 0.0004976042440584392, "loss": 3.0524, "step": 544 }, { "epoch": 0.07, "learning_rate": 0.0004975903174511846, "loss": 3.1787, "step": 545 }, { "epoch": 0.07, "learning_rate": 0.0004975763506792612, "loss": 3.1183, "step": 546 }, { "epoch": 0.07, "learning_rate": 0.0004975623437449345, "loss": 3.1033, "step": 547 }, { "epoch": 0.07, "learning_rate": 0.0004975482966504772, "loss": 3.1427, "step": 548 }, { "epoch": 0.07, "learning_rate": 0.0004975342093981675, "loss": 2.9235, "step": 549 }, { "epoch": 0.07, "learning_rate": 0.0004975200819902911, "loss": 3.2151, "step": 550 }, { "epoch": 0.07, "learning_rate": 0.0004975059144291394, "loss": 2.9228, "step": 551 }, { "epoch": 0.07, "learning_rate": 0.000497491706717011, "loss": 3.0136, "step": 552 }, { "epoch": 0.07, "learning_rate": 0.0004974774588562106, "loss": 2.909, "step": 553 }, { "epoch": 0.07, "learning_rate": 0.0004974631708490494, "loss": 2.9684, "step": 554 }, { "epoch": 0.07, "learning_rate": 0.0004974488426978454, "loss": 3.0433, "step": 555 }, { "epoch": 0.07, "learning_rate": 0.000497434474404923, "loss": 2.9545, "step": 556 }, { "epoch": 0.07, "learning_rate": 0.0004974200659726128, "loss": 3.082, "step": 557 }, { "epoch": 0.07, "learning_rate": 0.0004974056174032524, "loss": 3.128, "step": 558 }, { "epoch": 0.07, "learning_rate": 0.0004973911286991856, "loss": 3.0289, "step": 559 }, { "epoch": 0.07, "learning_rate": 0.0004973765998627628, "loss": 3.0358, "step": 560 }, { "epoch": 0.07, "learning_rate": 0.0004973620308963408, "loss": 3.1045, "step": 561 }, { "epoch": 0.07, "learning_rate": 0.0004973474218022832, "loss": 3.1759, "step": 562 }, { "epoch": 0.07, "learning_rate": 0.0004973327725829598, "loss": 2.9658, "step": 563 }, { "epoch": 0.07, "learning_rate": 0.0004973180832407472, "loss": 2.9602, "step": 564 }, { "epoch": 0.07, "learning_rate": 0.000497303353778028, "loss": 3.0346, "step": 565 }, { "epoch": 0.07, "learning_rate": 0.0004972885841971918, "loss": 3.0933, "step": 566 }, { "epoch": 0.07, "learning_rate": 0.0004972737745006347, "loss": 2.9927, "step": 567 }, { "epoch": 0.07, "learning_rate": 0.0004972589246907591, "loss": 3.1981, "step": 568 }, { "epoch": 0.07, "learning_rate": 0.0004972440347699738, "loss": 2.9546, "step": 569 }, { "epoch": 0.07, "learning_rate": 0.0004972291047406945, "loss": 2.9517, "step": 570 }, { "epoch": 0.07, "learning_rate": 0.000497214134605343, "loss": 2.9676, "step": 571 }, { "epoch": 0.07, "learning_rate": 0.000497199124366348, "loss": 2.7584, "step": 572 }, { "epoch": 0.07, "learning_rate": 0.0004971840740261443, "loss": 3.0269, "step": 573 }, { "epoch": 0.07, "learning_rate": 0.0004971689835871735, "loss": 2.8855, "step": 574 }, { "epoch": 0.07, "learning_rate": 0.0004971538530518835, "loss": 3.024, "step": 575 }, { "epoch": 0.07, "learning_rate": 0.000497138682422729, "loss": 2.8967, "step": 576 }, { "epoch": 0.07, "learning_rate": 0.0004971234717021708, "loss": 2.9493, "step": 577 }, { "epoch": 0.07, "learning_rate": 0.0004971082208926767, "loss": 2.9937, "step": 578 }, { "epoch": 0.07, "learning_rate": 0.0004970929299967205, "loss": 2.8745, "step": 579 }, { "epoch": 0.07, "learning_rate": 0.0004970775990167826, "loss": 2.8852, "step": 580 }, { "epoch": 0.07, "learning_rate": 0.0004970622279553503, "loss": 2.8725, "step": 581 }, { "epoch": 0.07, "learning_rate": 0.000497046816814917, "loss": 2.8722, "step": 582 }, { "epoch": 0.07, "learning_rate": 0.0004970313655979828, "loss": 2.9416, "step": 583 }, { "epoch": 0.07, "learning_rate": 0.0004970158743070541, "loss": 2.9411, "step": 584 }, { "epoch": 0.07, "learning_rate": 0.0004970003429446441, "loss": 2.9879, "step": 585 }, { "epoch": 0.07, "learning_rate": 0.0004969847715132722, "loss": 2.9693, "step": 586 }, { "epoch": 0.07, "learning_rate": 0.0004969691600154645, "loss": 2.869, "step": 587 }, { "epoch": 0.07, "learning_rate": 0.0004969535084537536, "loss": 3.0031, "step": 588 }, { "epoch": 0.07, "learning_rate": 0.0004969378168306784, "loss": 2.7986, "step": 589 }, { "epoch": 0.07, "learning_rate": 0.0004969220851487844, "loss": 2.8469, "step": 590 }, { "epoch": 0.07, "learning_rate": 0.0004969063134106238, "loss": 2.9182, "step": 591 }, { "epoch": 0.07, "learning_rate": 0.000496890501618755, "loss": 2.9953, "step": 592 }, { "epoch": 0.07, "learning_rate": 0.0004968746497757431, "loss": 2.9938, "step": 593 }, { "epoch": 0.07, "learning_rate": 0.0004968587578841596, "loss": 3.0611, "step": 594 }, { "epoch": 0.07, "learning_rate": 0.0004968428259465824, "loss": 2.8734, "step": 595 }, { "epoch": 0.07, "learning_rate": 0.0004968268539655963, "loss": 2.9621, "step": 596 }, { "epoch": 0.07, "learning_rate": 0.0004968108419437919, "loss": 2.9326, "step": 597 }, { "epoch": 0.07, "learning_rate": 0.0004967947898837671, "loss": 2.7563, "step": 598 }, { "epoch": 0.07, "learning_rate": 0.0004967786977881257, "loss": 2.7503, "step": 599 }, { "epoch": 0.07, "learning_rate": 0.0004967625656594782, "loss": 2.9304, "step": 600 }, { "epoch": 0.07, "eval_acceptance_rate": 0.4536316394805908, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.436086776771774, "eval_expected_tokens": 1.7951084709647176, "eval_loss": 2.9235150814056396, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.19877711774117937, "eval_runtime": 50.2894, "eval_samples_per_second": 9.684, "eval_steps_per_second": 0.318, "step": 600 }, { "epoch": 0.08, "learning_rate": 0.0004967463935004417, "loss": 3.1152, "step": 601 }, { "epoch": 0.08, "learning_rate": 0.0004967301813136396, "loss": 2.9501, "step": 602 }, { "epoch": 0.08, "learning_rate": 0.0004967139291017018, "loss": 2.786, "step": 603 }, { "epoch": 0.08, "learning_rate": 0.000496697636867265, "loss": 2.896, "step": 604 }, { "epoch": 0.08, "learning_rate": 0.0004966813046129719, "loss": 2.8646, "step": 605 }, { "epoch": 0.08, "learning_rate": 0.0004966649323414721, "loss": 2.8806, "step": 606 }, { "epoch": 0.08, "learning_rate": 0.0004966485200554216, "loss": 2.8751, "step": 607 }, { "epoch": 0.08, "learning_rate": 0.0004966320677574827, "loss": 2.7909, "step": 608 }, { "epoch": 0.08, "learning_rate": 0.0004966155754503245, "loss": 2.8605, "step": 609 }, { "epoch": 0.08, "learning_rate": 0.0004965990431366222, "loss": 2.9389, "step": 610 }, { "epoch": 0.08, "learning_rate": 0.0004965824708190579, "loss": 2.8089, "step": 611 }, { "epoch": 0.08, "learning_rate": 0.0004965658585003199, "loss": 2.7258, "step": 612 }, { "epoch": 0.08, "learning_rate": 0.0004965492061831031, "loss": 2.8952, "step": 613 }, { "epoch": 0.08, "learning_rate": 0.0004965325138701088, "loss": 2.8337, "step": 614 }, { "epoch": 0.08, "learning_rate": 0.0004965157815640453, "loss": 2.8844, "step": 615 }, { "epoch": 0.08, "learning_rate": 0.0004964990092676262, "loss": 2.888, "step": 616 }, { "epoch": 0.08, "learning_rate": 0.000496482196983573, "loss": 2.6942, "step": 617 }, { "epoch": 0.08, "learning_rate": 0.0004964653447146127, "loss": 2.7222, "step": 618 }, { "epoch": 0.08, "learning_rate": 0.0004964484524634791, "loss": 2.9703, "step": 619 }, { "epoch": 0.08, "learning_rate": 0.0004964315202329127, "loss": 2.7651, "step": 620 }, { "epoch": 0.08, "learning_rate": 0.0004964145480256603, "loss": 2.7926, "step": 621 }, { "epoch": 0.08, "learning_rate": 0.0004963975358444748, "loss": 2.7826, "step": 622 }, { "epoch": 0.08, "learning_rate": 0.0004963804836921163, "loss": 2.7276, "step": 623 }, { "epoch": 0.08, "learning_rate": 0.0004963633915713509, "loss": 2.6889, "step": 624 }, { "epoch": 0.08, "learning_rate": 0.0004963462594849515, "loss": 2.8571, "step": 625 }, { "epoch": 0.08, "learning_rate": 0.000496329087435697, "loss": 2.7437, "step": 626 }, { "epoch": 0.08, "learning_rate": 0.0004963118754263734, "loss": 2.8619, "step": 627 }, { "epoch": 0.08, "learning_rate": 0.0004962946234597726, "loss": 2.7902, "step": 628 }, { "epoch": 0.08, "learning_rate": 0.0004962773315386935, "loss": 2.7152, "step": 629 }, { "epoch": 0.08, "learning_rate": 0.000496259999665941, "loss": 2.7045, "step": 630 }, { "epoch": 0.08, "learning_rate": 0.0004962426278443269, "loss": 2.8674, "step": 631 }, { "epoch": 0.08, "learning_rate": 0.0004962252160766693, "loss": 2.8312, "step": 632 }, { "epoch": 0.08, "learning_rate": 0.0004962077643657926, "loss": 2.902, "step": 633 }, { "epoch": 0.08, "learning_rate": 0.0004961902727145281, "loss": 2.8576, "step": 634 }, { "epoch": 0.08, "learning_rate": 0.0004961727411257131, "loss": 2.7887, "step": 635 }, { "epoch": 0.08, "learning_rate": 0.0004961551696021917, "loss": 2.7413, "step": 636 }, { "epoch": 0.08, "learning_rate": 0.0004961375581468145, "loss": 2.7543, "step": 637 }, { "epoch": 0.08, "learning_rate": 0.0004961199067624383, "loss": 2.7452, "step": 638 }, { "epoch": 0.08, "learning_rate": 0.0004961022154519267, "loss": 2.8766, "step": 639 }, { "epoch": 0.08, "learning_rate": 0.0004960844842181495, "loss": 2.8699, "step": 640 }, { "epoch": 0.08, "learning_rate": 0.000496066713063983, "loss": 2.7637, "step": 641 }, { "epoch": 0.08, "learning_rate": 0.0004960489019923105, "loss": 2.6245, "step": 642 }, { "epoch": 0.08, "learning_rate": 0.000496031051006021, "loss": 2.684, "step": 643 }, { "epoch": 0.08, "learning_rate": 0.0004960131601080103, "loss": 2.6826, "step": 644 }, { "epoch": 0.08, "learning_rate": 0.000495995229301181, "loss": 2.6498, "step": 645 }, { "epoch": 0.08, "learning_rate": 0.0004959772585884416, "loss": 2.788, "step": 646 }, { "epoch": 0.08, "learning_rate": 0.0004959592479727074, "loss": 2.7488, "step": 647 }, { "epoch": 0.08, "learning_rate": 0.0004959411974569002, "loss": 2.8219, "step": 648 }, { "epoch": 0.08, "learning_rate": 0.0004959231070439481, "loss": 2.7923, "step": 649 }, { "epoch": 0.08, "learning_rate": 0.0004959049767367859, "loss": 2.9116, "step": 650 }, { "epoch": 0.08, "learning_rate": 0.0004958868065383546, "loss": 2.7565, "step": 651 }, { "epoch": 0.08, "learning_rate": 0.000495868596451602, "loss": 2.6272, "step": 652 }, { "epoch": 0.08, "learning_rate": 0.000495850346479482, "loss": 2.6872, "step": 653 }, { "epoch": 0.08, "learning_rate": 0.0004958320566249552, "loss": 2.803, "step": 654 }, { "epoch": 0.08, "learning_rate": 0.0004958137268909887, "loss": 2.8309, "step": 655 }, { "epoch": 0.08, "learning_rate": 0.0004957953572805558, "loss": 2.7153, "step": 656 }, { "epoch": 0.08, "learning_rate": 0.0004957769477966367, "loss": 2.7031, "step": 657 }, { "epoch": 0.08, "learning_rate": 0.0004957584984422176, "loss": 2.7299, "step": 658 }, { "epoch": 0.08, "learning_rate": 0.0004957400092202917, "loss": 2.8131, "step": 659 }, { "epoch": 0.08, "learning_rate": 0.0004957214801338581, "loss": 2.6643, "step": 660 }, { "epoch": 0.08, "learning_rate": 0.0004957029111859226, "loss": 2.6053, "step": 661 }, { "epoch": 0.08, "learning_rate": 0.0004956843023794977, "loss": 2.5798, "step": 662 }, { "epoch": 0.08, "learning_rate": 0.0004956656537176021, "loss": 2.6418, "step": 663 }, { "epoch": 0.08, "learning_rate": 0.0004956469652032609, "loss": 2.7437, "step": 664 }, { "epoch": 0.08, "learning_rate": 0.0004956282368395059, "loss": 2.8484, "step": 665 }, { "epoch": 0.08, "learning_rate": 0.0004956094686293754, "loss": 2.7785, "step": 666 }, { "epoch": 0.08, "learning_rate": 0.0004955906605759137, "loss": 2.7418, "step": 667 }, { "epoch": 0.08, "learning_rate": 0.0004955718126821722, "loss": 2.7575, "step": 668 }, { "epoch": 0.08, "learning_rate": 0.0004955529249512082, "loss": 2.7015, "step": 669 }, { "epoch": 0.08, "learning_rate": 0.0004955339973860859, "loss": 2.6193, "step": 670 }, { "epoch": 0.08, "learning_rate": 0.0004955150299898757, "loss": 2.7094, "step": 671 }, { "epoch": 0.08, "learning_rate": 0.0004954960227656544, "loss": 2.8789, "step": 672 }, { "epoch": 0.08, "learning_rate": 0.0004954769757165055, "loss": 2.6815, "step": 673 }, { "epoch": 0.08, "learning_rate": 0.000495457888845519, "loss": 2.6914, "step": 674 }, { "epoch": 0.08, "learning_rate": 0.0004954387621557911, "loss": 2.7484, "step": 675 }, { "epoch": 0.08, "learning_rate": 0.0004954195956504245, "loss": 2.7339, "step": 676 }, { "epoch": 0.08, "learning_rate": 0.0004954003893325285, "loss": 2.7496, "step": 677 }, { "epoch": 0.08, "learning_rate": 0.0004953811432052188, "loss": 2.7201, "step": 678 }, { "epoch": 0.08, "learning_rate": 0.0004953618572716176, "loss": 2.7006, "step": 679 }, { "epoch": 0.09, "learning_rate": 0.0004953425315348534, "loss": 2.8383, "step": 680 }, { "epoch": 0.09, "learning_rate": 0.0004953231659980613, "loss": 2.703, "step": 681 }, { "epoch": 0.09, "learning_rate": 0.0004953037606643828, "loss": 2.8211, "step": 682 }, { "epoch": 0.09, "learning_rate": 0.000495284315536966, "loss": 2.6505, "step": 683 }, { "epoch": 0.09, "learning_rate": 0.0004952648306189654, "loss": 2.7112, "step": 684 }, { "epoch": 0.09, "learning_rate": 0.0004952453059135415, "loss": 2.7088, "step": 685 }, { "epoch": 0.09, "learning_rate": 0.0004952257414238619, "loss": 2.7146, "step": 686 }, { "epoch": 0.09, "learning_rate": 0.0004952061371531005, "loss": 2.63, "step": 687 }, { "epoch": 0.09, "learning_rate": 0.0004951864931044374, "loss": 2.5934, "step": 688 }, { "epoch": 0.09, "learning_rate": 0.0004951668092810593, "loss": 2.6895, "step": 689 }, { "epoch": 0.09, "learning_rate": 0.0004951470856861594, "loss": 2.6641, "step": 690 }, { "epoch": 0.09, "learning_rate": 0.0004951273223229373, "loss": 2.6822, "step": 691 }, { "epoch": 0.09, "learning_rate": 0.0004951075191945989, "loss": 2.5915, "step": 692 }, { "epoch": 0.09, "learning_rate": 0.000495087676304357, "loss": 2.5861, "step": 693 }, { "epoch": 0.09, "learning_rate": 0.0004950677936554305, "loss": 2.4777, "step": 694 }, { "epoch": 0.09, "learning_rate": 0.0004950478712510446, "loss": 2.6084, "step": 695 }, { "epoch": 0.09, "learning_rate": 0.0004950279090944313, "loss": 2.6537, "step": 696 }, { "epoch": 0.09, "learning_rate": 0.0004950079071888289, "loss": 2.7556, "step": 697 }, { "epoch": 0.09, "learning_rate": 0.0004949878655374822, "loss": 2.6604, "step": 698 }, { "epoch": 0.09, "learning_rate": 0.0004949677841436423, "loss": 2.6879, "step": 699 }, { "epoch": 0.09, "learning_rate": 0.0004949476630105669, "loss": 2.7843, "step": 700 }, { "epoch": 0.09, "eval_acceptance_rate": 0.47293928265571594, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.4819383090386902, "eval_expected_tokens": 1.852422886298363, "eval_loss": 2.5964770317077637, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.21310572157459073, "eval_runtime": 50.3977, "eval_samples_per_second": 9.663, "eval_steps_per_second": 0.317, "step": 700 }, { "epoch": 0.09, "learning_rate": 0.0004949275021415201, "loss": 2.5617, "step": 701 }, { "epoch": 0.09, "learning_rate": 0.0004949073015397725, "loss": 2.646, "step": 702 }, { "epoch": 0.09, "learning_rate": 0.000494887061208601, "loss": 2.5507, "step": 703 }, { "epoch": 0.09, "learning_rate": 0.000494866781151289, "loss": 2.6283, "step": 704 }, { "epoch": 0.09, "learning_rate": 0.0004948464613711265, "loss": 2.757, "step": 705 }, { "epoch": 0.09, "learning_rate": 0.0004948261018714097, "loss": 2.6018, "step": 706 }, { "epoch": 0.09, "learning_rate": 0.0004948057026554415, "loss": 2.587, "step": 707 }, { "epoch": 0.09, "learning_rate": 0.0004947852637265309, "loss": 2.6221, "step": 708 }, { "epoch": 0.09, "learning_rate": 0.0004947647850879938, "loss": 2.6444, "step": 709 }, { "epoch": 0.09, "learning_rate": 0.0004947442667431522, "loss": 2.6644, "step": 710 }, { "epoch": 0.09, "learning_rate": 0.0004947237086953345, "loss": 2.5475, "step": 711 }, { "epoch": 0.09, "learning_rate": 0.0004947031109478758, "loss": 2.6931, "step": 712 }, { "epoch": 0.09, "learning_rate": 0.0004946824735041175, "loss": 2.4867, "step": 713 }, { "epoch": 0.09, "learning_rate": 0.0004946617963674075, "loss": 2.3883, "step": 714 }, { "epoch": 0.09, "learning_rate": 0.0004946410795410999, "loss": 2.52, "step": 715 }, { "epoch": 0.09, "learning_rate": 0.0004946203230285557, "loss": 2.5744, "step": 716 }, { "epoch": 0.09, "learning_rate": 0.0004945995268331419, "loss": 2.5173, "step": 717 }, { "epoch": 0.09, "learning_rate": 0.0004945786909582321, "loss": 2.5062, "step": 718 }, { "epoch": 0.09, "learning_rate": 0.0004945578154072066, "loss": 2.6243, "step": 719 }, { "epoch": 0.09, "learning_rate": 0.0004945369001834514, "loss": 2.6192, "step": 720 }, { "epoch": 0.09, "learning_rate": 0.0004945159452903598, "loss": 2.4807, "step": 721 }, { "epoch": 0.09, "learning_rate": 0.000494494950731331, "loss": 2.5573, "step": 722 }, { "epoch": 0.09, "learning_rate": 0.0004944739165097709, "loss": 2.6553, "step": 723 }, { "epoch": 0.09, "learning_rate": 0.0004944528426290915, "loss": 2.7179, "step": 724 }, { "epoch": 0.09, "learning_rate": 0.0004944317290927117, "loss": 2.6057, "step": 725 }, { "epoch": 0.09, "learning_rate": 0.0004944105759040565, "loss": 2.6474, "step": 726 }, { "epoch": 0.09, "learning_rate": 0.0004943893830665573, "loss": 2.5779, "step": 727 }, { "epoch": 0.09, "learning_rate": 0.0004943681505836522, "loss": 2.5995, "step": 728 }, { "epoch": 0.09, "learning_rate": 0.0004943468784587855, "loss": 2.6019, "step": 729 }, { "epoch": 0.09, "learning_rate": 0.0004943255666954082, "loss": 2.5746, "step": 730 }, { "epoch": 0.09, "learning_rate": 0.0004943042152969772, "loss": 2.6259, "step": 731 }, { "epoch": 0.09, "learning_rate": 0.0004942828242669564, "loss": 2.8715, "step": 732 }, { "epoch": 0.09, "learning_rate": 0.000494261393608816, "loss": 2.6768, "step": 733 }, { "epoch": 0.09, "learning_rate": 0.0004942399233260322, "loss": 2.5693, "step": 734 }, { "epoch": 0.09, "learning_rate": 0.0004942184134220882, "loss": 2.5384, "step": 735 }, { "epoch": 0.09, "learning_rate": 0.0004941968639004733, "loss": 2.5503, "step": 736 }, { "epoch": 0.09, "learning_rate": 0.0004941752747646835, "loss": 2.5304, "step": 737 }, { "epoch": 0.09, "learning_rate": 0.0004941536460182208, "loss": 2.631, "step": 738 }, { "epoch": 0.09, "learning_rate": 0.0004941319776645939, "loss": 2.6743, "step": 739 }, { "epoch": 0.09, "learning_rate": 0.0004941102697073181, "loss": 2.6151, "step": 740 }, { "epoch": 0.09, "learning_rate": 0.0004940885221499146, "loss": 2.5311, "step": 741 }, { "epoch": 0.09, "learning_rate": 0.0004940667349959115, "loss": 2.6299, "step": 742 }, { "epoch": 0.09, "learning_rate": 0.0004940449082488432, "loss": 2.5473, "step": 743 }, { "epoch": 0.09, "learning_rate": 0.0004940230419122505, "loss": 2.6455, "step": 744 }, { "epoch": 0.09, "learning_rate": 0.0004940011359896804, "loss": 2.4593, "step": 745 }, { "epoch": 0.09, "learning_rate": 0.0004939791904846869, "loss": 2.5738, "step": 746 }, { "epoch": 0.09, "learning_rate": 0.0004939572054008296, "loss": 2.5309, "step": 747 }, { "epoch": 0.09, "learning_rate": 0.0004939351807416753, "loss": 2.6884, "step": 748 }, { "epoch": 0.09, "learning_rate": 0.0004939131165107968, "loss": 2.7882, "step": 749 }, { "epoch": 0.09, "learning_rate": 0.0004938910127117735, "loss": 2.6583, "step": 750 }, { "epoch": 0.09, "learning_rate": 0.0004938688693481909, "loss": 2.6894, "step": 751 }, { "epoch": 0.09, "learning_rate": 0.0004938466864236413, "loss": 2.6613, "step": 752 }, { "epoch": 0.09, "learning_rate": 0.0004938244639417233, "loss": 2.6549, "step": 753 }, { "epoch": 0.09, "learning_rate": 0.0004938022019060418, "loss": 2.4998, "step": 754 }, { "epoch": 0.09, "learning_rate": 0.0004937799003202081, "loss": 2.5834, "step": 755 }, { "epoch": 0.09, "learning_rate": 0.0004937575591878403, "loss": 2.5732, "step": 756 }, { "epoch": 0.09, "learning_rate": 0.0004937351785125623, "loss": 2.5235, "step": 757 }, { "epoch": 0.09, "learning_rate": 0.0004937127582980051, "loss": 2.5183, "step": 758 }, { "epoch": 0.09, "learning_rate": 0.0004936902985478055, "loss": 2.5891, "step": 759 }, { "epoch": 0.1, "learning_rate": 0.000493667799265607, "loss": 2.5214, "step": 760 }, { "epoch": 0.1, "learning_rate": 0.0004936452604550596, "loss": 2.4142, "step": 761 }, { "epoch": 0.1, "learning_rate": 0.0004936226821198194, "loss": 2.5079, "step": 762 }, { "epoch": 0.1, "learning_rate": 0.0004936000642635493, "loss": 2.5518, "step": 763 }, { "epoch": 0.1, "learning_rate": 0.0004935774068899184, "loss": 2.6427, "step": 764 }, { "epoch": 0.1, "learning_rate": 0.0004935547100026021, "loss": 2.5739, "step": 765 }, { "epoch": 0.1, "learning_rate": 0.0004935319736052825, "loss": 2.6289, "step": 766 }, { "epoch": 0.1, "learning_rate": 0.0004935091977016479, "loss": 2.5594, "step": 767 }, { "epoch": 0.1, "learning_rate": 0.0004934863822953928, "loss": 2.5164, "step": 768 }, { "epoch": 0.1, "learning_rate": 0.0004934635273902189, "loss": 2.3813, "step": 769 }, { "epoch": 0.1, "learning_rate": 0.0004934406329898333, "loss": 2.3986, "step": 770 }, { "epoch": 0.1, "learning_rate": 0.0004934176990979503, "loss": 2.5889, "step": 771 }, { "epoch": 0.1, "learning_rate": 0.0004933947257182901, "loss": 2.515, "step": 772 }, { "epoch": 0.1, "learning_rate": 0.0004933717128545794, "loss": 2.4531, "step": 773 }, { "epoch": 0.1, "learning_rate": 0.0004933486605105518, "loss": 2.4527, "step": 774 }, { "epoch": 0.1, "learning_rate": 0.0004933255686899464, "loss": 2.4212, "step": 775 }, { "epoch": 0.1, "learning_rate": 0.0004933024373965097, "loss": 2.5189, "step": 776 }, { "epoch": 0.1, "learning_rate": 0.0004932792666339936, "loss": 2.5159, "step": 777 }, { "epoch": 0.1, "learning_rate": 0.0004932560564061574, "loss": 2.3209, "step": 778 }, { "epoch": 0.1, "learning_rate": 0.000493232806716766, "loss": 2.5327, "step": 779 }, { "epoch": 0.1, "learning_rate": 0.0004932095175695911, "loss": 2.5705, "step": 780 }, { "epoch": 0.1, "learning_rate": 0.0004931861889684109, "loss": 2.6271, "step": 781 }, { "epoch": 0.1, "learning_rate": 0.0004931628209170094, "loss": 2.4674, "step": 782 }, { "epoch": 0.1, "learning_rate": 0.0004931394134191778, "loss": 2.3985, "step": 783 }, { "epoch": 0.1, "learning_rate": 0.0004931159664787132, "loss": 2.3993, "step": 784 }, { "epoch": 0.1, "learning_rate": 0.0004930924800994192, "loss": 2.6282, "step": 785 }, { "epoch": 0.1, "learning_rate": 0.0004930689542851057, "loss": 2.4983, "step": 786 }, { "epoch": 0.1, "learning_rate": 0.0004930453890395894, "loss": 2.5847, "step": 787 }, { "epoch": 0.1, "learning_rate": 0.0004930217843666929, "loss": 2.5813, "step": 788 }, { "epoch": 0.1, "learning_rate": 0.0004929981402702453, "loss": 2.5221, "step": 789 }, { "epoch": 0.1, "learning_rate": 0.0004929744567540826, "loss": 2.5398, "step": 790 }, { "epoch": 0.1, "learning_rate": 0.0004929507338220463, "loss": 2.741, "step": 791 }, { "epoch": 0.1, "learning_rate": 0.0004929269714779852, "loss": 2.5653, "step": 792 }, { "epoch": 0.1, "learning_rate": 0.0004929031697257539, "loss": 2.4361, "step": 793 }, { "epoch": 0.1, "learning_rate": 0.0004928793285692135, "loss": 2.4863, "step": 794 }, { "epoch": 0.1, "learning_rate": 0.0004928554480122316, "loss": 2.5608, "step": 795 }, { "epoch": 0.1, "learning_rate": 0.0004928315280586824, "loss": 2.3688, "step": 796 }, { "epoch": 0.1, "learning_rate": 0.0004928075687124459, "loss": 2.5221, "step": 797 }, { "epoch": 0.1, "learning_rate": 0.000492783569977409, "loss": 2.3722, "step": 798 }, { "epoch": 0.1, "learning_rate": 0.0004927595318574649, "loss": 2.3937, "step": 799 }, { "epoch": 0.1, "learning_rate": 0.000492735454356513, "loss": 2.4252, "step": 800 }, { "epoch": 0.1, "eval_acceptance_rate": 0.4918650686740875, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.5290596567697794, "eval_expected_tokens": 1.9113245709622244, "eval_loss": 2.451101303100586, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2278311427405561, "eval_runtime": 50.2795, "eval_samples_per_second": 9.686, "eval_steps_per_second": 0.318, "step": 800 }, { "epoch": 0.1, "learning_rate": 0.0004927113374784593, "loss": 2.4826, "step": 801 }, { "epoch": 0.1, "learning_rate": 0.000492687181227216, "loss": 2.5162, "step": 802 }, { "epoch": 0.1, "learning_rate": 0.0004926629856067019, "loss": 2.5003, "step": 803 }, { "epoch": 0.1, "learning_rate": 0.000492638750620842, "loss": 2.4065, "step": 804 }, { "epoch": 0.1, "learning_rate": 0.0004926144762735676, "loss": 2.4071, "step": 805 }, { "epoch": 0.1, "learning_rate": 0.0004925901625688168, "loss": 2.5358, "step": 806 }, { "epoch": 0.1, "learning_rate": 0.0004925658095105338, "loss": 2.536, "step": 807 }, { "epoch": 0.1, "learning_rate": 0.0004925414171026691, "loss": 2.3607, "step": 808 }, { "epoch": 0.1, "learning_rate": 0.0004925169853491797, "loss": 2.3284, "step": 809 }, { "epoch": 0.1, "learning_rate": 0.000492492514254029, "loss": 2.5359, "step": 810 }, { "epoch": 0.1, "learning_rate": 0.0004924680038211868, "loss": 2.3465, "step": 811 }, { "epoch": 0.1, "learning_rate": 0.000492443454054629, "loss": 2.4557, "step": 812 }, { "epoch": 0.1, "learning_rate": 0.0004924188649583386, "loss": 2.438, "step": 813 }, { "epoch": 0.1, "learning_rate": 0.000492394236536304, "loss": 2.4429, "step": 814 }, { "epoch": 0.1, "learning_rate": 0.0004923695687925207, "loss": 2.4613, "step": 815 }, { "epoch": 0.1, "learning_rate": 0.0004923448617309905, "loss": 2.4507, "step": 816 }, { "epoch": 0.1, "learning_rate": 0.0004923201153557212, "loss": 2.4308, "step": 817 }, { "epoch": 0.1, "learning_rate": 0.0004922953296707273, "loss": 2.4297, "step": 818 }, { "epoch": 0.1, "learning_rate": 0.0004922705046800296, "loss": 2.4686, "step": 819 }, { "epoch": 0.1, "learning_rate": 0.0004922456403876552, "loss": 2.5606, "step": 820 }, { "epoch": 0.1, "learning_rate": 0.0004922207367976377, "loss": 2.4917, "step": 821 }, { "epoch": 0.1, "learning_rate": 0.0004921957939140171, "loss": 2.399, "step": 822 }, { "epoch": 0.1, "learning_rate": 0.0004921708117408396, "loss": 2.3154, "step": 823 }, { "epoch": 0.1, "learning_rate": 0.0004921457902821578, "loss": 2.5301, "step": 824 }, { "epoch": 0.1, "learning_rate": 0.0004921207295420308, "loss": 2.3834, "step": 825 }, { "epoch": 0.1, "learning_rate": 0.0004920956295245241, "loss": 2.5339, "step": 826 }, { "epoch": 0.1, "learning_rate": 0.0004920704902337095, "loss": 2.4119, "step": 827 }, { "epoch": 0.1, "learning_rate": 0.0004920453116736648, "loss": 2.4125, "step": 828 }, { "epoch": 0.1, "learning_rate": 0.000492020093848475, "loss": 2.4224, "step": 829 }, { "epoch": 0.1, "learning_rate": 0.0004919948367622307, "loss": 2.4276, "step": 830 }, { "epoch": 0.1, "learning_rate": 0.0004919695404190293, "loss": 2.3602, "step": 831 }, { "epoch": 0.1, "learning_rate": 0.0004919442048229744, "loss": 2.3083, "step": 832 }, { "epoch": 0.1, "learning_rate": 0.0004919188299781757, "loss": 2.299, "step": 833 }, { "epoch": 0.1, "learning_rate": 0.0004918934158887501, "loss": 2.5153, "step": 834 }, { "epoch": 0.1, "learning_rate": 0.00049186796255882, "loss": 2.4415, "step": 835 }, { "epoch": 0.1, "learning_rate": 0.0004918424699925146, "loss": 2.3808, "step": 836 }, { "epoch": 0.1, "learning_rate": 0.0004918169381939692, "loss": 2.4764, "step": 837 }, { "epoch": 0.1, "learning_rate": 0.0004917913671673258, "loss": 2.3688, "step": 838 }, { "epoch": 0.1, "learning_rate": 0.0004917657569167327, "loss": 2.6143, "step": 839 }, { "epoch": 0.1, "learning_rate": 0.0004917401074463441, "loss": 2.4445, "step": 840 }, { "epoch": 0.11, "learning_rate": 0.0004917144187603212, "loss": 2.4864, "step": 841 }, { "epoch": 0.11, "learning_rate": 0.0004916886908628311, "loss": 2.4569, "step": 842 }, { "epoch": 0.11, "learning_rate": 0.0004916629237580476, "loss": 2.4497, "step": 843 }, { "epoch": 0.11, "learning_rate": 0.0004916371174501507, "loss": 2.4009, "step": 844 }, { "epoch": 0.11, "learning_rate": 0.0004916112719433266, "loss": 2.2712, "step": 845 }, { "epoch": 0.11, "learning_rate": 0.0004915853872417681, "loss": 2.3457, "step": 846 }, { "epoch": 0.11, "learning_rate": 0.0004915594633496743, "loss": 2.412, "step": 847 }, { "epoch": 0.11, "learning_rate": 0.0004915335002712506, "loss": 2.2878, "step": 848 }, { "epoch": 0.11, "learning_rate": 0.0004915074980107088, "loss": 2.4579, "step": 849 }, { "epoch": 0.11, "learning_rate": 0.0004914814565722671, "loss": 2.4262, "step": 850 }, { "epoch": 0.11, "learning_rate": 0.0004914553759601499, "loss": 2.3061, "step": 851 }, { "epoch": 0.11, "learning_rate": 0.0004914292561785881, "loss": 2.4194, "step": 852 }, { "epoch": 0.11, "learning_rate": 0.000491403097231819, "loss": 2.3372, "step": 853 }, { "epoch": 0.11, "learning_rate": 0.000491376899124086, "loss": 2.4795, "step": 854 }, { "epoch": 0.11, "learning_rate": 0.000491350661859639, "loss": 2.3938, "step": 855 }, { "epoch": 0.11, "learning_rate": 0.0004913243854427346, "loss": 2.3977, "step": 856 }, { "epoch": 0.11, "learning_rate": 0.0004912980698776351, "loss": 2.3595, "step": 857 }, { "epoch": 0.11, "learning_rate": 0.0004912717151686095, "loss": 2.4194, "step": 858 }, { "epoch": 0.11, "learning_rate": 0.0004912453213199332, "loss": 2.3294, "step": 859 }, { "epoch": 0.11, "learning_rate": 0.0004912188883358879, "loss": 2.3449, "step": 860 }, { "epoch": 0.11, "learning_rate": 0.0004911924162207615, "loss": 2.5006, "step": 861 }, { "epoch": 0.11, "learning_rate": 0.0004911659049788485, "loss": 2.4769, "step": 862 }, { "epoch": 0.11, "learning_rate": 0.0004911393546144495, "loss": 2.4139, "step": 863 }, { "epoch": 0.11, "learning_rate": 0.0004911127651318717, "loss": 2.4264, "step": 864 }, { "epoch": 0.11, "learning_rate": 0.0004910861365354283, "loss": 2.4166, "step": 865 }, { "epoch": 0.11, "learning_rate": 0.0004910594688294393, "loss": 2.4977, "step": 866 }, { "epoch": 0.11, "learning_rate": 0.0004910327620182306, "loss": 2.2513, "step": 867 }, { "epoch": 0.11, "learning_rate": 0.0004910060161061348, "loss": 2.347, "step": 868 }, { "epoch": 0.11, "learning_rate": 0.0004909792310974904, "loss": 2.4697, "step": 869 }, { "epoch": 0.11, "learning_rate": 0.0004909524069966429, "loss": 2.3841, "step": 870 }, { "epoch": 0.11, "learning_rate": 0.0004909255438079435, "loss": 2.5078, "step": 871 }, { "epoch": 0.11, "learning_rate": 0.0004908986415357501, "loss": 2.3819, "step": 872 }, { "epoch": 0.11, "learning_rate": 0.0004908717001844269, "loss": 2.3767, "step": 873 }, { "epoch": 0.11, "learning_rate": 0.0004908447197583442, "loss": 2.4578, "step": 874 }, { "epoch": 0.11, "learning_rate": 0.0004908177002618791, "loss": 2.4608, "step": 875 }, { "epoch": 0.11, "learning_rate": 0.0004907906416994146, "loss": 2.4207, "step": 876 }, { "epoch": 0.11, "learning_rate": 0.0004907635440753401, "loss": 2.3665, "step": 877 }, { "epoch": 0.11, "learning_rate": 0.0004907364073940516, "loss": 2.4159, "step": 878 }, { "epoch": 0.11, "learning_rate": 0.0004907092316599514, "loss": 2.3949, "step": 879 }, { "epoch": 0.11, "learning_rate": 0.0004906820168774477, "loss": 2.3292, "step": 880 }, { "epoch": 0.11, "learning_rate": 0.0004906547630509555, "loss": 2.3614, "step": 881 }, { "epoch": 0.11, "learning_rate": 0.000490627470184896, "loss": 2.3685, "step": 882 }, { "epoch": 0.11, "learning_rate": 0.0004906001382836967, "loss": 2.3475, "step": 883 }, { "epoch": 0.11, "learning_rate": 0.0004905727673517914, "loss": 2.3197, "step": 884 }, { "epoch": 0.11, "learning_rate": 0.0004905453573936202, "loss": 2.3486, "step": 885 }, { "epoch": 0.11, "learning_rate": 0.0004905179084136299, "loss": 2.3722, "step": 886 }, { "epoch": 0.11, "learning_rate": 0.000490490420416273, "loss": 2.3114, "step": 887 }, { "epoch": 0.11, "learning_rate": 0.0004904628934060088, "loss": 2.3314, "step": 888 }, { "epoch": 0.11, "learning_rate": 0.0004904353273873028, "loss": 2.2205, "step": 889 }, { "epoch": 0.11, "learning_rate": 0.000490407722364627, "loss": 2.3302, "step": 890 }, { "epoch": 0.11, "learning_rate": 0.0004903800783424592, "loss": 2.3282, "step": 891 }, { "epoch": 0.11, "learning_rate": 0.0004903523953252841, "loss": 2.318, "step": 892 }, { "epoch": 0.11, "learning_rate": 0.0004903246733175924, "loss": 2.3199, "step": 893 }, { "epoch": 0.11, "learning_rate": 0.0004902969123238814, "loss": 2.2917, "step": 894 }, { "epoch": 0.11, "learning_rate": 0.0004902691123486544, "loss": 2.265, "step": 895 }, { "epoch": 0.11, "learning_rate": 0.0004902412733964212, "loss": 2.3483, "step": 896 }, { "epoch": 0.11, "learning_rate": 0.0004902133954716978, "loss": 2.4319, "step": 897 }, { "epoch": 0.11, "learning_rate": 0.0004901854785790068, "loss": 2.389, "step": 898 }, { "epoch": 0.11, "learning_rate": 0.0004901575227228769, "loss": 2.3591, "step": 899 }, { "epoch": 0.11, "learning_rate": 0.000490129527907843, "loss": 2.2699, "step": 900 }, { "epoch": 0.11, "eval_acceptance_rate": 0.5242845416069031, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.6150616227837744, "eval_expected_tokens": 2.018827028479718, "eval_loss": 2.317944288253784, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.25470675711992946, "eval_runtime": 50.2622, "eval_samples_per_second": 9.689, "eval_steps_per_second": 0.318, "step": 900 }, { "epoch": 0.11, "learning_rate": 0.0004901014941384468, "loss": 2.3721, "step": 901 }, { "epoch": 0.11, "learning_rate": 0.0004900734214192358, "loss": 2.3032, "step": 902 }, { "epoch": 0.11, "learning_rate": 0.000490045309754764, "loss": 2.6101, "step": 903 }, { "epoch": 0.11, "learning_rate": 0.0004900171591495918, "loss": 2.1929, "step": 904 }, { "epoch": 0.11, "learning_rate": 0.0004899889696082857, "loss": 2.273, "step": 905 }, { "epoch": 0.11, "learning_rate": 0.000489960741135419, "loss": 2.3064, "step": 906 }, { "epoch": 0.11, "learning_rate": 0.0004899324737355707, "loss": 2.3934, "step": 907 }, { "epoch": 0.11, "learning_rate": 0.0004899041674133265, "loss": 2.3381, "step": 908 }, { "epoch": 0.11, "learning_rate": 0.0004898758221732783, "loss": 2.44, "step": 909 }, { "epoch": 0.11, "learning_rate": 0.0004898474380200244, "loss": 2.2716, "step": 910 }, { "epoch": 0.11, "learning_rate": 0.0004898190149581693, "loss": 2.266, "step": 911 }, { "epoch": 0.11, "learning_rate": 0.0004897905529923237, "loss": 2.3759, "step": 912 }, { "epoch": 0.11, "learning_rate": 0.000489762052127105, "loss": 2.4271, "step": 913 }, { "epoch": 0.11, "learning_rate": 0.0004897335123671366, "loss": 2.3738, "step": 914 }, { "epoch": 0.11, "learning_rate": 0.0004897049337170483, "loss": 2.3245, "step": 915 }, { "epoch": 0.11, "learning_rate": 0.000489676316181476, "loss": 2.3624, "step": 916 }, { "epoch": 0.11, "learning_rate": 0.0004896476597650624, "loss": 2.3344, "step": 917 }, { "epoch": 0.11, "learning_rate": 0.0004896189644724562, "loss": 2.3084, "step": 918 }, { "epoch": 0.11, "learning_rate": 0.0004895902303083121, "loss": 2.2705, "step": 919 }, { "epoch": 0.12, "learning_rate": 0.0004895614572772916, "loss": 2.3792, "step": 920 }, { "epoch": 0.12, "learning_rate": 0.0004895326453840624, "loss": 2.296, "step": 921 }, { "epoch": 0.12, "learning_rate": 0.0004895037946332983, "loss": 2.3146, "step": 922 }, { "epoch": 0.12, "learning_rate": 0.0004894749050296796, "loss": 2.2718, "step": 923 }, { "epoch": 0.12, "learning_rate": 0.0004894459765778929, "loss": 2.2702, "step": 924 }, { "epoch": 0.12, "learning_rate": 0.000489417009282631, "loss": 2.1853, "step": 925 }, { "epoch": 0.12, "learning_rate": 0.000489388003148593, "loss": 2.3964, "step": 926 }, { "epoch": 0.12, "learning_rate": 0.0004893589581804844, "loss": 2.3072, "step": 927 }, { "epoch": 0.12, "learning_rate": 0.0004893298743830168, "loss": 2.2686, "step": 928 }, { "epoch": 0.12, "learning_rate": 0.0004893007517609084, "loss": 2.3391, "step": 929 }, { "epoch": 0.12, "learning_rate": 0.0004892715903188836, "loss": 2.292, "step": 930 }, { "epoch": 0.12, "learning_rate": 0.0004892423900616727, "loss": 2.2109, "step": 931 }, { "epoch": 0.12, "learning_rate": 0.0004892131509940129, "loss": 2.3222, "step": 932 }, { "epoch": 0.12, "learning_rate": 0.0004891838731206475, "loss": 2.3179, "step": 933 }, { "epoch": 0.12, "learning_rate": 0.0004891545564463259, "loss": 2.3498, "step": 934 }, { "epoch": 0.12, "learning_rate": 0.0004891252009758039, "loss": 2.4434, "step": 935 }, { "epoch": 0.12, "learning_rate": 0.0004890958067138436, "loss": 2.3111, "step": 936 }, { "epoch": 0.12, "learning_rate": 0.0004890663736652135, "loss": 2.3946, "step": 937 }, { "epoch": 0.12, "learning_rate": 0.0004890369018346882, "loss": 2.2936, "step": 938 }, { "epoch": 0.12, "learning_rate": 0.0004890073912270487, "loss": 2.2816, "step": 939 }, { "epoch": 0.12, "learning_rate": 0.0004889778418470823, "loss": 2.3206, "step": 940 }, { "epoch": 0.12, "learning_rate": 0.0004889482536995825, "loss": 2.2007, "step": 941 }, { "epoch": 0.12, "learning_rate": 0.0004889186267893493, "loss": 2.3888, "step": 942 }, { "epoch": 0.12, "learning_rate": 0.0004888889611211889, "loss": 2.3029, "step": 943 }, { "epoch": 0.12, "learning_rate": 0.0004888592566999134, "loss": 2.2509, "step": 944 }, { "epoch": 0.12, "learning_rate": 0.0004888295135303418, "loss": 2.2415, "step": 945 }, { "epoch": 0.12, "learning_rate": 0.000488799731617299, "loss": 2.2378, "step": 946 }, { "epoch": 0.12, "learning_rate": 0.0004887699109656164, "loss": 2.3195, "step": 947 }, { "epoch": 0.12, "learning_rate": 0.0004887400515801314, "loss": 2.3061, "step": 948 }, { "epoch": 0.12, "learning_rate": 0.0004887101534656881, "loss": 2.2976, "step": 949 }, { "epoch": 0.12, "learning_rate": 0.0004886802166271364, "loss": 2.32, "step": 950 }, { "epoch": 0.12, "learning_rate": 0.0004886502410693329, "loss": 2.3052, "step": 951 }, { "epoch": 0.12, "learning_rate": 0.00048862022679714, "loss": 2.1541, "step": 952 }, { "epoch": 0.12, "learning_rate": 0.0004885901738154271, "loss": 2.3202, "step": 953 }, { "epoch": 0.12, "learning_rate": 0.0004885600821290692, "loss": 2.2045, "step": 954 }, { "epoch": 0.12, "learning_rate": 0.0004885299517429479, "loss": 2.1987, "step": 955 }, { "epoch": 0.12, "learning_rate": 0.0004884997826619511, "loss": 2.1913, "step": 956 }, { "epoch": 0.12, "learning_rate": 0.0004884695748909728, "loss": 2.2817, "step": 957 }, { "epoch": 0.12, "learning_rate": 0.0004884393284349134, "loss": 2.1885, "step": 958 }, { "epoch": 0.12, "learning_rate": 0.0004884090432986795, "loss": 2.1436, "step": 959 }, { "epoch": 0.12, "learning_rate": 0.0004883787194871841, "loss": 2.3033, "step": 960 }, { "epoch": 0.12, "learning_rate": 0.0004883483570053464, "loss": 2.2636, "step": 961 }, { "epoch": 0.12, "learning_rate": 0.0004883179558580918, "loss": 2.2126, "step": 962 }, { "epoch": 0.12, "learning_rate": 0.0004882875160503521, "loss": 2.195, "step": 963 }, { "epoch": 0.12, "learning_rate": 0.0004882570375870653, "loss": 2.2457, "step": 964 }, { "epoch": 0.12, "learning_rate": 0.0004882265204731757, "loss": 2.2694, "step": 965 }, { "epoch": 0.12, "learning_rate": 0.00048819596471363387, "loss": 2.2109, "step": 966 }, { "epoch": 0.12, "learning_rate": 0.0004881653703133966, "loss": 2.2042, "step": 967 }, { "epoch": 0.12, "learning_rate": 0.000488134737277427, "loss": 2.1759, "step": 968 }, { "epoch": 0.12, "learning_rate": 0.0004881040656106944, "loss": 2.1361, "step": 969 }, { "epoch": 0.12, "learning_rate": 0.0004880733553181744, "loss": 2.2487, "step": 970 }, { "epoch": 0.12, "learning_rate": 0.000488042606404849, "loss": 2.2892, "step": 971 }, { "epoch": 0.12, "learning_rate": 0.0004880118188757064, "loss": 2.1841, "step": 972 }, { "epoch": 0.12, "learning_rate": 0.00048798099273574083, "loss": 2.2648, "step": 973 }, { "epoch": 0.12, "learning_rate": 0.00048795012798995307, "loss": 2.2312, "step": 974 }, { "epoch": 0.12, "learning_rate": 0.00048791922464335016, "loss": 2.1314, "step": 975 }, { "epoch": 0.12, "learning_rate": 0.0004878882827009452, "loss": 2.1926, "step": 976 }, { "epoch": 0.12, "learning_rate": 0.00048785730216775763, "loss": 2.259, "step": 977 }, { "epoch": 0.12, "learning_rate": 0.00048782628304881324, "loss": 2.1977, "step": 978 }, { "epoch": 0.12, "learning_rate": 0.0004877952253491441, "loss": 2.2437, "step": 979 }, { "epoch": 0.12, "learning_rate": 0.0004877641290737884, "loss": 2.2619, "step": 980 }, { "epoch": 0.12, "learning_rate": 0.0004877329942277906, "loss": 2.2439, "step": 981 }, { "epoch": 0.12, "learning_rate": 0.0004877018208162015, "loss": 2.2498, "step": 982 }, { "epoch": 0.12, "learning_rate": 0.0004876706088440782, "loss": 2.3278, "step": 983 }, { "epoch": 0.12, "learning_rate": 0.00048763935831648374, "loss": 2.2618, "step": 984 }, { "epoch": 0.12, "learning_rate": 0.000487608069238488, "loss": 2.2946, "step": 985 }, { "epoch": 0.12, "learning_rate": 0.0004875767416151665, "loss": 2.2909, "step": 986 }, { "epoch": 0.12, "learning_rate": 0.0004875453754516014, "loss": 2.2101, "step": 987 }, { "epoch": 0.12, "learning_rate": 0.0004875139707528808, "loss": 2.3741, "step": 988 }, { "epoch": 0.12, "learning_rate": 0.0004874825275240996, "loss": 2.2475, "step": 989 }, { "epoch": 0.12, "learning_rate": 0.0004874510457703583, "loss": 2.2518, "step": 990 }, { "epoch": 0.12, "learning_rate": 0.00048741952549676405, "loss": 2.2793, "step": 991 }, { "epoch": 0.12, "learning_rate": 0.00048738796670843004, "loss": 2.2289, "step": 992 }, { "epoch": 0.12, "learning_rate": 0.000487356369410476, "loss": 2.2139, "step": 993 }, { "epoch": 0.12, "learning_rate": 0.00048732473360802754, "loss": 2.158, "step": 994 }, { "epoch": 0.12, "learning_rate": 0.00048729305930621676, "loss": 2.2756, "step": 995 }, { "epoch": 0.12, "learning_rate": 0.00048726134651018194, "loss": 2.2387, "step": 996 }, { "epoch": 0.12, "learning_rate": 0.0004872295952250676, "loss": 2.2838, "step": 997 }, { "epoch": 0.12, "learning_rate": 0.00048719780545602455, "loss": 2.1813, "step": 998 }, { "epoch": 0.12, "learning_rate": 0.0004871659772082097, "loss": 2.1755, "step": 999 }, { "epoch": 0.12, "learning_rate": 0.00048713411048678635, "loss": 2.2461, "step": 1000 }, { "epoch": 0.12, "eval_acceptance_rate": 0.5451273918151855, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.6740718399352261, "eval_expected_tokens": 2.0925897999190326, "eval_loss": 2.1608245372772217, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2731474499797582, "eval_runtime": 50.3019, "eval_samples_per_second": 9.682, "eval_steps_per_second": 0.318, "step": 1000 }, { "epoch": 0.13, "learning_rate": 0.00048710220529692405, "loss": 2.3047, "step": 1001 }, { "epoch": 0.13, "learning_rate": 0.00048707026164379846, "loss": 2.3149, "step": 1002 }, { "epoch": 0.13, "learning_rate": 0.0004870382795325916, "loss": 2.208, "step": 1003 }, { "epoch": 0.13, "learning_rate": 0.0004870062589684916, "loss": 2.2635, "step": 1004 }, { "epoch": 0.13, "learning_rate": 0.000486974199956693, "loss": 2.1919, "step": 1005 }, { "epoch": 0.13, "learning_rate": 0.00048694210250239646, "loss": 2.1671, "step": 1006 }, { "epoch": 0.13, "learning_rate": 0.00048690996661080886, "loss": 2.2721, "step": 1007 }, { "epoch": 0.13, "learning_rate": 0.0004868777922871434, "loss": 2.1514, "step": 1008 }, { "epoch": 0.13, "learning_rate": 0.00048684557953661945, "loss": 2.3194, "step": 1009 }, { "epoch": 0.13, "learning_rate": 0.0004868133283644627, "loss": 2.2897, "step": 1010 }, { "epoch": 0.13, "learning_rate": 0.0004867810387759049, "loss": 2.3235, "step": 1011 }, { "epoch": 0.13, "learning_rate": 0.00048674871077618423, "loss": 2.1683, "step": 1012 }, { "epoch": 0.13, "learning_rate": 0.00048671634437054504, "loss": 2.2024, "step": 1013 }, { "epoch": 0.13, "learning_rate": 0.00048668393956423773, "loss": 2.2384, "step": 1014 }, { "epoch": 0.13, "learning_rate": 0.00048665149636251923, "loss": 2.2905, "step": 1015 }, { "epoch": 0.13, "learning_rate": 0.00048661901477065246, "loss": 2.2377, "step": 1016 }, { "epoch": 0.13, "learning_rate": 0.0004865864947939067, "loss": 2.2592, "step": 1017 }, { "epoch": 0.13, "learning_rate": 0.0004865539364375574, "loss": 2.2395, "step": 1018 }, { "epoch": 0.13, "learning_rate": 0.00048652133970688633, "loss": 2.2923, "step": 1019 }, { "epoch": 0.13, "learning_rate": 0.0004864887046071813, "loss": 2.1268, "step": 1020 }, { "epoch": 0.13, "learning_rate": 0.0004864560311437365, "loss": 2.2603, "step": 1021 }, { "epoch": 0.13, "learning_rate": 0.00048642331932185226, "loss": 2.3932, "step": 1022 }, { "epoch": 0.13, "learning_rate": 0.0004863905691468352, "loss": 2.122, "step": 1023 }, { "epoch": 0.13, "learning_rate": 0.00048635778062399815, "loss": 2.3259, "step": 1024 }, { "epoch": 0.13, "learning_rate": 0.00048632495375866005, "loss": 2.2195, "step": 1025 }, { "epoch": 0.13, "learning_rate": 0.0004862920885561463, "loss": 2.2162, "step": 1026 }, { "epoch": 0.13, "learning_rate": 0.00048625918502178815, "loss": 2.1973, "step": 1027 }, { "epoch": 0.13, "learning_rate": 0.0004862262431609235, "loss": 2.1951, "step": 1028 }, { "epoch": 0.13, "learning_rate": 0.00048619326297889614, "loss": 2.3542, "step": 1029 }, { "epoch": 0.13, "learning_rate": 0.00048616024448105624, "loss": 2.1825, "step": 1030 }, { "epoch": 0.13, "learning_rate": 0.0004861271876727601, "loss": 2.1902, "step": 1031 }, { "epoch": 0.13, "learning_rate": 0.0004860940925593703, "loss": 2.0909, "step": 1032 }, { "epoch": 0.13, "learning_rate": 0.0004860609591462556, "loss": 2.2213, "step": 1033 }, { "epoch": 0.13, "learning_rate": 0.0004860277874387909, "loss": 2.2008, "step": 1034 }, { "epoch": 0.13, "learning_rate": 0.0004859945774423575, "loss": 2.3241, "step": 1035 }, { "epoch": 0.13, "learning_rate": 0.00048596132916234276, "loss": 2.1045, "step": 1036 }, { "epoch": 0.13, "learning_rate": 0.00048592804260414027, "loss": 2.1322, "step": 1037 }, { "epoch": 0.13, "learning_rate": 0.0004858947177731499, "loss": 2.3078, "step": 1038 }, { "epoch": 0.13, "learning_rate": 0.00048586135467477764, "loss": 2.0971, "step": 1039 }, { "epoch": 0.13, "learning_rate": 0.00048582795331443573, "loss": 2.1726, "step": 1040 }, { "epoch": 0.13, "learning_rate": 0.0004857945136975427, "loss": 2.2521, "step": 1041 }, { "epoch": 0.13, "learning_rate": 0.00048576103582952313, "loss": 2.1355, "step": 1042 }, { "epoch": 0.13, "learning_rate": 0.0004857275197158078, "loss": 2.1986, "step": 1043 }, { "epoch": 0.13, "learning_rate": 0.00048569396536183386, "loss": 2.1379, "step": 1044 }, { "epoch": 0.13, "learning_rate": 0.00048566037277304465, "loss": 2.2648, "step": 1045 }, { "epoch": 0.13, "learning_rate": 0.0004856267419548894, "loss": 2.2322, "step": 1046 }, { "epoch": 0.13, "learning_rate": 0.00048559307291282404, "loss": 2.2792, "step": 1047 }, { "epoch": 0.13, "learning_rate": 0.0004855593656523103, "loss": 2.165, "step": 1048 }, { "epoch": 0.13, "learning_rate": 0.0004855256201788163, "loss": 2.2136, "step": 1049 }, { "epoch": 0.13, "learning_rate": 0.0004854918364978163, "loss": 2.2543, "step": 1050 }, { "epoch": 0.13, "learning_rate": 0.0004854580146147907, "loss": 2.249, "step": 1051 }, { "epoch": 0.13, "learning_rate": 0.0004854241545352262, "loss": 2.1837, "step": 1052 }, { "epoch": 0.13, "learning_rate": 0.00048539025626461563, "loss": 2.0697, "step": 1053 }, { "epoch": 0.13, "learning_rate": 0.0004853563198084582, "loss": 2.133, "step": 1054 }, { "epoch": 0.13, "learning_rate": 0.00048532234517225894, "loss": 2.1667, "step": 1055 }, { "epoch": 0.13, "learning_rate": 0.00048528833236152947, "loss": 2.1446, "step": 1056 }, { "epoch": 0.13, "learning_rate": 0.0004852542813817873, "loss": 2.1837, "step": 1057 }, { "epoch": 0.13, "learning_rate": 0.0004852201922385564, "loss": 2.2129, "step": 1058 }, { "epoch": 0.13, "learning_rate": 0.0004851860649373666, "loss": 2.2304, "step": 1059 }, { "epoch": 0.13, "learning_rate": 0.0004851518994837543, "loss": 2.133, "step": 1060 }, { "epoch": 0.13, "learning_rate": 0.00048511769588326185, "loss": 2.2201, "step": 1061 }, { "epoch": 0.13, "learning_rate": 0.00048508345414143765, "loss": 2.2103, "step": 1062 }, { "epoch": 0.13, "learning_rate": 0.0004850491742638367, "loss": 2.1534, "step": 1063 }, { "epoch": 0.13, "learning_rate": 0.00048501485625601995, "loss": 2.1387, "step": 1064 }, { "epoch": 0.13, "learning_rate": 0.00048498050012355437, "loss": 2.1209, "step": 1065 }, { "epoch": 0.13, "learning_rate": 0.00048494610587201346, "loss": 2.3126, "step": 1066 }, { "epoch": 0.13, "learning_rate": 0.0004849116735069766, "loss": 2.2354, "step": 1067 }, { "epoch": 0.13, "learning_rate": 0.0004848772030340297, "loss": 2.1696, "step": 1068 }, { "epoch": 0.13, "learning_rate": 0.0004848426944587644, "loss": 2.167, "step": 1069 }, { "epoch": 0.13, "learning_rate": 0.00048480814778677884, "loss": 2.1708, "step": 1070 }, { "epoch": 0.13, "learning_rate": 0.00048477356302367724, "loss": 2.1646, "step": 1071 }, { "epoch": 0.13, "learning_rate": 0.00048473894017507014, "loss": 2.1108, "step": 1072 }, { "epoch": 0.13, "learning_rate": 0.00048470427924657413, "loss": 2.1804, "step": 1073 }, { "epoch": 0.13, "learning_rate": 0.0004846695802438118, "loss": 2.1705, "step": 1074 }, { "epoch": 0.13, "learning_rate": 0.0004846348431724122, "loss": 2.1434, "step": 1075 }, { "epoch": 0.13, "learning_rate": 0.00048460006803801057, "loss": 2.1162, "step": 1076 }, { "epoch": 0.13, "learning_rate": 0.000484565254846248, "loss": 2.0987, "step": 1077 }, { "epoch": 0.13, "learning_rate": 0.00048453040360277213, "loss": 2.317, "step": 1078 }, { "epoch": 0.13, "learning_rate": 0.0004844955143132366, "loss": 2.1789, "step": 1079 }, { "epoch": 0.14, "learning_rate": 0.00048446058698330114, "loss": 2.1346, "step": 1080 }, { "epoch": 0.14, "learning_rate": 0.0004844256216186318, "loss": 2.1527, "step": 1081 }, { "epoch": 0.14, "learning_rate": 0.0004843906182249007, "loss": 2.1008, "step": 1082 }, { "epoch": 0.14, "learning_rate": 0.0004843555768077862, "loss": 2.0671, "step": 1083 }, { "epoch": 0.14, "learning_rate": 0.0004843204973729729, "loss": 2.236, "step": 1084 }, { "epoch": 0.14, "learning_rate": 0.0004842853799261513, "loss": 2.182, "step": 1085 }, { "epoch": 0.14, "learning_rate": 0.0004842502244730183, "loss": 2.2142, "step": 1086 }, { "epoch": 0.14, "learning_rate": 0.0004842150310192769, "loss": 2.1236, "step": 1087 }, { "epoch": 0.14, "learning_rate": 0.00048417979957063626, "loss": 2.1313, "step": 1088 }, { "epoch": 0.14, "learning_rate": 0.00048414453013281166, "loss": 2.1698, "step": 1089 }, { "epoch": 0.14, "learning_rate": 0.0004841092227115247, "loss": 2.2115, "step": 1090 }, { "epoch": 0.14, "learning_rate": 0.000484073877312503, "loss": 2.1329, "step": 1091 }, { "epoch": 0.14, "learning_rate": 0.0004840384939414803, "loss": 2.054, "step": 1092 }, { "epoch": 0.14, "learning_rate": 0.0004840030726041966, "loss": 2.0262, "step": 1093 }, { "epoch": 0.14, "learning_rate": 0.000483967613306398, "loss": 2.275, "step": 1094 }, { "epoch": 0.14, "learning_rate": 0.0004839321160538369, "loss": 2.0441, "step": 1095 }, { "epoch": 0.14, "learning_rate": 0.0004838965808522716, "loss": 2.2063, "step": 1096 }, { "epoch": 0.14, "learning_rate": 0.00048386100770746686, "loss": 2.0632, "step": 1097 }, { "epoch": 0.14, "learning_rate": 0.0004838253966251933, "loss": 2.2152, "step": 1098 }, { "epoch": 0.14, "learning_rate": 0.0004837897476112279, "loss": 2.2413, "step": 1099 }, { "epoch": 0.14, "learning_rate": 0.00048375406067135377, "loss": 2.1063, "step": 1100 }, { "epoch": 0.14, "eval_acceptance_rate": 0.5404502749443054, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.6605676215958924, "eval_expected_tokens": 2.0757095269948653, "eval_loss": 2.1191511154174805, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2689273817487164, "eval_runtime": 50.2722, "eval_samples_per_second": 9.687, "eval_steps_per_second": 0.318, "step": 1100 }, { "epoch": 0.14, "learning_rate": 0.00048371833581135995, "loss": 2.1968, "step": 1101 }, { "epoch": 0.14, "learning_rate": 0.000483682573037042, "loss": 2.1992, "step": 1102 }, { "epoch": 0.14, "learning_rate": 0.00048364677235420133, "loss": 1.9741, "step": 1103 }, { "epoch": 0.14, "learning_rate": 0.00048361093376864567, "loss": 2.1328, "step": 1104 }, { "epoch": 0.14, "learning_rate": 0.00048357505728618884, "loss": 2.1406, "step": 1105 }, { "epoch": 0.14, "learning_rate": 0.00048353914291265074, "loss": 2.097, "step": 1106 }, { "epoch": 0.14, "learning_rate": 0.0004835031906538575, "loss": 2.0423, "step": 1107 }, { "epoch": 0.14, "learning_rate": 0.00048346720051564143, "loss": 2.2075, "step": 1108 }, { "epoch": 0.14, "learning_rate": 0.0004834311725038409, "loss": 2.1808, "step": 1109 }, { "epoch": 0.14, "learning_rate": 0.00048339510662430044, "loss": 2.1907, "step": 1110 }, { "epoch": 0.14, "learning_rate": 0.00048335900288287075, "loss": 2.1036, "step": 1111 }, { "epoch": 0.14, "learning_rate": 0.0004833228612854087, "loss": 2.1252, "step": 1112 }, { "epoch": 0.14, "learning_rate": 0.00048328668183777725, "loss": 2.1676, "step": 1113 }, { "epoch": 0.14, "learning_rate": 0.00048325046454584543, "loss": 1.9986, "step": 1114 }, { "epoch": 0.14, "learning_rate": 0.00048321420941548857, "loss": 2.1838, "step": 1115 }, { "epoch": 0.14, "learning_rate": 0.0004831779164525881, "loss": 2.0705, "step": 1116 }, { "epoch": 0.14, "learning_rate": 0.0004831415856630314, "loss": 2.1621, "step": 1117 }, { "epoch": 0.14, "learning_rate": 0.0004831052170527123, "loss": 2.0965, "step": 1118 }, { "epoch": 0.14, "learning_rate": 0.0004830688106275305, "loss": 2.2221, "step": 1119 }, { "epoch": 0.14, "learning_rate": 0.00048303236639339196, "loss": 2.2025, "step": 1120 }, { "epoch": 0.14, "learning_rate": 0.00048299588435620877, "loss": 2.3216, "step": 1121 }, { "epoch": 0.14, "learning_rate": 0.00048295936452189903, "loss": 2.1471, "step": 1122 }, { "epoch": 0.14, "learning_rate": 0.0004829228068963872, "loss": 2.0402, "step": 1123 }, { "epoch": 0.14, "learning_rate": 0.00048288621148560374, "loss": 2.1771, "step": 1124 }, { "epoch": 0.14, "learning_rate": 0.00048284957829548516, "loss": 2.099, "step": 1125 }, { "epoch": 0.14, "learning_rate": 0.0004828129073319742, "loss": 2.1549, "step": 1126 }, { "epoch": 0.14, "learning_rate": 0.0004827761986010197, "loss": 2.1278, "step": 1127 }, { "epoch": 0.14, "learning_rate": 0.00048273945210857675, "loss": 2.0914, "step": 1128 }, { "epoch": 0.14, "learning_rate": 0.00048270266786060623, "loss": 2.1401, "step": 1129 }, { "epoch": 0.14, "learning_rate": 0.00048266584586307556, "loss": 2.1636, "step": 1130 }, { "epoch": 0.14, "learning_rate": 0.000482628986121958, "loss": 2.0262, "step": 1131 }, { "epoch": 0.14, "learning_rate": 0.000482592088643233, "loss": 2.0639, "step": 1132 }, { "epoch": 0.14, "learning_rate": 0.00048255515343288626, "loss": 2.0687, "step": 1133 }, { "epoch": 0.14, "learning_rate": 0.00048251818049690944, "loss": 2.0686, "step": 1134 }, { "epoch": 0.14, "learning_rate": 0.0004824811698413003, "loss": 2.0325, "step": 1135 }, { "epoch": 0.14, "learning_rate": 0.00048244412147206283, "loss": 1.998, "step": 1136 }, { "epoch": 0.14, "learning_rate": 0.00048240703539520724, "loss": 2.0777, "step": 1137 }, { "epoch": 0.14, "learning_rate": 0.0004823699116167495, "loss": 2.1605, "step": 1138 }, { "epoch": 0.14, "learning_rate": 0.00048233275014271203, "loss": 2.065, "step": 1139 }, { "epoch": 0.14, "learning_rate": 0.0004822955509791233, "loss": 2.0954, "step": 1140 }, { "epoch": 0.14, "learning_rate": 0.00048225831413201776, "loss": 2.0351, "step": 1141 }, { "epoch": 0.14, "learning_rate": 0.0004822210396074361, "loss": 2.1041, "step": 1142 }, { "epoch": 0.14, "learning_rate": 0.000482183727411425, "loss": 2.0967, "step": 1143 }, { "epoch": 0.14, "learning_rate": 0.00048214637755003743, "loss": 2.0417, "step": 1144 }, { "epoch": 0.14, "learning_rate": 0.00048210899002933236, "loss": 2.0904, "step": 1145 }, { "epoch": 0.14, "learning_rate": 0.0004820715648553748, "loss": 2.1614, "step": 1146 }, { "epoch": 0.14, "learning_rate": 0.00048203410203423604, "loss": 2.127, "step": 1147 }, { "epoch": 0.14, "learning_rate": 0.0004819966015719933, "loss": 2.1713, "step": 1148 }, { "epoch": 0.14, "learning_rate": 0.00048195906347473, "loss": 2.0859, "step": 1149 }, { "epoch": 0.14, "learning_rate": 0.00048192148774853574, "loss": 2.0235, "step": 1150 }, { "epoch": 0.14, "learning_rate": 0.0004818838743995061, "loss": 2.0495, "step": 1151 }, { "epoch": 0.14, "learning_rate": 0.0004818462234337427, "loss": 1.992, "step": 1152 }, { "epoch": 0.14, "learning_rate": 0.0004818085348573535, "loss": 2.1533, "step": 1153 }, { "epoch": 0.14, "learning_rate": 0.0004817708086764525, "loss": 2.1414, "step": 1154 }, { "epoch": 0.14, "learning_rate": 0.00048173304489715953, "loss": 2.2, "step": 1155 }, { "epoch": 0.14, "learning_rate": 0.00048169524352560076, "loss": 2.0505, "step": 1156 }, { "epoch": 0.14, "learning_rate": 0.00048165740456790844, "loss": 2.1224, "step": 1157 }, { "epoch": 0.14, "learning_rate": 0.000481619528030221, "loss": 2.1192, "step": 1158 }, { "epoch": 0.14, "learning_rate": 0.00048158161391868273, "loss": 2.0284, "step": 1159 }, { "epoch": 0.14, "learning_rate": 0.00048154366223944413, "loss": 2.1084, "step": 1160 }, { "epoch": 0.15, "learning_rate": 0.00048150567299866186, "loss": 2.0968, "step": 1161 }, { "epoch": 0.15, "learning_rate": 0.0004814676462024987, "loss": 2.0901, "step": 1162 }, { "epoch": 0.15, "learning_rate": 0.0004814295818571233, "loss": 2.0596, "step": 1163 }, { "epoch": 0.15, "learning_rate": 0.0004813914799687107, "loss": 2.0871, "step": 1164 }, { "epoch": 0.15, "learning_rate": 0.00048135334054344166, "loss": 2.0385, "step": 1165 }, { "epoch": 0.15, "learning_rate": 0.0004813151635875034, "loss": 2.0631, "step": 1166 }, { "epoch": 0.15, "learning_rate": 0.0004812769491070891, "loss": 1.9622, "step": 1167 }, { "epoch": 0.15, "learning_rate": 0.0004812386971083979, "loss": 2.0503, "step": 1168 }, { "epoch": 0.15, "learning_rate": 0.00048120040759763516, "loss": 2.0793, "step": 1169 }, { "epoch": 0.15, "learning_rate": 0.00048116208058101233, "loss": 2.1533, "step": 1170 }, { "epoch": 0.15, "learning_rate": 0.0004811237160647469, "loss": 2.3342, "step": 1171 }, { "epoch": 0.15, "learning_rate": 0.0004810853140550624, "loss": 2.1438, "step": 1172 }, { "epoch": 0.15, "learning_rate": 0.0004810468745581885, "loss": 2.1084, "step": 1173 }, { "epoch": 0.15, "learning_rate": 0.00048100839758036105, "loss": 2.1341, "step": 1174 }, { "epoch": 0.15, "learning_rate": 0.0004809698831278217, "loss": 2.1526, "step": 1175 }, { "epoch": 0.15, "learning_rate": 0.00048093133120681846, "loss": 2.1378, "step": 1176 }, { "epoch": 0.15, "learning_rate": 0.0004808927418236053, "loss": 2.0359, "step": 1177 }, { "epoch": 0.15, "learning_rate": 0.00048085411498444234, "loss": 2.0386, "step": 1178 }, { "epoch": 0.15, "learning_rate": 0.00048081545069559553, "loss": 2.0492, "step": 1179 }, { "epoch": 0.15, "learning_rate": 0.00048077674896333724, "loss": 2.1828, "step": 1180 }, { "epoch": 0.15, "learning_rate": 0.0004807380097939457, "loss": 2.0856, "step": 1181 }, { "epoch": 0.15, "learning_rate": 0.0004806992331937053, "loss": 2.0714, "step": 1182 }, { "epoch": 0.15, "learning_rate": 0.0004806604191689064, "loss": 2.0501, "step": 1183 }, { "epoch": 0.15, "learning_rate": 0.0004806215677258456, "loss": 2.2008, "step": 1184 }, { "epoch": 0.15, "learning_rate": 0.0004805826788708254, "loss": 2.1604, "step": 1185 }, { "epoch": 0.15, "learning_rate": 0.00048054375261015437, "loss": 2.0781, "step": 1186 }, { "epoch": 0.15, "learning_rate": 0.0004805047889501474, "loss": 2.2349, "step": 1187 }, { "epoch": 0.15, "learning_rate": 0.00048046578789712516, "loss": 2.062, "step": 1188 }, { "epoch": 0.15, "learning_rate": 0.0004804267494574145, "loss": 2.176, "step": 1189 }, { "epoch": 0.15, "learning_rate": 0.00048038767363734825, "loss": 2.1391, "step": 1190 }, { "epoch": 0.15, "learning_rate": 0.0004803485604432655, "loss": 2.1047, "step": 1191 }, { "epoch": 0.15, "learning_rate": 0.00048030940988151116, "loss": 2.1586, "step": 1192 }, { "epoch": 0.15, "learning_rate": 0.00048027022195843653, "loss": 2.1072, "step": 1193 }, { "epoch": 0.15, "learning_rate": 0.0004802309966803985, "loss": 2.1146, "step": 1194 }, { "epoch": 0.15, "learning_rate": 0.00048019173405376047, "loss": 2.1164, "step": 1195 }, { "epoch": 0.15, "learning_rate": 0.0004801524340848917, "loss": 2.0718, "step": 1196 }, { "epoch": 0.15, "learning_rate": 0.0004801130967801674, "loss": 2.1046, "step": 1197 }, { "epoch": 0.15, "learning_rate": 0.00048007372214596913, "loss": 1.9869, "step": 1198 }, { "epoch": 0.15, "learning_rate": 0.00048003431018868415, "loss": 2.0022, "step": 1199 }, { "epoch": 0.15, "learning_rate": 0.0004799948609147061, "loss": 2.1194, "step": 1200 }, { "epoch": 0.15, "eval_acceptance_rate": 0.5648292303085327, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.73267310176981, "eval_expected_tokens": 2.1658413772122627, "eval_loss": 2.0305747985839844, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2914603443030656, "eval_runtime": 50.2264, "eval_samples_per_second": 9.696, "eval_steps_per_second": 0.319, "step": 1200 }, { "epoch": 0.15, "learning_rate": 0.00047995537433043444, "loss": 2.0256, "step": 1201 }, { "epoch": 0.15, "learning_rate": 0.0004799158504422748, "loss": 2.04, "step": 1202 }, { "epoch": 0.15, "learning_rate": 0.0004798762892566389, "loss": 2.0342, "step": 1203 }, { "epoch": 0.15, "learning_rate": 0.0004798366907799444, "loss": 2.0947, "step": 1204 }, { "epoch": 0.15, "learning_rate": 0.00047979705501861506, "loss": 2.0264, "step": 1205 }, { "epoch": 0.15, "learning_rate": 0.00047975738197908056, "loss": 2.0105, "step": 1206 }, { "epoch": 0.15, "learning_rate": 0.000479717671667777, "loss": 2.0927, "step": 1207 }, { "epoch": 0.15, "learning_rate": 0.0004796779240911461, "loss": 1.8679, "step": 1208 }, { "epoch": 0.15, "learning_rate": 0.00047963813925563584, "loss": 2.0474, "step": 1209 }, { "epoch": 0.15, "learning_rate": 0.00047959831716770017, "loss": 1.9859, "step": 1210 }, { "epoch": 0.15, "learning_rate": 0.00047955845783379924, "loss": 1.9933, "step": 1211 }, { "epoch": 0.15, "learning_rate": 0.00047951856126039906, "loss": 2.0194, "step": 1212 }, { "epoch": 0.15, "learning_rate": 0.0004794786274539716, "loss": 2.0944, "step": 1213 }, { "epoch": 0.15, "learning_rate": 0.00047943865642099525, "loss": 2.1093, "step": 1214 }, { "epoch": 0.15, "learning_rate": 0.0004793986481679541, "loss": 2.0461, "step": 1215 }, { "epoch": 0.15, "learning_rate": 0.0004793586027013384, "loss": 2.0354, "step": 1216 }, { "epoch": 0.15, "learning_rate": 0.0004793185200276444, "loss": 2.0811, "step": 1217 }, { "epoch": 0.15, "learning_rate": 0.00047927840015337433, "loss": 2.0299, "step": 1218 }, { "epoch": 0.15, "learning_rate": 0.00047923824308503663, "loss": 2.1113, "step": 1219 }, { "epoch": 0.15, "learning_rate": 0.00047919804882914566, "loss": 2.035, "step": 1220 }, { "epoch": 0.15, "learning_rate": 0.00047915781739222175, "loss": 1.9477, "step": 1221 }, { "epoch": 0.15, "learning_rate": 0.0004791175487807914, "loss": 2.0153, "step": 1222 }, { "epoch": 0.15, "learning_rate": 0.0004790772430013871, "loss": 2.1353, "step": 1223 }, { "epoch": 0.15, "learning_rate": 0.0004790369000605472, "loss": 2.0747, "step": 1224 }, { "epoch": 0.15, "learning_rate": 0.0004789965199648165, "loss": 1.9526, "step": 1225 }, { "epoch": 0.15, "learning_rate": 0.0004789561027207453, "loss": 2.0274, "step": 1226 }, { "epoch": 0.15, "learning_rate": 0.00047891564833489034, "loss": 2.0296, "step": 1227 }, { "epoch": 0.15, "learning_rate": 0.000478875156813814, "loss": 2.0726, "step": 1228 }, { "epoch": 0.15, "learning_rate": 0.0004788346281640852, "loss": 2.1075, "step": 1229 }, { "epoch": 0.15, "learning_rate": 0.00047879406239227835, "loss": 2.1311, "step": 1230 }, { "epoch": 0.15, "learning_rate": 0.0004787534595049743, "loss": 2.0783, "step": 1231 }, { "epoch": 0.15, "learning_rate": 0.0004787128195087596, "loss": 2.059, "step": 1232 }, { "epoch": 0.15, "learning_rate": 0.00047867214241022703, "loss": 2.0808, "step": 1233 }, { "epoch": 0.15, "learning_rate": 0.00047863142821597535, "loss": 2.165, "step": 1234 }, { "epoch": 0.15, "learning_rate": 0.00047859067693260923, "loss": 2.1418, "step": 1235 }, { "epoch": 0.15, "learning_rate": 0.00047854988856673954, "loss": 2.0415, "step": 1236 }, { "epoch": 0.15, "learning_rate": 0.00047850906312498295, "loss": 2.0898, "step": 1237 }, { "epoch": 0.15, "learning_rate": 0.00047846820061396236, "loss": 2.0037, "step": 1238 }, { "epoch": 0.15, "learning_rate": 0.0004784273010403065, "loss": 2.0028, "step": 1239 }, { "epoch": 0.15, "learning_rate": 0.0004783863644106502, "loss": 2.0091, "step": 1240 }, { "epoch": 0.16, "learning_rate": 0.0004783453907316344, "loss": 2.1649, "step": 1241 }, { "epoch": 0.16, "learning_rate": 0.00047830438000990585, "loss": 1.9596, "step": 1242 }, { "epoch": 0.16, "learning_rate": 0.0004782633322521174, "loss": 1.9687, "step": 1243 }, { "epoch": 0.16, "learning_rate": 0.0004782222474649279, "loss": 2.1166, "step": 1244 }, { "epoch": 0.16, "learning_rate": 0.0004781811256550023, "loss": 1.9471, "step": 1245 }, { "epoch": 0.16, "learning_rate": 0.00047813996682901137, "loss": 1.989, "step": 1246 }, { "epoch": 0.16, "learning_rate": 0.00047809877099363204, "loss": 2.0513, "step": 1247 }, { "epoch": 0.16, "learning_rate": 0.0004780575381555472, "loss": 1.9824, "step": 1248 }, { "epoch": 0.16, "learning_rate": 0.00047801626832144574, "loss": 2.0096, "step": 1249 }, { "epoch": 0.16, "learning_rate": 0.00047797496149802257, "loss": 1.9854, "step": 1250 }, { "epoch": 0.16, "learning_rate": 0.00047793361769197845, "loss": 2.0235, "step": 1251 }, { "epoch": 0.16, "learning_rate": 0.0004778922369100204, "loss": 1.9868, "step": 1252 }, { "epoch": 0.16, "learning_rate": 0.0004778508191588613, "loss": 2.0503, "step": 1253 }, { "epoch": 0.16, "learning_rate": 0.00047780936444522003, "loss": 2.0683, "step": 1254 }, { "epoch": 0.16, "learning_rate": 0.0004777678727758213, "loss": 2.0631, "step": 1255 }, { "epoch": 0.16, "learning_rate": 0.00047772634415739625, "loss": 1.9552, "step": 1256 }, { "epoch": 0.16, "learning_rate": 0.00047768477859668157, "loss": 1.9891, "step": 1257 }, { "epoch": 0.16, "learning_rate": 0.0004776431761004202, "loss": 1.9841, "step": 1258 }, { "epoch": 0.16, "learning_rate": 0.0004776015366753609, "loss": 2.0524, "step": 1259 }, { "epoch": 0.16, "learning_rate": 0.0004775598603282587, "loss": 1.9877, "step": 1260 }, { "epoch": 0.16, "learning_rate": 0.00047751814706587415, "loss": 2.0575, "step": 1261 }, { "epoch": 0.16, "learning_rate": 0.00047747639689497436, "loss": 1.9819, "step": 1262 }, { "epoch": 0.16, "learning_rate": 0.0004774346098223319, "loss": 1.9907, "step": 1263 }, { "epoch": 0.16, "learning_rate": 0.00047739278585472573, "loss": 1.9739, "step": 1264 }, { "epoch": 0.16, "learning_rate": 0.00047735092499894054, "loss": 2.0292, "step": 1265 }, { "epoch": 0.16, "learning_rate": 0.00047730902726176715, "loss": 1.9804, "step": 1266 }, { "epoch": 0.16, "learning_rate": 0.0004772670926500022, "loss": 2.0023, "step": 1267 }, { "epoch": 0.16, "learning_rate": 0.00047722512117044863, "loss": 1.9715, "step": 1268 }, { "epoch": 0.16, "learning_rate": 0.000477183112829915, "loss": 1.9929, "step": 1269 }, { "epoch": 0.16, "learning_rate": 0.00047714106763521594, "loss": 2.0079, "step": 1270 }, { "epoch": 0.16, "learning_rate": 0.0004770989855931722, "loss": 2.0267, "step": 1271 }, { "epoch": 0.16, "learning_rate": 0.00047705686671061057, "loss": 1.9706, "step": 1272 }, { "epoch": 0.16, "learning_rate": 0.0004770147109943633, "loss": 2.0705, "step": 1273 }, { "epoch": 0.16, "learning_rate": 0.00047697251845126937, "loss": 1.9794, "step": 1274 }, { "epoch": 0.16, "learning_rate": 0.00047693028908817313, "loss": 1.9062, "step": 1275 }, { "epoch": 0.16, "learning_rate": 0.00047688802291192525, "loss": 1.9609, "step": 1276 }, { "epoch": 0.16, "learning_rate": 0.0004768457199293821, "loss": 2.047, "step": 1277 }, { "epoch": 0.16, "learning_rate": 0.0004768033801474063, "loss": 2.1037, "step": 1278 }, { "epoch": 0.16, "learning_rate": 0.0004767610035728662, "loss": 1.9554, "step": 1279 }, { "epoch": 0.16, "learning_rate": 0.00047671859021263636, "loss": 1.9949, "step": 1280 }, { "epoch": 0.16, "learning_rate": 0.0004766761400735971, "loss": 1.9734, "step": 1281 }, { "epoch": 0.16, "learning_rate": 0.0004766336531626347, "loss": 1.9667, "step": 1282 }, { "epoch": 0.16, "learning_rate": 0.0004765911294866416, "loss": 1.9572, "step": 1283 }, { "epoch": 0.16, "learning_rate": 0.00047654856905251597, "loss": 1.9425, "step": 1284 }, { "epoch": 0.16, "learning_rate": 0.0004765059718671623, "loss": 1.8457, "step": 1285 }, { "epoch": 0.16, "learning_rate": 0.0004764633379374905, "loss": 2.0096, "step": 1286 }, { "epoch": 0.16, "learning_rate": 0.0004764206672704169, "loss": 1.9966, "step": 1287 }, { "epoch": 0.16, "learning_rate": 0.0004763779598728636, "loss": 2.1996, "step": 1288 }, { "epoch": 0.16, "learning_rate": 0.00047633521575175873, "loss": 2.135, "step": 1289 }, { "epoch": 0.16, "learning_rate": 0.00047629243491403633, "loss": 2.0688, "step": 1290 }, { "epoch": 0.16, "learning_rate": 0.0004762496173666364, "loss": 1.947, "step": 1291 }, { "epoch": 0.16, "learning_rate": 0.0004762067631165049, "loss": 1.9984, "step": 1292 }, { "epoch": 0.16, "learning_rate": 0.00047616387217059376, "loss": 2.0485, "step": 1293 }, { "epoch": 0.16, "learning_rate": 0.00047612094453586073, "loss": 1.9649, "step": 1294 }, { "epoch": 0.16, "learning_rate": 0.0004760779802192698, "loss": 2.0221, "step": 1295 }, { "epoch": 0.16, "learning_rate": 0.0004760349792277906, "loss": 2.0103, "step": 1296 }, { "epoch": 0.16, "learning_rate": 0.000475991941568399, "loss": 1.9843, "step": 1297 }, { "epoch": 0.16, "learning_rate": 0.0004759488672480765, "loss": 2.0034, "step": 1298 }, { "epoch": 0.16, "learning_rate": 0.0004759057562738109, "loss": 1.9382, "step": 1299 }, { "epoch": 0.16, "learning_rate": 0.0004758626086525956, "loss": 1.9052, "step": 1300 }, { "epoch": 0.16, "eval_acceptance_rate": 0.5704622268676758, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.7499488868543056, "eval_expected_tokens": 2.187436108567882, "eval_loss": 2.0047531127929688, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2968590271419705, "eval_runtime": 50.2662, "eval_samples_per_second": 9.688, "eval_steps_per_second": 0.318, "step": 1300 }, { "epoch": 0.16, "learning_rate": 0.0004758194243914301, "loss": 2.0366, "step": 1301 }, { "epoch": 0.16, "learning_rate": 0.00047577620349732, "loss": 1.9801, "step": 1302 }, { "epoch": 0.16, "learning_rate": 0.0004757329459772766, "loss": 1.9413, "step": 1303 }, { "epoch": 0.16, "learning_rate": 0.00047568965183831727, "loss": 1.9294, "step": 1304 }, { "epoch": 0.16, "learning_rate": 0.0004756463210874652, "loss": 2.0207, "step": 1305 }, { "epoch": 0.16, "learning_rate": 0.00047560295373174965, "loss": 1.9945, "step": 1306 }, { "epoch": 0.16, "learning_rate": 0.00047555954977820584, "loss": 2.0022, "step": 1307 }, { "epoch": 0.16, "learning_rate": 0.00047551610923387465, "loss": 2.0403, "step": 1308 }, { "epoch": 0.16, "learning_rate": 0.00047547263210580336, "loss": 2.044, "step": 1309 }, { "epoch": 0.16, "learning_rate": 0.00047542911840104486, "loss": 1.9592, "step": 1310 }, { "epoch": 0.16, "learning_rate": 0.000475385568126658, "loss": 1.9635, "step": 1311 }, { "epoch": 0.16, "learning_rate": 0.0004753419812897076, "loss": 1.8858, "step": 1312 }, { "epoch": 0.16, "learning_rate": 0.00047529835789726437, "loss": 2.0137, "step": 1313 }, { "epoch": 0.16, "learning_rate": 0.00047525469795640506, "loss": 1.9357, "step": 1314 }, { "epoch": 0.16, "learning_rate": 0.00047521100147421237, "loss": 2.0566, "step": 1315 }, { "epoch": 0.16, "learning_rate": 0.0004751672684577747, "loss": 2.0275, "step": 1316 }, { "epoch": 0.16, "learning_rate": 0.0004751234989141866, "loss": 2.0624, "step": 1317 }, { "epoch": 0.16, "learning_rate": 0.0004750796928505484, "loss": 2.0306, "step": 1318 }, { "epoch": 0.16, "learning_rate": 0.0004750358502739665, "loss": 2.0308, "step": 1319 }, { "epoch": 0.17, "learning_rate": 0.00047499197119155303, "loss": 1.9405, "step": 1320 }, { "epoch": 0.17, "learning_rate": 0.0004749480556104263, "loss": 1.9651, "step": 1321 }, { "epoch": 0.17, "learning_rate": 0.0004749041035377103, "loss": 2.0002, "step": 1322 }, { "epoch": 0.17, "learning_rate": 0.00047486011498053504, "loss": 1.9308, "step": 1323 }, { "epoch": 0.17, "learning_rate": 0.00047481608994603654, "loss": 1.9959, "step": 1324 }, { "epoch": 0.17, "learning_rate": 0.00047477202844135646, "loss": 1.928, "step": 1325 }, { "epoch": 0.17, "learning_rate": 0.0004747279304736428, "loss": 1.8967, "step": 1326 }, { "epoch": 0.17, "learning_rate": 0.00047468379605004895, "loss": 2.0226, "step": 1327 }, { "epoch": 0.17, "learning_rate": 0.00047463962517773473, "loss": 2.0008, "step": 1328 }, { "epoch": 0.17, "learning_rate": 0.0004745954178638656, "loss": 1.9727, "step": 1329 }, { "epoch": 0.17, "learning_rate": 0.0004745511741156129, "loss": 2.0532, "step": 1330 }, { "epoch": 0.17, "learning_rate": 0.0004745068939401539, "loss": 2.0905, "step": 1331 }, { "epoch": 0.17, "learning_rate": 0.000474462577344672, "loss": 1.9939, "step": 1332 }, { "epoch": 0.17, "learning_rate": 0.00047441822433635624, "loss": 1.9316, "step": 1333 }, { "epoch": 0.17, "learning_rate": 0.0004743738349224016, "loss": 2.0206, "step": 1334 }, { "epoch": 0.17, "learning_rate": 0.0004743294091100092, "loss": 1.9107, "step": 1335 }, { "epoch": 0.17, "learning_rate": 0.00047428494690638577, "loss": 1.9791, "step": 1336 }, { "epoch": 0.17, "learning_rate": 0.0004742404483187441, "loss": 1.9755, "step": 1337 }, { "epoch": 0.17, "learning_rate": 0.0004741959133543029, "loss": 2.0302, "step": 1338 }, { "epoch": 0.17, "learning_rate": 0.0004741513420202867, "loss": 1.994, "step": 1339 }, { "epoch": 0.17, "learning_rate": 0.0004741067343239259, "loss": 2.0295, "step": 1340 }, { "epoch": 0.17, "learning_rate": 0.000474062090272457, "loss": 1.9746, "step": 1341 }, { "epoch": 0.17, "learning_rate": 0.0004740174098731221, "loss": 1.9155, "step": 1342 }, { "epoch": 0.17, "learning_rate": 0.00047397269313316953, "loss": 2.0722, "step": 1343 }, { "epoch": 0.17, "learning_rate": 0.0004739279400598532, "loss": 2.0063, "step": 1344 }, { "epoch": 0.17, "learning_rate": 0.00047388315066043316, "loss": 1.9951, "step": 1345 }, { "epoch": 0.17, "learning_rate": 0.00047383832494217514, "loss": 2.0392, "step": 1346 }, { "epoch": 0.17, "learning_rate": 0.00047379346291235094, "loss": 2.0369, "step": 1347 }, { "epoch": 0.17, "learning_rate": 0.0004737485645782382, "loss": 1.8495, "step": 1348 }, { "epoch": 0.17, "learning_rate": 0.00047370362994712036, "loss": 1.9573, "step": 1349 }, { "epoch": 0.17, "learning_rate": 0.00047365865902628684, "loss": 1.9017, "step": 1350 }, { "epoch": 0.17, "learning_rate": 0.00047361365182303295, "loss": 1.9573, "step": 1351 }, { "epoch": 0.17, "learning_rate": 0.0004735686083446599, "loss": 1.952, "step": 1352 }, { "epoch": 0.17, "learning_rate": 0.00047352352859847467, "loss": 1.9735, "step": 1353 }, { "epoch": 0.17, "learning_rate": 0.00047347841259179026, "loss": 1.9469, "step": 1354 }, { "epoch": 0.17, "learning_rate": 0.00047343326033192544, "loss": 1.9768, "step": 1355 }, { "epoch": 0.17, "learning_rate": 0.00047338807182620497, "loss": 1.9265, "step": 1356 }, { "epoch": 0.17, "learning_rate": 0.0004733428470819594, "loss": 1.973, "step": 1357 }, { "epoch": 0.17, "learning_rate": 0.0004732975861065252, "loss": 1.8367, "step": 1358 }, { "epoch": 0.17, "learning_rate": 0.00047325228890724466, "loss": 1.9677, "step": 1359 }, { "epoch": 0.17, "learning_rate": 0.00047320695549146605, "loss": 1.9306, "step": 1360 }, { "epoch": 0.17, "learning_rate": 0.0004731615858665435, "loss": 1.9733, "step": 1361 }, { "epoch": 0.17, "learning_rate": 0.0004731161800398368, "loss": 1.9889, "step": 1362 }, { "epoch": 0.17, "learning_rate": 0.00047307073801871203, "loss": 2.008, "step": 1363 }, { "epoch": 0.17, "learning_rate": 0.00047302525981054067, "loss": 2.094, "step": 1364 }, { "epoch": 0.17, "learning_rate": 0.0004729797454227005, "loss": 2.0486, "step": 1365 }, { "epoch": 0.17, "learning_rate": 0.00047293419486257486, "loss": 2.0539, "step": 1366 }, { "epoch": 0.17, "learning_rate": 0.00047288860813755304, "loss": 1.9142, "step": 1367 }, { "epoch": 0.17, "learning_rate": 0.0004728429852550303, "loss": 2.0448, "step": 1368 }, { "epoch": 0.17, "learning_rate": 0.0004727973262224076, "loss": 1.9976, "step": 1369 }, { "epoch": 0.17, "learning_rate": 0.00047275163104709196, "loss": 2.0063, "step": 1370 }, { "epoch": 0.17, "learning_rate": 0.00047270589973649616, "loss": 1.9192, "step": 1371 }, { "epoch": 0.17, "learning_rate": 0.00047266013229803864, "loss": 2.0313, "step": 1372 }, { "epoch": 0.17, "learning_rate": 0.00047261432873914414, "loss": 1.992, "step": 1373 }, { "epoch": 0.17, "learning_rate": 0.0004725684890672429, "loss": 1.9564, "step": 1374 }, { "epoch": 0.17, "learning_rate": 0.00047252261328977114, "loss": 2.0166, "step": 1375 }, { "epoch": 0.17, "learning_rate": 0.000472476701414171, "loss": 1.9403, "step": 1376 }, { "epoch": 0.17, "learning_rate": 0.00047243075344789026, "loss": 2.0356, "step": 1377 }, { "epoch": 0.17, "learning_rate": 0.0004723847693983828, "loss": 1.9637, "step": 1378 }, { "epoch": 0.17, "learning_rate": 0.0004723387492731084, "loss": 1.9293, "step": 1379 }, { "epoch": 0.17, "learning_rate": 0.0004722926930795323, "loss": 1.9436, "step": 1380 }, { "epoch": 0.17, "learning_rate": 0.000472246600825126, "loss": 1.919, "step": 1381 }, { "epoch": 0.17, "learning_rate": 0.00047220047251736664, "loss": 2.0956, "step": 1382 }, { "epoch": 0.17, "learning_rate": 0.0004721543081637372, "loss": 1.9784, "step": 1383 }, { "epoch": 0.17, "learning_rate": 0.0004721081077717267, "loss": 2.0714, "step": 1384 }, { "epoch": 0.17, "learning_rate": 0.00047206187134882975, "loss": 1.82, "step": 1385 }, { "epoch": 0.17, "learning_rate": 0.000472015598902547, "loss": 1.9254, "step": 1386 }, { "epoch": 0.17, "learning_rate": 0.0004719692904403849, "loss": 1.9169, "step": 1387 }, { "epoch": 0.17, "learning_rate": 0.0004719229459698556, "loss": 1.9157, "step": 1388 }, { "epoch": 0.17, "learning_rate": 0.00047187656549847734, "loss": 1.8864, "step": 1389 }, { "epoch": 0.17, "learning_rate": 0.000471830149033774, "loss": 1.958, "step": 1390 }, { "epoch": 0.17, "learning_rate": 0.00047178369658327536, "loss": 1.958, "step": 1391 }, { "epoch": 0.17, "learning_rate": 0.00047173720815451697, "loss": 1.9931, "step": 1392 }, { "epoch": 0.17, "learning_rate": 0.0004716906837550404, "loss": 1.9802, "step": 1393 }, { "epoch": 0.17, "learning_rate": 0.000471644123392393, "loss": 1.9512, "step": 1394 }, { "epoch": 0.17, "learning_rate": 0.00047159752707412776, "loss": 1.9706, "step": 1395 }, { "epoch": 0.17, "learning_rate": 0.00047155089480780364, "loss": 2.0195, "step": 1396 }, { "epoch": 0.17, "learning_rate": 0.00047150422660098557, "loss": 2.0977, "step": 1397 }, { "epoch": 0.17, "learning_rate": 0.0004714575224612441, "loss": 2.0333, "step": 1398 }, { "epoch": 0.17, "learning_rate": 0.00047141078239615564, "loss": 1.9539, "step": 1399 }, { "epoch": 0.17, "learning_rate": 0.00047136400641330245, "loss": 2.0168, "step": 1400 }, { "epoch": 0.17, "eval_acceptance_rate": 0.5731274485588074, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.7582051792518991, "eval_expected_tokens": 2.197756474064874, "eval_loss": 1.9467182159423828, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.2994391185162185, "eval_runtime": 50.2859, "eval_samples_per_second": 9.685, "eval_steps_per_second": 0.318, "step": 1400 }, { "epoch": 0.18, "learning_rate": 0.0004713171945202728, "loss": 2.0137, "step": 1401 }, { "epoch": 0.18, "learning_rate": 0.0004712703467246604, "loss": 1.9679, "step": 1402 }, { "epoch": 0.18, "learning_rate": 0.0004712234630340652, "loss": 1.9564, "step": 1403 }, { "epoch": 0.18, "learning_rate": 0.00047117654345609275, "loss": 2.0406, "step": 1404 }, { "epoch": 0.18, "learning_rate": 0.00047112958799835425, "loss": 1.9698, "step": 1405 }, { "epoch": 0.18, "learning_rate": 0.0004710825966684672, "loss": 1.9562, "step": 1406 }, { "epoch": 0.18, "learning_rate": 0.00047103556947405447, "loss": 1.9787, "step": 1407 }, { "epoch": 0.18, "learning_rate": 0.0004709885064227449, "loss": 2.0634, "step": 1408 }, { "epoch": 0.18, "learning_rate": 0.0004709414075221734, "loss": 2.019, "step": 1409 }, { "epoch": 0.18, "learning_rate": 0.00047089427277998016, "loss": 1.8711, "step": 1410 }, { "epoch": 0.18, "learning_rate": 0.00047084710220381164, "loss": 1.9645, "step": 1411 }, { "epoch": 0.18, "learning_rate": 0.00047079989580132, "loss": 1.897, "step": 1412 }, { "epoch": 0.18, "learning_rate": 0.00047075265358016303, "loss": 1.9659, "step": 1413 }, { "epoch": 0.18, "learning_rate": 0.00047070537554800456, "loss": 2.043, "step": 1414 }, { "epoch": 0.18, "learning_rate": 0.0004706580617125141, "loss": 1.994, "step": 1415 }, { "epoch": 0.18, "learning_rate": 0.0004706107120813671, "loss": 1.9603, "step": 1416 }, { "epoch": 0.18, "learning_rate": 0.0004705633266622447, "loss": 1.9907, "step": 1417 }, { "epoch": 0.18, "learning_rate": 0.0004705159054628338, "loss": 1.9258, "step": 1418 }, { "epoch": 0.18, "learning_rate": 0.0004704684484908272, "loss": 1.9418, "step": 1419 }, { "epoch": 0.18, "learning_rate": 0.0004704209557539235, "loss": 1.9022, "step": 1420 }, { "epoch": 0.18, "learning_rate": 0.000470373427259827, "loss": 1.9411, "step": 1421 }, { "epoch": 0.18, "learning_rate": 0.00047032586301624804, "loss": 2.0018, "step": 1422 }, { "epoch": 0.18, "learning_rate": 0.0004702782630309025, "loss": 1.9083, "step": 1423 }, { "epoch": 0.18, "learning_rate": 0.00047023062731151217, "loss": 1.9834, "step": 1424 }, { "epoch": 0.18, "learning_rate": 0.0004701829558658046, "loss": 2.0062, "step": 1425 }, { "epoch": 0.18, "learning_rate": 0.00047013524870151327, "loss": 1.8565, "step": 1426 }, { "epoch": 0.18, "learning_rate": 0.00047008750582637715, "loss": 1.8599, "step": 1427 }, { "epoch": 0.18, "learning_rate": 0.0004700397272481414, "loss": 2.0151, "step": 1428 }, { "epoch": 0.18, "learning_rate": 0.0004699919129745567, "loss": 2.0318, "step": 1429 }, { "epoch": 0.18, "learning_rate": 0.0004699440630133794, "loss": 1.9837, "step": 1430 }, { "epoch": 0.18, "learning_rate": 0.00046989617737237224, "loss": 1.9034, "step": 1431 }, { "epoch": 0.18, "learning_rate": 0.0004698482560593029, "loss": 1.9688, "step": 1432 }, { "epoch": 0.18, "learning_rate": 0.00046980029908194566, "loss": 1.9322, "step": 1433 }, { "epoch": 0.18, "learning_rate": 0.00046975230644808, "loss": 1.9215, "step": 1434 }, { "epoch": 0.18, "learning_rate": 0.00046970427816549133, "loss": 1.8351, "step": 1435 }, { "epoch": 0.18, "learning_rate": 0.00046965621424197116, "loss": 1.921, "step": 1436 }, { "epoch": 0.18, "learning_rate": 0.0004696081146853163, "loss": 1.9332, "step": 1437 }, { "epoch": 0.18, "learning_rate": 0.0004695599795033296, "loss": 1.9315, "step": 1438 }, { "epoch": 0.18, "learning_rate": 0.00046951180870381977, "loss": 1.978, "step": 1439 }, { "epoch": 0.18, "learning_rate": 0.00046946360229460117, "loss": 1.9053, "step": 1440 }, { "epoch": 0.18, "learning_rate": 0.00046941536028349384, "loss": 1.9573, "step": 1441 }, { "epoch": 0.18, "learning_rate": 0.0004693670826783238, "loss": 1.9311, "step": 1442 }, { "epoch": 0.18, "learning_rate": 0.0004693187694869228, "loss": 1.9775, "step": 1443 }, { "epoch": 0.18, "learning_rate": 0.0004692704207171282, "loss": 1.9462, "step": 1444 }, { "epoch": 0.18, "learning_rate": 0.0004692220363767833, "loss": 1.8953, "step": 1445 }, { "epoch": 0.18, "learning_rate": 0.0004691736164737371, "loss": 1.9797, "step": 1446 }, { "epoch": 0.18, "learning_rate": 0.00046912516101584446, "loss": 1.9418, "step": 1447 }, { "epoch": 0.18, "learning_rate": 0.0004690766700109659, "loss": 1.9342, "step": 1448 }, { "epoch": 0.18, "learning_rate": 0.0004690281434669677, "loss": 1.927, "step": 1449 }, { "epoch": 0.18, "learning_rate": 0.000468979581391722, "loss": 2.0168, "step": 1450 }, { "epoch": 0.18, "learning_rate": 0.00046893098379310655, "loss": 1.8734, "step": 1451 }, { "epoch": 0.18, "learning_rate": 0.0004688823506790051, "loss": 1.9379, "step": 1452 }, { "epoch": 0.18, "learning_rate": 0.00046883368205730694, "loss": 1.8334, "step": 1453 }, { "epoch": 0.18, "learning_rate": 0.00046878497793590725, "loss": 1.954, "step": 1454 }, { "epoch": 0.18, "learning_rate": 0.00046873623832270687, "loss": 1.8611, "step": 1455 }, { "epoch": 0.18, "learning_rate": 0.0004686874632256125, "loss": 1.8822, "step": 1456 }, { "epoch": 0.18, "learning_rate": 0.00046863865265253657, "loss": 1.8923, "step": 1457 }, { "epoch": 0.18, "learning_rate": 0.00046858980661139714, "loss": 1.8669, "step": 1458 }, { "epoch": 0.18, "learning_rate": 0.00046854092511011827, "loss": 1.9491, "step": 1459 }, { "epoch": 0.18, "learning_rate": 0.0004684920081566295, "loss": 1.9622, "step": 1460 }, { "epoch": 0.18, "learning_rate": 0.00046844305575886636, "loss": 1.9597, "step": 1461 }, { "epoch": 0.18, "learning_rate": 0.0004683940679247699, "loss": 1.8697, "step": 1462 }, { "epoch": 0.18, "learning_rate": 0.0004683450446622871, "loss": 1.9062, "step": 1463 }, { "epoch": 0.18, "learning_rate": 0.0004682959859793706, "loss": 1.8335, "step": 1464 }, { "epoch": 0.18, "learning_rate": 0.00046824689188397896, "loss": 2.0114, "step": 1465 }, { "epoch": 0.18, "learning_rate": 0.00046819776238407607, "loss": 1.7627, "step": 1466 }, { "epoch": 0.18, "learning_rate": 0.00046814859748763206, "loss": 1.9966, "step": 1467 }, { "epoch": 0.18, "learning_rate": 0.0004680993972026225, "loss": 1.8939, "step": 1468 }, { "epoch": 0.18, "learning_rate": 0.00046805016153702873, "loss": 1.8613, "step": 1469 }, { "epoch": 0.18, "learning_rate": 0.0004680008904988379, "loss": 1.8248, "step": 1470 }, { "epoch": 0.18, "learning_rate": 0.0004679515840960429, "loss": 1.9473, "step": 1471 }, { "epoch": 0.18, "learning_rate": 0.00046790224233664236, "loss": 1.8705, "step": 1472 }, { "epoch": 0.18, "learning_rate": 0.00046785286522864046, "loss": 1.8827, "step": 1473 }, { "epoch": 0.18, "learning_rate": 0.0004678034527800474, "loss": 1.8407, "step": 1474 }, { "epoch": 0.18, "learning_rate": 0.0004677540049988789, "loss": 1.8601, "step": 1475 }, { "epoch": 0.18, "learning_rate": 0.0004677045218931566, "loss": 1.8635, "step": 1476 }, { "epoch": 0.18, "learning_rate": 0.0004676550034709076, "loss": 1.9133, "step": 1477 }, { "epoch": 0.18, "learning_rate": 0.0004676054497401651, "loss": 1.8963, "step": 1478 }, { "epoch": 0.18, "learning_rate": 0.0004675558607089677, "loss": 1.9315, "step": 1479 }, { "epoch": 0.18, "learning_rate": 0.0004675062363853598, "loss": 1.9823, "step": 1480 }, { "epoch": 0.19, "learning_rate": 0.0004674565767773917, "loss": 1.9238, "step": 1481 }, { "epoch": 0.19, "learning_rate": 0.0004674068818931192, "loss": 1.8179, "step": 1482 }, { "epoch": 0.19, "learning_rate": 0.00046735715174060386, "loss": 1.8615, "step": 1483 }, { "epoch": 0.19, "learning_rate": 0.00046730738632791327, "loss": 1.9472, "step": 1484 }, { "epoch": 0.19, "learning_rate": 0.00046725758566312016, "loss": 1.8436, "step": 1485 }, { "epoch": 0.19, "learning_rate": 0.00046720774975430354, "loss": 1.8437, "step": 1486 }, { "epoch": 0.19, "learning_rate": 0.00046715787860954785, "loss": 1.8435, "step": 1487 }, { "epoch": 0.19, "learning_rate": 0.0004671079722369432, "loss": 1.8304, "step": 1488 }, { "epoch": 0.19, "learning_rate": 0.00046705803064458573, "loss": 2.0278, "step": 1489 }, { "epoch": 0.19, "learning_rate": 0.0004670080538405769, "loss": 1.9103, "step": 1490 }, { "epoch": 0.19, "learning_rate": 0.00046695804183302403, "loss": 2.0192, "step": 1491 }, { "epoch": 0.19, "learning_rate": 0.00046690799463004035, "loss": 1.8774, "step": 1492 }, { "epoch": 0.19, "learning_rate": 0.00046685791223974457, "loss": 1.8489, "step": 1493 }, { "epoch": 0.19, "learning_rate": 0.00046680779467026115, "loss": 2.0066, "step": 1494 }, { "epoch": 0.19, "learning_rate": 0.0004667576419297202, "loss": 2.0027, "step": 1495 }, { "epoch": 0.19, "learning_rate": 0.0004667074540262577, "loss": 1.879, "step": 1496 }, { "epoch": 0.19, "learning_rate": 0.00046665723096801526, "loss": 1.9749, "step": 1497 }, { "epoch": 0.19, "learning_rate": 0.0004666069727631401, "loss": 1.906, "step": 1498 }, { "epoch": 0.19, "learning_rate": 0.00046655667941978543, "loss": 1.9068, "step": 1499 }, { "epoch": 0.19, "learning_rate": 0.00046650635094610973, "loss": 2.0436, "step": 1500 }, { "epoch": 0.19, "eval_acceptance_rate": 0.584759533405304, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.7948668742830456, "eval_expected_tokens": 2.243583592853807, "eval_loss": 1.8825244903564453, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3108958982134517, "eval_runtime": 50.277, "eval_samples_per_second": 9.686, "eval_steps_per_second": 0.318, "step": 1500 }, { "epoch": 0.19, "learning_rate": 0.0004664559873502774, "loss": 1.9757, "step": 1501 }, { "epoch": 0.19, "learning_rate": 0.0004664055886404586, "loss": 1.9346, "step": 1502 }, { "epoch": 0.19, "learning_rate": 0.0004663551548248292, "loss": 1.9135, "step": 1503 }, { "epoch": 0.19, "learning_rate": 0.0004663046859115706, "loss": 1.9771, "step": 1504 }, { "epoch": 0.19, "learning_rate": 0.00046625418190887, "loss": 1.8901, "step": 1505 }, { "epoch": 0.19, "learning_rate": 0.00046620364282492024, "loss": 1.9423, "step": 1506 }, { "epoch": 0.19, "learning_rate": 0.00046615306866791995, "loss": 1.8635, "step": 1507 }, { "epoch": 0.19, "learning_rate": 0.00046610245944607334, "loss": 1.9227, "step": 1508 }, { "epoch": 0.19, "learning_rate": 0.0004660518151675904, "loss": 1.8608, "step": 1509 }, { "epoch": 0.19, "learning_rate": 0.0004660011358406867, "loss": 1.9608, "step": 1510 }, { "epoch": 0.19, "learning_rate": 0.0004659504214735837, "loss": 1.9144, "step": 1511 }, { "epoch": 0.19, "learning_rate": 0.0004658996720745081, "loss": 1.8957, "step": 1512 }, { "epoch": 0.19, "learning_rate": 0.0004658488876516929, "loss": 1.929, "step": 1513 }, { "epoch": 0.19, "learning_rate": 0.0004657980682133763, "loss": 1.8177, "step": 1514 }, { "epoch": 0.19, "learning_rate": 0.0004657472137678024, "loss": 1.8487, "step": 1515 }, { "epoch": 0.19, "learning_rate": 0.00046569632432322093, "loss": 1.804, "step": 1516 }, { "epoch": 0.19, "learning_rate": 0.0004656453998878873, "loss": 1.9499, "step": 1517 }, { "epoch": 0.19, "learning_rate": 0.0004655944404700625, "loss": 1.8798, "step": 1518 }, { "epoch": 0.19, "learning_rate": 0.00046554344607801335, "loss": 1.8885, "step": 1519 }, { "epoch": 0.19, "learning_rate": 0.0004654924167200123, "loss": 1.879, "step": 1520 }, { "epoch": 0.19, "learning_rate": 0.00046544135240433746, "loss": 1.9308, "step": 1521 }, { "epoch": 0.19, "learning_rate": 0.0004653902531392725, "loss": 1.8947, "step": 1522 }, { "epoch": 0.19, "learning_rate": 0.00046533911893310685, "loss": 1.9088, "step": 1523 }, { "epoch": 0.19, "learning_rate": 0.0004652879497941358, "loss": 1.9055, "step": 1524 }, { "epoch": 0.19, "learning_rate": 0.00046523674573066, "loss": 1.8658, "step": 1525 }, { "epoch": 0.19, "learning_rate": 0.0004651855067509859, "loss": 1.8399, "step": 1526 }, { "epoch": 0.19, "learning_rate": 0.0004651342328634256, "loss": 1.9078, "step": 1527 }, { "epoch": 0.19, "learning_rate": 0.00046508292407629694, "loss": 1.8406, "step": 1528 }, { "epoch": 0.19, "learning_rate": 0.0004650315803979232, "loss": 1.9243, "step": 1529 }, { "epoch": 0.19, "learning_rate": 0.0004649802018366336, "loss": 1.9493, "step": 1530 }, { "epoch": 0.19, "learning_rate": 0.0004649287884007628, "loss": 1.8346, "step": 1531 }, { "epoch": 0.19, "learning_rate": 0.00046487734009865127, "loss": 1.9422, "step": 1532 }, { "epoch": 0.19, "learning_rate": 0.0004648258569386451, "loss": 1.9317, "step": 1533 }, { "epoch": 0.19, "learning_rate": 0.00046477433892909594, "loss": 1.8978, "step": 1534 }, { "epoch": 0.19, "learning_rate": 0.00046472278607836114, "loss": 2.0078, "step": 1535 }, { "epoch": 0.19, "learning_rate": 0.0004646711983948038, "loss": 2.0032, "step": 1536 }, { "epoch": 0.19, "learning_rate": 0.0004646195758867925, "loss": 1.9498, "step": 1537 }, { "epoch": 0.19, "learning_rate": 0.00046456791856270164, "loss": 1.9174, "step": 1538 }, { "epoch": 0.19, "learning_rate": 0.0004645162264309112, "loss": 2.1297, "step": 1539 }, { "epoch": 0.19, "learning_rate": 0.00046446449949980663, "loss": 2.1021, "step": 1540 }, { "epoch": 0.19, "learning_rate": 0.0004644127377777795, "loss": 1.9366, "step": 1541 }, { "epoch": 0.19, "learning_rate": 0.0004643609412732264, "loss": 2.02, "step": 1542 }, { "epoch": 0.19, "learning_rate": 0.00046430910999455, "loss": 1.955, "step": 1543 }, { "epoch": 0.19, "learning_rate": 0.0004642572439501586, "loss": 1.8538, "step": 1544 }, { "epoch": 0.19, "learning_rate": 0.00046420534314846584, "loss": 1.9431, "step": 1545 }, { "epoch": 0.19, "learning_rate": 0.0004641534075978914, "loss": 1.8388, "step": 1546 }, { "epoch": 0.19, "learning_rate": 0.0004641014373068602, "loss": 1.8887, "step": 1547 }, { "epoch": 0.19, "learning_rate": 0.00046404943228380295, "loss": 1.852, "step": 1548 }, { "epoch": 0.19, "learning_rate": 0.0004639973925371562, "loss": 1.8544, "step": 1549 }, { "epoch": 0.19, "learning_rate": 0.0004639453180753619, "loss": 1.9328, "step": 1550 }, { "epoch": 0.19, "learning_rate": 0.0004638932089068676, "loss": 1.9745, "step": 1551 }, { "epoch": 0.19, "learning_rate": 0.0004638410650401267, "loss": 1.9244, "step": 1552 }, { "epoch": 0.19, "learning_rate": 0.000463788886483598, "loss": 1.9435, "step": 1553 }, { "epoch": 0.19, "learning_rate": 0.00046373667324574596, "loss": 1.8567, "step": 1554 }, { "epoch": 0.19, "learning_rate": 0.00046368442533504097, "loss": 1.9193, "step": 1555 }, { "epoch": 0.19, "learning_rate": 0.00046363214275995854, "loss": 1.9358, "step": 1556 }, { "epoch": 0.19, "learning_rate": 0.00046357982552898023, "loss": 1.9655, "step": 1557 }, { "epoch": 0.19, "learning_rate": 0.000463527473650593, "loss": 1.8789, "step": 1558 }, { "epoch": 0.19, "learning_rate": 0.00046347508713328955, "loss": 1.801, "step": 1559 }, { "epoch": 0.2, "learning_rate": 0.0004634226659855681, "loss": 1.9001, "step": 1560 }, { "epoch": 0.2, "learning_rate": 0.0004633702102159325, "loss": 1.8342, "step": 1561 }, { "epoch": 0.2, "learning_rate": 0.0004633177198328923, "loss": 1.8697, "step": 1562 }, { "epoch": 0.2, "learning_rate": 0.00046326519484496257, "loss": 1.8461, "step": 1563 }, { "epoch": 0.2, "learning_rate": 0.000463212635260664, "loss": 1.8795, "step": 1564 }, { "epoch": 0.2, "learning_rate": 0.00046316004108852305, "loss": 1.839, "step": 1565 }, { "epoch": 0.2, "learning_rate": 0.0004631074123370716, "loss": 1.8251, "step": 1566 }, { "epoch": 0.2, "learning_rate": 0.0004630547490148472, "loss": 1.9547, "step": 1567 }, { "epoch": 0.2, "learning_rate": 0.00046300205113039295, "loss": 1.8697, "step": 1568 }, { "epoch": 0.2, "learning_rate": 0.00046294931869225774, "loss": 1.8272, "step": 1569 }, { "epoch": 0.2, "learning_rate": 0.0004628965517089959, "loss": 1.9093, "step": 1570 }, { "epoch": 0.2, "learning_rate": 0.00046284375018916735, "loss": 1.8611, "step": 1571 }, { "epoch": 0.2, "learning_rate": 0.0004627909141413378, "loss": 1.8419, "step": 1572 }, { "epoch": 0.2, "learning_rate": 0.0004627380435740783, "loss": 1.9394, "step": 1573 }, { "epoch": 0.2, "learning_rate": 0.00046268513849596573, "loss": 1.7579, "step": 1574 }, { "epoch": 0.2, "learning_rate": 0.0004626321989155824, "loss": 1.7989, "step": 1575 }, { "epoch": 0.2, "learning_rate": 0.00046257922484151635, "loss": 1.778, "step": 1576 }, { "epoch": 0.2, "learning_rate": 0.0004625262162823612, "loss": 1.8479, "step": 1577 }, { "epoch": 0.2, "learning_rate": 0.000462473173246716, "loss": 1.8874, "step": 1578 }, { "epoch": 0.2, "learning_rate": 0.0004624200957431856, "loss": 1.9194, "step": 1579 }, { "epoch": 0.2, "learning_rate": 0.00046236698378038026, "loss": 1.9618, "step": 1580 }, { "epoch": 0.2, "learning_rate": 0.000462313837366916, "loss": 1.9189, "step": 1581 }, { "epoch": 0.2, "learning_rate": 0.0004622606565114144, "loss": 1.8566, "step": 1582 }, { "epoch": 0.2, "learning_rate": 0.00046220744122250245, "loss": 1.9019, "step": 1583 }, { "epoch": 0.2, "learning_rate": 0.000462154191508813, "loss": 1.8408, "step": 1584 }, { "epoch": 0.2, "learning_rate": 0.00046210090737898424, "loss": 1.899, "step": 1585 }, { "epoch": 0.2, "learning_rate": 0.0004620475888416601, "loss": 1.8556, "step": 1586 }, { "epoch": 0.2, "learning_rate": 0.00046199423590548993, "loss": 1.9595, "step": 1587 }, { "epoch": 0.2, "learning_rate": 0.0004619408485791289, "loss": 1.9473, "step": 1588 }, { "epoch": 0.2, "learning_rate": 0.00046188742687123754, "loss": 1.8717, "step": 1589 }, { "epoch": 0.2, "learning_rate": 0.00046183397079048205, "loss": 1.8556, "step": 1590 }, { "epoch": 0.2, "learning_rate": 0.0004617804803455343, "loss": 1.889, "step": 1591 }, { "epoch": 0.2, "learning_rate": 0.0004617269555450715, "loss": 1.9476, "step": 1592 }, { "epoch": 0.2, "learning_rate": 0.0004616733963977766, "loss": 1.9332, "step": 1593 }, { "epoch": 0.2, "learning_rate": 0.00046161980291233815, "loss": 1.908, "step": 1594 }, { "epoch": 0.2, "learning_rate": 0.00046156617509745016, "loss": 1.9393, "step": 1595 }, { "epoch": 0.2, "learning_rate": 0.00046151251296181216, "loss": 1.8375, "step": 1596 }, { "epoch": 0.2, "learning_rate": 0.0004614588165141295, "loss": 1.8301, "step": 1597 }, { "epoch": 0.2, "learning_rate": 0.0004614050857631129, "loss": 1.8083, "step": 1598 }, { "epoch": 0.2, "learning_rate": 0.00046135132071747865, "loss": 1.8573, "step": 1599 }, { "epoch": 0.2, "learning_rate": 0.00046129752138594874, "loss": 1.8361, "step": 1600 }, { "epoch": 0.2, "eval_acceptance_rate": 0.5956124067306519, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.8300104092115197, "eval_expected_tokens": 2.2875130115144, "eval_loss": 1.8316420316696167, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3218782528785999, "eval_runtime": 50.4935, "eval_samples_per_second": 9.645, "eval_steps_per_second": 0.317, "step": 1600 }, { "epoch": 0.2, "learning_rate": 0.0004612436877772505, "loss": 1.7997, "step": 1601 }, { "epoch": 0.2, "learning_rate": 0.00046118981990011685, "loss": 1.7848, "step": 1602 }, { "epoch": 0.2, "learning_rate": 0.00046113591776328664, "loss": 1.8236, "step": 1603 }, { "epoch": 0.2, "learning_rate": 0.00046108198137550377, "loss": 1.8721, "step": 1604 }, { "epoch": 0.2, "learning_rate": 0.000461028010745518, "loss": 1.8716, "step": 1605 }, { "epoch": 0.2, "learning_rate": 0.00046097400588208465, "loss": 1.9317, "step": 1606 }, { "epoch": 0.2, "learning_rate": 0.0004609199667939644, "loss": 1.9022, "step": 1607 }, { "epoch": 0.2, "learning_rate": 0.00046086589348992357, "loss": 1.8723, "step": 1608 }, { "epoch": 0.2, "learning_rate": 0.0004608117859787342, "loss": 1.8215, "step": 1609 }, { "epoch": 0.2, "learning_rate": 0.00046075764426917366, "loss": 1.9034, "step": 1610 }, { "epoch": 0.2, "learning_rate": 0.00046070346837002486, "loss": 1.9132, "step": 1611 }, { "epoch": 0.2, "learning_rate": 0.00046064925829007643, "loss": 1.805, "step": 1612 }, { "epoch": 0.2, "learning_rate": 0.0004605950140381225, "loss": 1.8821, "step": 1613 }, { "epoch": 0.2, "learning_rate": 0.00046054073562296256, "loss": 1.8427, "step": 1614 }, { "epoch": 0.2, "learning_rate": 0.00046048642305340185, "loss": 1.847, "step": 1615 }, { "epoch": 0.2, "learning_rate": 0.00046043207633825114, "loss": 1.8571, "step": 1616 }, { "epoch": 0.2, "learning_rate": 0.00046037769548632656, "loss": 1.8505, "step": 1617 }, { "epoch": 0.2, "learning_rate": 0.00046032328050645, "loss": 1.8634, "step": 1618 }, { "epoch": 0.2, "learning_rate": 0.0004602688314074487, "loss": 1.8441, "step": 1619 }, { "epoch": 0.2, "learning_rate": 0.00046021434819815555, "loss": 1.8783, "step": 1620 }, { "epoch": 0.2, "learning_rate": 0.000460159830887409, "loss": 1.8118, "step": 1621 }, { "epoch": 0.2, "learning_rate": 0.00046010527948405286, "loss": 1.8471, "step": 1622 }, { "epoch": 0.2, "learning_rate": 0.0004600506939969367, "loss": 1.8657, "step": 1623 }, { "epoch": 0.2, "learning_rate": 0.0004599960744349154, "loss": 1.8668, "step": 1624 }, { "epoch": 0.2, "learning_rate": 0.00045994142080684955, "loss": 1.917, "step": 1625 }, { "epoch": 0.2, "learning_rate": 0.00045988673312160504, "loss": 1.8069, "step": 1626 }, { "epoch": 0.2, "learning_rate": 0.00045983201138805367, "loss": 1.7872, "step": 1627 }, { "epoch": 0.2, "learning_rate": 0.0004597772556150723, "loss": 1.9209, "step": 1628 }, { "epoch": 0.2, "learning_rate": 0.00045972246581154377, "loss": 2.0839, "step": 1629 }, { "epoch": 0.2, "learning_rate": 0.00045966764198635603, "loss": 1.9217, "step": 1630 }, { "epoch": 0.2, "learning_rate": 0.00045961278414840276, "loss": 1.9862, "step": 1631 }, { "epoch": 0.2, "learning_rate": 0.0004595578923065831, "loss": 1.9634, "step": 1632 }, { "epoch": 0.2, "learning_rate": 0.00045950296646980185, "loss": 1.8418, "step": 1633 }, { "epoch": 0.2, "learning_rate": 0.000459448006646969, "loss": 1.8941, "step": 1634 }, { "epoch": 0.2, "learning_rate": 0.0004593930128470005, "loss": 1.882, "step": 1635 }, { "epoch": 0.2, "learning_rate": 0.00045933798507881744, "loss": 1.9789, "step": 1636 }, { "epoch": 0.2, "learning_rate": 0.0004592829233513465, "loss": 1.8618, "step": 1637 }, { "epoch": 0.2, "learning_rate": 0.00045922782767351993, "loss": 1.8689, "step": 1638 }, { "epoch": 0.2, "learning_rate": 0.00045917269805427567, "loss": 1.8927, "step": 1639 }, { "epoch": 0.2, "learning_rate": 0.0004591175345025567, "loss": 1.9084, "step": 1640 }, { "epoch": 0.21, "learning_rate": 0.000459062337027312, "loss": 1.7666, "step": 1641 }, { "epoch": 0.21, "learning_rate": 0.0004590071056374957, "loss": 1.8075, "step": 1642 }, { "epoch": 0.21, "learning_rate": 0.0004589518403420676, "loss": 1.885, "step": 1643 }, { "epoch": 0.21, "learning_rate": 0.00045889654114999296, "loss": 1.895, "step": 1644 }, { "epoch": 0.21, "learning_rate": 0.00045884120807024255, "loss": 1.9295, "step": 1645 }, { "epoch": 0.21, "learning_rate": 0.00045878584111179265, "loss": 1.9281, "step": 1646 }, { "epoch": 0.21, "learning_rate": 0.000458730440283625, "loss": 1.884, "step": 1647 }, { "epoch": 0.21, "learning_rate": 0.00045867500559472674, "loss": 1.7789, "step": 1648 }, { "epoch": 0.21, "learning_rate": 0.0004586195370540908, "loss": 1.88, "step": 1649 }, { "epoch": 0.21, "learning_rate": 0.0004585640346707153, "loss": 1.9131, "step": 1650 }, { "epoch": 0.21, "learning_rate": 0.000458508498453604, "loss": 1.8571, "step": 1651 }, { "epoch": 0.21, "learning_rate": 0.0004584529284117662, "loss": 1.8671, "step": 1652 }, { "epoch": 0.21, "learning_rate": 0.00045839732455421633, "loss": 1.8716, "step": 1653 }, { "epoch": 0.21, "learning_rate": 0.0004583416868899748, "loss": 1.8088, "step": 1654 }, { "epoch": 0.21, "learning_rate": 0.0004582860154280674, "loss": 1.8182, "step": 1655 }, { "epoch": 0.21, "learning_rate": 0.00045823031017752484, "loss": 1.9309, "step": 1656 }, { "epoch": 0.21, "learning_rate": 0.0004581745711473842, "loss": 1.8478, "step": 1657 }, { "epoch": 0.21, "learning_rate": 0.00045811879834668735, "loss": 1.8, "step": 1658 }, { "epoch": 0.21, "learning_rate": 0.000458062991784482, "loss": 1.7953, "step": 1659 }, { "epoch": 0.21, "learning_rate": 0.0004580071514698211, "loss": 1.8386, "step": 1660 }, { "epoch": 0.21, "learning_rate": 0.0004579512774117632, "loss": 1.8176, "step": 1661 }, { "epoch": 0.21, "learning_rate": 0.0004578953696193725, "loss": 1.7866, "step": 1662 }, { "epoch": 0.21, "learning_rate": 0.0004578394281017184, "loss": 1.8334, "step": 1663 }, { "epoch": 0.21, "learning_rate": 0.0004577834528678757, "loss": 1.9157, "step": 1664 }, { "epoch": 0.21, "learning_rate": 0.000457727443926925, "loss": 1.8048, "step": 1665 }, { "epoch": 0.21, "learning_rate": 0.00045767140128795224, "loss": 1.8259, "step": 1666 }, { "epoch": 0.21, "learning_rate": 0.00045761532496004865, "loss": 1.8419, "step": 1667 }, { "epoch": 0.21, "learning_rate": 0.00045755921495231107, "loss": 1.9529, "step": 1668 }, { "epoch": 0.21, "learning_rate": 0.00045750307127384186, "loss": 1.8994, "step": 1669 }, { "epoch": 0.21, "learning_rate": 0.00045744689393374877, "loss": 1.7243, "step": 1670 }, { "epoch": 0.21, "learning_rate": 0.00045739068294114505, "loss": 1.8299, "step": 1671 }, { "epoch": 0.21, "learning_rate": 0.0004573344383051492, "loss": 1.8468, "step": 1672 }, { "epoch": 0.21, "learning_rate": 0.0004572781600348856, "loss": 1.7974, "step": 1673 }, { "epoch": 0.21, "learning_rate": 0.00045722184813948367, "loss": 1.8746, "step": 1674 }, { "epoch": 0.21, "learning_rate": 0.0004571655026280785, "loss": 1.7644, "step": 1675 }, { "epoch": 0.21, "learning_rate": 0.0004571091235098106, "loss": 1.91, "step": 1676 }, { "epoch": 0.21, "learning_rate": 0.0004570527107938259, "loss": 1.8583, "step": 1677 }, { "epoch": 0.21, "learning_rate": 0.0004569962644892759, "loss": 1.806, "step": 1678 }, { "epoch": 0.21, "learning_rate": 0.00045693978460531725, "loss": 1.7943, "step": 1679 }, { "epoch": 0.21, "learning_rate": 0.0004568832711511125, "loss": 1.7915, "step": 1680 }, { "epoch": 0.21, "learning_rate": 0.0004568267241358292, "loss": 1.7468, "step": 1681 }, { "epoch": 0.21, "learning_rate": 0.00045677014356864043, "loss": 1.7388, "step": 1682 }, { "epoch": 0.21, "learning_rate": 0.00045671352945872516, "loss": 1.9001, "step": 1683 }, { "epoch": 0.21, "learning_rate": 0.00045665688181526724, "loss": 1.8809, "step": 1684 }, { "epoch": 0.21, "learning_rate": 0.00045660020064745615, "loss": 1.7992, "step": 1685 }, { "epoch": 0.21, "learning_rate": 0.0004565434859644869, "loss": 1.7877, "step": 1686 }, { "epoch": 0.21, "learning_rate": 0.0004564867377755599, "loss": 1.8866, "step": 1687 }, { "epoch": 0.21, "learning_rate": 0.00045642995608988104, "loss": 1.7801, "step": 1688 }, { "epoch": 0.21, "learning_rate": 0.00045637314091666147, "loss": 1.8812, "step": 1689 }, { "epoch": 0.21, "learning_rate": 0.00045631629226511783, "loss": 1.8314, "step": 1690 }, { "epoch": 0.21, "learning_rate": 0.00045625941014447235, "loss": 1.7532, "step": 1691 }, { "epoch": 0.21, "learning_rate": 0.0004562024945639525, "loss": 1.8831, "step": 1692 }, { "epoch": 0.21, "learning_rate": 0.00045614554553279134, "loss": 1.7603, "step": 1693 }, { "epoch": 0.21, "learning_rate": 0.0004560885630602272, "loss": 1.8936, "step": 1694 }, { "epoch": 0.21, "learning_rate": 0.0004560315471555039, "loss": 1.8536, "step": 1695 }, { "epoch": 0.21, "learning_rate": 0.0004559744978278708, "loss": 1.7792, "step": 1696 }, { "epoch": 0.21, "learning_rate": 0.0004559174150865824, "loss": 1.8648, "step": 1697 }, { "epoch": 0.21, "learning_rate": 0.00045586029894089893, "loss": 1.6934, "step": 1698 }, { "epoch": 0.21, "learning_rate": 0.00045580314940008585, "loss": 1.8875, "step": 1699 }, { "epoch": 0.21, "learning_rate": 0.0004557459664734141, "loss": 1.8532, "step": 1700 }, { "epoch": 0.21, "eval_acceptance_rate": 0.6013785600662231, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.8490577503581684, "eval_expected_tokens": 2.3113221879477104, "eval_loss": 1.8134979009628296, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.32783054698692765, "eval_runtime": 50.2958, "eval_samples_per_second": 9.683, "eval_steps_per_second": 0.318, "step": 1700 }, { "epoch": 0.21, "learning_rate": 0.00045568875017016003, "loss": 1.8148, "step": 1701 }, { "epoch": 0.21, "learning_rate": 0.00045563150049960537, "loss": 1.7799, "step": 1702 }, { "epoch": 0.21, "learning_rate": 0.0004555742174710373, "loss": 1.7711, "step": 1703 }, { "epoch": 0.21, "learning_rate": 0.0004555169010937485, "loss": 1.8592, "step": 1704 }, { "epoch": 0.21, "learning_rate": 0.0004554595513770369, "loss": 1.8332, "step": 1705 }, { "epoch": 0.21, "learning_rate": 0.0004554021683302058, "loss": 1.8135, "step": 1706 }, { "epoch": 0.21, "learning_rate": 0.0004553447519625641, "loss": 1.8641, "step": 1707 }, { "epoch": 0.21, "learning_rate": 0.000455287302283426, "loss": 1.7251, "step": 1708 }, { "epoch": 0.21, "learning_rate": 0.00045522981930211116, "loss": 1.8427, "step": 1709 }, { "epoch": 0.21, "learning_rate": 0.0004551723030279446, "loss": 1.7573, "step": 1710 }, { "epoch": 0.21, "learning_rate": 0.0004551147534702567, "loss": 1.8289, "step": 1711 }, { "epoch": 0.21, "learning_rate": 0.00045505717063838323, "loss": 1.7774, "step": 1712 }, { "epoch": 0.21, "learning_rate": 0.0004549995545416655, "loss": 1.8295, "step": 1713 }, { "epoch": 0.21, "learning_rate": 0.0004549419051894502, "loss": 1.8295, "step": 1714 }, { "epoch": 0.21, "learning_rate": 0.0004548842225910891, "loss": 1.83, "step": 1715 }, { "epoch": 0.21, "learning_rate": 0.00045482650675593986, "loss": 1.8456, "step": 1716 }, { "epoch": 0.21, "learning_rate": 0.00045476875769336506, "loss": 1.75, "step": 1717 }, { "epoch": 0.21, "learning_rate": 0.000454710975412733, "loss": 1.8283, "step": 1718 }, { "epoch": 0.21, "learning_rate": 0.0004546531599234173, "loss": 1.8214, "step": 1719 }, { "epoch": 0.21, "learning_rate": 0.00045459531123479674, "loss": 1.8402, "step": 1720 }, { "epoch": 0.22, "learning_rate": 0.0004545374293562559, "loss": 1.795, "step": 1721 }, { "epoch": 0.22, "learning_rate": 0.0004544795142971843, "loss": 1.8204, "step": 1722 }, { "epoch": 0.22, "learning_rate": 0.0004544215660669772, "loss": 1.9177, "step": 1723 }, { "epoch": 0.22, "learning_rate": 0.000454363584675035, "loss": 1.791, "step": 1724 }, { "epoch": 0.22, "learning_rate": 0.0004543055701307637, "loss": 1.8547, "step": 1725 }, { "epoch": 0.22, "learning_rate": 0.0004542475224435744, "loss": 1.7836, "step": 1726 }, { "epoch": 0.22, "learning_rate": 0.00045418944162288385, "loss": 1.762, "step": 1727 }, { "epoch": 0.22, "learning_rate": 0.000454131327678114, "loss": 1.8402, "step": 1728 }, { "epoch": 0.22, "learning_rate": 0.0004540731806186922, "loss": 1.7921, "step": 1729 }, { "epoch": 0.22, "learning_rate": 0.0004540150004540512, "loss": 1.9008, "step": 1730 }, { "epoch": 0.22, "learning_rate": 0.0004539567871936292, "loss": 1.7633, "step": 1731 }, { "epoch": 0.22, "learning_rate": 0.0004538985408468697, "loss": 1.7928, "step": 1732 }, { "epoch": 0.22, "learning_rate": 0.00045384026142322135, "loss": 1.8073, "step": 1733 }, { "epoch": 0.22, "learning_rate": 0.00045378194893213854, "loss": 1.897, "step": 1734 }, { "epoch": 0.22, "learning_rate": 0.0004537236033830809, "loss": 1.7801, "step": 1735 }, { "epoch": 0.22, "learning_rate": 0.0004536652247855133, "loss": 1.8728, "step": 1736 }, { "epoch": 0.22, "learning_rate": 0.000453606813148906, "loss": 1.8147, "step": 1737 }, { "epoch": 0.22, "learning_rate": 0.0004535483684827348, "loss": 1.752, "step": 1738 }, { "epoch": 0.22, "learning_rate": 0.00045348989079648064, "loss": 1.8072, "step": 1739 }, { "epoch": 0.22, "learning_rate": 0.0004534313800996299, "loss": 1.8528, "step": 1740 }, { "epoch": 0.22, "learning_rate": 0.0004533728364016744, "loss": 1.9427, "step": 1741 }, { "epoch": 0.22, "learning_rate": 0.00045331425971211105, "loss": 1.7968, "step": 1742 }, { "epoch": 0.22, "learning_rate": 0.0004532556500404425, "loss": 1.7815, "step": 1743 }, { "epoch": 0.22, "learning_rate": 0.0004531970073961764, "loss": 1.7777, "step": 1744 }, { "epoch": 0.22, "learning_rate": 0.0004531383317888261, "loss": 1.8461, "step": 1745 }, { "epoch": 0.22, "learning_rate": 0.0004530796232279099, "loss": 1.7983, "step": 1746 }, { "epoch": 0.22, "learning_rate": 0.0004530208817229516, "loss": 1.7319, "step": 1747 }, { "epoch": 0.22, "learning_rate": 0.0004529621072834805, "loss": 1.791, "step": 1748 }, { "epoch": 0.22, "learning_rate": 0.0004529032999190311, "loss": 1.7327, "step": 1749 }, { "epoch": 0.22, "learning_rate": 0.0004528444596391433, "loss": 1.7708, "step": 1750 }, { "epoch": 0.22, "learning_rate": 0.0004527855864533622, "loss": 1.7948, "step": 1751 }, { "epoch": 0.22, "learning_rate": 0.00045272668037123844, "loss": 1.8422, "step": 1752 }, { "epoch": 0.22, "learning_rate": 0.00045266774140232784, "loss": 1.7263, "step": 1753 }, { "epoch": 0.22, "learning_rate": 0.0004526087695561917, "loss": 1.7666, "step": 1754 }, { "epoch": 0.22, "learning_rate": 0.0004525497648423964, "loss": 1.8271, "step": 1755 }, { "epoch": 0.22, "learning_rate": 0.00045249072727051405, "loss": 1.8114, "step": 1756 }, { "epoch": 0.22, "learning_rate": 0.00045243165685012165, "loss": 1.6831, "step": 1757 }, { "epoch": 0.22, "learning_rate": 0.0004523725535908019, "loss": 1.8839, "step": 1758 }, { "epoch": 0.22, "learning_rate": 0.00045231341750214254, "loss": 1.8326, "step": 1759 }, { "epoch": 0.22, "learning_rate": 0.0004522542485937369, "loss": 1.913, "step": 1760 }, { "epoch": 0.22, "learning_rate": 0.00045219504687518333, "loss": 1.832, "step": 1761 }, { "epoch": 0.22, "learning_rate": 0.00045213581235608577, "loss": 1.8752, "step": 1762 }, { "epoch": 0.22, "learning_rate": 0.0004520765450460533, "loss": 1.8513, "step": 1763 }, { "epoch": 0.22, "learning_rate": 0.00045201724495470055, "loss": 1.7727, "step": 1764 }, { "epoch": 0.22, "learning_rate": 0.0004519579120916472, "loss": 1.8146, "step": 1765 }, { "epoch": 0.22, "learning_rate": 0.0004518985464665184, "loss": 1.7966, "step": 1766 }, { "epoch": 0.22, "learning_rate": 0.00045183914808894453, "loss": 1.7688, "step": 1767 }, { "epoch": 0.22, "learning_rate": 0.0004517797169685614, "loss": 1.8228, "step": 1768 }, { "epoch": 0.22, "learning_rate": 0.00045172025311501004, "loss": 1.7047, "step": 1769 }, { "epoch": 0.22, "learning_rate": 0.00045166075653793673, "loss": 1.8551, "step": 1770 }, { "epoch": 0.22, "learning_rate": 0.00045160122724699324, "loss": 1.8557, "step": 1771 }, { "epoch": 0.22, "learning_rate": 0.0004515416652518366, "loss": 1.7559, "step": 1772 }, { "epoch": 0.22, "learning_rate": 0.00045148207056212896, "loss": 1.8822, "step": 1773 }, { "epoch": 0.22, "learning_rate": 0.000451422443187538, "loss": 1.8399, "step": 1774 }, { "epoch": 0.22, "learning_rate": 0.0004513627831377365, "loss": 1.8161, "step": 1775 }, { "epoch": 0.22, "learning_rate": 0.00045130309042240276, "loss": 1.836, "step": 1776 }, { "epoch": 0.22, "learning_rate": 0.0004512433650512202, "loss": 1.7774, "step": 1777 }, { "epoch": 0.22, "learning_rate": 0.0004511836070338777, "loss": 1.7987, "step": 1778 }, { "epoch": 0.22, "learning_rate": 0.00045112381638006923, "loss": 1.78, "step": 1779 }, { "epoch": 0.22, "learning_rate": 0.0004510639930994942, "loss": 1.8323, "step": 1780 }, { "epoch": 0.22, "learning_rate": 0.00045100413720185737, "loss": 1.7627, "step": 1781 }, { "epoch": 0.22, "learning_rate": 0.00045094424869686854, "loss": 1.7157, "step": 1782 }, { "epoch": 0.22, "learning_rate": 0.00045088432759424316, "loss": 1.7658, "step": 1783 }, { "epoch": 0.22, "learning_rate": 0.00045082437390370157, "loss": 1.8858, "step": 1784 }, { "epoch": 0.22, "learning_rate": 0.00045076438763496973, "loss": 1.8037, "step": 1785 }, { "epoch": 0.22, "learning_rate": 0.0004507043687977787, "loss": 1.806, "step": 1786 }, { "epoch": 0.22, "learning_rate": 0.00045064431740186476, "loss": 1.8681, "step": 1787 }, { "epoch": 0.22, "learning_rate": 0.00045058423345696987, "loss": 1.8311, "step": 1788 }, { "epoch": 0.22, "learning_rate": 0.0004505241169728407, "loss": 1.7284, "step": 1789 }, { "epoch": 0.22, "learning_rate": 0.0004504639679592296, "loss": 1.732, "step": 1790 }, { "epoch": 0.22, "learning_rate": 0.0004504037864258942, "loss": 1.7064, "step": 1791 }, { "epoch": 0.22, "learning_rate": 0.00045034357238259707, "loss": 1.7993, "step": 1792 }, { "epoch": 0.22, "learning_rate": 0.00045028332583910635, "loss": 1.8334, "step": 1793 }, { "epoch": 0.22, "learning_rate": 0.00045022304680519544, "loss": 1.7628, "step": 1794 }, { "epoch": 0.22, "learning_rate": 0.00045016273529064287, "loss": 1.7516, "step": 1795 }, { "epoch": 0.22, "learning_rate": 0.00045010239130523254, "loss": 1.8016, "step": 1796 }, { "epoch": 0.22, "learning_rate": 0.00045004201485875355, "loss": 1.8251, "step": 1797 }, { "epoch": 0.22, "learning_rate": 0.00044998160596100036, "loss": 1.809, "step": 1798 }, { "epoch": 0.22, "learning_rate": 0.0004499211646217727, "loss": 1.766, "step": 1799 }, { "epoch": 0.23, "learning_rate": 0.0004498606908508753, "loss": 1.7688, "step": 1800 }, { "epoch": 0.23, "eval_acceptance_rate": 0.6064440011978149, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.8660086537910723, "eval_expected_tokens": 2.3325108172388402, "eval_loss": 1.770708680152893, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.33312770430971006, "eval_runtime": 50.3316, "eval_samples_per_second": 9.676, "eval_steps_per_second": 0.318, "step": 1800 }, { "epoch": 0.23, "learning_rate": 0.0004498001846581186, "loss": 1.8255, "step": 1801 }, { "epoch": 0.23, "learning_rate": 0.0004497396460533178, "loss": 1.8218, "step": 1802 }, { "epoch": 0.23, "learning_rate": 0.0004496790750462938, "loss": 1.7952, "step": 1803 }, { "epoch": 0.23, "learning_rate": 0.00044961847164687254, "loss": 1.8359, "step": 1804 }, { "epoch": 0.23, "learning_rate": 0.0004495578358648851, "loss": 1.7646, "step": 1805 }, { "epoch": 0.23, "learning_rate": 0.0004494971677101681, "loss": 1.7323, "step": 1806 }, { "epoch": 0.23, "learning_rate": 0.0004494364671925634, "loss": 1.6835, "step": 1807 }, { "epoch": 0.23, "learning_rate": 0.00044937573432191763, "loss": 1.741, "step": 1808 }, { "epoch": 0.23, "learning_rate": 0.00044931496910808325, "loss": 1.8193, "step": 1809 }, { "epoch": 0.23, "learning_rate": 0.0004492541715609177, "loss": 1.7813, "step": 1810 }, { "epoch": 0.23, "learning_rate": 0.0004491933416902836, "loss": 1.7655, "step": 1811 }, { "epoch": 0.23, "learning_rate": 0.00044913247950604905, "loss": 1.7757, "step": 1812 }, { "epoch": 0.23, "learning_rate": 0.0004490715850180872, "loss": 1.8469, "step": 1813 }, { "epoch": 0.23, "learning_rate": 0.00044901065823627653, "loss": 1.8334, "step": 1814 }, { "epoch": 0.23, "learning_rate": 0.00044894969917050055, "loss": 1.7579, "step": 1815 }, { "epoch": 0.23, "learning_rate": 0.0004488887078306484, "loss": 1.7476, "step": 1816 }, { "epoch": 0.23, "learning_rate": 0.00044882768422661406, "loss": 1.7776, "step": 1817 }, { "epoch": 0.23, "learning_rate": 0.0004487666283682971, "loss": 1.7549, "step": 1818 }, { "epoch": 0.23, "learning_rate": 0.00044870554026560193, "loss": 1.8384, "step": 1819 }, { "epoch": 0.23, "learning_rate": 0.0004486444199284386, "loss": 1.8284, "step": 1820 }, { "epoch": 0.23, "learning_rate": 0.00044858326736672204, "loss": 1.7807, "step": 1821 }, { "epoch": 0.23, "learning_rate": 0.00044852208259037266, "loss": 1.7852, "step": 1822 }, { "epoch": 0.23, "learning_rate": 0.0004484608656093159, "loss": 1.7944, "step": 1823 }, { "epoch": 0.23, "learning_rate": 0.0004483996164334827, "loss": 1.8377, "step": 1824 }, { "epoch": 0.23, "learning_rate": 0.0004483383350728088, "loss": 1.7487, "step": 1825 }, { "epoch": 0.23, "learning_rate": 0.0004482770215372356, "loss": 1.8405, "step": 1826 }, { "epoch": 0.23, "learning_rate": 0.0004482156758367094, "loss": 1.6938, "step": 1827 }, { "epoch": 0.23, "learning_rate": 0.00044815429798118186, "loss": 1.7501, "step": 1828 }, { "epoch": 0.23, "learning_rate": 0.00044809288798060987, "loss": 1.7907, "step": 1829 }, { "epoch": 0.23, "learning_rate": 0.00044803144584495546, "loss": 1.7868, "step": 1830 }, { "epoch": 0.23, "learning_rate": 0.00044796997158418604, "loss": 1.7912, "step": 1831 }, { "epoch": 0.23, "learning_rate": 0.00044790846520827393, "loss": 1.7528, "step": 1832 }, { "epoch": 0.23, "learning_rate": 0.00044784692672719694, "loss": 1.8441, "step": 1833 }, { "epoch": 0.23, "learning_rate": 0.000447785356150938, "loss": 1.8089, "step": 1834 }, { "epoch": 0.23, "learning_rate": 0.00044772375348948513, "loss": 1.8473, "step": 1835 }, { "epoch": 0.23, "learning_rate": 0.0004476621187528317, "loss": 1.7993, "step": 1836 }, { "epoch": 0.23, "learning_rate": 0.0004476004519509763, "loss": 1.7497, "step": 1837 }, { "epoch": 0.23, "learning_rate": 0.0004475387530939226, "loss": 1.6748, "step": 1838 }, { "epoch": 0.23, "learning_rate": 0.00044747702219167954, "loss": 1.7493, "step": 1839 }, { "epoch": 0.23, "learning_rate": 0.00044741525925426123, "loss": 1.7654, "step": 1840 }, { "epoch": 0.23, "learning_rate": 0.00044735346429168707, "loss": 1.6776, "step": 1841 }, { "epoch": 0.23, "learning_rate": 0.0004472916373139815, "loss": 1.8266, "step": 1842 }, { "epoch": 0.23, "learning_rate": 0.0004472297783311742, "loss": 1.7684, "step": 1843 }, { "epoch": 0.23, "learning_rate": 0.00044716788735330016, "loss": 1.7309, "step": 1844 }, { "epoch": 0.23, "learning_rate": 0.0004471059643903995, "loss": 1.8402, "step": 1845 }, { "epoch": 0.23, "learning_rate": 0.00044704400945251745, "loss": 1.7633, "step": 1846 }, { "epoch": 0.23, "learning_rate": 0.0004469820225497044, "loss": 1.7873, "step": 1847 }, { "epoch": 0.23, "learning_rate": 0.00044692000369201627, "loss": 1.7537, "step": 1848 }, { "epoch": 0.23, "learning_rate": 0.0004468579528895136, "loss": 1.7849, "step": 1849 }, { "epoch": 0.23, "learning_rate": 0.0004467958701522625, "loss": 1.7489, "step": 1850 }, { "epoch": 0.23, "learning_rate": 0.00044673375549033435, "loss": 1.7832, "step": 1851 }, { "epoch": 0.23, "learning_rate": 0.00044667160891380533, "loss": 1.7843, "step": 1852 }, { "epoch": 0.23, "learning_rate": 0.00044660943043275703, "loss": 1.7313, "step": 1853 }, { "epoch": 0.23, "learning_rate": 0.0004465472200572763, "loss": 1.7434, "step": 1854 }, { "epoch": 0.23, "learning_rate": 0.00044648497779745496, "loss": 1.7284, "step": 1855 }, { "epoch": 0.23, "learning_rate": 0.00044642270366339006, "loss": 1.7358, "step": 1856 }, { "epoch": 0.23, "learning_rate": 0.00044636039766518396, "loss": 1.7601, "step": 1857 }, { "epoch": 0.23, "learning_rate": 0.0004462980598129439, "loss": 1.8453, "step": 1858 }, { "epoch": 0.23, "learning_rate": 0.00044623569011678274, "loss": 1.8102, "step": 1859 }, { "epoch": 0.23, "learning_rate": 0.000446173288586818, "loss": 1.7934, "step": 1860 }, { "epoch": 0.23, "learning_rate": 0.0004461108552331727, "loss": 1.8354, "step": 1861 }, { "epoch": 0.23, "learning_rate": 0.00044604839006597485, "loss": 1.7628, "step": 1862 }, { "epoch": 0.23, "learning_rate": 0.00044598589309535785, "loss": 1.8149, "step": 1863 }, { "epoch": 0.23, "learning_rate": 0.00044592336433145995, "loss": 1.7371, "step": 1864 }, { "epoch": 0.23, "learning_rate": 0.0004458608037844248, "loss": 1.7444, "step": 1865 }, { "epoch": 0.23, "learning_rate": 0.000445798211464401, "loss": 1.718, "step": 1866 }, { "epoch": 0.23, "learning_rate": 0.0004457355873815425, "loss": 1.7121, "step": 1867 }, { "epoch": 0.23, "learning_rate": 0.0004456729315460084, "loss": 1.8205, "step": 1868 }, { "epoch": 0.23, "learning_rate": 0.00044561024396796273, "loss": 1.7279, "step": 1869 }, { "epoch": 0.23, "learning_rate": 0.00044554752465757484, "loss": 1.844, "step": 1870 }, { "epoch": 0.23, "learning_rate": 0.00044548477362501936, "loss": 1.7317, "step": 1871 }, { "epoch": 0.23, "learning_rate": 0.0004454219908804757, "loss": 1.7204, "step": 1872 }, { "epoch": 0.23, "learning_rate": 0.0004453591764341287, "loss": 1.7691, "step": 1873 }, { "epoch": 0.23, "learning_rate": 0.0004452963302961682, "loss": 1.7381, "step": 1874 }, { "epoch": 0.23, "learning_rate": 0.00044523345247678945, "loss": 1.7204, "step": 1875 }, { "epoch": 0.23, "learning_rate": 0.00044517054298619235, "loss": 1.7286, "step": 1876 }, { "epoch": 0.23, "learning_rate": 0.0004451076018345824, "loss": 1.7826, "step": 1877 }, { "epoch": 0.23, "learning_rate": 0.0004450446290321701, "loss": 1.7835, "step": 1878 }, { "epoch": 0.23, "learning_rate": 0.0004449816245891709, "loss": 1.7789, "step": 1879 }, { "epoch": 0.23, "learning_rate": 0.00044491858851580557, "loss": 1.7385, "step": 1880 }, { "epoch": 0.24, "learning_rate": 0.0004448555208223001, "loss": 1.7047, "step": 1881 }, { "epoch": 0.24, "learning_rate": 0.00044479242151888523, "loss": 1.7814, "step": 1882 }, { "epoch": 0.24, "learning_rate": 0.00044472929061579726, "loss": 1.7479, "step": 1883 }, { "epoch": 0.24, "learning_rate": 0.0004446661281232773, "loss": 1.7751, "step": 1884 }, { "epoch": 0.24, "learning_rate": 0.0004446029340515719, "loss": 1.8893, "step": 1885 }, { "epoch": 0.24, "learning_rate": 0.00044453970841093243, "loss": 1.7728, "step": 1886 }, { "epoch": 0.24, "learning_rate": 0.0004444764512116155, "loss": 1.7662, "step": 1887 }, { "epoch": 0.24, "learning_rate": 0.0004444131624638828, "loss": 1.7729, "step": 1888 }, { "epoch": 0.24, "learning_rate": 0.00044434984217800124, "loss": 1.8582, "step": 1889 }, { "epoch": 0.24, "learning_rate": 0.0004442864903642427, "loss": 1.7881, "step": 1890 }, { "epoch": 0.24, "learning_rate": 0.0004442231070328844, "loss": 1.8505, "step": 1891 }, { "epoch": 0.24, "learning_rate": 0.00044415969219420843, "loss": 1.6834, "step": 1892 }, { "epoch": 0.24, "learning_rate": 0.00044409624585850207, "loss": 1.731, "step": 1893 }, { "epoch": 0.24, "learning_rate": 0.0004440327680360578, "loss": 1.7449, "step": 1894 }, { "epoch": 0.24, "learning_rate": 0.0004439692587371731, "loss": 1.7512, "step": 1895 }, { "epoch": 0.24, "learning_rate": 0.0004439057179721506, "loss": 1.7449, "step": 1896 }, { "epoch": 0.24, "learning_rate": 0.000443842145751298, "loss": 1.7525, "step": 1897 }, { "epoch": 0.24, "learning_rate": 0.0004437785420849282, "loss": 1.771, "step": 1898 }, { "epoch": 0.24, "learning_rate": 0.000443714906983359, "loss": 1.815, "step": 1899 }, { "epoch": 0.24, "learning_rate": 0.00044365124045691364, "loss": 1.7898, "step": 1900 }, { "epoch": 0.24, "eval_acceptance_rate": 0.6103216409683228, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.8791240401378635, "eval_expected_tokens": 2.348905050172329, "eval_loss": 1.7430039644241333, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.33722626254308224, "eval_runtime": 50.3039, "eval_samples_per_second": 9.681, "eval_steps_per_second": 0.318, "step": 1900 }, { "epoch": 0.24, "learning_rate": 0.0004435875425159201, "loss": 1.7243, "step": 1901 }, { "epoch": 0.24, "learning_rate": 0.0004435238131707117, "loss": 1.8093, "step": 1902 }, { "epoch": 0.24, "learning_rate": 0.00044346005243162654, "loss": 1.7632, "step": 1903 }, { "epoch": 0.24, "learning_rate": 0.00044339626030900827, "loss": 1.7968, "step": 1904 }, { "epoch": 0.24, "learning_rate": 0.00044333243681320535, "loss": 1.7331, "step": 1905 }, { "epoch": 0.24, "learning_rate": 0.0004432685819545713, "loss": 1.8075, "step": 1906 }, { "epoch": 0.24, "learning_rate": 0.0004432046957434649, "loss": 1.7912, "step": 1907 }, { "epoch": 0.24, "learning_rate": 0.0004431407781902498, "loss": 1.7402, "step": 1908 }, { "epoch": 0.24, "learning_rate": 0.00044307682930529494, "loss": 1.7463, "step": 1909 }, { "epoch": 0.24, "learning_rate": 0.0004430128490989742, "loss": 1.7934, "step": 1910 }, { "epoch": 0.24, "learning_rate": 0.00044294883758166665, "loss": 1.7678, "step": 1911 }, { "epoch": 0.24, "learning_rate": 0.0004428847947637563, "loss": 1.723, "step": 1912 }, { "epoch": 0.24, "learning_rate": 0.0004428207206556325, "loss": 1.7674, "step": 1913 }, { "epoch": 0.24, "learning_rate": 0.00044275661526768927, "loss": 1.7004, "step": 1914 }, { "epoch": 0.24, "learning_rate": 0.00044269247861032613, "loss": 1.78, "step": 1915 }, { "epoch": 0.24, "learning_rate": 0.0004426283106939473, "loss": 1.7969, "step": 1916 }, { "epoch": 0.24, "learning_rate": 0.0004425641115289624, "loss": 1.7692, "step": 1917 }, { "epoch": 0.24, "learning_rate": 0.0004424998811257859, "loss": 1.7692, "step": 1918 }, { "epoch": 0.24, "learning_rate": 0.00044243561949483735, "loss": 1.7814, "step": 1919 }, { "epoch": 0.24, "learning_rate": 0.0004423713266465415, "loss": 1.8063, "step": 1920 }, { "epoch": 0.24, "learning_rate": 0.00044230700259132805, "loss": 1.743, "step": 1921 }, { "epoch": 0.24, "learning_rate": 0.0004422426473396318, "loss": 1.7868, "step": 1922 }, { "epoch": 0.24, "learning_rate": 0.00044217826090189264, "loss": 1.7059, "step": 1923 }, { "epoch": 0.24, "learning_rate": 0.0004421138432885554, "loss": 1.7557, "step": 1924 }, { "epoch": 0.24, "learning_rate": 0.00044204939451007014, "loss": 1.7128, "step": 1925 }, { "epoch": 0.24, "learning_rate": 0.00044198491457689184, "loss": 1.7974, "step": 1926 }, { "epoch": 0.24, "learning_rate": 0.00044192040349948057, "loss": 1.8619, "step": 1927 }, { "epoch": 0.24, "learning_rate": 0.0004418558612883016, "loss": 1.8133, "step": 1928 }, { "epoch": 0.24, "learning_rate": 0.00044179128795382493, "loss": 1.8199, "step": 1929 }, { "epoch": 0.24, "learning_rate": 0.0004417266835065259, "loss": 1.782, "step": 1930 }, { "epoch": 0.24, "learning_rate": 0.0004416620479568847, "loss": 1.7258, "step": 1931 }, { "epoch": 0.24, "learning_rate": 0.0004415973813153867, "loss": 1.7545, "step": 1932 }, { "epoch": 0.24, "learning_rate": 0.0004415326835925224, "loss": 1.7351, "step": 1933 }, { "epoch": 0.24, "learning_rate": 0.00044146795479878704, "loss": 1.7824, "step": 1934 }, { "epoch": 0.24, "learning_rate": 0.00044140319494468115, "loss": 1.8424, "step": 1935 }, { "epoch": 0.24, "learning_rate": 0.0004413384040407102, "loss": 1.9422, "step": 1936 }, { "epoch": 0.24, "learning_rate": 0.0004412735820973848, "loss": 1.7457, "step": 1937 }, { "epoch": 0.24, "learning_rate": 0.0004412087291252204, "loss": 1.8716, "step": 1938 }, { "epoch": 0.24, "learning_rate": 0.0004411438451347376, "loss": 1.7323, "step": 1939 }, { "epoch": 0.24, "learning_rate": 0.00044107893013646207, "loss": 1.7613, "step": 1940 }, { "epoch": 0.24, "learning_rate": 0.00044101398414092454, "loss": 1.7984, "step": 1941 }, { "epoch": 0.24, "learning_rate": 0.0004409490071586606, "loss": 1.7757, "step": 1942 }, { "epoch": 0.24, "learning_rate": 0.00044088399920021095, "loss": 1.7394, "step": 1943 }, { "epoch": 0.24, "learning_rate": 0.0004408189602761214, "loss": 1.7261, "step": 1944 }, { "epoch": 0.24, "learning_rate": 0.0004407538903969427, "loss": 1.7435, "step": 1945 }, { "epoch": 0.24, "learning_rate": 0.00044068878957323053, "loss": 1.7857, "step": 1946 }, { "epoch": 0.24, "learning_rate": 0.0004406236578155459, "loss": 1.7979, "step": 1947 }, { "epoch": 0.24, "learning_rate": 0.0004405584951344545, "loss": 1.8303, "step": 1948 }, { "epoch": 0.24, "learning_rate": 0.00044049330154052717, "loss": 1.724, "step": 1949 }, { "epoch": 0.24, "learning_rate": 0.0004404280770443398, "loss": 1.6813, "step": 1950 }, { "epoch": 0.24, "learning_rate": 0.00044036282165647324, "loss": 1.7398, "step": 1951 }, { "epoch": 0.24, "learning_rate": 0.0004402975353875134, "loss": 1.761, "step": 1952 }, { "epoch": 0.24, "learning_rate": 0.000440232218248051, "loss": 1.7434, "step": 1953 }, { "epoch": 0.24, "learning_rate": 0.00044016687024868217, "loss": 1.7387, "step": 1954 }, { "epoch": 0.24, "learning_rate": 0.0004401014914000078, "loss": 1.7493, "step": 1955 }, { "epoch": 0.24, "learning_rate": 0.00044003608171263355, "loss": 1.6685, "step": 1956 }, { "epoch": 0.24, "learning_rate": 0.0004399706411971706, "loss": 1.7172, "step": 1957 }, { "epoch": 0.24, "learning_rate": 0.00043990516986423475, "loss": 1.7316, "step": 1958 }, { "epoch": 0.24, "learning_rate": 0.00043983966772444684, "loss": 1.8152, "step": 1959 }, { "epoch": 0.24, "learning_rate": 0.0004397741347884329, "loss": 1.7582, "step": 1960 }, { "epoch": 0.25, "learning_rate": 0.0004397085710668238, "loss": 1.6645, "step": 1961 }, { "epoch": 0.25, "learning_rate": 0.0004396429765702553, "loss": 1.7704, "step": 1962 }, { "epoch": 0.25, "learning_rate": 0.0004395773513093686, "loss": 1.6949, "step": 1963 }, { "epoch": 0.25, "learning_rate": 0.00043951169529480926, "loss": 1.6751, "step": 1964 }, { "epoch": 0.25, "learning_rate": 0.00043944600853722835, "loss": 1.6999, "step": 1965 }, { "epoch": 0.25, "learning_rate": 0.0004393802910472816, "loss": 1.8077, "step": 1966 }, { "epoch": 0.25, "learning_rate": 0.00043931454283562986, "loss": 1.6914, "step": 1967 }, { "epoch": 0.25, "learning_rate": 0.0004392487639129391, "loss": 1.8236, "step": 1968 }, { "epoch": 0.25, "learning_rate": 0.00043918295428988, "loss": 1.844, "step": 1969 }, { "epoch": 0.25, "learning_rate": 0.0004391171139771284, "loss": 1.7175, "step": 1970 }, { "epoch": 0.25, "learning_rate": 0.00043905124298536505, "loss": 1.7329, "step": 1971 }, { "epoch": 0.25, "learning_rate": 0.00043898534132527563, "loss": 1.717, "step": 1972 }, { "epoch": 0.25, "learning_rate": 0.00043891940900755094, "loss": 1.7629, "step": 1973 }, { "epoch": 0.25, "learning_rate": 0.00043885344604288666, "loss": 1.7932, "step": 1974 }, { "epoch": 0.25, "learning_rate": 0.0004387874524419834, "loss": 1.6646, "step": 1975 }, { "epoch": 0.25, "learning_rate": 0.0004387214282155469, "loss": 1.7201, "step": 1976 }, { "epoch": 0.25, "learning_rate": 0.00043865537337428763, "loss": 1.7479, "step": 1977 }, { "epoch": 0.25, "learning_rate": 0.0004385892879289213, "loss": 1.7926, "step": 1978 }, { "epoch": 0.25, "learning_rate": 0.00043852317189016835, "loss": 1.7216, "step": 1979 }, { "epoch": 0.25, "learning_rate": 0.0004384570252687542, "loss": 1.6968, "step": 1980 }, { "epoch": 0.25, "learning_rate": 0.0004383908480754095, "loss": 1.733, "step": 1981 }, { "epoch": 0.25, "learning_rate": 0.0004383246403208695, "loss": 1.738, "step": 1982 }, { "epoch": 0.25, "learning_rate": 0.00043825840201587465, "loss": 1.8332, "step": 1983 }, { "epoch": 0.25, "learning_rate": 0.00043819213317117033, "loss": 1.6687, "step": 1984 }, { "epoch": 0.25, "learning_rate": 0.0004381258337975067, "loss": 1.7074, "step": 1985 }, { "epoch": 0.25, "learning_rate": 0.0004380595039056391, "loss": 1.6845, "step": 1986 }, { "epoch": 0.25, "learning_rate": 0.00043799314350632755, "loss": 1.6647, "step": 1987 }, { "epoch": 0.25, "learning_rate": 0.00043792675261033744, "loss": 1.7532, "step": 1988 }, { "epoch": 0.25, "learning_rate": 0.0004378603312284387, "loss": 1.7243, "step": 1989 }, { "epoch": 0.25, "learning_rate": 0.0004377938793714064, "loss": 1.7487, "step": 1990 }, { "epoch": 0.25, "learning_rate": 0.0004377273970500205, "loss": 1.6642, "step": 1991 }, { "epoch": 0.25, "learning_rate": 0.00043766088427506586, "loss": 1.7835, "step": 1992 }, { "epoch": 0.25, "learning_rate": 0.0004375943410573324, "loss": 1.7625, "step": 1993 }, { "epoch": 0.25, "learning_rate": 0.0004375277674076149, "loss": 1.7148, "step": 1994 }, { "epoch": 0.25, "learning_rate": 0.0004374611633367131, "loss": 1.7285, "step": 1995 }, { "epoch": 0.25, "learning_rate": 0.00043739452885543164, "loss": 1.7722, "step": 1996 }, { "epoch": 0.25, "learning_rate": 0.00043732786397458013, "loss": 1.7535, "step": 1997 }, { "epoch": 0.25, "learning_rate": 0.0004372611687049731, "loss": 1.7941, "step": 1998 }, { "epoch": 0.25, "learning_rate": 0.00043719444305743006, "loss": 1.7218, "step": 1999 }, { "epoch": 0.25, "learning_rate": 0.00043712768704277526, "loss": 1.7106, "step": 2000 }, { "epoch": 0.25, "eval_acceptance_rate": 0.622637152671814, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.9215918157748901, "eval_expected_tokens": 2.4019897697186128, "eval_loss": 1.701086401939392, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.35049744242965314, "eval_runtime": 50.3922, "eval_samples_per_second": 9.664, "eval_steps_per_second": 0.318, "step": 2000 }, { "epoch": 0.25, "learning_rate": 0.0004370609006718382, "loss": 1.6948, "step": 2001 }, { "epoch": 0.25, "learning_rate": 0.00043699408395545295, "loss": 1.7093, "step": 2002 }, { "epoch": 0.25, "learning_rate": 0.0004369272369044588, "loss": 1.6816, "step": 2003 }, { "epoch": 0.25, "learning_rate": 0.0004368603595296997, "loss": 1.7069, "step": 2004 }, { "epoch": 0.25, "learning_rate": 0.0004367934518420248, "loss": 1.7411, "step": 2005 }, { "epoch": 0.25, "learning_rate": 0.0004367265138522879, "loss": 1.6686, "step": 2006 }, { "epoch": 0.25, "learning_rate": 0.00043665954557134786, "loss": 1.7378, "step": 2007 }, { "epoch": 0.25, "learning_rate": 0.00043659254701006845, "loss": 1.6698, "step": 2008 }, { "epoch": 0.25, "learning_rate": 0.00043652551817931835, "loss": 1.7346, "step": 2009 }, { "epoch": 0.25, "learning_rate": 0.0004364584590899711, "loss": 1.7643, "step": 2010 }, { "epoch": 0.25, "learning_rate": 0.00043639136975290516, "loss": 1.7794, "step": 2011 }, { "epoch": 0.25, "learning_rate": 0.000436324250179004, "loss": 1.7416, "step": 2012 }, { "epoch": 0.25, "learning_rate": 0.0004362571003791558, "loss": 1.6935, "step": 2013 }, { "epoch": 0.25, "learning_rate": 0.00043618992036425385, "loss": 1.7353, "step": 2014 }, { "epoch": 0.25, "learning_rate": 0.0004361227101451961, "loss": 1.6948, "step": 2015 }, { "epoch": 0.25, "learning_rate": 0.0004360554697328858, "loss": 1.7607, "step": 2016 }, { "epoch": 0.25, "learning_rate": 0.00043598819913823065, "loss": 1.7017, "step": 2017 }, { "epoch": 0.25, "learning_rate": 0.0004359208983721434, "loss": 1.6548, "step": 2018 }, { "epoch": 0.25, "learning_rate": 0.0004358535674455418, "loss": 1.7532, "step": 2019 }, { "epoch": 0.25, "learning_rate": 0.00043578620636934855, "loss": 1.7324, "step": 2020 }, { "epoch": 0.25, "learning_rate": 0.000435718815154491, "loss": 1.6539, "step": 2021 }, { "epoch": 0.25, "learning_rate": 0.0004356513938119015, "loss": 1.8003, "step": 2022 }, { "epoch": 0.25, "learning_rate": 0.00043558394235251733, "loss": 1.6835, "step": 2023 }, { "epoch": 0.25, "learning_rate": 0.00043551646078728056, "loss": 1.5837, "step": 2024 }, { "epoch": 0.25, "learning_rate": 0.0004354489491271383, "loss": 1.6512, "step": 2025 }, { "epoch": 0.25, "learning_rate": 0.0004353814073830423, "loss": 1.7802, "step": 2026 }, { "epoch": 0.25, "learning_rate": 0.0004353138355659495, "loss": 1.7792, "step": 2027 }, { "epoch": 0.25, "learning_rate": 0.0004352462336868215, "loss": 1.7428, "step": 2028 }, { "epoch": 0.25, "learning_rate": 0.0004351786017566248, "loss": 1.7399, "step": 2029 }, { "epoch": 0.25, "learning_rate": 0.0004351109397863309, "loss": 1.6974, "step": 2030 }, { "epoch": 0.25, "learning_rate": 0.00043504324778691585, "loss": 1.7256, "step": 2031 }, { "epoch": 0.25, "learning_rate": 0.00043497552576936103, "loss": 1.7456, "step": 2032 }, { "epoch": 0.25, "learning_rate": 0.00043490777374465244, "loss": 1.7266, "step": 2033 }, { "epoch": 0.25, "learning_rate": 0.00043483999172378086, "loss": 1.7307, "step": 2034 }, { "epoch": 0.25, "learning_rate": 0.00043477217971774214, "loss": 1.705, "step": 2035 }, { "epoch": 0.25, "learning_rate": 0.00043470433773753683, "loss": 1.7067, "step": 2036 }, { "epoch": 0.25, "learning_rate": 0.00043463646579417045, "loss": 1.8021, "step": 2037 }, { "epoch": 0.25, "learning_rate": 0.0004345685638986533, "loss": 1.6816, "step": 2038 }, { "epoch": 0.25, "learning_rate": 0.0004345006320620007, "loss": 1.6752, "step": 2039 }, { "epoch": 0.26, "learning_rate": 0.00043443267029523257, "loss": 1.7499, "step": 2040 }, { "epoch": 0.26, "learning_rate": 0.0004343646786093739, "loss": 1.7566, "step": 2041 }, { "epoch": 0.26, "learning_rate": 0.0004342966570154544, "loss": 1.7761, "step": 2042 }, { "epoch": 0.26, "learning_rate": 0.0004342286055245087, "loss": 1.7804, "step": 2043 }, { "epoch": 0.26, "learning_rate": 0.00043416052414757637, "loss": 1.6911, "step": 2044 }, { "epoch": 0.26, "learning_rate": 0.0004340924128957016, "loss": 1.7574, "step": 2045 }, { "epoch": 0.26, "learning_rate": 0.00043402427177993366, "loss": 1.7085, "step": 2046 }, { "epoch": 0.26, "learning_rate": 0.0004339561008113264, "loss": 1.7308, "step": 2047 }, { "epoch": 0.26, "learning_rate": 0.0004338879000009388, "loss": 1.6626, "step": 2048 }, { "epoch": 0.26, "learning_rate": 0.00043381966935983454, "loss": 1.695, "step": 2049 }, { "epoch": 0.26, "learning_rate": 0.00043375140889908214, "loss": 1.783, "step": 2050 }, { "epoch": 0.26, "learning_rate": 0.000433683118629755, "loss": 1.7368, "step": 2051 }, { "epoch": 0.26, "learning_rate": 0.00043361479856293125, "loss": 1.7412, "step": 2052 }, { "epoch": 0.26, "learning_rate": 0.00043354644870969395, "loss": 1.741, "step": 2053 }, { "epoch": 0.26, "learning_rate": 0.00043347806908113095, "loss": 1.7774, "step": 2054 }, { "epoch": 0.26, "learning_rate": 0.0004334096596883351, "loss": 1.8034, "step": 2055 }, { "epoch": 0.26, "learning_rate": 0.00043334122054240374, "loss": 1.7977, "step": 2056 }, { "epoch": 0.26, "learning_rate": 0.00043327275165443933, "loss": 1.7155, "step": 2057 }, { "epoch": 0.26, "learning_rate": 0.00043320425303554897, "loss": 1.7115, "step": 2058 }, { "epoch": 0.26, "learning_rate": 0.0004331357246968447, "loss": 1.6347, "step": 2059 }, { "epoch": 0.26, "learning_rate": 0.0004330671666494434, "loss": 1.6627, "step": 2060 }, { "epoch": 0.26, "learning_rate": 0.00043299857890446666, "loss": 1.694, "step": 2061 }, { "epoch": 0.26, "learning_rate": 0.0004329299614730409, "loss": 1.6681, "step": 2062 }, { "epoch": 0.26, "learning_rate": 0.0004328613143662975, "loss": 1.6811, "step": 2063 }, { "epoch": 0.26, "learning_rate": 0.00043279263759537254, "loss": 1.647, "step": 2064 }, { "epoch": 0.26, "learning_rate": 0.0004327239311714068, "loss": 1.7224, "step": 2065 }, { "epoch": 0.26, "learning_rate": 0.00043265519510554614, "loss": 1.7289, "step": 2066 }, { "epoch": 0.26, "learning_rate": 0.000432586429408941, "loss": 1.7639, "step": 2067 }, { "epoch": 0.26, "learning_rate": 0.00043251763409274667, "loss": 1.7454, "step": 2068 }, { "epoch": 0.26, "learning_rate": 0.00043244880916812334, "loss": 1.7409, "step": 2069 }, { "epoch": 0.26, "learning_rate": 0.000432379954646236, "loss": 1.7201, "step": 2070 }, { "epoch": 0.26, "learning_rate": 0.0004323110705382544, "loss": 1.6656, "step": 2071 }, { "epoch": 0.26, "learning_rate": 0.00043224215685535287, "loss": 1.6814, "step": 2072 }, { "epoch": 0.26, "learning_rate": 0.00043217321360871096, "loss": 1.7249, "step": 2073 }, { "epoch": 0.26, "learning_rate": 0.00043210424080951273, "loss": 1.7847, "step": 2074 }, { "epoch": 0.26, "learning_rate": 0.0004320352384689471, "loss": 1.6869, "step": 2075 }, { "epoch": 0.26, "learning_rate": 0.00043196620659820777, "loss": 1.6828, "step": 2076 }, { "epoch": 0.26, "learning_rate": 0.00043189714520849333, "loss": 1.6802, "step": 2077 }, { "epoch": 0.26, "learning_rate": 0.000431828054311007, "loss": 1.6933, "step": 2078 }, { "epoch": 0.26, "learning_rate": 0.00043175893391695687, "loss": 1.6974, "step": 2079 }, { "epoch": 0.26, "learning_rate": 0.0004316897840375558, "loss": 1.7784, "step": 2080 }, { "epoch": 0.26, "learning_rate": 0.00043162060468402146, "loss": 1.7128, "step": 2081 }, { "epoch": 0.26, "learning_rate": 0.00043155139586757637, "loss": 1.7489, "step": 2082 }, { "epoch": 0.26, "learning_rate": 0.0004314821575994476, "loss": 1.7336, "step": 2083 }, { "epoch": 0.26, "learning_rate": 0.0004314128898908672, "loss": 1.7228, "step": 2084 }, { "epoch": 0.26, "learning_rate": 0.0004313435927530719, "loss": 1.637, "step": 2085 }, { "epoch": 0.26, "learning_rate": 0.00043127426619730334, "loss": 1.678, "step": 2086 }, { "epoch": 0.26, "learning_rate": 0.00043120491023480775, "loss": 1.6831, "step": 2087 }, { "epoch": 0.26, "learning_rate": 0.00043113552487683625, "loss": 1.6568, "step": 2088 }, { "epoch": 0.26, "learning_rate": 0.00043106611013464465, "loss": 1.7165, "step": 2089 }, { "epoch": 0.26, "learning_rate": 0.00043099666601949363, "loss": 1.7664, "step": 2090 }, { "epoch": 0.26, "learning_rate": 0.00043092719254264844, "loss": 1.7215, "step": 2091 }, { "epoch": 0.26, "learning_rate": 0.0004308576897153794, "loss": 1.6337, "step": 2092 }, { "epoch": 0.26, "learning_rate": 0.0004307881575489613, "loss": 1.6675, "step": 2093 }, { "epoch": 0.26, "learning_rate": 0.0004307185960546739, "loss": 1.6531, "step": 2094 }, { "epoch": 0.26, "learning_rate": 0.00043064900524380156, "loss": 1.6319, "step": 2095 }, { "epoch": 0.26, "learning_rate": 0.0004305793851276335, "loss": 1.6461, "step": 2096 }, { "epoch": 0.26, "learning_rate": 0.00043050973571746355, "loss": 1.7559, "step": 2097 }, { "epoch": 0.26, "learning_rate": 0.00043044005702459054, "loss": 1.7602, "step": 2098 }, { "epoch": 0.26, "learning_rate": 0.0004303703490603179, "loss": 1.6938, "step": 2099 }, { "epoch": 0.26, "learning_rate": 0.00043030061183595364, "loss": 1.6795, "step": 2100 }, { "epoch": 0.26, "eval_acceptance_rate": 0.6290252208709717, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.9441143862936243, "eval_expected_tokens": 2.4301429828670305, "eval_loss": 1.6663103103637695, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3575357457167577, "eval_runtime": 50.4211, "eval_samples_per_second": 9.659, "eval_steps_per_second": 0.317, "step": 2100 }, { "epoch": 0.26, "learning_rate": 0.0004302308453628109, "loss": 1.6755, "step": 2101 }, { "epoch": 0.26, "learning_rate": 0.0004301610496522072, "loss": 1.712, "step": 2102 }, { "epoch": 0.26, "learning_rate": 0.000430091224715465, "loss": 1.6497, "step": 2103 }, { "epoch": 0.26, "learning_rate": 0.00043002137056391164, "loss": 1.7129, "step": 2104 }, { "epoch": 0.26, "learning_rate": 0.0004299514872088788, "loss": 1.6712, "step": 2105 }, { "epoch": 0.26, "learning_rate": 0.00042988157466170316, "loss": 1.7578, "step": 2106 }, { "epoch": 0.26, "learning_rate": 0.00042981163293372615, "loss": 1.772, "step": 2107 }, { "epoch": 0.26, "learning_rate": 0.0004297416620362938, "loss": 1.7162, "step": 2108 }, { "epoch": 0.26, "learning_rate": 0.000429671661980757, "loss": 1.6124, "step": 2109 }, { "epoch": 0.26, "learning_rate": 0.00042960163277847144, "loss": 1.7302, "step": 2110 }, { "epoch": 0.26, "learning_rate": 0.0004295315744407972, "loss": 1.6518, "step": 2111 }, { "epoch": 0.26, "learning_rate": 0.0004294614869790994, "loss": 1.6598, "step": 2112 }, { "epoch": 0.26, "learning_rate": 0.00042939137040474783, "loss": 1.7372, "step": 2113 }, { "epoch": 0.26, "learning_rate": 0.00042932122472911697, "loss": 1.719, "step": 2114 }, { "epoch": 0.26, "learning_rate": 0.0004292510499635859, "loss": 1.6348, "step": 2115 }, { "epoch": 0.26, "learning_rate": 0.00042918084611953856, "loss": 1.7112, "step": 2116 }, { "epoch": 0.26, "learning_rate": 0.00042911061320836366, "loss": 1.6714, "step": 2117 }, { "epoch": 0.26, "learning_rate": 0.0004290403512414545, "loss": 1.6605, "step": 2118 }, { "epoch": 0.26, "learning_rate": 0.00042897006023020915, "loss": 1.5995, "step": 2119 }, { "epoch": 0.27, "learning_rate": 0.0004288997401860303, "loss": 1.686, "step": 2120 }, { "epoch": 0.27, "learning_rate": 0.0004288293911203255, "loss": 1.7017, "step": 2121 }, { "epoch": 0.27, "learning_rate": 0.000428759013044507, "loss": 1.6498, "step": 2122 }, { "epoch": 0.27, "learning_rate": 0.00042868860596999156, "loss": 1.7289, "step": 2123 }, { "epoch": 0.27, "learning_rate": 0.00042861816990820087, "loss": 1.7313, "step": 2124 }, { "epoch": 0.27, "learning_rate": 0.00042854770487056116, "loss": 1.7514, "step": 2125 }, { "epoch": 0.27, "learning_rate": 0.00042847721086850356, "loss": 1.7904, "step": 2126 }, { "epoch": 0.27, "learning_rate": 0.0004284066879134636, "loss": 1.7368, "step": 2127 }, { "epoch": 0.27, "learning_rate": 0.0004283361360168817, "loss": 1.7385, "step": 2128 }, { "epoch": 0.27, "learning_rate": 0.0004282655551902031, "loss": 1.6101, "step": 2129 }, { "epoch": 0.27, "learning_rate": 0.00042819494544487746, "loss": 1.7064, "step": 2130 }, { "epoch": 0.27, "learning_rate": 0.0004281243067923593, "loss": 1.7181, "step": 2131 }, { "epoch": 0.27, "learning_rate": 0.00042805363924410775, "loss": 1.8269, "step": 2132 }, { "epoch": 0.27, "learning_rate": 0.00042798294281158665, "loss": 1.679, "step": 2133 }, { "epoch": 0.27, "learning_rate": 0.00042791221750626465, "loss": 1.6968, "step": 2134 }, { "epoch": 0.27, "learning_rate": 0.0004278414633396148, "loss": 1.7042, "step": 2135 }, { "epoch": 0.27, "learning_rate": 0.00042777068032311507, "loss": 1.7526, "step": 2136 }, { "epoch": 0.27, "learning_rate": 0.00042769986846824813, "loss": 1.7064, "step": 2137 }, { "epoch": 0.27, "learning_rate": 0.00042762902778650114, "loss": 1.7433, "step": 2138 }, { "epoch": 0.27, "learning_rate": 0.00042755815828936607, "loss": 1.7266, "step": 2139 }, { "epoch": 0.27, "learning_rate": 0.0004274872599883396, "loss": 1.6931, "step": 2140 }, { "epoch": 0.27, "learning_rate": 0.00042741633289492285, "loss": 1.7351, "step": 2141 }, { "epoch": 0.27, "learning_rate": 0.00042734537702062195, "loss": 1.6824, "step": 2142 }, { "epoch": 0.27, "learning_rate": 0.00042727439237694745, "loss": 1.6474, "step": 2143 }, { "epoch": 0.27, "learning_rate": 0.00042720337897541455, "loss": 1.691, "step": 2144 }, { "epoch": 0.27, "learning_rate": 0.0004271323368275434, "loss": 1.6519, "step": 2145 }, { "epoch": 0.27, "learning_rate": 0.0004270612659448584, "loss": 1.7471, "step": 2146 }, { "epoch": 0.27, "learning_rate": 0.00042699016633888904, "loss": 1.7167, "step": 2147 }, { "epoch": 0.27, "learning_rate": 0.00042691903802116916, "loss": 1.8413, "step": 2148 }, { "epoch": 0.27, "learning_rate": 0.0004268478810032373, "loss": 1.7382, "step": 2149 }, { "epoch": 0.27, "learning_rate": 0.00042677669529663686, "loss": 1.6667, "step": 2150 }, { "epoch": 0.27, "learning_rate": 0.0004267054809129157, "loss": 1.717, "step": 2151 }, { "epoch": 0.27, "learning_rate": 0.0004266342378636263, "loss": 1.7532, "step": 2152 }, { "epoch": 0.27, "learning_rate": 0.00042656296616032596, "loss": 1.6846, "step": 2153 }, { "epoch": 0.27, "learning_rate": 0.0004264916658145764, "loss": 1.6722, "step": 2154 }, { "epoch": 0.27, "learning_rate": 0.00042642033683794436, "loss": 1.7531, "step": 2155 }, { "epoch": 0.27, "learning_rate": 0.00042634897924200075, "loss": 1.6681, "step": 2156 }, { "epoch": 0.27, "learning_rate": 0.00042627759303832155, "loss": 1.7109, "step": 2157 }, { "epoch": 0.27, "learning_rate": 0.00042620617823848706, "loss": 1.7894, "step": 2158 }, { "epoch": 0.27, "learning_rate": 0.00042613473485408237, "loss": 1.7302, "step": 2159 }, { "epoch": 0.27, "learning_rate": 0.0004260632628966974, "loss": 1.6197, "step": 2160 }, { "epoch": 0.27, "learning_rate": 0.00042599176237792616, "loss": 1.7899, "step": 2161 }, { "epoch": 0.27, "learning_rate": 0.0004259202333093678, "loss": 1.7292, "step": 2162 }, { "epoch": 0.27, "learning_rate": 0.00042584867570262595, "loss": 1.7845, "step": 2163 }, { "epoch": 0.27, "learning_rate": 0.00042577708956930885, "loss": 1.7422, "step": 2164 }, { "epoch": 0.27, "learning_rate": 0.0004257054749210293, "loss": 1.7146, "step": 2165 }, { "epoch": 0.27, "learning_rate": 0.00042563383176940484, "loss": 1.6662, "step": 2166 }, { "epoch": 0.27, "learning_rate": 0.00042556216012605753, "loss": 1.6577, "step": 2167 }, { "epoch": 0.27, "learning_rate": 0.0004254904600026143, "loss": 1.5976, "step": 2168 }, { "epoch": 0.27, "learning_rate": 0.0004254187314107062, "loss": 1.7067, "step": 2169 }, { "epoch": 0.27, "learning_rate": 0.0004253469743619694, "loss": 1.7213, "step": 2170 }, { "epoch": 0.27, "learning_rate": 0.00042527518886804444, "loss": 1.6923, "step": 2171 }, { "epoch": 0.27, "learning_rate": 0.0004252033749405766, "loss": 1.573, "step": 2172 }, { "epoch": 0.27, "learning_rate": 0.00042513153259121563, "loss": 1.7107, "step": 2173 }, { "epoch": 0.27, "learning_rate": 0.0004250596618316159, "loss": 1.7033, "step": 2174 }, { "epoch": 0.27, "learning_rate": 0.00042498776267343665, "loss": 1.6029, "step": 2175 }, { "epoch": 0.27, "learning_rate": 0.00042491583512834137, "loss": 1.7068, "step": 2176 }, { "epoch": 0.27, "learning_rate": 0.00042484387920799834, "loss": 1.6662, "step": 2177 }, { "epoch": 0.27, "learning_rate": 0.00042477189492408043, "loss": 1.618, "step": 2178 }, { "epoch": 0.27, "learning_rate": 0.0004246998822882651, "loss": 1.7221, "step": 2179 }, { "epoch": 0.27, "learning_rate": 0.0004246278413122343, "loss": 1.8062, "step": 2180 }, { "epoch": 0.27, "learning_rate": 0.00042455577200767476, "loss": 1.6587, "step": 2181 }, { "epoch": 0.27, "learning_rate": 0.0004244836743862778, "loss": 1.6582, "step": 2182 }, { "epoch": 0.27, "learning_rate": 0.00042441154845973924, "loss": 1.6994, "step": 2183 }, { "epoch": 0.27, "learning_rate": 0.0004243393942397594, "loss": 1.6697, "step": 2184 }, { "epoch": 0.27, "learning_rate": 0.00042426721173804337, "loss": 1.6462, "step": 2185 }, { "epoch": 0.27, "learning_rate": 0.0004241950009663008, "loss": 1.6787, "step": 2186 }, { "epoch": 0.27, "learning_rate": 0.0004241227619362459, "loss": 1.6767, "step": 2187 }, { "epoch": 0.27, "learning_rate": 0.00042405049465959733, "loss": 1.6686, "step": 2188 }, { "epoch": 0.27, "learning_rate": 0.00042397819914807855, "loss": 1.632, "step": 2189 }, { "epoch": 0.27, "learning_rate": 0.0004239058754134175, "loss": 1.7152, "step": 2190 }, { "epoch": 0.27, "learning_rate": 0.0004238335234673467, "loss": 1.6709, "step": 2191 }, { "epoch": 0.27, "learning_rate": 0.0004237611433216032, "loss": 1.6322, "step": 2192 }, { "epoch": 0.27, "learning_rate": 0.00042368873498792873, "loss": 1.7328, "step": 2193 }, { "epoch": 0.27, "learning_rate": 0.0004236162984780695, "loss": 1.6872, "step": 2194 }, { "epoch": 0.27, "learning_rate": 0.00042354383380377637, "loss": 1.6864, "step": 2195 }, { "epoch": 0.27, "learning_rate": 0.0004234713409768046, "loss": 1.7664, "step": 2196 }, { "epoch": 0.27, "learning_rate": 0.0004233988200089144, "loss": 1.5771, "step": 2197 }, { "epoch": 0.27, "learning_rate": 0.00042332627091186994, "loss": 1.6838, "step": 2198 }, { "epoch": 0.27, "learning_rate": 0.0004232536936974406, "loss": 1.7028, "step": 2199 }, { "epoch": 0.28, "learning_rate": 0.00042318108837739987, "loss": 1.7166, "step": 2200 }, { "epoch": 0.28, "eval_acceptance_rate": 0.6291758418083191, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.9446495622023905, "eval_expected_tokens": 2.430811952752988, "eval_loss": 1.6534298658370972, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.357702988188247, "eval_runtime": 50.5533, "eval_samples_per_second": 9.633, "eval_steps_per_second": 0.316, "step": 2200 }, { "epoch": 0.28, "learning_rate": 0.00042310845496352595, "loss": 1.7137, "step": 2201 }, { "epoch": 0.28, "learning_rate": 0.0004230357934676017, "loss": 1.6591, "step": 2202 }, { "epoch": 0.28, "learning_rate": 0.0004229631039014143, "loss": 1.6642, "step": 2203 }, { "epoch": 0.28, "learning_rate": 0.0004228903862767558, "loss": 1.632, "step": 2204 }, { "epoch": 0.28, "learning_rate": 0.00042281764060542247, "loss": 1.6714, "step": 2205 }, { "epoch": 0.28, "learning_rate": 0.0004227448668992153, "loss": 1.6152, "step": 2206 }, { "epoch": 0.28, "learning_rate": 0.00042267206516993983, "loss": 1.7141, "step": 2207 }, { "epoch": 0.28, "learning_rate": 0.00042259923542940607, "loss": 1.6423, "step": 2208 }, { "epoch": 0.28, "learning_rate": 0.0004225263776894288, "loss": 1.6191, "step": 2209 }, { "epoch": 0.28, "learning_rate": 0.0004224534919618269, "loss": 1.7114, "step": 2210 }, { "epoch": 0.28, "learning_rate": 0.00042238057825842436, "loss": 1.7083, "step": 2211 }, { "epoch": 0.28, "learning_rate": 0.00042230763659104916, "loss": 1.6898, "step": 2212 }, { "epoch": 0.28, "learning_rate": 0.0004222346669715341, "loss": 1.6333, "step": 2213 }, { "epoch": 0.28, "learning_rate": 0.00042216166941171665, "loss": 1.6913, "step": 2214 }, { "epoch": 0.28, "learning_rate": 0.0004220886439234385, "loss": 1.7309, "step": 2215 }, { "epoch": 0.28, "learning_rate": 0.00042201559051854604, "loss": 1.629, "step": 2216 }, { "epoch": 0.28, "learning_rate": 0.0004219425092088902, "loss": 1.6484, "step": 2217 }, { "epoch": 0.28, "learning_rate": 0.0004218694000063263, "loss": 1.6581, "step": 2218 }, { "epoch": 0.28, "learning_rate": 0.00042179626292271436, "loss": 1.5903, "step": 2219 }, { "epoch": 0.28, "learning_rate": 0.0004217230979699188, "loss": 1.6482, "step": 2220 }, { "epoch": 0.28, "learning_rate": 0.0004216499051598087, "loss": 1.6011, "step": 2221 }, { "epoch": 0.28, "learning_rate": 0.0004215766845042575, "loss": 1.5299, "step": 2222 }, { "epoch": 0.28, "learning_rate": 0.00042150343601514315, "loss": 1.6053, "step": 2223 }, { "epoch": 0.28, "learning_rate": 0.0004214301597043483, "loss": 1.5546, "step": 2224 }, { "epoch": 0.28, "learning_rate": 0.00042135685558375994, "loss": 1.6283, "step": 2225 }, { "epoch": 0.28, "learning_rate": 0.0004212835236652697, "loss": 1.7549, "step": 2226 }, { "epoch": 0.28, "learning_rate": 0.0004212101639607736, "loss": 1.695, "step": 2227 }, { "epoch": 0.28, "learning_rate": 0.0004211367764821722, "loss": 1.6595, "step": 2228 }, { "epoch": 0.28, "learning_rate": 0.0004210633612413706, "loss": 1.594, "step": 2229 }, { "epoch": 0.28, "learning_rate": 0.0004209899182502784, "loss": 1.6158, "step": 2230 }, { "epoch": 0.28, "learning_rate": 0.0004209164475208097, "loss": 1.6358, "step": 2231 }, { "epoch": 0.28, "learning_rate": 0.00042084294906488315, "loss": 1.6476, "step": 2232 }, { "epoch": 0.28, "learning_rate": 0.00042076942289442167, "loss": 1.7546, "step": 2233 }, { "epoch": 0.28, "learning_rate": 0.000420695869021353, "loss": 1.5845, "step": 2234 }, { "epoch": 0.28, "learning_rate": 0.0004206222874576091, "loss": 1.7102, "step": 2235 }, { "epoch": 0.28, "learning_rate": 0.00042054867821512666, "loss": 1.557, "step": 2236 }, { "epoch": 0.28, "learning_rate": 0.0004204750413058466, "loss": 1.7081, "step": 2237 }, { "epoch": 0.28, "learning_rate": 0.00042040137674171464, "loss": 1.6591, "step": 2238 }, { "epoch": 0.28, "learning_rate": 0.00042032768453468075, "loss": 1.6418, "step": 2239 }, { "epoch": 0.28, "learning_rate": 0.0004202539646966993, "loss": 1.5861, "step": 2240 }, { "epoch": 0.28, "learning_rate": 0.0004201802172397295, "loss": 1.7026, "step": 2241 }, { "epoch": 0.28, "learning_rate": 0.00042010644217573467, "loss": 1.6399, "step": 2242 }, { "epoch": 0.28, "learning_rate": 0.0004200326395166829, "loss": 1.6094, "step": 2243 }, { "epoch": 0.28, "learning_rate": 0.0004199588092745465, "loss": 1.6158, "step": 2244 }, { "epoch": 0.28, "learning_rate": 0.0004198849514613026, "loss": 1.5855, "step": 2245 }, { "epoch": 0.28, "learning_rate": 0.00041981106608893226, "loss": 1.689, "step": 2246 }, { "epoch": 0.28, "learning_rate": 0.00041973715316942164, "loss": 1.6365, "step": 2247 }, { "epoch": 0.28, "learning_rate": 0.00041966321271476087, "loss": 1.6274, "step": 2248 }, { "epoch": 0.28, "learning_rate": 0.00041958924473694485, "loss": 1.6606, "step": 2249 }, { "epoch": 0.28, "learning_rate": 0.00041951524924797276, "loss": 1.6488, "step": 2250 }, { "epoch": 0.28, "learning_rate": 0.0004194412262598483, "loss": 1.7022, "step": 2251 }, { "epoch": 0.28, "learning_rate": 0.00041936717578457973, "loss": 1.6493, "step": 2252 }, { "epoch": 0.28, "learning_rate": 0.00041929309783417967, "loss": 1.7199, "step": 2253 }, { "epoch": 0.28, "learning_rate": 0.0004192189924206652, "loss": 1.6263, "step": 2254 }, { "epoch": 0.28, "learning_rate": 0.0004191448595560579, "loss": 1.6337, "step": 2255 }, { "epoch": 0.28, "learning_rate": 0.0004190706992523836, "loss": 1.6429, "step": 2256 }, { "epoch": 0.28, "learning_rate": 0.00041899651152167304, "loss": 1.6571, "step": 2257 }, { "epoch": 0.28, "learning_rate": 0.0004189222963759609, "loss": 1.6362, "step": 2258 }, { "epoch": 0.28, "learning_rate": 0.00041884805382728663, "loss": 1.6156, "step": 2259 }, { "epoch": 0.28, "learning_rate": 0.00041877378388769406, "loss": 1.6725, "step": 2260 }, { "epoch": 0.28, "learning_rate": 0.0004186994865692313, "loss": 1.6665, "step": 2261 }, { "epoch": 0.28, "learning_rate": 0.0004186251618839512, "loss": 1.647, "step": 2262 }, { "epoch": 0.28, "learning_rate": 0.0004185508098439107, "loss": 1.6392, "step": 2263 }, { "epoch": 0.28, "learning_rate": 0.00041847643046117146, "loss": 1.7006, "step": 2264 }, { "epoch": 0.28, "learning_rate": 0.00041840202374779947, "loss": 1.6838, "step": 2265 }, { "epoch": 0.28, "learning_rate": 0.00041832758971586514, "loss": 1.6867, "step": 2266 }, { "epoch": 0.28, "learning_rate": 0.00041825312837744333, "loss": 1.6177, "step": 2267 }, { "epoch": 0.28, "learning_rate": 0.00041817863974461336, "loss": 1.7357, "step": 2268 }, { "epoch": 0.28, "learning_rate": 0.00041810412382945893, "loss": 1.6729, "step": 2269 }, { "epoch": 0.28, "learning_rate": 0.00041802958064406814, "loss": 1.6719, "step": 2270 }, { "epoch": 0.28, "learning_rate": 0.00041795501020053363, "loss": 1.6988, "step": 2271 }, { "epoch": 0.28, "learning_rate": 0.00041788041251095235, "loss": 1.659, "step": 2272 }, { "epoch": 0.28, "learning_rate": 0.0004178057875874257, "loss": 1.7126, "step": 2273 }, { "epoch": 0.28, "learning_rate": 0.00041773113544205955, "loss": 1.6657, "step": 2274 }, { "epoch": 0.28, "learning_rate": 0.000417656456086964, "loss": 1.632, "step": 2275 }, { "epoch": 0.28, "learning_rate": 0.00041758174953425393, "loss": 1.6493, "step": 2276 }, { "epoch": 0.28, "learning_rate": 0.00041750701579604824, "loss": 1.6342, "step": 2277 }, { "epoch": 0.28, "learning_rate": 0.00041743225488447043, "loss": 1.6773, "step": 2278 }, { "epoch": 0.28, "learning_rate": 0.00041735746681164845, "loss": 1.6433, "step": 2279 }, { "epoch": 0.28, "learning_rate": 0.0004172826515897146, "loss": 1.6678, "step": 2280 }, { "epoch": 0.29, "learning_rate": 0.00041720780923080547, "loss": 1.6118, "step": 2281 }, { "epoch": 0.29, "learning_rate": 0.00041713293974706225, "loss": 1.6691, "step": 2282 }, { "epoch": 0.29, "learning_rate": 0.00041705804315063047, "loss": 1.6804, "step": 2283 }, { "epoch": 0.29, "learning_rate": 0.00041698311945365986, "loss": 1.6551, "step": 2284 }, { "epoch": 0.29, "learning_rate": 0.00041690816866830487, "loss": 1.6113, "step": 2285 }, { "epoch": 0.29, "learning_rate": 0.0004168331908067241, "loss": 1.5583, "step": 2286 }, { "epoch": 0.29, "learning_rate": 0.0004167581858810808, "loss": 1.6429, "step": 2287 }, { "epoch": 0.29, "learning_rate": 0.0004166831539035423, "loss": 1.6444, "step": 2288 }, { "epoch": 0.29, "learning_rate": 0.0004166080948862803, "loss": 1.5981, "step": 2289 }, { "epoch": 0.29, "learning_rate": 0.0004165330088414714, "loss": 1.655, "step": 2290 }, { "epoch": 0.29, "learning_rate": 0.00041645789578129594, "loss": 1.6164, "step": 2291 }, { "epoch": 0.29, "learning_rate": 0.0004163827557179391, "loss": 1.6708, "step": 2292 }, { "epoch": 0.29, "learning_rate": 0.0004163075886635902, "loss": 1.6595, "step": 2293 }, { "epoch": 0.29, "learning_rate": 0.00041623239463044306, "loss": 1.5913, "step": 2294 }, { "epoch": 0.29, "learning_rate": 0.00041615717363069577, "loss": 1.6297, "step": 2295 }, { "epoch": 0.29, "learning_rate": 0.0004160819256765509, "loss": 1.5271, "step": 2296 }, { "epoch": 0.29, "learning_rate": 0.00041600665078021527, "loss": 1.6199, "step": 2297 }, { "epoch": 0.29, "learning_rate": 0.00041593134895390025, "loss": 1.6282, "step": 2298 }, { "epoch": 0.29, "learning_rate": 0.00041585602020982143, "loss": 1.6112, "step": 2299 }, { "epoch": 0.29, "learning_rate": 0.0004157806645601988, "loss": 1.6856, "step": 2300 }, { "epoch": 0.29, "eval_acceptance_rate": 0.6328938007354736, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.9579206571819718, "eval_expected_tokens": 2.4474008214774647, "eval_loss": 1.6235488653182983, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.36185020536936624, "eval_runtime": 50.3983, "eval_samples_per_second": 9.663, "eval_steps_per_second": 0.317, "step": 2300 }, { "epoch": 0.29, "learning_rate": 0.00041570528201725676, "loss": 1.6308, "step": 2301 }, { "epoch": 0.29, "learning_rate": 0.000415629872593224, "loss": 1.6505, "step": 2302 }, { "epoch": 0.29, "learning_rate": 0.00041555443630033364, "loss": 1.6583, "step": 2303 }, { "epoch": 0.29, "learning_rate": 0.000415478973150823, "loss": 1.6233, "step": 2304 }, { "epoch": 0.29, "learning_rate": 0.0004154034831569341, "loss": 1.6816, "step": 2305 }, { "epoch": 0.29, "learning_rate": 0.00041532796633091297, "loss": 1.5908, "step": 2306 }, { "epoch": 0.29, "learning_rate": 0.00041525242268501016, "loss": 1.6625, "step": 2307 }, { "epoch": 0.29, "learning_rate": 0.00041517685223148055, "loss": 1.6984, "step": 2308 }, { "epoch": 0.29, "learning_rate": 0.00041510125498258325, "loss": 1.5562, "step": 2309 }, { "epoch": 0.29, "learning_rate": 0.000415025630950582, "loss": 1.6011, "step": 2310 }, { "epoch": 0.29, "learning_rate": 0.0004149499801477445, "loss": 1.6397, "step": 2311 }, { "epoch": 0.29, "learning_rate": 0.0004148743025863431, "loss": 1.5897, "step": 2312 }, { "epoch": 0.29, "learning_rate": 0.0004147985982786544, "loss": 1.7205, "step": 2313 }, { "epoch": 0.29, "learning_rate": 0.00041472286723695935, "loss": 1.6188, "step": 2314 }, { "epoch": 0.29, "learning_rate": 0.0004146471094735431, "loss": 1.5269, "step": 2315 }, { "epoch": 0.29, "learning_rate": 0.0004145713250006954, "loss": 1.558, "step": 2316 }, { "epoch": 0.29, "learning_rate": 0.00041449551383071004, "loss": 1.6426, "step": 2317 }, { "epoch": 0.29, "learning_rate": 0.00041441967597588537, "loss": 1.6281, "step": 2318 }, { "epoch": 0.29, "learning_rate": 0.00041434381144852395, "loss": 1.6142, "step": 2319 }, { "epoch": 0.29, "learning_rate": 0.00041426792026093273, "loss": 1.6118, "step": 2320 }, { "epoch": 0.29, "learning_rate": 0.0004141920024254229, "loss": 1.6081, "step": 2321 }, { "epoch": 0.29, "learning_rate": 0.00041411605795431, "loss": 1.5377, "step": 2322 }, { "epoch": 0.29, "learning_rate": 0.000414040086859914, "loss": 1.6788, "step": 2323 }, { "epoch": 0.29, "learning_rate": 0.00041396408915455907, "loss": 1.6845, "step": 2324 }, { "epoch": 0.29, "learning_rate": 0.00041388806485057376, "loss": 1.6557, "step": 2325 }, { "epoch": 0.29, "learning_rate": 0.00041381201396029077, "loss": 1.6255, "step": 2326 }, { "epoch": 0.29, "learning_rate": 0.00041373593649604747, "loss": 1.62, "step": 2327 }, { "epoch": 0.29, "learning_rate": 0.0004136598324701851, "loss": 1.5707, "step": 2328 }, { "epoch": 0.29, "learning_rate": 0.0004135837018950497, "loss": 1.716, "step": 2329 }, { "epoch": 0.29, "learning_rate": 0.00041350754478299115, "loss": 1.678, "step": 2330 }, { "epoch": 0.29, "learning_rate": 0.00041343136114636376, "loss": 1.5937, "step": 2331 }, { "epoch": 0.29, "learning_rate": 0.0004133551509975264, "loss": 1.6434, "step": 2332 }, { "epoch": 0.29, "learning_rate": 0.0004132789143488419, "loss": 1.6835, "step": 2333 }, { "epoch": 0.29, "learning_rate": 0.0004132026512126777, "loss": 1.6804, "step": 2334 }, { "epoch": 0.29, "learning_rate": 0.0004131263616014053, "loss": 1.6724, "step": 2335 }, { "epoch": 0.29, "learning_rate": 0.0004130500455274005, "loss": 1.6702, "step": 2336 }, { "epoch": 0.29, "learning_rate": 0.0004129737030030436, "loss": 1.632, "step": 2337 }, { "epoch": 0.29, "learning_rate": 0.00041289733404071893, "loss": 1.6282, "step": 2338 }, { "epoch": 0.29, "learning_rate": 0.00041282093865281546, "loss": 1.5825, "step": 2339 }, { "epoch": 0.29, "learning_rate": 0.00041274451685172596, "loss": 1.5939, "step": 2340 }, { "epoch": 0.29, "learning_rate": 0.00041266806864984784, "loss": 1.5935, "step": 2341 }, { "epoch": 0.29, "learning_rate": 0.0004125915940595828, "loss": 1.6602, "step": 2342 }, { "epoch": 0.29, "learning_rate": 0.00041251509309333666, "loss": 1.6461, "step": 2343 }, { "epoch": 0.29, "learning_rate": 0.0004124385657635196, "loss": 1.6029, "step": 2344 }, { "epoch": 0.29, "learning_rate": 0.0004123620120825459, "loss": 1.6043, "step": 2345 }, { "epoch": 0.29, "learning_rate": 0.00041228543206283454, "loss": 1.6294, "step": 2346 }, { "epoch": 0.29, "learning_rate": 0.0004122088257168083, "loss": 1.5493, "step": 2347 }, { "epoch": 0.29, "learning_rate": 0.00041213219305689453, "loss": 1.533, "step": 2348 }, { "epoch": 0.29, "learning_rate": 0.0004120555340955248, "loss": 1.6411, "step": 2349 }, { "epoch": 0.29, "learning_rate": 0.0004119788488451347, "loss": 1.6157, "step": 2350 }, { "epoch": 0.29, "learning_rate": 0.0004119021373181645, "loss": 1.6626, "step": 2351 }, { "epoch": 0.29, "learning_rate": 0.00041182539952705846, "loss": 1.6269, "step": 2352 }, { "epoch": 0.29, "learning_rate": 0.0004117486354842651, "loss": 1.6684, "step": 2353 }, { "epoch": 0.29, "learning_rate": 0.0004116718452022373, "loss": 1.6098, "step": 2354 }, { "epoch": 0.29, "learning_rate": 0.00041159502869343224, "loss": 1.6003, "step": 2355 }, { "epoch": 0.29, "learning_rate": 0.00041151818597031113, "loss": 1.5991, "step": 2356 }, { "epoch": 0.29, "learning_rate": 0.0004114413170453397, "loss": 1.6037, "step": 2357 }, { "epoch": 0.29, "learning_rate": 0.00041136442193098765, "loss": 1.5489, "step": 2358 }, { "epoch": 0.29, "learning_rate": 0.0004112875006397292, "loss": 1.6745, "step": 2359 }, { "epoch": 0.29, "learning_rate": 0.00041121055318404264, "loss": 1.7314, "step": 2360 }, { "epoch": 0.3, "learning_rate": 0.0004111335795764106, "loss": 1.6187, "step": 2361 }, { "epoch": 0.3, "learning_rate": 0.00041105657982932, "loss": 1.6406, "step": 2362 }, { "epoch": 0.3, "learning_rate": 0.0004109795539552617, "loss": 1.5979, "step": 2363 }, { "epoch": 0.3, "learning_rate": 0.0004109025019667312, "loss": 1.5596, "step": 2364 }, { "epoch": 0.3, "learning_rate": 0.0004108254238762279, "loss": 1.6079, "step": 2365 }, { "epoch": 0.3, "learning_rate": 0.00041074831969625575, "loss": 1.6637, "step": 2366 }, { "epoch": 0.3, "learning_rate": 0.0004106711894393226, "loss": 1.5853, "step": 2367 }, { "epoch": 0.3, "learning_rate": 0.00041059403311794086, "loss": 1.722, "step": 2368 }, { "epoch": 0.3, "learning_rate": 0.00041051685074462697, "loss": 1.6229, "step": 2369 }, { "epoch": 0.3, "learning_rate": 0.00041043964233190145, "loss": 1.5854, "step": 2370 }, { "epoch": 0.3, "learning_rate": 0.0004103624078922895, "loss": 1.5529, "step": 2371 }, { "epoch": 0.3, "learning_rate": 0.00041028514743832, "loss": 1.5839, "step": 2372 }, { "epoch": 0.3, "learning_rate": 0.0004102078609825266, "loss": 1.651, "step": 2373 }, { "epoch": 0.3, "learning_rate": 0.0004101305485374466, "loss": 1.6679, "step": 2374 }, { "epoch": 0.3, "learning_rate": 0.000410053210115622, "loss": 1.6386, "step": 2375 }, { "epoch": 0.3, "learning_rate": 0.0004099758457295988, "loss": 1.6938, "step": 2376 }, { "epoch": 0.3, "learning_rate": 0.00040989845539192703, "loss": 1.5893, "step": 2377 }, { "epoch": 0.3, "learning_rate": 0.00040982103911516146, "loss": 1.5791, "step": 2378 }, { "epoch": 0.3, "learning_rate": 0.0004097435969118605, "loss": 1.6051, "step": 2379 }, { "epoch": 0.3, "learning_rate": 0.00040966612879458697, "loss": 1.587, "step": 2380 }, { "epoch": 0.3, "learning_rate": 0.0004095886347759081, "loss": 1.629, "step": 2381 }, { "epoch": 0.3, "learning_rate": 0.00040951111486839506, "loss": 1.6396, "step": 2382 }, { "epoch": 0.3, "learning_rate": 0.00040943356908462324, "loss": 1.7139, "step": 2383 }, { "epoch": 0.3, "learning_rate": 0.00040935599743717243, "loss": 1.5764, "step": 2384 }, { "epoch": 0.3, "learning_rate": 0.00040927839993862644, "loss": 1.6403, "step": 2385 }, { "epoch": 0.3, "learning_rate": 0.0004092007766015732, "loss": 1.5783, "step": 2386 }, { "epoch": 0.3, "learning_rate": 0.00040912312743860503, "loss": 1.6467, "step": 2387 }, { "epoch": 0.3, "learning_rate": 0.0004090454524623185, "loss": 1.6617, "step": 2388 }, { "epoch": 0.3, "learning_rate": 0.0004089677516853139, "loss": 1.6343, "step": 2389 }, { "epoch": 0.3, "learning_rate": 0.00040889002512019637, "loss": 1.6783, "step": 2390 }, { "epoch": 0.3, "learning_rate": 0.00040881227277957463, "loss": 1.7407, "step": 2391 }, { "epoch": 0.3, "learning_rate": 0.00040873449467606195, "loss": 1.573, "step": 2392 }, { "epoch": 0.3, "learning_rate": 0.00040865669082227566, "loss": 1.6608, "step": 2393 }, { "epoch": 0.3, "learning_rate": 0.00040857886123083734, "loss": 1.5642, "step": 2394 }, { "epoch": 0.3, "learning_rate": 0.00040850100591437253, "loss": 1.6355, "step": 2395 }, { "epoch": 0.3, "learning_rate": 0.0004084231248855113, "loss": 1.5586, "step": 2396 }, { "epoch": 0.3, "learning_rate": 0.00040834521815688753, "loss": 1.5569, "step": 2397 }, { "epoch": 0.3, "learning_rate": 0.00040826728574113954, "loss": 1.7387, "step": 2398 }, { "epoch": 0.3, "learning_rate": 0.0004081893276509096, "loss": 1.6014, "step": 2399 }, { "epoch": 0.3, "learning_rate": 0.0004081113438988443, "loss": 1.5661, "step": 2400 }, { "epoch": 0.3, "eval_acceptance_rate": 0.6370581984519958, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.972924569980603, "eval_expected_tokens": 2.466155712475754, "eval_loss": 1.605292558670044, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.36653892811893846, "eval_runtime": 50.4133, "eval_samples_per_second": 9.66, "eval_steps_per_second": 0.317, "step": 2400 }, { "epoch": 0.3, "learning_rate": 0.00040803333449759437, "loss": 1.5836, "step": 2401 }, { "epoch": 0.3, "learning_rate": 0.0004079552994598147, "loss": 1.5769, "step": 2402 }, { "epoch": 0.3, "learning_rate": 0.0004078772387981642, "loss": 1.5831, "step": 2403 }, { "epoch": 0.3, "learning_rate": 0.0004077991525253061, "loss": 1.6808, "step": 2404 }, { "epoch": 0.3, "learning_rate": 0.0004077210406539077, "loss": 1.6332, "step": 2405 }, { "epoch": 0.3, "learning_rate": 0.0004076429031966405, "loss": 1.5713, "step": 2406 }, { "epoch": 0.3, "learning_rate": 0.0004075647401661803, "loss": 1.5955, "step": 2407 }, { "epoch": 0.3, "learning_rate": 0.00040748655157520677, "loss": 1.5891, "step": 2408 }, { "epoch": 0.3, "learning_rate": 0.0004074083374364037, "loss": 1.5259, "step": 2409 }, { "epoch": 0.3, "learning_rate": 0.0004073300977624594, "loss": 1.5035, "step": 2410 }, { "epoch": 0.3, "learning_rate": 0.0004072518325660659, "loss": 1.5783, "step": 2411 }, { "epoch": 0.3, "learning_rate": 0.00040717354185991966, "loss": 1.5936, "step": 2412 }, { "epoch": 0.3, "learning_rate": 0.0004070952256567212, "loss": 1.5666, "step": 2413 }, { "epoch": 0.3, "learning_rate": 0.0004070168839691751, "loss": 1.5878, "step": 2414 }, { "epoch": 0.3, "learning_rate": 0.0004069385168099902, "loss": 1.5481, "step": 2415 }, { "epoch": 0.3, "learning_rate": 0.00040686012419187916, "loss": 1.6408, "step": 2416 }, { "epoch": 0.3, "learning_rate": 0.00040678170612755936, "loss": 1.5974, "step": 2417 }, { "epoch": 0.3, "learning_rate": 0.0004067032626297518, "loss": 1.6753, "step": 2418 }, { "epoch": 0.3, "learning_rate": 0.0004066247937111816, "loss": 1.6161, "step": 2419 }, { "epoch": 0.3, "learning_rate": 0.0004065462993845784, "loss": 1.5718, "step": 2420 }, { "epoch": 0.3, "learning_rate": 0.00040646777966267566, "loss": 1.5814, "step": 2421 }, { "epoch": 0.3, "learning_rate": 0.00040638923455821107, "loss": 1.5673, "step": 2422 }, { "epoch": 0.3, "learning_rate": 0.00040631066408392636, "loss": 1.6631, "step": 2423 }, { "epoch": 0.3, "learning_rate": 0.0004062320682525673, "loss": 1.6682, "step": 2424 }, { "epoch": 0.3, "learning_rate": 0.00040615344707688406, "loss": 1.5653, "step": 2425 }, { "epoch": 0.3, "learning_rate": 0.00040607480056963064, "loss": 1.5873, "step": 2426 }, { "epoch": 0.3, "learning_rate": 0.0004059961287435653, "loss": 1.5605, "step": 2427 }, { "epoch": 0.3, "learning_rate": 0.00040591743161145035, "loss": 1.6178, "step": 2428 }, { "epoch": 0.3, "learning_rate": 0.0004058387091860522, "loss": 1.5019, "step": 2429 }, { "epoch": 0.3, "learning_rate": 0.0004057599614801415, "loss": 1.5831, "step": 2430 }, { "epoch": 0.3, "learning_rate": 0.00040568118850649266, "loss": 1.6025, "step": 2431 }, { "epoch": 0.3, "learning_rate": 0.0004056023902778846, "loss": 1.5523, "step": 2432 }, { "epoch": 0.3, "learning_rate": 0.00040552356680710014, "loss": 1.6063, "step": 2433 }, { "epoch": 0.3, "learning_rate": 0.00040544471810692606, "loss": 1.6136, "step": 2434 }, { "epoch": 0.3, "learning_rate": 0.0004053658441901535, "loss": 1.5486, "step": 2435 }, { "epoch": 0.3, "learning_rate": 0.0004052869450695776, "loss": 1.6541, "step": 2436 }, { "epoch": 0.3, "learning_rate": 0.0004052080207579974, "loss": 1.5319, "step": 2437 }, { "epoch": 0.3, "learning_rate": 0.00040512907126821643, "loss": 1.5741, "step": 2438 }, { "epoch": 0.3, "learning_rate": 0.0004050500966130418, "loss": 1.5699, "step": 2439 }, { "epoch": 0.3, "learning_rate": 0.0004049710968052851, "loss": 1.5402, "step": 2440 }, { "epoch": 0.31, "learning_rate": 0.0004048920718577618, "loss": 1.584, "step": 2441 }, { "epoch": 0.31, "learning_rate": 0.0004048130217832915, "loss": 1.6441, "step": 2442 }, { "epoch": 0.31, "learning_rate": 0.000404733946594698, "loss": 1.5863, "step": 2443 }, { "epoch": 0.31, "learning_rate": 0.00040465484630480884, "loss": 1.5653, "step": 2444 }, { "epoch": 0.31, "learning_rate": 0.0004045757209264562, "loss": 1.6206, "step": 2445 }, { "epoch": 0.31, "learning_rate": 0.0004044965704724756, "loss": 1.5854, "step": 2446 }, { "epoch": 0.31, "learning_rate": 0.0004044173949557072, "loss": 1.5314, "step": 2447 }, { "epoch": 0.31, "learning_rate": 0.000404338194388995, "loss": 1.5735, "step": 2448 }, { "epoch": 0.31, "learning_rate": 0.0004042589687851872, "loss": 1.6022, "step": 2449 }, { "epoch": 0.31, "learning_rate": 0.0004041797181571358, "loss": 1.5767, "step": 2450 }, { "epoch": 0.31, "learning_rate": 0.0004041004425176971, "loss": 1.592, "step": 2451 }, { "epoch": 0.31, "learning_rate": 0.00040402114187973134, "loss": 1.595, "step": 2452 }, { "epoch": 0.31, "learning_rate": 0.0004039418162561028, "loss": 1.6381, "step": 2453 }, { "epoch": 0.31, "learning_rate": 0.00040386246565968004, "loss": 1.6076, "step": 2454 }, { "epoch": 0.31, "learning_rate": 0.00040378309010333534, "loss": 1.6206, "step": 2455 }, { "epoch": 0.31, "learning_rate": 0.0004037036895999453, "loss": 1.5163, "step": 2456 }, { "epoch": 0.31, "learning_rate": 0.00040362426416239036, "loss": 1.5438, "step": 2457 }, { "epoch": 0.31, "learning_rate": 0.0004035448138035551, "loss": 1.6184, "step": 2458 }, { "epoch": 0.31, "learning_rate": 0.0004034653385363282, "loss": 1.6726, "step": 2459 }, { "epoch": 0.31, "learning_rate": 0.00040338583837360223, "loss": 1.5663, "step": 2460 }, { "epoch": 0.31, "learning_rate": 0.00040330631332827395, "loss": 1.5538, "step": 2461 }, { "epoch": 0.31, "learning_rate": 0.00040322676341324415, "loss": 1.6197, "step": 2462 }, { "epoch": 0.31, "learning_rate": 0.00040314718864141745, "loss": 1.65, "step": 2463 }, { "epoch": 0.31, "learning_rate": 0.00040306758902570284, "loss": 1.6107, "step": 2464 }, { "epoch": 0.31, "learning_rate": 0.000402987964579013, "loss": 1.5757, "step": 2465 }, { "epoch": 0.31, "learning_rate": 0.0004029083153142648, "loss": 1.4945, "step": 2466 }, { "epoch": 0.31, "learning_rate": 0.00040282864124437925, "loss": 1.5916, "step": 2467 }, { "epoch": 0.31, "learning_rate": 0.0004027489423822811, "loss": 1.6213, "step": 2468 }, { "epoch": 0.31, "learning_rate": 0.0004026692187408994, "loss": 1.5622, "step": 2469 }, { "epoch": 0.31, "learning_rate": 0.00040258947033316707, "loss": 1.5483, "step": 2470 }, { "epoch": 0.31, "learning_rate": 0.000402509697172021, "loss": 1.6144, "step": 2471 }, { "epoch": 0.31, "learning_rate": 0.00040242989927040233, "loss": 1.5733, "step": 2472 }, { "epoch": 0.31, "learning_rate": 0.00040235007664125595, "loss": 1.5921, "step": 2473 }, { "epoch": 0.31, "learning_rate": 0.00040227022929753087, "loss": 1.6119, "step": 2474 }, { "epoch": 0.31, "learning_rate": 0.0004021903572521802, "loss": 1.5174, "step": 2475 }, { "epoch": 0.31, "learning_rate": 0.0004021104605181608, "loss": 1.6221, "step": 2476 }, { "epoch": 0.31, "learning_rate": 0.0004020305391084339, "loss": 1.6048, "step": 2477 }, { "epoch": 0.31, "learning_rate": 0.0004019505930359644, "loss": 1.5807, "step": 2478 }, { "epoch": 0.31, "learning_rate": 0.00040187062231372136, "loss": 1.5899, "step": 2479 }, { "epoch": 0.31, "learning_rate": 0.0004017906269546778, "loss": 1.5965, "step": 2480 }, { "epoch": 0.31, "learning_rate": 0.0004017106069718108, "loss": 1.5789, "step": 2481 }, { "epoch": 0.31, "learning_rate": 0.0004016305623781014, "loss": 1.5477, "step": 2482 }, { "epoch": 0.31, "learning_rate": 0.0004015504931865345, "loss": 1.648, "step": 2483 }, { "epoch": 0.31, "learning_rate": 0.00040147039941009925, "loss": 1.5975, "step": 2484 }, { "epoch": 0.31, "learning_rate": 0.00040139028106178854, "loss": 1.5973, "step": 2485 }, { "epoch": 0.31, "learning_rate": 0.0004013101381545995, "loss": 1.5399, "step": 2486 }, { "epoch": 0.31, "learning_rate": 0.00040122997070153284, "loss": 1.5594, "step": 2487 }, { "epoch": 0.31, "learning_rate": 0.00040114977871559375, "loss": 1.5678, "step": 2488 }, { "epoch": 0.31, "learning_rate": 0.0004010695622097911, "loss": 1.5966, "step": 2489 }, { "epoch": 0.31, "learning_rate": 0.0004009893211971376, "loss": 1.5772, "step": 2490 }, { "epoch": 0.31, "learning_rate": 0.0004009090556906504, "loss": 1.702, "step": 2491 }, { "epoch": 0.31, "learning_rate": 0.00040082876570335027, "loss": 1.5641, "step": 2492 }, { "epoch": 0.31, "learning_rate": 0.000400748451248262, "loss": 1.5554, "step": 2493 }, { "epoch": 0.31, "learning_rate": 0.00040066811233841436, "loss": 1.6235, "step": 2494 }, { "epoch": 0.31, "learning_rate": 0.00040058774898684015, "loss": 1.5975, "step": 2495 }, { "epoch": 0.31, "learning_rate": 0.00040050736120657613, "loss": 1.5835, "step": 2496 }, { "epoch": 0.31, "learning_rate": 0.0004004269490106629, "loss": 1.6439, "step": 2497 }, { "epoch": 0.31, "learning_rate": 0.00040034651241214526, "loss": 1.6723, "step": 2498 }, { "epoch": 0.31, "learning_rate": 0.0004002660514240717, "loss": 1.6311, "step": 2499 }, { "epoch": 0.31, "learning_rate": 0.00040018556605949475, "loss": 1.5022, "step": 2500 }, { "epoch": 0.31, "eval_acceptance_rate": 0.6333799958229065, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.9596647610675915, "eval_expected_tokens": 2.4495809513344895, "eval_loss": 1.6241973638534546, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.36239523783362226, "eval_runtime": 50.3276, "eval_samples_per_second": 9.677, "eval_steps_per_second": 0.318, "step": 2500 }, { "epoch": 0.31, "learning_rate": 0.00040010505633147106, "loss": 1.6175, "step": 2501 }, { "epoch": 0.31, "learning_rate": 0.00040002452225306096, "loss": 1.613, "step": 2502 }, { "epoch": 0.31, "learning_rate": 0.000399943963837329, "loss": 1.6043, "step": 2503 }, { "epoch": 0.31, "learning_rate": 0.0003998633810973435, "loss": 1.6493, "step": 2504 }, { "epoch": 0.31, "learning_rate": 0.0003997827740461768, "loss": 1.5762, "step": 2505 }, { "epoch": 0.31, "learning_rate": 0.00039970214269690506, "loss": 1.6805, "step": 2506 }, { "epoch": 0.31, "learning_rate": 0.00039962148706260863, "loss": 1.6468, "step": 2507 }, { "epoch": 0.31, "learning_rate": 0.00039954080715637157, "loss": 1.655, "step": 2508 }, { "epoch": 0.31, "learning_rate": 0.0003994601029912819, "loss": 1.5229, "step": 2509 }, { "epoch": 0.31, "learning_rate": 0.00039937937458043183, "loss": 1.5378, "step": 2510 }, { "epoch": 0.31, "learning_rate": 0.00039929862193691714, "loss": 1.5881, "step": 2511 }, { "epoch": 0.31, "learning_rate": 0.0003992178450738377, "loss": 1.6899, "step": 2512 }, { "epoch": 0.31, "learning_rate": 0.0003991370440042974, "loss": 1.6229, "step": 2513 }, { "epoch": 0.31, "learning_rate": 0.00039905621874140396, "loss": 1.5308, "step": 2514 }, { "epoch": 0.31, "learning_rate": 0.000398975369298269, "loss": 1.5859, "step": 2515 }, { "epoch": 0.31, "learning_rate": 0.00039889449568800817, "loss": 1.6293, "step": 2516 }, { "epoch": 0.31, "learning_rate": 0.0003988135979237408, "loss": 1.5499, "step": 2517 }, { "epoch": 0.31, "learning_rate": 0.00039873267601859066, "loss": 1.578, "step": 2518 }, { "epoch": 0.31, "learning_rate": 0.00039865172998568467, "loss": 1.523, "step": 2519 }, { "epoch": 0.32, "learning_rate": 0.00039857075983815437, "loss": 1.5835, "step": 2520 }, { "epoch": 0.32, "learning_rate": 0.00039848976558913487, "loss": 1.5585, "step": 2521 }, { "epoch": 0.32, "learning_rate": 0.0003984087472517651, "loss": 1.6102, "step": 2522 }, { "epoch": 0.32, "learning_rate": 0.0003983277048391882, "loss": 1.5837, "step": 2523 }, { "epoch": 0.32, "learning_rate": 0.00039824663836455094, "loss": 1.6019, "step": 2524 }, { "epoch": 0.32, "learning_rate": 0.00039816554784100424, "loss": 1.6343, "step": 2525 }, { "epoch": 0.32, "learning_rate": 0.00039808443328170266, "loss": 1.6666, "step": 2526 }, { "epoch": 0.32, "learning_rate": 0.0003980032946998049, "loss": 1.5897, "step": 2527 }, { "epoch": 0.32, "learning_rate": 0.0003979221321084734, "loss": 1.5845, "step": 2528 }, { "epoch": 0.32, "learning_rate": 0.0003978409455208745, "loss": 1.6338, "step": 2529 }, { "epoch": 0.32, "learning_rate": 0.00039775973495017854, "loss": 1.5878, "step": 2530 }, { "epoch": 0.32, "learning_rate": 0.00039767850040955954, "loss": 1.6087, "step": 2531 }, { "epoch": 0.32, "learning_rate": 0.0003975972419121958, "loss": 1.6043, "step": 2532 }, { "epoch": 0.32, "learning_rate": 0.0003975159594712692, "loss": 1.5275, "step": 2533 }, { "epoch": 0.32, "learning_rate": 0.0003974346530999653, "loss": 1.57, "step": 2534 }, { "epoch": 0.32, "learning_rate": 0.0003973533228114742, "loss": 1.6053, "step": 2535 }, { "epoch": 0.32, "learning_rate": 0.0003972719686189891, "loss": 1.571, "step": 2536 }, { "epoch": 0.32, "learning_rate": 0.0003971905905357077, "loss": 1.5452, "step": 2537 }, { "epoch": 0.32, "learning_rate": 0.00039710918857483135, "loss": 1.7221, "step": 2538 }, { "epoch": 0.32, "learning_rate": 0.0003970277627495652, "loss": 1.5381, "step": 2539 }, { "epoch": 0.32, "learning_rate": 0.0003969463130731183, "loss": 1.5802, "step": 2540 }, { "epoch": 0.32, "learning_rate": 0.0003968648395587036, "loss": 1.5686, "step": 2541 }, { "epoch": 0.32, "learning_rate": 0.000396783342219538, "loss": 1.562, "step": 2542 }, { "epoch": 0.32, "learning_rate": 0.0003967018210688422, "loss": 1.6361, "step": 2543 }, { "epoch": 0.32, "learning_rate": 0.0003966202761198406, "loss": 1.5518, "step": 2544 }, { "epoch": 0.32, "learning_rate": 0.0003965387073857618, "loss": 1.5426, "step": 2545 }, { "epoch": 0.32, "learning_rate": 0.0003964571148798378, "loss": 1.533, "step": 2546 }, { "epoch": 0.32, "learning_rate": 0.0003963754986153051, "loss": 1.5979, "step": 2547 }, { "epoch": 0.32, "learning_rate": 0.00039629385860540334, "loss": 1.5975, "step": 2548 }, { "epoch": 0.32, "learning_rate": 0.0003962121948633765, "loss": 1.5953, "step": 2549 }, { "epoch": 0.32, "learning_rate": 0.00039613050740247227, "loss": 1.5475, "step": 2550 }, { "epoch": 0.32, "learning_rate": 0.000396048796235942, "loss": 1.5192, "step": 2551 }, { "epoch": 0.32, "learning_rate": 0.0003959670613770414, "loss": 1.5809, "step": 2552 }, { "epoch": 0.32, "learning_rate": 0.00039588530283902936, "loss": 1.6647, "step": 2553 }, { "epoch": 0.32, "learning_rate": 0.0003958035206351691, "loss": 1.5771, "step": 2554 }, { "epoch": 0.32, "learning_rate": 0.0003957217147787276, "loss": 1.7383, "step": 2555 }, { "epoch": 0.32, "learning_rate": 0.00039563988528297533, "loss": 1.5665, "step": 2556 }, { "epoch": 0.32, "learning_rate": 0.00039555803216118713, "loss": 1.6753, "step": 2557 }, { "epoch": 0.32, "learning_rate": 0.0003954761554266412, "loss": 1.9139, "step": 2558 }, { "epoch": 0.32, "learning_rate": 0.00039539425509261985, "loss": 1.7007, "step": 2559 }, { "epoch": 0.32, "learning_rate": 0.00039531233117240914, "loss": 1.6421, "step": 2560 }, { "epoch": 0.32, "learning_rate": 0.000395230383679299, "loss": 1.6468, "step": 2561 }, { "epoch": 0.32, "learning_rate": 0.0003951484126265831, "loss": 1.7528, "step": 2562 }, { "epoch": 0.32, "learning_rate": 0.0003950664180275589, "loss": 1.886, "step": 2563 }, { "epoch": 0.32, "learning_rate": 0.0003949843998955279, "loss": 1.5726, "step": 2564 }, { "epoch": 0.32, "learning_rate": 0.0003949023582437952, "loss": 1.6308, "step": 2565 }, { "epoch": 0.32, "learning_rate": 0.0003948202930856697, "loss": 1.6967, "step": 2566 }, { "epoch": 0.32, "learning_rate": 0.00039473820443446437, "loss": 1.6768, "step": 2567 }, { "epoch": 0.32, "learning_rate": 0.00039465609230349564, "loss": 1.5496, "step": 2568 }, { "epoch": 0.32, "learning_rate": 0.00039457395670608404, "loss": 1.6734, "step": 2569 }, { "epoch": 0.32, "learning_rate": 0.00039449179765555376, "loss": 1.5644, "step": 2570 }, { "epoch": 0.32, "learning_rate": 0.0003944096151652329, "loss": 1.5761, "step": 2571 }, { "epoch": 0.32, "learning_rate": 0.00039432740924845314, "loss": 1.4992, "step": 2572 }, { "epoch": 0.32, "learning_rate": 0.0003942451799185502, "loss": 1.5556, "step": 2573 }, { "epoch": 0.32, "learning_rate": 0.00039416292718886354, "loss": 1.5962, "step": 2574 }, { "epoch": 0.32, "learning_rate": 0.0003940806510727364, "loss": 1.5867, "step": 2575 }, { "epoch": 0.32, "learning_rate": 0.00039399835158351567, "loss": 1.5556, "step": 2576 }, { "epoch": 0.32, "learning_rate": 0.00039391602873455235, "loss": 1.6418, "step": 2577 }, { "epoch": 0.32, "learning_rate": 0.00039383368253920084, "loss": 1.5636, "step": 2578 }, { "epoch": 0.32, "learning_rate": 0.0003937513130108197, "loss": 1.5955, "step": 2579 }, { "epoch": 0.32, "learning_rate": 0.00039366892016277097, "loss": 1.5864, "step": 2580 }, { "epoch": 0.32, "learning_rate": 0.0003935865040084207, "loss": 1.6082, "step": 2581 }, { "epoch": 0.32, "learning_rate": 0.00039350406456113864, "loss": 1.6525, "step": 2582 }, { "epoch": 0.32, "learning_rate": 0.0003934216018342982, "loss": 1.5897, "step": 2583 }, { "epoch": 0.32, "learning_rate": 0.00039333911584127677, "loss": 1.6249, "step": 2584 }, { "epoch": 0.32, "learning_rate": 0.0003932566065954553, "loss": 1.6498, "step": 2585 }, { "epoch": 0.32, "learning_rate": 0.0003931740741102188, "loss": 1.6301, "step": 2586 }, { "epoch": 0.32, "learning_rate": 0.0003930915183989558, "loss": 1.6231, "step": 2587 }, { "epoch": 0.32, "learning_rate": 0.0003930089394750586, "loss": 1.6172, "step": 2588 }, { "epoch": 0.32, "learning_rate": 0.0003929263373519235, "loss": 1.6573, "step": 2589 }, { "epoch": 0.32, "learning_rate": 0.00039284371204295024, "loss": 1.6387, "step": 2590 }, { "epoch": 0.32, "learning_rate": 0.00039276106356154263, "loss": 1.5817, "step": 2591 }, { "epoch": 0.32, "learning_rate": 0.000392678391921108, "loss": 1.6298, "step": 2592 }, { "epoch": 0.32, "learning_rate": 0.0003925956971350575, "loss": 1.5109, "step": 2593 }, { "epoch": 0.32, "learning_rate": 0.0003925129792168062, "loss": 1.5319, "step": 2594 }, { "epoch": 0.32, "learning_rate": 0.00039243023817977265, "loss": 1.6055, "step": 2595 }, { "epoch": 0.32, "learning_rate": 0.0003923474740373794, "loss": 1.5865, "step": 2596 }, { "epoch": 0.32, "learning_rate": 0.0003922646868030526, "loss": 1.609, "step": 2597 }, { "epoch": 0.32, "learning_rate": 0.00039218187649022216, "loss": 1.5284, "step": 2598 }, { "epoch": 0.32, "learning_rate": 0.00039209904311232185, "loss": 1.549, "step": 2599 }, { "epoch": 0.33, "learning_rate": 0.00039201618668278893, "loss": 1.6325, "step": 2600 }, { "epoch": 0.33, "eval_acceptance_rate": 0.6340993642807007, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 1.962248999050358, "eval_expected_tokens": 2.4528112488129477, "eval_loss": 1.5888440608978271, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.36320281220323686, "eval_runtime": 50.3539, "eval_samples_per_second": 9.672, "eval_steps_per_second": 0.318, "step": 2600 }, { "epoch": 0.33, "learning_rate": 0.0003919333072150647, "loss": 1.5493, "step": 2601 }, { "epoch": 0.33, "learning_rate": 0.00039185040472259405, "loss": 1.5493, "step": 2602 }, { "epoch": 0.33, "learning_rate": 0.0003917674792188255, "loss": 1.496, "step": 2603 }, { "epoch": 0.33, "learning_rate": 0.00039168453071721155, "loss": 1.6347, "step": 2604 }, { "epoch": 0.33, "learning_rate": 0.0003916015592312082, "loss": 1.5869, "step": 2605 }, { "epoch": 0.33, "learning_rate": 0.00039151856477427535, "loss": 1.6161, "step": 2606 }, { "epoch": 0.33, "learning_rate": 0.0003914355473598765, "loss": 1.6142, "step": 2607 }, { "epoch": 0.33, "learning_rate": 0.00039135250700147886, "loss": 1.6539, "step": 2608 }, { "epoch": 0.33, "learning_rate": 0.0003912694437125536, "loss": 1.6692, "step": 2609 }, { "epoch": 0.33, "learning_rate": 0.00039118635750657525, "loss": 1.6316, "step": 2610 }, { "epoch": 0.33, "learning_rate": 0.00039110324839702236, "loss": 1.5522, "step": 2611 }, { "epoch": 0.33, "learning_rate": 0.0003910201163973771, "loss": 1.5264, "step": 2612 }, { "epoch": 0.33, "learning_rate": 0.00039093696152112524, "loss": 1.5605, "step": 2613 }, { "epoch": 0.33, "learning_rate": 0.00039085378378175644, "loss": 1.5496, "step": 2614 }, { "epoch": 0.33, "learning_rate": 0.0003907705831927639, "loss": 1.525, "step": 2615 }, { "epoch": 0.33, "learning_rate": 0.00039068735976764473, "loss": 1.5405, "step": 2616 }, { "epoch": 0.33, "learning_rate": 0.00039060411351989944, "loss": 1.5338, "step": 2617 }, { "epoch": 0.33, "learning_rate": 0.00039052084446303264, "loss": 1.5385, "step": 2618 }, { "epoch": 0.33, "learning_rate": 0.00039043755261055236, "loss": 1.4733, "step": 2619 }, { "epoch": 0.33, "learning_rate": 0.00039035423797597024, "loss": 1.5793, "step": 2620 }, { "epoch": 0.33, "learning_rate": 0.00039027090057280195, "loss": 1.6237, "step": 2621 }, { "epoch": 0.33, "learning_rate": 0.0003901875404145666, "loss": 1.5852, "step": 2622 }, { "epoch": 0.33, "learning_rate": 0.00039010415751478713, "loss": 1.567, "step": 2623 }, { "epoch": 0.33, "learning_rate": 0.00039002075188699004, "loss": 1.6295, "step": 2624 }, { "epoch": 0.33, "learning_rate": 0.00038993732354470554, "loss": 1.5905, "step": 2625 }, { "epoch": 0.33, "learning_rate": 0.0003898538725014676, "loss": 1.5325, "step": 2626 }, { "epoch": 0.33, "learning_rate": 0.000389770398770814, "loss": 1.467, "step": 2627 }, { "epoch": 0.33, "learning_rate": 0.00038968690236628584, "loss": 1.5284, "step": 2628 }, { "epoch": 0.33, "learning_rate": 0.00038960338330142817, "loss": 1.5908, "step": 2629 }, { "epoch": 0.33, "learning_rate": 0.00038951984158978957, "loss": 1.5236, "step": 2630 }, { "epoch": 0.33, "learning_rate": 0.0003894362772449226, "loss": 1.5108, "step": 2631 }, { "epoch": 0.33, "learning_rate": 0.00038935269028038297, "loss": 1.531, "step": 2632 }, { "epoch": 0.33, "learning_rate": 0.0003892690807097305, "loss": 1.5305, "step": 2633 }, { "epoch": 0.33, "learning_rate": 0.0003891854485465285, "loss": 1.6002, "step": 2634 }, { "epoch": 0.33, "learning_rate": 0.00038910179380434397, "loss": 1.5064, "step": 2635 }, { "epoch": 0.33, "learning_rate": 0.00038901811649674756, "loss": 1.5549, "step": 2636 }, { "epoch": 0.33, "learning_rate": 0.00038893441663731364, "loss": 1.5541, "step": 2637 }, { "epoch": 0.33, "learning_rate": 0.00038885069423962016, "loss": 1.5581, "step": 2638 }, { "epoch": 0.33, "learning_rate": 0.00038876694931724875, "loss": 1.5995, "step": 2639 }, { "epoch": 0.33, "learning_rate": 0.0003886831818837847, "loss": 1.5495, "step": 2640 }, { "epoch": 0.33, "learning_rate": 0.00038859939195281704, "loss": 1.4942, "step": 2641 }, { "epoch": 0.33, "learning_rate": 0.00038851557953793823, "loss": 1.5412, "step": 2642 }, { "epoch": 0.33, "learning_rate": 0.0003884317446527446, "loss": 1.606, "step": 2643 }, { "epoch": 0.33, "learning_rate": 0.000388347887310836, "loss": 1.566, "step": 2644 }, { "epoch": 0.33, "learning_rate": 0.00038826400752581605, "loss": 1.5706, "step": 2645 }, { "epoch": 0.33, "learning_rate": 0.00038818010531129176, "loss": 1.645, "step": 2646 }, { "epoch": 0.33, "learning_rate": 0.00038809618068087406, "loss": 1.6149, "step": 2647 }, { "epoch": 0.33, "learning_rate": 0.0003880122336481774, "loss": 1.638, "step": 2648 }, { "epoch": 0.33, "learning_rate": 0.00038792826422681973, "loss": 1.6203, "step": 2649 }, { "epoch": 0.33, "learning_rate": 0.000387844272430423, "loss": 1.5951, "step": 2650 }, { "epoch": 0.33, "learning_rate": 0.00038776025827261224, "loss": 1.5742, "step": 2651 }, { "epoch": 0.33, "learning_rate": 0.0003876762217670167, "loss": 1.6361, "step": 2652 }, { "epoch": 0.33, "learning_rate": 0.00038759216292726887, "loss": 1.5263, "step": 2653 }, { "epoch": 0.33, "learning_rate": 0.0003875080817670048, "loss": 1.6173, "step": 2654 }, { "epoch": 0.33, "learning_rate": 0.00038742397829986464, "loss": 1.5058, "step": 2655 }, { "epoch": 0.33, "learning_rate": 0.0003873398525394917, "loss": 1.5485, "step": 2656 }, { "epoch": 0.33, "learning_rate": 0.00038725570449953296, "loss": 1.5769, "step": 2657 }, { "epoch": 0.33, "learning_rate": 0.00038717153419363926, "loss": 1.5667, "step": 2658 }, { "epoch": 0.33, "learning_rate": 0.00038708734163546475, "loss": 1.5203, "step": 2659 }, { "epoch": 0.33, "learning_rate": 0.0003870031268386676, "loss": 1.6019, "step": 2660 }, { "epoch": 0.33, "learning_rate": 0.00038691888981690894, "loss": 1.5684, "step": 2661 }, { "epoch": 0.33, "learning_rate": 0.0003868346305838542, "loss": 1.5629, "step": 2662 }, { "epoch": 0.33, "learning_rate": 0.0003867503491531721, "loss": 1.5892, "step": 2663 }, { "epoch": 0.33, "learning_rate": 0.00038666604553853467, "loss": 1.611, "step": 2664 }, { "epoch": 0.33, "learning_rate": 0.00038658171975361824, "loss": 1.5848, "step": 2665 }, { "epoch": 0.33, "learning_rate": 0.000386497371812102, "loss": 1.526, "step": 2666 }, { "epoch": 0.33, "learning_rate": 0.0003864130017276693, "loss": 1.5245, "step": 2667 }, { "epoch": 0.33, "learning_rate": 0.0003863286095140067, "loss": 1.5655, "step": 2668 }, { "epoch": 0.33, "learning_rate": 0.00038624419518480457, "loss": 1.5222, "step": 2669 }, { "epoch": 0.33, "learning_rate": 0.00038615975875375683, "loss": 1.5827, "step": 2670 }, { "epoch": 0.33, "learning_rate": 0.0003860753002345608, "loss": 1.5388, "step": 2671 }, { "epoch": 0.33, "learning_rate": 0.00038599081964091765, "loss": 1.5446, "step": 2672 }, { "epoch": 0.33, "learning_rate": 0.00038590631698653206, "loss": 1.5799, "step": 2673 }, { "epoch": 0.33, "learning_rate": 0.0003858217922851121, "loss": 1.6389, "step": 2674 }, { "epoch": 0.33, "learning_rate": 0.00038573724555036973, "loss": 1.5342, "step": 2675 }, { "epoch": 0.33, "learning_rate": 0.0003856526767960201, "loss": 1.5233, "step": 2676 }, { "epoch": 0.33, "learning_rate": 0.0003855680860357824, "loss": 1.5882, "step": 2677 }, { "epoch": 0.33, "learning_rate": 0.0003854834732833789, "loss": 1.5924, "step": 2678 }, { "epoch": 0.33, "learning_rate": 0.0003853988385525359, "loss": 1.524, "step": 2679 }, { "epoch": 0.34, "learning_rate": 0.00038531418185698286, "loss": 1.4649, "step": 2680 }, { "epoch": 0.34, "learning_rate": 0.000385229503210453, "loss": 1.5278, "step": 2681 }, { "epoch": 0.34, "learning_rate": 0.0003851448026266833, "loss": 1.6019, "step": 2682 }, { "epoch": 0.34, "learning_rate": 0.00038506008011941376, "loss": 1.5023, "step": 2683 }, { "epoch": 0.34, "learning_rate": 0.0003849753357023885, "loss": 1.6445, "step": 2684 }, { "epoch": 0.34, "learning_rate": 0.0003848905693893549, "loss": 1.6102, "step": 2685 }, { "epoch": 0.34, "learning_rate": 0.0003848057811940638, "loss": 1.5054, "step": 2686 }, { "epoch": 0.34, "learning_rate": 0.00038472097113026986, "loss": 1.474, "step": 2687 }, { "epoch": 0.34, "learning_rate": 0.0003846361392117312, "loss": 1.5573, "step": 2688 }, { "epoch": 0.34, "learning_rate": 0.0003845512854522094, "loss": 1.4621, "step": 2689 }, { "epoch": 0.34, "learning_rate": 0.00038446640986546966, "loss": 1.5071, "step": 2690 }, { "epoch": 0.34, "learning_rate": 0.00038438151246528065, "loss": 1.5106, "step": 2691 }, { "epoch": 0.34, "learning_rate": 0.0003842965932654146, "loss": 1.5596, "step": 2692 }, { "epoch": 0.34, "learning_rate": 0.00038421165227964726, "loss": 1.5497, "step": 2693 }, { "epoch": 0.34, "learning_rate": 0.0003841266895217582, "loss": 1.537, "step": 2694 }, { "epoch": 0.34, "learning_rate": 0.00038404170500553003, "loss": 1.5348, "step": 2695 }, { "epoch": 0.34, "learning_rate": 0.00038395669874474915, "loss": 1.5387, "step": 2696 }, { "epoch": 0.34, "learning_rate": 0.0003838716707532056, "loss": 1.6179, "step": 2697 }, { "epoch": 0.34, "learning_rate": 0.0003837866210446926, "loss": 1.622, "step": 2698 }, { "epoch": 0.34, "learning_rate": 0.0003837015496330074, "loss": 1.5634, "step": 2699 }, { "epoch": 0.34, "learning_rate": 0.00038361645653195025, "loss": 1.5716, "step": 2700 }, { "epoch": 0.34, "eval_acceptance_rate": 0.6505028605461121, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0223831878007745, "eval_expected_tokens": 2.5279789847509684, "eval_loss": 1.549546718597412, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.38199474618774204, "eval_runtime": 50.4193, "eval_samples_per_second": 9.659, "eval_steps_per_second": 0.317, "step": 2700 }, { "epoch": 0.34, "learning_rate": 0.00038353134175532526, "loss": 1.5192, "step": 2701 }, { "epoch": 0.34, "learning_rate": 0.0003834462053169399, "loss": 1.5255, "step": 2702 }, { "epoch": 0.34, "learning_rate": 0.00038336104723060515, "loss": 1.4745, "step": 2703 }, { "epoch": 0.34, "learning_rate": 0.00038327586751013575, "loss": 1.5181, "step": 2704 }, { "epoch": 0.34, "learning_rate": 0.0003831906661693494, "loss": 1.5153, "step": 2705 }, { "epoch": 0.34, "learning_rate": 0.00038310544322206805, "loss": 1.4602, "step": 2706 }, { "epoch": 0.34, "learning_rate": 0.00038302019868211647, "loss": 1.6344, "step": 2707 }, { "epoch": 0.34, "learning_rate": 0.00038293493256332336, "loss": 1.5218, "step": 2708 }, { "epoch": 0.34, "learning_rate": 0.0003828496448795207, "loss": 1.5357, "step": 2709 }, { "epoch": 0.34, "learning_rate": 0.0003827643356445442, "loss": 1.4725, "step": 2710 }, { "epoch": 0.34, "learning_rate": 0.00038267900487223274, "loss": 1.5116, "step": 2711 }, { "epoch": 0.34, "learning_rate": 0.000382593652576429, "loss": 1.6234, "step": 2712 }, { "epoch": 0.34, "learning_rate": 0.00038250827877097906, "loss": 1.5167, "step": 2713 }, { "epoch": 0.34, "learning_rate": 0.00038242288346973226, "loss": 1.5548, "step": 2714 }, { "epoch": 0.34, "learning_rate": 0.0003823374666865418, "loss": 1.457, "step": 2715 }, { "epoch": 0.34, "learning_rate": 0.0003822520284352641, "loss": 1.5372, "step": 2716 }, { "epoch": 0.34, "learning_rate": 0.00038216656872975917, "loss": 1.5529, "step": 2717 }, { "epoch": 0.34, "learning_rate": 0.0003820810875838905, "loss": 1.5203, "step": 2718 }, { "epoch": 0.34, "learning_rate": 0.0003819955850115251, "loss": 1.5417, "step": 2719 }, { "epoch": 0.34, "learning_rate": 0.00038191006102653317, "loss": 1.5433, "step": 2720 }, { "epoch": 0.34, "learning_rate": 0.00038182451564278884, "loss": 1.5736, "step": 2721 }, { "epoch": 0.34, "learning_rate": 0.00038173894887416946, "loss": 1.5077, "step": 2722 }, { "epoch": 0.34, "learning_rate": 0.0003816533607345557, "loss": 1.4828, "step": 2723 }, { "epoch": 0.34, "learning_rate": 0.00038156775123783195, "loss": 1.4882, "step": 2724 }, { "epoch": 0.34, "learning_rate": 0.000381482120397886, "loss": 1.4827, "step": 2725 }, { "epoch": 0.34, "learning_rate": 0.0003813964682286092, "loss": 1.579, "step": 2726 }, { "epoch": 0.34, "learning_rate": 0.0003813107947438959, "loss": 1.5646, "step": 2727 }, { "epoch": 0.34, "learning_rate": 0.00038122509995764466, "loss": 1.4532, "step": 2728 }, { "epoch": 0.34, "learning_rate": 0.00038113938388375677, "loss": 1.56, "step": 2729 }, { "epoch": 0.34, "learning_rate": 0.00038105364653613747, "loss": 1.5336, "step": 2730 }, { "epoch": 0.34, "learning_rate": 0.0003809678879286952, "loss": 1.4788, "step": 2731 }, { "epoch": 0.34, "learning_rate": 0.00038088210807534184, "loss": 1.616, "step": 2732 }, { "epoch": 0.34, "learning_rate": 0.000380796306989993, "loss": 1.5281, "step": 2733 }, { "epoch": 0.34, "learning_rate": 0.00038071048468656734, "loss": 1.5822, "step": 2734 }, { "epoch": 0.34, "learning_rate": 0.0003806246411789872, "loss": 1.5462, "step": 2735 }, { "epoch": 0.34, "learning_rate": 0.0003805387764811784, "loss": 1.577, "step": 2736 }, { "epoch": 0.34, "learning_rate": 0.00038045289060706995, "loss": 1.5153, "step": 2737 }, { "epoch": 0.34, "learning_rate": 0.00038036698357059464, "loss": 1.5392, "step": 2738 }, { "epoch": 0.34, "learning_rate": 0.00038028105538568833, "loss": 1.5447, "step": 2739 }, { "epoch": 0.34, "learning_rate": 0.00038019510606629057, "loss": 1.5466, "step": 2740 }, { "epoch": 0.34, "learning_rate": 0.0003801091356263443, "loss": 1.5527, "step": 2741 }, { "epoch": 0.34, "learning_rate": 0.0003800231440797958, "loss": 1.4778, "step": 2742 }, { "epoch": 0.34, "learning_rate": 0.0003799371314405947, "loss": 1.5992, "step": 2743 }, { "epoch": 0.34, "learning_rate": 0.0003798510977226943, "loss": 1.5786, "step": 2744 }, { "epoch": 0.34, "learning_rate": 0.0003797650429400512, "loss": 1.482, "step": 2745 }, { "epoch": 0.34, "learning_rate": 0.00037967896710662535, "loss": 1.4984, "step": 2746 }, { "epoch": 0.34, "learning_rate": 0.00037959287023638024, "loss": 1.4748, "step": 2747 }, { "epoch": 0.34, "learning_rate": 0.00037950675234328256, "loss": 1.5326, "step": 2748 }, { "epoch": 0.34, "learning_rate": 0.00037942061344130263, "loss": 1.5283, "step": 2749 }, { "epoch": 0.34, "learning_rate": 0.0003793344535444142, "loss": 1.4895, "step": 2750 }, { "epoch": 0.34, "learning_rate": 0.0003792482726665942, "loss": 1.6076, "step": 2751 }, { "epoch": 0.34, "learning_rate": 0.00037916207082182307, "loss": 1.4505, "step": 2752 }, { "epoch": 0.34, "learning_rate": 0.0003790758480240848, "loss": 1.4863, "step": 2753 }, { "epoch": 0.34, "learning_rate": 0.0003789896042873665, "loss": 1.5273, "step": 2754 }, { "epoch": 0.34, "learning_rate": 0.0003789033396256589, "loss": 1.6206, "step": 2755 }, { "epoch": 0.34, "learning_rate": 0.00037881705405295616, "loss": 1.5901, "step": 2756 }, { "epoch": 0.34, "learning_rate": 0.00037873074758325553, "loss": 1.5477, "step": 2757 }, { "epoch": 0.34, "learning_rate": 0.000378644420230558, "loss": 1.5051, "step": 2758 }, { "epoch": 0.34, "learning_rate": 0.0003785580720088676, "loss": 1.5434, "step": 2759 }, { "epoch": 0.34, "learning_rate": 0.0003784717029321922, "loss": 1.5766, "step": 2760 }, { "epoch": 0.35, "learning_rate": 0.0003783853130145425, "loss": 1.5837, "step": 2761 }, { "epoch": 0.35, "learning_rate": 0.00037829890226993314, "loss": 1.4923, "step": 2762 }, { "epoch": 0.35, "learning_rate": 0.0003782124707123817, "loss": 1.4846, "step": 2763 }, { "epoch": 0.35, "learning_rate": 0.0003781260183559093, "loss": 1.5495, "step": 2764 }, { "epoch": 0.35, "learning_rate": 0.00037803954521454055, "loss": 1.5847, "step": 2765 }, { "epoch": 0.35, "learning_rate": 0.00037795305130230327, "loss": 1.5571, "step": 2766 }, { "epoch": 0.35, "learning_rate": 0.00037786653663322866, "loss": 1.6485, "step": 2767 }, { "epoch": 0.35, "learning_rate": 0.00037778000122135134, "loss": 1.4649, "step": 2768 }, { "epoch": 0.35, "learning_rate": 0.00037769344508070934, "loss": 1.5065, "step": 2769 }, { "epoch": 0.35, "learning_rate": 0.00037760686822534394, "loss": 1.5971, "step": 2770 }, { "epoch": 0.35, "learning_rate": 0.0003775202706692999, "loss": 1.568, "step": 2771 }, { "epoch": 0.35, "learning_rate": 0.0003774336524266252, "loss": 1.5655, "step": 2772 }, { "epoch": 0.35, "learning_rate": 0.00037734701351137127, "loss": 1.5908, "step": 2773 }, { "epoch": 0.35, "learning_rate": 0.00037726035393759286, "loss": 1.549, "step": 2774 }, { "epoch": 0.35, "learning_rate": 0.00037717367371934807, "loss": 1.5522, "step": 2775 }, { "epoch": 0.35, "learning_rate": 0.0003770869728706985, "loss": 1.5527, "step": 2776 }, { "epoch": 0.35, "learning_rate": 0.0003770002514057088, "loss": 1.5858, "step": 2777 }, { "epoch": 0.35, "learning_rate": 0.0003769135093384473, "loss": 1.5849, "step": 2778 }, { "epoch": 0.35, "learning_rate": 0.0003768267466829852, "loss": 1.5622, "step": 2779 }, { "epoch": 0.35, "learning_rate": 0.00037673996345339764, "loss": 1.5288, "step": 2780 }, { "epoch": 0.35, "learning_rate": 0.0003766531596637626, "loss": 1.5524, "step": 2781 }, { "epoch": 0.35, "learning_rate": 0.0003765663353281618, "loss": 1.5578, "step": 2782 }, { "epoch": 0.35, "learning_rate": 0.00037647949046067976, "loss": 1.4847, "step": 2783 }, { "epoch": 0.35, "learning_rate": 0.0003763926250754049, "loss": 1.5276, "step": 2784 }, { "epoch": 0.35, "learning_rate": 0.00037630573918642876, "loss": 1.5905, "step": 2785 }, { "epoch": 0.35, "learning_rate": 0.00037621883280784596, "loss": 1.5728, "step": 2786 }, { "epoch": 0.35, "learning_rate": 0.0003761319059537548, "loss": 1.5161, "step": 2787 }, { "epoch": 0.35, "learning_rate": 0.0003760449586382568, "loss": 1.4862, "step": 2788 }, { "epoch": 0.35, "learning_rate": 0.00037595799087545656, "loss": 1.52, "step": 2789 }, { "epoch": 0.35, "learning_rate": 0.0003758710026794623, "loss": 1.5325, "step": 2790 }, { "epoch": 0.35, "learning_rate": 0.00037578399406438547, "loss": 1.5377, "step": 2791 }, { "epoch": 0.35, "learning_rate": 0.0003756969650443408, "loss": 1.5969, "step": 2792 }, { "epoch": 0.35, "learning_rate": 0.0003756099156334463, "loss": 1.5447, "step": 2793 }, { "epoch": 0.35, "learning_rate": 0.00037552284584582333, "loss": 1.567, "step": 2794 }, { "epoch": 0.35, "learning_rate": 0.0003754357556955966, "loss": 1.5538, "step": 2795 }, { "epoch": 0.35, "learning_rate": 0.0003753486451968939, "loss": 1.5394, "step": 2796 }, { "epoch": 0.35, "learning_rate": 0.0003752615143638468, "loss": 1.493, "step": 2797 }, { "epoch": 0.35, "learning_rate": 0.0003751743632105896, "loss": 1.5163, "step": 2798 }, { "epoch": 0.35, "learning_rate": 0.0003750871917512603, "loss": 1.533, "step": 2799 }, { "epoch": 0.35, "learning_rate": 0.000375, "loss": 1.5362, "step": 2800 }, { "epoch": 0.35, "eval_acceptance_rate": 0.652835488319397, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0311246560003218, "eval_expected_tokens": 2.538905820000402, "eval_loss": 1.5316981077194214, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3847264550001006, "eval_runtime": 50.3451, "eval_samples_per_second": 9.673, "eval_steps_per_second": 0.318, "step": 2800 }, { "epoch": 0.35, "learning_rate": 0.00037491278797095317, "loss": 1.4889, "step": 2801 }, { "epoch": 0.35, "learning_rate": 0.00037482555567826747, "loss": 1.5093, "step": 2802 }, { "epoch": 0.35, "learning_rate": 0.00037473830313609403, "loss": 1.4711, "step": 2803 }, { "epoch": 0.35, "learning_rate": 0.00037465103035858716, "loss": 1.5297, "step": 2804 }, { "epoch": 0.35, "learning_rate": 0.00037456373735990427, "loss": 1.5646, "step": 2805 }, { "epoch": 0.35, "learning_rate": 0.00037447642415420644, "loss": 1.5836, "step": 2806 }, { "epoch": 0.35, "learning_rate": 0.00037438909075565775, "loss": 1.5902, "step": 2807 }, { "epoch": 0.35, "learning_rate": 0.00037430173717842557, "loss": 1.5598, "step": 2808 }, { "epoch": 0.35, "learning_rate": 0.0003742143634366807, "loss": 1.563, "step": 2809 }, { "epoch": 0.35, "learning_rate": 0.000374126969544597, "loss": 1.619, "step": 2810 }, { "epoch": 0.35, "learning_rate": 0.0003740395555163517, "loss": 1.5056, "step": 2811 }, { "epoch": 0.35, "learning_rate": 0.0003739521213661254, "loss": 1.5273, "step": 2812 }, { "epoch": 0.35, "learning_rate": 0.0003738646671081019, "loss": 1.538, "step": 2813 }, { "epoch": 0.35, "learning_rate": 0.00037377719275646817, "loss": 1.4952, "step": 2814 }, { "epoch": 0.35, "learning_rate": 0.0003736896983254144, "loss": 1.8387, "step": 2815 }, { "epoch": 0.35, "learning_rate": 0.00037360218382913427, "loss": 1.5139, "step": 2816 }, { "epoch": 0.35, "learning_rate": 0.0003735146492818245, "loss": 1.6709, "step": 2817 }, { "epoch": 0.35, "learning_rate": 0.00037342709469768515, "loss": 1.626, "step": 2818 }, { "epoch": 0.35, "learning_rate": 0.00037333952009091964, "loss": 1.6815, "step": 2819 }, { "epoch": 0.35, "learning_rate": 0.0003732519254757344, "loss": 1.5276, "step": 2820 }, { "epoch": 0.35, "learning_rate": 0.0003731643108663393, "loss": 1.6923, "step": 2821 }, { "epoch": 0.35, "learning_rate": 0.00037307667627694733, "loss": 1.6159, "step": 2822 }, { "epoch": 0.35, "learning_rate": 0.00037298902172177477, "loss": 1.9019, "step": 2823 }, { "epoch": 0.35, "learning_rate": 0.0003729013472150412, "loss": 1.5523, "step": 2824 }, { "epoch": 0.35, "learning_rate": 0.00037281365277096933, "loss": 1.6829, "step": 2825 }, { "epoch": 0.35, "learning_rate": 0.0003727259384037852, "loss": 1.5471, "step": 2826 }, { "epoch": 0.35, "learning_rate": 0.00037263820412771804, "loss": 1.6037, "step": 2827 }, { "epoch": 0.35, "learning_rate": 0.00037255044995700024, "loss": 1.6108, "step": 2828 }, { "epoch": 0.35, "learning_rate": 0.0003724626759058675, "loss": 1.6874, "step": 2829 }, { "epoch": 0.35, "learning_rate": 0.0003723748819885587, "loss": 1.5978, "step": 2830 }, { "epoch": 0.35, "learning_rate": 0.0003722870682193161, "loss": 1.6469, "step": 2831 }, { "epoch": 0.35, "learning_rate": 0.00037219923461238505, "loss": 1.5597, "step": 2832 }, { "epoch": 0.35, "learning_rate": 0.0003721113811820138, "loss": 1.5594, "step": 2833 }, { "epoch": 0.35, "learning_rate": 0.00037202350794245456, "loss": 1.5374, "step": 2834 }, { "epoch": 0.35, "learning_rate": 0.00037193561490796206, "loss": 1.6569, "step": 2835 }, { "epoch": 0.35, "learning_rate": 0.0003718477020927946, "loss": 1.5193, "step": 2836 }, { "epoch": 0.35, "learning_rate": 0.0003717597695112136, "loss": 1.5608, "step": 2837 }, { "epoch": 0.35, "learning_rate": 0.0003716718171774837, "loss": 1.6204, "step": 2838 }, { "epoch": 0.35, "learning_rate": 0.00037158384510587263, "loss": 1.5749, "step": 2839 }, { "epoch": 0.35, "learning_rate": 0.0003714958533106515, "loss": 1.5888, "step": 2840 }, { "epoch": 0.36, "learning_rate": 0.0003714078418060946, "loss": 1.5608, "step": 2841 }, { "epoch": 0.36, "learning_rate": 0.00037131981060647924, "loss": 1.5297, "step": 2842 }, { "epoch": 0.36, "learning_rate": 0.0003712317597260861, "loss": 1.6097, "step": 2843 }, { "epoch": 0.36, "learning_rate": 0.00037114368917919905, "loss": 1.5565, "step": 2844 }, { "epoch": 0.36, "learning_rate": 0.00037105559898010495, "loss": 1.5307, "step": 2845 }, { "epoch": 0.36, "learning_rate": 0.00037096748914309407, "loss": 1.6201, "step": 2846 }, { "epoch": 0.36, "learning_rate": 0.0003708793596824598, "loss": 1.5084, "step": 2847 }, { "epoch": 0.36, "learning_rate": 0.00037079121061249883, "loss": 1.5526, "step": 2848 }, { "epoch": 0.36, "learning_rate": 0.00037070304194751063, "loss": 1.5338, "step": 2849 }, { "epoch": 0.36, "learning_rate": 0.0003706148537017984, "loss": 1.4914, "step": 2850 }, { "epoch": 0.36, "learning_rate": 0.000370526645889668, "loss": 1.5454, "step": 2851 }, { "epoch": 0.36, "learning_rate": 0.0003704384185254288, "loss": 1.4858, "step": 2852 }, { "epoch": 0.36, "learning_rate": 0.00037035017162339337, "loss": 1.5569, "step": 2853 }, { "epoch": 0.36, "learning_rate": 0.00037026190519787714, "loss": 1.5124, "step": 2854 }, { "epoch": 0.36, "learning_rate": 0.00037017361926319894, "loss": 1.5097, "step": 2855 }, { "epoch": 0.36, "learning_rate": 0.00037008531383368075, "loss": 1.5106, "step": 2856 }, { "epoch": 0.36, "learning_rate": 0.00036999698892364776, "loss": 1.468, "step": 2857 }, { "epoch": 0.36, "learning_rate": 0.000369908644547428, "loss": 1.5831, "step": 2858 }, { "epoch": 0.36, "learning_rate": 0.0003698202807193532, "loss": 1.6045, "step": 2859 }, { "epoch": 0.36, "learning_rate": 0.00036973189745375767, "loss": 1.5817, "step": 2860 }, { "epoch": 0.36, "learning_rate": 0.0003696434947649793, "loss": 1.4926, "step": 2861 }, { "epoch": 0.36, "learning_rate": 0.000369555072667359, "loss": 1.5141, "step": 2862 }, { "epoch": 0.36, "learning_rate": 0.0003694666311752407, "loss": 1.5293, "step": 2863 }, { "epoch": 0.36, "learning_rate": 0.0003693781703029716, "loss": 1.5151, "step": 2864 }, { "epoch": 0.36, "learning_rate": 0.0003692896900649021, "loss": 1.4987, "step": 2865 }, { "epoch": 0.36, "learning_rate": 0.0003692011904753856, "loss": 1.4606, "step": 2866 }, { "epoch": 0.36, "learning_rate": 0.00036911267154877877, "loss": 1.602, "step": 2867 }, { "epoch": 0.36, "learning_rate": 0.0003690241332994413, "loss": 1.5906, "step": 2868 }, { "epoch": 0.36, "learning_rate": 0.0003689355757417361, "loss": 1.5057, "step": 2869 }, { "epoch": 0.36, "learning_rate": 0.00036884699889002905, "loss": 1.4504, "step": 2870 }, { "epoch": 0.36, "learning_rate": 0.0003687584027586894, "loss": 1.4981, "step": 2871 }, { "epoch": 0.36, "learning_rate": 0.00036866978736208955, "loss": 1.4479, "step": 2872 }, { "epoch": 0.36, "learning_rate": 0.00036858115271460473, "loss": 1.5978, "step": 2873 }, { "epoch": 0.36, "learning_rate": 0.00036849249883061337, "loss": 1.5018, "step": 2874 }, { "epoch": 0.36, "learning_rate": 0.0003684038257244973, "loss": 1.5504, "step": 2875 }, { "epoch": 0.36, "learning_rate": 0.0003683151334106413, "loss": 1.4841, "step": 2876 }, { "epoch": 0.36, "learning_rate": 0.0003682264219034329, "loss": 1.575, "step": 2877 }, { "epoch": 0.36, "learning_rate": 0.00036813769121726354, "loss": 1.4849, "step": 2878 }, { "epoch": 0.36, "learning_rate": 0.000368048941366527, "loss": 1.5506, "step": 2879 }, { "epoch": 0.36, "learning_rate": 0.0003679601723656205, "loss": 1.5119, "step": 2880 }, { "epoch": 0.36, "learning_rate": 0.00036787138422894464, "loss": 1.4818, "step": 2881 }, { "epoch": 0.36, "learning_rate": 0.00036778257697090246, "loss": 1.5532, "step": 2882 }, { "epoch": 0.36, "learning_rate": 0.00036769375060590067, "loss": 1.4728, "step": 2883 }, { "epoch": 0.36, "learning_rate": 0.0003676049051483489, "loss": 1.49, "step": 2884 }, { "epoch": 0.36, "learning_rate": 0.0003675160406126598, "loss": 1.4453, "step": 2885 }, { "epoch": 0.36, "learning_rate": 0.00036742715701324917, "loss": 1.5675, "step": 2886 }, { "epoch": 0.36, "learning_rate": 0.00036733825436453593, "loss": 1.4956, "step": 2887 }, { "epoch": 0.36, "learning_rate": 0.0003672493326809422, "loss": 1.5169, "step": 2888 }, { "epoch": 0.36, "learning_rate": 0.0003671603919768929, "loss": 1.5199, "step": 2889 }, { "epoch": 0.36, "learning_rate": 0.00036707143226681616, "loss": 1.4053, "step": 2890 }, { "epoch": 0.36, "learning_rate": 0.00036698245356514336, "loss": 1.5018, "step": 2891 }, { "epoch": 0.36, "learning_rate": 0.0003668934558863088, "loss": 1.5326, "step": 2892 }, { "epoch": 0.36, "learning_rate": 0.0003668044392447498, "loss": 1.518, "step": 2893 }, { "epoch": 0.36, "learning_rate": 0.00036671540365490696, "loss": 1.4355, "step": 2894 }, { "epoch": 0.36, "learning_rate": 0.00036662634913122376, "loss": 1.5172, "step": 2895 }, { "epoch": 0.36, "learning_rate": 0.0003665372756881468, "loss": 1.5154, "step": 2896 }, { "epoch": 0.36, "learning_rate": 0.00036644818334012597, "loss": 1.5045, "step": 2897 }, { "epoch": 0.36, "learning_rate": 0.00036635907210161387, "loss": 1.5031, "step": 2898 }, { "epoch": 0.36, "learning_rate": 0.0003662699419870662, "loss": 1.5228, "step": 2899 }, { "epoch": 0.36, "learning_rate": 0.00036618079301094214, "loss": 1.4965, "step": 2900 }, { "epoch": 0.36, "eval_acceptance_rate": 0.6553004384040833, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0404141852906226, "eval_expected_tokens": 2.550517731613278, "eval_loss": 1.5254207849502563, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3876294329033195, "eval_runtime": 50.4335, "eval_samples_per_second": 9.656, "eval_steps_per_second": 0.317, "step": 2900 }, { "epoch": 0.36, "learning_rate": 0.0003660916251877035, "loss": 1.5141, "step": 2901 }, { "epoch": 0.36, "learning_rate": 0.00036600243853181523, "loss": 1.4734, "step": 2902 }, { "epoch": 0.36, "learning_rate": 0.0003659132330577456, "loss": 1.5819, "step": 2903 }, { "epoch": 0.36, "learning_rate": 0.00036582400877996547, "loss": 1.5104, "step": 2904 }, { "epoch": 0.36, "learning_rate": 0.0003657347657129491, "loss": 1.5961, "step": 2905 }, { "epoch": 0.36, "learning_rate": 0.0003656455038711738, "loss": 1.5192, "step": 2906 }, { "epoch": 0.36, "learning_rate": 0.0003655562232691197, "loss": 1.5038, "step": 2907 }, { "epoch": 0.36, "learning_rate": 0.0003654669239212702, "loss": 1.5355, "step": 2908 }, { "epoch": 0.36, "learning_rate": 0.0003653776058421117, "loss": 1.5507, "step": 2909 }, { "epoch": 0.36, "learning_rate": 0.00036528826904613337, "loss": 1.4749, "step": 2910 }, { "epoch": 0.36, "learning_rate": 0.0003651989135478278, "loss": 1.4285, "step": 2911 }, { "epoch": 0.36, "learning_rate": 0.00036510953936169036, "loss": 1.5098, "step": 2912 }, { "epoch": 0.36, "learning_rate": 0.00036502014650221967, "loss": 1.5059, "step": 2913 }, { "epoch": 0.36, "learning_rate": 0.00036493073498391706, "loss": 1.462, "step": 2914 }, { "epoch": 0.36, "learning_rate": 0.00036484130482128716, "loss": 1.5424, "step": 2915 }, { "epoch": 0.36, "learning_rate": 0.00036475185602883746, "loss": 1.505, "step": 2916 }, { "epoch": 0.36, "learning_rate": 0.0003646623886210788, "loss": 1.5411, "step": 2917 }, { "epoch": 0.36, "learning_rate": 0.0003645729026125244, "loss": 1.4481, "step": 2918 }, { "epoch": 0.36, "learning_rate": 0.00036448339801769126, "loss": 1.5828, "step": 2919 }, { "epoch": 0.36, "learning_rate": 0.00036439387485109886, "loss": 1.4536, "step": 2920 }, { "epoch": 0.37, "learning_rate": 0.0003643043331272697, "loss": 1.5398, "step": 2921 }, { "epoch": 0.37, "learning_rate": 0.00036421477286072975, "loss": 1.438, "step": 2922 }, { "epoch": 0.37, "learning_rate": 0.00036412519406600743, "loss": 1.4857, "step": 2923 }, { "epoch": 0.37, "learning_rate": 0.00036403559675763454, "loss": 1.5563, "step": 2924 }, { "epoch": 0.37, "learning_rate": 0.0003639459809501457, "loss": 1.5194, "step": 2925 }, { "epoch": 0.37, "learning_rate": 0.00036385634665807873, "loss": 1.5214, "step": 2926 }, { "epoch": 0.37, "learning_rate": 0.0003637666938959741, "loss": 1.5253, "step": 2927 }, { "epoch": 0.37, "learning_rate": 0.0003636770226783756, "loss": 1.5508, "step": 2928 }, { "epoch": 0.37, "learning_rate": 0.00036358733301982993, "loss": 1.4211, "step": 2929 }, { "epoch": 0.37, "learning_rate": 0.00036349762493488667, "loss": 1.4692, "step": 2930 }, { "epoch": 0.37, "learning_rate": 0.00036340789843809855, "loss": 1.5513, "step": 2931 }, { "epoch": 0.37, "learning_rate": 0.0003633181535440212, "loss": 1.484, "step": 2932 }, { "epoch": 0.37, "learning_rate": 0.00036322839026721324, "loss": 1.4703, "step": 2933 }, { "epoch": 0.37, "learning_rate": 0.00036313860862223615, "loss": 1.5162, "step": 2934 }, { "epoch": 0.37, "learning_rate": 0.0003630488086236547, "loss": 1.4729, "step": 2935 }, { "epoch": 0.37, "learning_rate": 0.0003629589902860363, "loss": 1.5051, "step": 2936 }, { "epoch": 0.37, "learning_rate": 0.00036286915362395154, "loss": 1.5442, "step": 2937 }, { "epoch": 0.37, "learning_rate": 0.00036277929865197395, "loss": 1.4918, "step": 2938 }, { "epoch": 0.37, "learning_rate": 0.0003626894253846801, "loss": 1.4675, "step": 2939 }, { "epoch": 0.37, "learning_rate": 0.0003625995338366492, "loss": 1.4764, "step": 2940 }, { "epoch": 0.37, "learning_rate": 0.0003625096240224639, "loss": 1.4548, "step": 2941 }, { "epoch": 0.37, "learning_rate": 0.0003624196959567094, "loss": 1.5105, "step": 2942 }, { "epoch": 0.37, "learning_rate": 0.00036232974965397414, "loss": 1.5533, "step": 2943 }, { "epoch": 0.37, "learning_rate": 0.00036223978512884945, "loss": 1.4844, "step": 2944 }, { "epoch": 0.37, "learning_rate": 0.0003621498023959294, "loss": 1.523, "step": 2945 }, { "epoch": 0.37, "learning_rate": 0.00036205980146981144, "loss": 1.4633, "step": 2946 }, { "epoch": 0.37, "learning_rate": 0.0003619697823650956, "loss": 1.5325, "step": 2947 }, { "epoch": 0.37, "learning_rate": 0.00036187974509638494, "loss": 1.522, "step": 2948 }, { "epoch": 0.37, "learning_rate": 0.0003617896896782855, "loss": 1.4764, "step": 2949 }, { "epoch": 0.37, "learning_rate": 0.0003616996161254065, "loss": 1.4078, "step": 2950 }, { "epoch": 0.37, "learning_rate": 0.0003616095244523596, "loss": 1.5061, "step": 2951 }, { "epoch": 0.37, "learning_rate": 0.0003615194146737598, "loss": 1.471, "step": 2952 }, { "epoch": 0.37, "learning_rate": 0.0003614292868042249, "loss": 1.4789, "step": 2953 }, { "epoch": 0.37, "learning_rate": 0.0003613391408583757, "loss": 1.5333, "step": 2954 }, { "epoch": 0.37, "learning_rate": 0.0003612489768508358, "loss": 1.4675, "step": 2955 }, { "epoch": 0.37, "learning_rate": 0.00036115879479623185, "loss": 1.5808, "step": 2956 }, { "epoch": 0.37, "learning_rate": 0.0003610685947091935, "loss": 1.5407, "step": 2957 }, { "epoch": 0.37, "learning_rate": 0.000360978376604353, "loss": 1.4611, "step": 2958 }, { "epoch": 0.37, "learning_rate": 0.0003608881404963458, "loss": 1.5418, "step": 2959 }, { "epoch": 0.37, "learning_rate": 0.00036079788639981037, "loss": 1.4784, "step": 2960 }, { "epoch": 0.37, "learning_rate": 0.0003607076143293877, "loss": 1.3893, "step": 2961 }, { "epoch": 0.37, "learning_rate": 0.00036061732429972215, "loss": 1.5373, "step": 2962 }, { "epoch": 0.37, "learning_rate": 0.00036052701632546065, "loss": 1.5015, "step": 2963 }, { "epoch": 0.37, "learning_rate": 0.00036043669042125316, "loss": 1.4724, "step": 2964 }, { "epoch": 0.37, "learning_rate": 0.00036034634660175256, "loss": 1.4904, "step": 2965 }, { "epoch": 0.37, "learning_rate": 0.0003602559848816148, "loss": 1.5265, "step": 2966 }, { "epoch": 0.37, "learning_rate": 0.00036016560527549836, "loss": 1.5055, "step": 2967 }, { "epoch": 0.37, "learning_rate": 0.0003600752077980648, "loss": 1.5233, "step": 2968 }, { "epoch": 0.37, "learning_rate": 0.0003599847924639788, "loss": 1.4615, "step": 2969 }, { "epoch": 0.37, "learning_rate": 0.00035989435928790757, "loss": 1.544, "step": 2970 }, { "epoch": 0.37, "learning_rate": 0.00035980390828452155, "loss": 1.4803, "step": 2971 }, { "epoch": 0.37, "learning_rate": 0.0003597134394684937, "loss": 1.4362, "step": 2972 }, { "epoch": 0.37, "learning_rate": 0.0003596229528545003, "loss": 1.494, "step": 2973 }, { "epoch": 0.37, "learning_rate": 0.00035953244845722017, "loss": 1.4628, "step": 2974 }, { "epoch": 0.37, "learning_rate": 0.0003594419262913351, "loss": 1.5333, "step": 2975 }, { "epoch": 0.37, "learning_rate": 0.00035935138637153, "loss": 1.5321, "step": 2976 }, { "epoch": 0.37, "learning_rate": 0.00035926082871249216, "loss": 1.5266, "step": 2977 }, { "epoch": 0.37, "learning_rate": 0.00035917025332891234, "loss": 1.5789, "step": 2978 }, { "epoch": 0.37, "learning_rate": 0.0003590796602354838, "loss": 1.5835, "step": 2979 }, { "epoch": 0.37, "learning_rate": 0.0003589890494469025, "loss": 1.4702, "step": 2980 }, { "epoch": 0.37, "learning_rate": 0.000358898420977868, "loss": 1.433, "step": 2981 }, { "epoch": 0.37, "learning_rate": 0.0003588077748430819, "loss": 1.5016, "step": 2982 }, { "epoch": 0.37, "learning_rate": 0.00035871711105724916, "loss": 1.4691, "step": 2983 }, { "epoch": 0.37, "learning_rate": 0.0003586264296350775, "loss": 1.5151, "step": 2984 }, { "epoch": 0.37, "learning_rate": 0.0003585357305912773, "loss": 1.5114, "step": 2985 }, { "epoch": 0.37, "learning_rate": 0.0003584450139405621, "loss": 1.5024, "step": 2986 }, { "epoch": 0.37, "learning_rate": 0.00035835427969764814, "loss": 1.5146, "step": 2987 }, { "epoch": 0.37, "learning_rate": 0.0003582635278772547, "loss": 1.4853, "step": 2988 }, { "epoch": 0.37, "learning_rate": 0.00035817275849410343, "loss": 1.4963, "step": 2989 }, { "epoch": 0.37, "learning_rate": 0.0003580819715629193, "loss": 1.5146, "step": 2990 }, { "epoch": 0.37, "learning_rate": 0.00035799116709843003, "loss": 1.409, "step": 2991 }, { "epoch": 0.37, "learning_rate": 0.00035790034511536605, "loss": 1.4388, "step": 2992 }, { "epoch": 0.37, "learning_rate": 0.00035780950562846076, "loss": 1.476, "step": 2993 }, { "epoch": 0.37, "learning_rate": 0.00035771864865245036, "loss": 1.4745, "step": 2994 }, { "epoch": 0.37, "learning_rate": 0.0003576277742020738, "loss": 1.5345, "step": 2995 }, { "epoch": 0.37, "learning_rate": 0.000357536882292073, "loss": 1.4594, "step": 2996 }, { "epoch": 0.37, "learning_rate": 0.00035744597293719263, "loss": 1.4523, "step": 2997 }, { "epoch": 0.37, "learning_rate": 0.0003573550461521803, "loss": 1.4993, "step": 2998 }, { "epoch": 0.37, "learning_rate": 0.0003572641019517861, "loss": 1.494, "step": 2999 }, { "epoch": 0.38, "learning_rate": 0.0003571731403507635, "loss": 1.4864, "step": 3000 }, { "epoch": 0.38, "eval_acceptance_rate": 0.658433735370636, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0523002882668773, "eval_expected_tokens": 2.5653753603335967, "eval_loss": 1.5153906345367432, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3913438400833992, "eval_runtime": 50.3215, "eval_samples_per_second": 9.678, "eval_steps_per_second": 0.318, "step": 3000 }, { "epoch": 0.38, "learning_rate": 0.0003570821613638685, "loss": 1.4816, "step": 3001 }, { "epoch": 0.38, "learning_rate": 0.00035699116500585964, "loss": 1.4605, "step": 3002 }, { "epoch": 0.38, "learning_rate": 0.0003569001512914988, "loss": 1.4874, "step": 3003 }, { "epoch": 0.38, "learning_rate": 0.0003568091202355504, "loss": 1.5066, "step": 3004 }, { "epoch": 0.38, "learning_rate": 0.0003567180718527816, "loss": 1.5037, "step": 3005 }, { "epoch": 0.38, "learning_rate": 0.00035662700615796263, "loss": 1.551, "step": 3006 }, { "epoch": 0.38, "learning_rate": 0.00035653592316586623, "loss": 1.4547, "step": 3007 }, { "epoch": 0.38, "learning_rate": 0.0003564448228912682, "loss": 1.4503, "step": 3008 }, { "epoch": 0.38, "learning_rate": 0.0003563537053489469, "loss": 1.4565, "step": 3009 }, { "epoch": 0.38, "learning_rate": 0.0003562625705536838, "loss": 1.4586, "step": 3010 }, { "epoch": 0.38, "learning_rate": 0.0003561714185202629, "loss": 1.5117, "step": 3011 }, { "epoch": 0.38, "learning_rate": 0.0003560802492634711, "loss": 1.4936, "step": 3012 }, { "epoch": 0.38, "learning_rate": 0.0003559890627980981, "loss": 1.5415, "step": 3013 }, { "epoch": 0.38, "learning_rate": 0.00035589785913893636, "loss": 1.4521, "step": 3014 }, { "epoch": 0.38, "learning_rate": 0.00035580663830078106, "loss": 1.5016, "step": 3015 }, { "epoch": 0.38, "learning_rate": 0.00035571540029843046, "loss": 1.5153, "step": 3016 }, { "epoch": 0.38, "learning_rate": 0.0003556241451466853, "loss": 1.4224, "step": 3017 }, { "epoch": 0.38, "learning_rate": 0.0003555328728603491, "loss": 1.5181, "step": 3018 }, { "epoch": 0.38, "learning_rate": 0.0003554415834542284, "loss": 1.4636, "step": 3019 }, { "epoch": 0.38, "learning_rate": 0.0003553502769431323, "loss": 1.4913, "step": 3020 }, { "epoch": 0.38, "learning_rate": 0.00035525895334187274, "loss": 1.4786, "step": 3021 }, { "epoch": 0.38, "learning_rate": 0.00035516761266526453, "loss": 1.4965, "step": 3022 }, { "epoch": 0.38, "learning_rate": 0.0003550762549281251, "loss": 1.5014, "step": 3023 }, { "epoch": 0.38, "learning_rate": 0.00035498488014527467, "loss": 1.4242, "step": 3024 }, { "epoch": 0.38, "learning_rate": 0.00035489348833153645, "loss": 1.427, "step": 3025 }, { "epoch": 0.38, "learning_rate": 0.00035480207950173596, "loss": 1.4909, "step": 3026 }, { "epoch": 0.38, "learning_rate": 0.00035471065367070193, "loss": 1.519, "step": 3027 }, { "epoch": 0.38, "learning_rate": 0.0003546192108532657, "loss": 1.5157, "step": 3028 }, { "epoch": 0.38, "learning_rate": 0.0003545277510642612, "loss": 1.5667, "step": 3029 }, { "epoch": 0.38, "learning_rate": 0.00035443627431852527, "loss": 1.4812, "step": 3030 }, { "epoch": 0.38, "learning_rate": 0.0003543447806308975, "loss": 1.5202, "step": 3031 }, { "epoch": 0.38, "learning_rate": 0.00035425327001622033, "loss": 1.4821, "step": 3032 }, { "epoch": 0.38, "learning_rate": 0.00035416174248933865, "loss": 1.5155, "step": 3033 }, { "epoch": 0.38, "learning_rate": 0.00035407019806510035, "loss": 1.4186, "step": 3034 }, { "epoch": 0.38, "learning_rate": 0.0003539786367583559, "loss": 1.4709, "step": 3035 }, { "epoch": 0.38, "learning_rate": 0.00035388705858395864, "loss": 1.4227, "step": 3036 }, { "epoch": 0.38, "learning_rate": 0.0003537954635567646, "loss": 1.5363, "step": 3037 }, { "epoch": 0.38, "learning_rate": 0.0003537038516916326, "loss": 1.4275, "step": 3038 }, { "epoch": 0.38, "learning_rate": 0.000353612223003424, "loss": 1.5077, "step": 3039 }, { "epoch": 0.38, "learning_rate": 0.000353520577507003, "loss": 1.4938, "step": 3040 }, { "epoch": 0.38, "learning_rate": 0.00035342891521723666, "loss": 1.4941, "step": 3041 }, { "epoch": 0.38, "learning_rate": 0.0003533372361489947, "loss": 1.4863, "step": 3042 }, { "epoch": 0.38, "learning_rate": 0.0003532455403171492, "loss": 1.4751, "step": 3043 }, { "epoch": 0.38, "learning_rate": 0.0003531538277365756, "loss": 1.5338, "step": 3044 }, { "epoch": 0.38, "learning_rate": 0.00035306209842215163, "loss": 1.5711, "step": 3045 }, { "epoch": 0.38, "learning_rate": 0.0003529703523887577, "loss": 1.4279, "step": 3046 }, { "epoch": 0.38, "learning_rate": 0.00035287858965127723, "loss": 1.5576, "step": 3047 }, { "epoch": 0.38, "learning_rate": 0.0003527868102245961, "loss": 1.505, "step": 3048 }, { "epoch": 0.38, "learning_rate": 0.00035269501412360293, "loss": 1.5292, "step": 3049 }, { "epoch": 0.38, "learning_rate": 0.00035260320136318925, "loss": 1.4985, "step": 3050 }, { "epoch": 0.38, "learning_rate": 0.00035251137195824894, "loss": 1.4725, "step": 3051 }, { "epoch": 0.38, "learning_rate": 0.00035241952592367893, "loss": 1.5538, "step": 3052 }, { "epoch": 0.38, "learning_rate": 0.00035232766327437866, "loss": 1.4696, "step": 3053 }, { "epoch": 0.38, "learning_rate": 0.00035223578402525027, "loss": 1.4989, "step": 3054 }, { "epoch": 0.38, "learning_rate": 0.0003521438881911986, "loss": 1.4829, "step": 3055 }, { "epoch": 0.38, "learning_rate": 0.00035205197578713125, "loss": 1.4279, "step": 3056 }, { "epoch": 0.38, "learning_rate": 0.0003519600468279585, "loss": 1.4437, "step": 3057 }, { "epoch": 0.38, "learning_rate": 0.00035186810132859317, "loss": 1.5056, "step": 3058 }, { "epoch": 0.38, "learning_rate": 0.0003517761393039509, "loss": 1.3891, "step": 3059 }, { "epoch": 0.38, "learning_rate": 0.0003516841607689501, "loss": 1.5153, "step": 3060 }, { "epoch": 0.38, "learning_rate": 0.0003515921657385116, "loss": 1.4955, "step": 3061 }, { "epoch": 0.38, "learning_rate": 0.00035150015422755903, "loss": 1.4732, "step": 3062 }, { "epoch": 0.38, "learning_rate": 0.0003514081262510188, "loss": 1.4521, "step": 3063 }, { "epoch": 0.38, "learning_rate": 0.00035131608182381986, "loss": 1.4444, "step": 3064 }, { "epoch": 0.38, "learning_rate": 0.00035122402096089385, "loss": 1.4597, "step": 3065 }, { "epoch": 0.38, "learning_rate": 0.00035113194367717516, "loss": 1.4279, "step": 3066 }, { "epoch": 0.38, "learning_rate": 0.00035103984998760074, "loss": 1.4718, "step": 3067 }, { "epoch": 0.38, "learning_rate": 0.0003509477399071102, "loss": 1.4422, "step": 3068 }, { "epoch": 0.38, "learning_rate": 0.0003508556134506459, "loss": 1.4222, "step": 3069 }, { "epoch": 0.38, "learning_rate": 0.0003507634706331528, "loss": 1.5136, "step": 3070 }, { "epoch": 0.38, "learning_rate": 0.00035067131146957847, "loss": 1.4773, "step": 3071 }, { "epoch": 0.38, "learning_rate": 0.0003505791359748733, "loss": 1.505, "step": 3072 }, { "epoch": 0.38, "learning_rate": 0.00035048694416399005, "loss": 1.514, "step": 3073 }, { "epoch": 0.38, "learning_rate": 0.0003503947360518844, "loss": 1.5107, "step": 3074 }, { "epoch": 0.38, "learning_rate": 0.00035030251165351446, "loss": 1.4944, "step": 3075 }, { "epoch": 0.38, "learning_rate": 0.00035021027098384127, "loss": 1.4726, "step": 3076 }, { "epoch": 0.38, "learning_rate": 0.0003501180140578281, "loss": 1.5151, "step": 3077 }, { "epoch": 0.38, "learning_rate": 0.0003500257408904412, "loss": 1.416, "step": 3078 }, { "epoch": 0.38, "learning_rate": 0.0003499334514966494, "loss": 1.4722, "step": 3079 }, { "epoch": 0.39, "learning_rate": 0.0003498411458914239, "loss": 1.464, "step": 3080 }, { "epoch": 0.39, "learning_rate": 0.0003497488240897389, "loss": 1.575, "step": 3081 }, { "epoch": 0.39, "learning_rate": 0.000349656486106571, "loss": 1.4349, "step": 3082 }, { "epoch": 0.39, "learning_rate": 0.0003495641319568994, "loss": 1.4515, "step": 3083 }, { "epoch": 0.39, "learning_rate": 0.00034947176165570613, "loss": 1.4182, "step": 3084 }, { "epoch": 0.39, "learning_rate": 0.0003493793752179757, "loss": 1.4878, "step": 3085 }, { "epoch": 0.39, "learning_rate": 0.00034928697265869515, "loss": 1.5032, "step": 3086 }, { "epoch": 0.39, "learning_rate": 0.0003491945539928544, "loss": 1.5108, "step": 3087 }, { "epoch": 0.39, "learning_rate": 0.0003491021192354455, "loss": 1.4879, "step": 3088 }, { "epoch": 0.39, "learning_rate": 0.0003490096684014638, "loss": 1.5255, "step": 3089 }, { "epoch": 0.39, "learning_rate": 0.00034891720150590665, "loss": 1.4632, "step": 3090 }, { "epoch": 0.39, "learning_rate": 0.0003488247185637744, "loss": 1.5492, "step": 3091 }, { "epoch": 0.39, "learning_rate": 0.00034873221959006974, "loss": 1.5477, "step": 3092 }, { "epoch": 0.39, "learning_rate": 0.00034863970459979803, "loss": 1.5219, "step": 3093 }, { "epoch": 0.39, "learning_rate": 0.0003485471736079674, "loss": 1.5089, "step": 3094 }, { "epoch": 0.39, "learning_rate": 0.00034845462662958833, "loss": 1.5163, "step": 3095 }, { "epoch": 0.39, "learning_rate": 0.000348362063679674, "loss": 1.4597, "step": 3096 }, { "epoch": 0.39, "learning_rate": 0.0003482694847732402, "loss": 1.4521, "step": 3097 }, { "epoch": 0.39, "learning_rate": 0.0003481768899253054, "loss": 1.5019, "step": 3098 }, { "epoch": 0.39, "learning_rate": 0.00034808427915089036, "loss": 1.5121, "step": 3099 }, { "epoch": 0.39, "learning_rate": 0.0003479916524650188, "loss": 1.4038, "step": 3100 }, { "epoch": 0.39, "eval_acceptance_rate": 0.6597695350646973, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0573942206711076, "eval_expected_tokens": 2.571742775838884, "eval_loss": 1.5045864582061768, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.3929356939597211, "eval_runtime": 50.4351, "eval_samples_per_second": 9.656, "eval_steps_per_second": 0.317, "step": 3100 }, { "epoch": 0.39, "learning_rate": 0.0003478990098827167, "loss": 1.4174, "step": 3101 }, { "epoch": 0.39, "learning_rate": 0.00034780635141901274, "loss": 1.4896, "step": 3102 }, { "epoch": 0.39, "learning_rate": 0.00034771367708893834, "loss": 1.4411, "step": 3103 }, { "epoch": 0.39, "learning_rate": 0.00034762098690752725, "loss": 1.42, "step": 3104 }, { "epoch": 0.39, "learning_rate": 0.0003475282808898158, "loss": 1.5294, "step": 3105 }, { "epoch": 0.39, "learning_rate": 0.0003474355590508431, "loss": 1.4416, "step": 3106 }, { "epoch": 0.39, "learning_rate": 0.00034734282140565063, "loss": 1.4733, "step": 3107 }, { "epoch": 0.39, "learning_rate": 0.00034725006796928256, "loss": 1.4276, "step": 3108 }, { "epoch": 0.39, "learning_rate": 0.00034715729875678554, "loss": 1.449, "step": 3109 }, { "epoch": 0.39, "learning_rate": 0.00034706451378320875, "loss": 1.4643, "step": 3110 }, { "epoch": 0.39, "learning_rate": 0.00034697171306360405, "loss": 1.4975, "step": 3111 }, { "epoch": 0.39, "learning_rate": 0.00034687889661302575, "loss": 1.4695, "step": 3112 }, { "epoch": 0.39, "learning_rate": 0.0003467860644465308, "loss": 1.4699, "step": 3113 }, { "epoch": 0.39, "learning_rate": 0.00034669321657917855, "loss": 1.4272, "step": 3114 }, { "epoch": 0.39, "learning_rate": 0.00034660035302603106, "loss": 1.4269, "step": 3115 }, { "epoch": 0.39, "learning_rate": 0.0003465074738021529, "loss": 1.4562, "step": 3116 }, { "epoch": 0.39, "learning_rate": 0.0003464145789226111, "loss": 1.4712, "step": 3117 }, { "epoch": 0.39, "learning_rate": 0.00034632166840247526, "loss": 1.4525, "step": 3118 }, { "epoch": 0.39, "learning_rate": 0.00034622874225681755, "loss": 1.4619, "step": 3119 }, { "epoch": 0.39, "learning_rate": 0.0003461358005007128, "loss": 1.4903, "step": 3120 }, { "epoch": 0.39, "learning_rate": 0.00034604284314923795, "loss": 1.4013, "step": 3121 }, { "epoch": 0.39, "learning_rate": 0.00034594987021747297, "loss": 1.4175, "step": 3122 }, { "epoch": 0.39, "learning_rate": 0.0003458568817205002, "loss": 1.485, "step": 3123 }, { "epoch": 0.39, "learning_rate": 0.00034576387767340413, "loss": 1.4561, "step": 3124 }, { "epoch": 0.39, "learning_rate": 0.0003456708580912725, "loss": 1.4355, "step": 3125 }, { "epoch": 0.39, "learning_rate": 0.0003455778229891949, "loss": 1.4858, "step": 3126 }, { "epoch": 0.39, "learning_rate": 0.00034548477238226363, "loss": 1.4279, "step": 3127 }, { "epoch": 0.39, "learning_rate": 0.0003453917062855738, "loss": 1.5028, "step": 3128 }, { "epoch": 0.39, "learning_rate": 0.00034529862471422275, "loss": 1.4175, "step": 3129 }, { "epoch": 0.39, "learning_rate": 0.00034520552768331025, "loss": 1.4645, "step": 3130 }, { "epoch": 0.39, "learning_rate": 0.0003451124152079389, "loss": 1.4634, "step": 3131 }, { "epoch": 0.39, "learning_rate": 0.0003450192873032134, "loss": 1.4297, "step": 3132 }, { "epoch": 0.39, "learning_rate": 0.00034492614398424137, "loss": 1.4268, "step": 3133 }, { "epoch": 0.39, "learning_rate": 0.00034483298526613267, "loss": 1.4475, "step": 3134 }, { "epoch": 0.39, "learning_rate": 0.00034473981116399973, "loss": 1.5312, "step": 3135 }, { "epoch": 0.39, "learning_rate": 0.00034464662169295736, "loss": 1.472, "step": 3136 }, { "epoch": 0.39, "learning_rate": 0.0003445534168681231, "loss": 1.4195, "step": 3137 }, { "epoch": 0.39, "learning_rate": 0.00034446019670461683, "loss": 1.4211, "step": 3138 }, { "epoch": 0.39, "learning_rate": 0.0003443669612175608, "loss": 1.426, "step": 3139 }, { "epoch": 0.39, "learning_rate": 0.00034427371042208013, "loss": 1.4523, "step": 3140 }, { "epoch": 0.39, "learning_rate": 0.000344180444333302, "loss": 1.3888, "step": 3141 }, { "epoch": 0.39, "learning_rate": 0.0003440871629663563, "loss": 1.3526, "step": 3142 }, { "epoch": 0.39, "learning_rate": 0.0003439938663363753, "loss": 1.4256, "step": 3143 }, { "epoch": 0.39, "learning_rate": 0.0003439005544584939, "loss": 1.4597, "step": 3144 }, { "epoch": 0.39, "learning_rate": 0.0003438072273478493, "loss": 1.5085, "step": 3145 }, { "epoch": 0.39, "learning_rate": 0.0003437138850195811, "loss": 1.5251, "step": 3146 }, { "epoch": 0.39, "learning_rate": 0.0003436205274888318, "loss": 1.5569, "step": 3147 }, { "epoch": 0.39, "learning_rate": 0.0003435271547707459, "loss": 1.45, "step": 3148 }, { "epoch": 0.39, "learning_rate": 0.00034343376688047045, "loss": 1.4223, "step": 3149 }, { "epoch": 0.39, "learning_rate": 0.00034334036383315526, "loss": 1.4847, "step": 3150 }, { "epoch": 0.39, "learning_rate": 0.00034324694564395226, "loss": 1.5004, "step": 3151 }, { "epoch": 0.39, "learning_rate": 0.00034315351232801596, "loss": 1.4594, "step": 3152 }, { "epoch": 0.39, "learning_rate": 0.0003430600639005034, "loss": 1.4059, "step": 3153 }, { "epoch": 0.39, "learning_rate": 0.00034296660037657386, "loss": 1.4438, "step": 3154 }, { "epoch": 0.39, "learning_rate": 0.0003428731217713893, "loss": 1.4634, "step": 3155 }, { "epoch": 0.39, "learning_rate": 0.00034277962810011405, "loss": 1.4489, "step": 3156 }, { "epoch": 0.39, "learning_rate": 0.00034268611937791487, "loss": 1.4263, "step": 3157 }, { "epoch": 0.39, "learning_rate": 0.0003425925956199609, "loss": 1.4412, "step": 3158 }, { "epoch": 0.39, "learning_rate": 0.00034249905684142384, "loss": 1.48, "step": 3159 }, { "epoch": 0.4, "learning_rate": 0.0003424055030574777, "loss": 1.4014, "step": 3160 }, { "epoch": 0.4, "learning_rate": 0.0003423119342832991, "loss": 1.4027, "step": 3161 }, { "epoch": 0.4, "learning_rate": 0.00034221835053406687, "loss": 1.485, "step": 3162 }, { "epoch": 0.4, "learning_rate": 0.0003421247518249624, "loss": 1.45, "step": 3163 }, { "epoch": 0.4, "learning_rate": 0.00034203113817116957, "loss": 1.472, "step": 3164 }, { "epoch": 0.4, "learning_rate": 0.0003419375095878744, "loss": 1.5099, "step": 3165 }, { "epoch": 0.4, "learning_rate": 0.00034184386609026574, "loss": 1.47, "step": 3166 }, { "epoch": 0.4, "learning_rate": 0.00034175020769353467, "loss": 1.4729, "step": 3167 }, { "epoch": 0.4, "learning_rate": 0.00034165653441287444, "loss": 1.4125, "step": 3168 }, { "epoch": 0.4, "learning_rate": 0.0003415628462634812, "loss": 1.4643, "step": 3169 }, { "epoch": 0.4, "learning_rate": 0.0003414691432605531, "loss": 1.4848, "step": 3170 }, { "epoch": 0.4, "learning_rate": 0.0003413754254192908, "loss": 1.4192, "step": 3171 }, { "epoch": 0.4, "learning_rate": 0.0003412816927548977, "loss": 1.4505, "step": 3172 }, { "epoch": 0.4, "learning_rate": 0.00034118794528257906, "loss": 1.5289, "step": 3173 }, { "epoch": 0.4, "learning_rate": 0.0003410941830175429, "loss": 1.4213, "step": 3174 }, { "epoch": 0.4, "learning_rate": 0.0003410004059749996, "loss": 1.5298, "step": 3175 }, { "epoch": 0.4, "learning_rate": 0.00034090661417016175, "loss": 1.4904, "step": 3176 }, { "epoch": 0.4, "learning_rate": 0.00034081280761824465, "loss": 1.4417, "step": 3177 }, { "epoch": 0.4, "learning_rate": 0.00034071898633446565, "loss": 1.4343, "step": 3178 }, { "epoch": 0.4, "learning_rate": 0.00034062515033404475, "loss": 1.479, "step": 3179 }, { "epoch": 0.4, "learning_rate": 0.0003405312996322042, "loss": 1.4362, "step": 3180 }, { "epoch": 0.4, "learning_rate": 0.0003404374342441687, "loss": 1.4548, "step": 3181 }, { "epoch": 0.4, "learning_rate": 0.0003403435541851653, "loss": 1.4662, "step": 3182 }, { "epoch": 0.4, "learning_rate": 0.00034024965947042346, "loss": 1.3877, "step": 3183 }, { "epoch": 0.4, "learning_rate": 0.00034015575011517497, "loss": 1.4323, "step": 3184 }, { "epoch": 0.4, "learning_rate": 0.0003400618261346541, "loss": 1.4801, "step": 3185 }, { "epoch": 0.4, "learning_rate": 0.0003399678875440972, "loss": 1.5006, "step": 3186 }, { "epoch": 0.4, "learning_rate": 0.00033987393435874335, "loss": 1.4387, "step": 3187 }, { "epoch": 0.4, "learning_rate": 0.0003397799665938339, "loss": 1.5442, "step": 3188 }, { "epoch": 0.4, "learning_rate": 0.0003396859842646125, "loss": 1.4973, "step": 3189 }, { "epoch": 0.4, "learning_rate": 0.00033959198738632503, "loss": 1.4541, "step": 3190 }, { "epoch": 0.4, "learning_rate": 0.00033949797597422013, "loss": 1.4278, "step": 3191 }, { "epoch": 0.4, "learning_rate": 0.0003394039500435483, "loss": 1.4275, "step": 3192 }, { "epoch": 0.4, "learning_rate": 0.0003393099096095628, "loss": 1.4466, "step": 3193 }, { "epoch": 0.4, "learning_rate": 0.000339215854687519, "loss": 1.445, "step": 3194 }, { "epoch": 0.4, "learning_rate": 0.00033912178529267475, "loss": 1.476, "step": 3195 }, { "epoch": 0.4, "learning_rate": 0.0003390277014402902, "loss": 1.5404, "step": 3196 }, { "epoch": 0.4, "learning_rate": 0.0003389336031456278, "loss": 1.4606, "step": 3197 }, { "epoch": 0.4, "learning_rate": 0.00033883949042395244, "loss": 1.5522, "step": 3198 }, { "epoch": 0.4, "learning_rate": 0.0003387453632905313, "loss": 1.4393, "step": 3199 }, { "epoch": 0.4, "learning_rate": 0.0003386512217606339, "loss": 1.5365, "step": 3200 }, { "epoch": 0.4, "eval_acceptance_rate": 0.6613964438438416, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0636198306849116, "eval_expected_tokens": 2.579524788356139, "eval_loss": 1.49101722240448, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.39488119708903485, "eval_runtime": 50.3157, "eval_samples_per_second": 9.679, "eval_steps_per_second": 0.318, "step": 3200 }, { "epoch": 0.4, "learning_rate": 0.0003385570658495321, "loss": 1.512, "step": 3201 }, { "epoch": 0.4, "learning_rate": 0.00033846289557250004, "loss": 1.4758, "step": 3202 }, { "epoch": 0.4, "learning_rate": 0.00033836871094481433, "loss": 1.5407, "step": 3203 }, { "epoch": 0.4, "learning_rate": 0.00033827451198175377, "loss": 1.497, "step": 3204 }, { "epoch": 0.4, "learning_rate": 0.0003381802986985993, "loss": 1.4613, "step": 3205 }, { "epoch": 0.4, "learning_rate": 0.0003380860711106348, "loss": 1.4525, "step": 3206 }, { "epoch": 0.4, "learning_rate": 0.00033799182923314596, "loss": 1.3873, "step": 3207 }, { "epoch": 0.4, "learning_rate": 0.0003378975730814208, "loss": 1.3926, "step": 3208 }, { "epoch": 0.4, "learning_rate": 0.00033780330267074986, "loss": 1.463, "step": 3209 }, { "epoch": 0.4, "learning_rate": 0.00033770901801642597, "loss": 1.4553, "step": 3210 }, { "epoch": 0.4, "learning_rate": 0.00033761471913374395, "loss": 1.4904, "step": 3211 }, { "epoch": 0.4, "learning_rate": 0.00033752040603800146, "loss": 1.4632, "step": 3212 }, { "epoch": 0.4, "learning_rate": 0.00033742607874449806, "loss": 1.4727, "step": 3213 }, { "epoch": 0.4, "learning_rate": 0.00033733173726853573, "loss": 1.461, "step": 3214 }, { "epoch": 0.4, "learning_rate": 0.00033723738162541876, "loss": 1.3746, "step": 3215 }, { "epoch": 0.4, "learning_rate": 0.0003371430118304538, "loss": 1.3594, "step": 3216 }, { "epoch": 0.4, "learning_rate": 0.00033704862789894974, "loss": 1.4552, "step": 3217 }, { "epoch": 0.4, "learning_rate": 0.0003369542298462176, "loss": 1.4413, "step": 3218 }, { "epoch": 0.4, "learning_rate": 0.00033685981768757114, "loss": 1.4602, "step": 3219 }, { "epoch": 0.4, "learning_rate": 0.00033676539143832576, "loss": 1.5451, "step": 3220 }, { "epoch": 0.4, "learning_rate": 0.00033667095111379977, "loss": 1.5053, "step": 3221 }, { "epoch": 0.4, "learning_rate": 0.00033657649672931333, "loss": 1.4745, "step": 3222 }, { "epoch": 0.4, "learning_rate": 0.00033648202830018927, "loss": 1.5274, "step": 3223 }, { "epoch": 0.4, "learning_rate": 0.00033638754584175217, "loss": 1.4594, "step": 3224 }, { "epoch": 0.4, "learning_rate": 0.00033629304936932947, "loss": 1.4399, "step": 3225 }, { "epoch": 0.4, "learning_rate": 0.00033619853889825047, "loss": 1.5057, "step": 3226 }, { "epoch": 0.4, "learning_rate": 0.0003361040144438468, "loss": 1.4641, "step": 3227 }, { "epoch": 0.4, "learning_rate": 0.0003360094760214526, "loss": 1.4274, "step": 3228 }, { "epoch": 0.4, "learning_rate": 0.0003359149236464041, "loss": 1.4677, "step": 3229 }, { "epoch": 0.4, "learning_rate": 0.0003358203573340396, "loss": 1.4664, "step": 3230 }, { "epoch": 0.4, "learning_rate": 0.0003357257770997001, "loss": 1.4543, "step": 3231 }, { "epoch": 0.4, "learning_rate": 0.00033563118295872843, "loss": 1.4294, "step": 3232 }, { "epoch": 0.4, "learning_rate": 0.00033553657492647, "loss": 1.487, "step": 3233 }, { "epoch": 0.4, "learning_rate": 0.0003354419530182724, "loss": 1.4266, "step": 3234 }, { "epoch": 0.4, "learning_rate": 0.0003353473172494852, "loss": 1.4806, "step": 3235 }, { "epoch": 0.4, "learning_rate": 0.0003352526676354606, "loss": 1.4677, "step": 3236 }, { "epoch": 0.4, "learning_rate": 0.00033515800419155276, "loss": 1.4506, "step": 3237 }, { "epoch": 0.4, "learning_rate": 0.00033506332693311835, "loss": 1.4416, "step": 3238 }, { "epoch": 0.4, "learning_rate": 0.00033496863587551603, "loss": 1.5016, "step": 3239 }, { "epoch": 0.41, "learning_rate": 0.0003348739310341068, "loss": 1.5041, "step": 3240 }, { "epoch": 0.41, "learning_rate": 0.00033477921242425397, "loss": 1.4894, "step": 3241 }, { "epoch": 0.41, "learning_rate": 0.0003346844800613229, "loss": 1.448, "step": 3242 }, { "epoch": 0.41, "learning_rate": 0.00033458973396068134, "loss": 1.4059, "step": 3243 }, { "epoch": 0.41, "learning_rate": 0.0003344949741376992, "loss": 1.4508, "step": 3244 }, { "epoch": 0.41, "learning_rate": 0.0003344002006077487, "loss": 1.4652, "step": 3245 }, { "epoch": 0.41, "learning_rate": 0.00033430541338620425, "loss": 1.3811, "step": 3246 }, { "epoch": 0.41, "learning_rate": 0.0003342106124884423, "loss": 1.4273, "step": 3247 }, { "epoch": 0.41, "learning_rate": 0.0003341157979298418, "loss": 1.5032, "step": 3248 }, { "epoch": 0.41, "learning_rate": 0.0003340209697257836, "loss": 1.4164, "step": 3249 }, { "epoch": 0.41, "learning_rate": 0.0003339261278916512, "loss": 1.5259, "step": 3250 }, { "epoch": 0.41, "learning_rate": 0.00033383127244282994, "loss": 1.4469, "step": 3251 }, { "epoch": 0.41, "learning_rate": 0.00033373640339470745, "loss": 1.4354, "step": 3252 }, { "epoch": 0.41, "learning_rate": 0.00033364152076267373, "loss": 1.4711, "step": 3253 }, { "epoch": 0.41, "learning_rate": 0.0003335466245621207, "loss": 1.5358, "step": 3254 }, { "epoch": 0.41, "learning_rate": 0.0003334517148084427, "loss": 1.5142, "step": 3255 }, { "epoch": 0.41, "learning_rate": 0.00033335679151703636, "loss": 1.4359, "step": 3256 }, { "epoch": 0.41, "learning_rate": 0.00033326185470330006, "loss": 1.4629, "step": 3257 }, { "epoch": 0.41, "learning_rate": 0.00033316690438263484, "loss": 1.5344, "step": 3258 }, { "epoch": 0.41, "learning_rate": 0.0003330719405704438, "loss": 1.4635, "step": 3259 }, { "epoch": 0.41, "learning_rate": 0.00033297696328213214, "loss": 1.5008, "step": 3260 }, { "epoch": 0.41, "learning_rate": 0.00033288197253310726, "loss": 1.4544, "step": 3261 }, { "epoch": 0.41, "learning_rate": 0.00033278696833877874, "loss": 1.5386, "step": 3262 }, { "epoch": 0.41, "learning_rate": 0.00033269195071455854, "loss": 1.3978, "step": 3263 }, { "epoch": 0.41, "learning_rate": 0.0003325969196758605, "loss": 1.4431, "step": 3264 }, { "epoch": 0.41, "learning_rate": 0.0003325018752381007, "loss": 1.4765, "step": 3265 }, { "epoch": 0.41, "learning_rate": 0.0003324068174166977, "loss": 1.5397, "step": 3266 }, { "epoch": 0.41, "learning_rate": 0.0003323117462270719, "loss": 1.4937, "step": 3267 }, { "epoch": 0.41, "learning_rate": 0.0003322166616846458, "loss": 1.4085, "step": 3268 }, { "epoch": 0.41, "learning_rate": 0.0003321215638048445, "loss": 1.4679, "step": 3269 }, { "epoch": 0.41, "learning_rate": 0.00033202645260309484, "loss": 1.4062, "step": 3270 }, { "epoch": 0.41, "learning_rate": 0.0003319313280948259, "loss": 1.5476, "step": 3271 }, { "epoch": 0.41, "learning_rate": 0.0003318361902954692, "loss": 1.5015, "step": 3272 }, { "epoch": 0.41, "learning_rate": 0.00033174103922045813, "loss": 1.5034, "step": 3273 }, { "epoch": 0.41, "learning_rate": 0.00033164587488522816, "loss": 1.4361, "step": 3274 }, { "epoch": 0.41, "learning_rate": 0.0003315506973052174, "loss": 1.5098, "step": 3275 }, { "epoch": 0.41, "learning_rate": 0.0003314555064958654, "loss": 1.4733, "step": 3276 }, { "epoch": 0.41, "learning_rate": 0.0003313603024726145, "loss": 1.4524, "step": 3277 }, { "epoch": 0.41, "learning_rate": 0.0003312650852509089, "loss": 1.4231, "step": 3278 }, { "epoch": 0.41, "learning_rate": 0.0003311698548461947, "loss": 1.4935, "step": 3279 }, { "epoch": 0.41, "learning_rate": 0.0003310746112739207, "loss": 1.4736, "step": 3280 }, { "epoch": 0.41, "learning_rate": 0.00033097935454953737, "loss": 1.4024, "step": 3281 }, { "epoch": 0.41, "learning_rate": 0.0003308840846884975, "loss": 1.4918, "step": 3282 }, { "epoch": 0.41, "learning_rate": 0.000330788801706256, "loss": 1.4788, "step": 3283 }, { "epoch": 0.41, "learning_rate": 0.0003306935056182699, "loss": 1.4143, "step": 3284 }, { "epoch": 0.41, "learning_rate": 0.00033059819643999836, "loss": 1.435, "step": 3285 }, { "epoch": 0.41, "learning_rate": 0.0003305028741869026, "loss": 1.5693, "step": 3286 }, { "epoch": 0.41, "learning_rate": 0.00033040753887444605, "loss": 1.5271, "step": 3287 }, { "epoch": 0.41, "learning_rate": 0.0003303121905180942, "loss": 1.445, "step": 3288 }, { "epoch": 0.41, "learning_rate": 0.00033021682913331483, "loss": 1.4442, "step": 3289 }, { "epoch": 0.41, "learning_rate": 0.00033012145473557743, "loss": 1.4777, "step": 3290 }, { "epoch": 0.41, "learning_rate": 0.000330026067340354, "loss": 1.4849, "step": 3291 }, { "epoch": 0.41, "learning_rate": 0.00032993066696311845, "loss": 1.448, "step": 3292 }, { "epoch": 0.41, "learning_rate": 0.0003298352536193469, "loss": 1.4411, "step": 3293 }, { "epoch": 0.41, "learning_rate": 0.0003297398273245175, "loss": 1.4291, "step": 3294 }, { "epoch": 0.41, "learning_rate": 0.0003296443880941106, "loss": 1.4355, "step": 3295 }, { "epoch": 0.41, "learning_rate": 0.0003295489359436083, "loss": 1.4332, "step": 3296 }, { "epoch": 0.41, "learning_rate": 0.0003294534708884954, "loss": 1.5339, "step": 3297 }, { "epoch": 0.41, "learning_rate": 0.0003293579929442583, "loss": 1.3916, "step": 3298 }, { "epoch": 0.41, "learning_rate": 0.0003292625021263856, "loss": 1.5073, "step": 3299 }, { "epoch": 0.41, "learning_rate": 0.00032916699845036816, "loss": 1.4803, "step": 3300 }, { "epoch": 0.41, "eval_acceptance_rate": 0.6639133095741272, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0732978186234075, "eval_expected_tokens": 2.5916222732792593, "eval_loss": 1.4792804718017578, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.39790556831981483, "eval_runtime": 50.3918, "eval_samples_per_second": 9.664, "eval_steps_per_second": 0.318, "step": 3300 }, { "epoch": 0.41, "learning_rate": 0.0003290714819316987, "loss": 1.4599, "step": 3301 }, { "epoch": 0.41, "learning_rate": 0.00032897595258587223, "loss": 1.4359, "step": 3302 }, { "epoch": 0.41, "learning_rate": 0.0003288804104283857, "loss": 1.4738, "step": 3303 }, { "epoch": 0.41, "learning_rate": 0.00032878485547473813, "loss": 1.4543, "step": 3304 }, { "epoch": 0.41, "learning_rate": 0.0003286892877404307, "loss": 1.47, "step": 3305 }, { "epoch": 0.41, "learning_rate": 0.00032859370724096665, "loss": 1.5189, "step": 3306 }, { "epoch": 0.41, "learning_rate": 0.00032849811399185126, "loss": 1.4696, "step": 3307 }, { "epoch": 0.41, "learning_rate": 0.00032840250800859184, "loss": 1.5052, "step": 3308 }, { "epoch": 0.41, "learning_rate": 0.00032830688930669783, "loss": 1.3963, "step": 3309 }, { "epoch": 0.41, "learning_rate": 0.00032821125790168077, "loss": 1.4451, "step": 3310 }, { "epoch": 0.41, "learning_rate": 0.0003281156138090541, "loss": 1.4053, "step": 3311 }, { "epoch": 0.41, "learning_rate": 0.0003280199570443336, "loss": 1.4638, "step": 3312 }, { "epoch": 0.41, "learning_rate": 0.0003279242876230367, "loss": 1.4236, "step": 3313 }, { "epoch": 0.41, "learning_rate": 0.0003278286055606833, "loss": 1.4712, "step": 3314 }, { "epoch": 0.41, "learning_rate": 0.0003277329108727951, "loss": 1.4001, "step": 3315 }, { "epoch": 0.41, "learning_rate": 0.0003276372035748959, "loss": 1.4354, "step": 3316 }, { "epoch": 0.41, "learning_rate": 0.0003275414836825115, "loss": 1.4782, "step": 3317 }, { "epoch": 0.41, "learning_rate": 0.0003274457512111699, "loss": 1.4643, "step": 3318 }, { "epoch": 0.41, "learning_rate": 0.0003273500061764011, "loss": 1.4561, "step": 3319 }, { "epoch": 0.41, "learning_rate": 0.00032725424859373687, "loss": 1.4187, "step": 3320 }, { "epoch": 0.42, "learning_rate": 0.0003271584784787114, "loss": 1.4463, "step": 3321 }, { "epoch": 0.42, "learning_rate": 0.00032706269584686064, "loss": 1.4073, "step": 3322 }, { "epoch": 0.42, "learning_rate": 0.0003269669007137227, "loss": 1.4982, "step": 3323 }, { "epoch": 0.42, "learning_rate": 0.0003268710930948376, "loss": 1.4216, "step": 3324 }, { "epoch": 0.42, "learning_rate": 0.00032677527300574774, "loss": 1.4939, "step": 3325 }, { "epoch": 0.42, "learning_rate": 0.00032667944046199703, "loss": 1.3813, "step": 3326 }, { "epoch": 0.42, "learning_rate": 0.0003265835954791317, "loss": 1.4599, "step": 3327 }, { "epoch": 0.42, "learning_rate": 0.00032648773807269994, "loss": 1.4405, "step": 3328 }, { "epoch": 0.42, "learning_rate": 0.00032639186825825196, "loss": 1.3919, "step": 3329 }, { "epoch": 0.42, "learning_rate": 0.00032629598605133996, "loss": 1.4301, "step": 3330 }, { "epoch": 0.42, "learning_rate": 0.00032620009146751825, "loss": 1.4315, "step": 3331 }, { "epoch": 0.42, "learning_rate": 0.0003261041845223431, "loss": 1.3964, "step": 3332 }, { "epoch": 0.42, "learning_rate": 0.00032600826523137254, "loss": 1.4021, "step": 3333 }, { "epoch": 0.42, "learning_rate": 0.0003259123336101671, "loss": 1.4558, "step": 3334 }, { "epoch": 0.42, "learning_rate": 0.0003258163896742888, "loss": 1.4077, "step": 3335 }, { "epoch": 0.42, "learning_rate": 0.0003257204334393019, "loss": 1.4405, "step": 3336 }, { "epoch": 0.42, "learning_rate": 0.0003256244649207728, "loss": 1.53, "step": 3337 }, { "epoch": 0.42, "learning_rate": 0.0003255284841342697, "loss": 1.487, "step": 3338 }, { "epoch": 0.42, "learning_rate": 0.0003254324910953627, "loss": 1.4101, "step": 3339 }, { "epoch": 0.42, "learning_rate": 0.000325336485819624, "loss": 1.4326, "step": 3340 }, { "epoch": 0.42, "learning_rate": 0.000325240468322628, "loss": 1.3825, "step": 3341 }, { "epoch": 0.42, "learning_rate": 0.0003251444386199506, "loss": 1.4639, "step": 3342 }, { "epoch": 0.42, "learning_rate": 0.00032504839672717025, "loss": 1.4713, "step": 3343 }, { "epoch": 0.42, "learning_rate": 0.0003249523426598669, "loss": 1.4062, "step": 3344 }, { "epoch": 0.42, "learning_rate": 0.0003248562764336226, "loss": 1.4015, "step": 3345 }, { "epoch": 0.42, "learning_rate": 0.0003247601980640217, "loss": 1.4139, "step": 3346 }, { "epoch": 0.42, "learning_rate": 0.00032466410756665, "loss": 1.5089, "step": 3347 }, { "epoch": 0.42, "learning_rate": 0.0003245680049570957, "loss": 1.469, "step": 3348 }, { "epoch": 0.42, "learning_rate": 0.0003244718902509486, "loss": 1.441, "step": 3349 }, { "epoch": 0.42, "learning_rate": 0.00032437576346380076, "loss": 1.4028, "step": 3350 }, { "epoch": 0.42, "learning_rate": 0.00032427962461124615, "loss": 1.4311, "step": 3351 }, { "epoch": 0.42, "learning_rate": 0.0003241834737088804, "loss": 1.5134, "step": 3352 }, { "epoch": 0.42, "learning_rate": 0.00032408731077230164, "loss": 1.4548, "step": 3353 }, { "epoch": 0.42, "learning_rate": 0.0003239911358171095, "loss": 1.4097, "step": 3354 }, { "epoch": 0.42, "learning_rate": 0.0003238949488589056, "loss": 1.5104, "step": 3355 }, { "epoch": 0.42, "learning_rate": 0.0003237987499132937, "loss": 1.3748, "step": 3356 }, { "epoch": 0.42, "learning_rate": 0.0003237025389958795, "loss": 1.5141, "step": 3357 }, { "epoch": 0.42, "learning_rate": 0.00032360631612227044, "loss": 1.4761, "step": 3358 }, { "epoch": 0.42, "learning_rate": 0.000323510081308076, "loss": 1.511, "step": 3359 }, { "epoch": 0.42, "learning_rate": 0.00032341383456890773, "loss": 1.4138, "step": 3360 }, { "epoch": 0.42, "learning_rate": 0.00032331757592037887, "loss": 1.4938, "step": 3361 }, { "epoch": 0.42, "learning_rate": 0.00032322130537810475, "loss": 1.4421, "step": 3362 }, { "epoch": 0.42, "learning_rate": 0.0003231250229577027, "loss": 1.464, "step": 3363 }, { "epoch": 0.42, "learning_rate": 0.00032302872867479173, "loss": 1.4587, "step": 3364 }, { "epoch": 0.42, "learning_rate": 0.000322932422544993, "loss": 1.4532, "step": 3365 }, { "epoch": 0.42, "learning_rate": 0.0003228361045839295, "loss": 1.4507, "step": 3366 }, { "epoch": 0.42, "learning_rate": 0.00032273977480722614, "loss": 1.4502, "step": 3367 }, { "epoch": 0.42, "learning_rate": 0.0003226434332305098, "loss": 1.4964, "step": 3368 }, { "epoch": 0.42, "learning_rate": 0.00032254707986940917, "loss": 1.456, "step": 3369 }, { "epoch": 0.42, "learning_rate": 0.000322450714739555, "loss": 1.4682, "step": 3370 }, { "epoch": 0.42, "learning_rate": 0.00032235433785657975, "loss": 1.4417, "step": 3371 }, { "epoch": 0.42, "learning_rate": 0.0003222579492361179, "loss": 1.3482, "step": 3372 }, { "epoch": 0.42, "learning_rate": 0.000322161548893806, "loss": 1.4136, "step": 3373 }, { "epoch": 0.42, "learning_rate": 0.00032206513684528203, "loss": 1.3767, "step": 3374 }, { "epoch": 0.42, "learning_rate": 0.0003219687131061865, "loss": 1.4802, "step": 3375 }, { "epoch": 0.42, "learning_rate": 0.00032187227769216143, "loss": 1.4386, "step": 3376 }, { "epoch": 0.42, "learning_rate": 0.0003217758306188505, "loss": 1.4303, "step": 3377 }, { "epoch": 0.42, "learning_rate": 0.0003216793719019, "loss": 1.4101, "step": 3378 }, { "epoch": 0.42, "learning_rate": 0.00032158290155695734, "loss": 1.4462, "step": 3379 }, { "epoch": 0.42, "learning_rate": 0.0003214864195996723, "loss": 1.4625, "step": 3380 }, { "epoch": 0.42, "learning_rate": 0.00032138992604569645, "loss": 1.4392, "step": 3381 }, { "epoch": 0.42, "learning_rate": 0.00032129342091068315, "loss": 1.4271, "step": 3382 }, { "epoch": 0.42, "learning_rate": 0.00032119690421028764, "loss": 1.4146, "step": 3383 }, { "epoch": 0.42, "learning_rate": 0.00032110037596016713, "loss": 1.4329, "step": 3384 }, { "epoch": 0.42, "learning_rate": 0.0003210038361759807, "loss": 1.3998, "step": 3385 }, { "epoch": 0.42, "learning_rate": 0.00032090728487338914, "loss": 1.4522, "step": 3386 }, { "epoch": 0.42, "learning_rate": 0.0003208107220680553, "loss": 1.434, "step": 3387 }, { "epoch": 0.42, "learning_rate": 0.00032071414777564377, "loss": 1.4245, "step": 3388 }, { "epoch": 0.42, "learning_rate": 0.0003206175620118211, "loss": 1.4386, "step": 3389 }, { "epoch": 0.42, "learning_rate": 0.00032052096479225566, "loss": 1.4854, "step": 3390 }, { "epoch": 0.42, "learning_rate": 0.0003204243561326176, "loss": 1.6848, "step": 3391 }, { "epoch": 0.42, "learning_rate": 0.0003203277360485791, "loss": 1.5059, "step": 3392 }, { "epoch": 0.42, "learning_rate": 0.00032023110455581407, "loss": 1.4912, "step": 3393 }, { "epoch": 0.42, "learning_rate": 0.0003201344616699982, "loss": 1.4804, "step": 3394 }, { "epoch": 0.42, "learning_rate": 0.00032003780740680923, "loss": 1.4314, "step": 3395 }, { "epoch": 0.42, "learning_rate": 0.0003199411417819265, "loss": 1.4483, "step": 3396 }, { "epoch": 0.42, "learning_rate": 0.0003198444648110315, "loss": 1.4999, "step": 3397 }, { "epoch": 0.42, "learning_rate": 0.00031974777650980735, "loss": 1.4701, "step": 3398 }, { "epoch": 0.42, "learning_rate": 0.0003196510768939389, "loss": 1.4467, "step": 3399 }, { "epoch": 0.42, "learning_rate": 0.00031955436597911317, "loss": 1.5224, "step": 3400 }, { "epoch": 0.42, "eval_acceptance_rate": 0.664512038230896, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.075608483241117, "eval_expected_tokens": 2.594510604051396, "eval_loss": 1.474923014640808, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.39862765101284897, "eval_runtime": 50.35, "eval_samples_per_second": 9.672, "eval_steps_per_second": 0.318, "step": 3400 }, { "epoch": 0.43, "learning_rate": 0.0003194576437810187, "loss": 1.4209, "step": 3401 }, { "epoch": 0.43, "learning_rate": 0.0003193609103153461, "loss": 1.4476, "step": 3402 }, { "epoch": 0.43, "learning_rate": 0.0003192641655977876, "loss": 1.4328, "step": 3403 }, { "epoch": 0.43, "learning_rate": 0.00031916740964403736, "loss": 1.4174, "step": 3404 }, { "epoch": 0.43, "learning_rate": 0.00031907064246979136, "loss": 1.4016, "step": 3405 }, { "epoch": 0.43, "learning_rate": 0.00031897386409074737, "loss": 1.4293, "step": 3406 }, { "epoch": 0.43, "learning_rate": 0.0003188770745226051, "loss": 1.3835, "step": 3407 }, { "epoch": 0.43, "learning_rate": 0.0003187802737810659, "loss": 1.4912, "step": 3408 }, { "epoch": 0.43, "learning_rate": 0.00031868346188183296, "loss": 1.483, "step": 3409 }, { "epoch": 0.43, "learning_rate": 0.0003185866388406115, "loss": 1.4741, "step": 3410 }, { "epoch": 0.43, "learning_rate": 0.0003184898046731082, "loss": 1.3867, "step": 3411 }, { "epoch": 0.43, "learning_rate": 0.0003183929593950317, "loss": 1.4124, "step": 3412 }, { "epoch": 0.43, "learning_rate": 0.0003182961030220926, "loss": 1.5512, "step": 3413 }, { "epoch": 0.43, "learning_rate": 0.0003181992355700032, "loss": 1.4588, "step": 3414 }, { "epoch": 0.43, "learning_rate": 0.0003181023570544773, "loss": 1.4104, "step": 3415 }, { "epoch": 0.43, "learning_rate": 0.00031800546749123106, "loss": 1.4213, "step": 3416 }, { "epoch": 0.43, "learning_rate": 0.00031790856689598197, "loss": 1.4769, "step": 3417 }, { "epoch": 0.43, "learning_rate": 0.0003178116552844494, "loss": 1.3923, "step": 3418 }, { "epoch": 0.43, "learning_rate": 0.0003177147326723547, "loss": 1.3949, "step": 3419 }, { "epoch": 0.43, "learning_rate": 0.000317617799075421, "loss": 1.4039, "step": 3420 }, { "epoch": 0.43, "learning_rate": 0.00031752085450937267, "loss": 1.4229, "step": 3421 }, { "epoch": 0.43, "learning_rate": 0.00031742389898993675, "loss": 1.3994, "step": 3422 }, { "epoch": 0.43, "learning_rate": 0.00031732693253284125, "loss": 1.3904, "step": 3423 }, { "epoch": 0.43, "learning_rate": 0.0003172299551538164, "loss": 1.4708, "step": 3424 }, { "epoch": 0.43, "learning_rate": 0.0003171329668685942, "loss": 1.4371, "step": 3425 }, { "epoch": 0.43, "learning_rate": 0.0003170359676929082, "loss": 1.4292, "step": 3426 }, { "epoch": 0.43, "learning_rate": 0.0003169389576424938, "loss": 1.3833, "step": 3427 }, { "epoch": 0.43, "learning_rate": 0.00031684193673308823, "loss": 1.4055, "step": 3428 }, { "epoch": 0.43, "learning_rate": 0.0003167449049804306, "loss": 1.4483, "step": 3429 }, { "epoch": 0.43, "learning_rate": 0.0003166478624002613, "loss": 1.4389, "step": 3430 }, { "epoch": 0.43, "learning_rate": 0.00031655080900832314, "loss": 1.3677, "step": 3431 }, { "epoch": 0.43, "learning_rate": 0.0003164537448203601, "loss": 1.3359, "step": 3432 }, { "epoch": 0.43, "learning_rate": 0.00031635666985211817, "loss": 1.4021, "step": 3433 }, { "epoch": 0.43, "learning_rate": 0.0003162595841193453, "loss": 1.3977, "step": 3434 }, { "epoch": 0.43, "learning_rate": 0.00031616248763779065, "loss": 1.4059, "step": 3435 }, { "epoch": 0.43, "learning_rate": 0.00031606538042320566, "loss": 1.4276, "step": 3436 }, { "epoch": 0.43, "learning_rate": 0.00031596826249134324, "loss": 1.4772, "step": 3437 }, { "epoch": 0.43, "learning_rate": 0.00031587113385795804, "loss": 1.4639, "step": 3438 }, { "epoch": 0.43, "learning_rate": 0.00031577399453880644, "loss": 1.4442, "step": 3439 }, { "epoch": 0.43, "learning_rate": 0.0003156768445496467, "loss": 1.4378, "step": 3440 }, { "epoch": 0.43, "learning_rate": 0.00031557968390623874, "loss": 1.4459, "step": 3441 }, { "epoch": 0.43, "learning_rate": 0.0003154825126243441, "loss": 1.4276, "step": 3442 }, { "epoch": 0.43, "learning_rate": 0.0003153853307197261, "loss": 1.4493, "step": 3443 }, { "epoch": 0.43, "learning_rate": 0.00031528813820814994, "loss": 1.4607, "step": 3444 }, { "epoch": 0.43, "learning_rate": 0.00031519093510538226, "loss": 1.4216, "step": 3445 }, { "epoch": 0.43, "learning_rate": 0.00031509372142719165, "loss": 1.5227, "step": 3446 }, { "epoch": 0.43, "learning_rate": 0.0003149964971893484, "loss": 1.4283, "step": 3447 }, { "epoch": 0.43, "learning_rate": 0.0003148992624076243, "loss": 1.4388, "step": 3448 }, { "epoch": 0.43, "learning_rate": 0.00031480201709779316, "loss": 1.4285, "step": 3449 }, { "epoch": 0.43, "learning_rate": 0.00031470476127563017, "loss": 1.479, "step": 3450 }, { "epoch": 0.43, "learning_rate": 0.0003146074949569126, "loss": 1.4696, "step": 3451 }, { "epoch": 0.43, "learning_rate": 0.000314510218157419, "loss": 1.4539, "step": 3452 }, { "epoch": 0.43, "learning_rate": 0.00031441293089292993, "loss": 1.4847, "step": 3453 }, { "epoch": 0.43, "learning_rate": 0.0003143156331792276, "loss": 1.3846, "step": 3454 }, { "epoch": 0.43, "learning_rate": 0.0003142183250320957, "loss": 1.4044, "step": 3455 }, { "epoch": 0.43, "learning_rate": 0.00031412100646732006, "loss": 1.3819, "step": 3456 }, { "epoch": 0.43, "learning_rate": 0.0003140236775006877, "loss": 1.4151, "step": 3457 }, { "epoch": 0.43, "learning_rate": 0.00031392633814798755, "loss": 1.4348, "step": 3458 }, { "epoch": 0.43, "learning_rate": 0.0003138289884250104, "loss": 1.3988, "step": 3459 }, { "epoch": 0.43, "learning_rate": 0.0003137316283475483, "loss": 1.3655, "step": 3460 }, { "epoch": 0.43, "learning_rate": 0.00031363425793139545, "loss": 1.4162, "step": 3461 }, { "epoch": 0.43, "learning_rate": 0.00031353687719234735, "loss": 1.4914, "step": 3462 }, { "epoch": 0.43, "learning_rate": 0.0003134394861462014, "loss": 1.3695, "step": 3463 }, { "epoch": 0.43, "learning_rate": 0.00031334208480875657, "loss": 1.4351, "step": 3464 }, { "epoch": 0.43, "learning_rate": 0.00031324467319581356, "loss": 1.4335, "step": 3465 }, { "epoch": 0.43, "learning_rate": 0.00031314725132317473, "loss": 1.4037, "step": 3466 }, { "epoch": 0.43, "learning_rate": 0.000313049819206644, "loss": 1.4365, "step": 3467 }, { "epoch": 0.43, "learning_rate": 0.000312952376862027, "loss": 1.4242, "step": 3468 }, { "epoch": 0.43, "learning_rate": 0.0003128549243051312, "loss": 1.4703, "step": 3469 }, { "epoch": 0.43, "learning_rate": 0.00031275746155176544, "loss": 1.422, "step": 3470 }, { "epoch": 0.43, "learning_rate": 0.0003126599886177404, "loss": 1.4437, "step": 3471 }, { "epoch": 0.43, "learning_rate": 0.00031256250551886836, "loss": 1.4455, "step": 3472 }, { "epoch": 0.43, "learning_rate": 0.0003124650122709634, "loss": 1.3889, "step": 3473 }, { "epoch": 0.43, "learning_rate": 0.0003123675088898408, "loss": 1.4173, "step": 3474 }, { "epoch": 0.43, "learning_rate": 0.00031226999539131793, "loss": 1.4244, "step": 3475 }, { "epoch": 0.43, "learning_rate": 0.0003121724717912138, "loss": 1.5111, "step": 3476 }, { "epoch": 0.43, "learning_rate": 0.0003120749381053486, "loss": 1.459, "step": 3477 }, { "epoch": 0.43, "learning_rate": 0.00031197739434954473, "loss": 1.3859, "step": 3478 }, { "epoch": 0.43, "learning_rate": 0.00031187984053962584, "loss": 1.3903, "step": 3479 }, { "epoch": 0.43, "learning_rate": 0.0003117822766914174, "loss": 1.4684, "step": 3480 }, { "epoch": 0.44, "learning_rate": 0.00031168470282074643, "loss": 1.3994, "step": 3481 }, { "epoch": 0.44, "learning_rate": 0.00031158711894344157, "loss": 1.4169, "step": 3482 }, { "epoch": 0.44, "learning_rate": 0.000311489525075333, "loss": 1.4564, "step": 3483 }, { "epoch": 0.44, "learning_rate": 0.0003113919212322528, "loss": 1.3753, "step": 3484 }, { "epoch": 0.44, "learning_rate": 0.0003112943074300344, "loss": 1.4256, "step": 3485 }, { "epoch": 0.44, "learning_rate": 0.00031119668368451294, "loss": 1.4125, "step": 3486 }, { "epoch": 0.44, "learning_rate": 0.0003110990500115251, "loss": 1.609, "step": 3487 }, { "epoch": 0.44, "learning_rate": 0.00031100140642690937, "loss": 1.4783, "step": 3488 }, { "epoch": 0.44, "learning_rate": 0.0003109037529465056, "loss": 1.5161, "step": 3489 }, { "epoch": 0.44, "learning_rate": 0.00031080608958615545, "loss": 1.4571, "step": 3490 }, { "epoch": 0.44, "learning_rate": 0.0003107084163617021, "loss": 1.3861, "step": 3491 }, { "epoch": 0.44, "learning_rate": 0.00031061073328899025, "loss": 1.4327, "step": 3492 }, { "epoch": 0.44, "learning_rate": 0.0003105130403838663, "loss": 1.4836, "step": 3493 }, { "epoch": 0.44, "learning_rate": 0.0003104153376621782, "loss": 1.5222, "step": 3494 }, { "epoch": 0.44, "learning_rate": 0.00031031762513977556, "loss": 1.378, "step": 3495 }, { "epoch": 0.44, "learning_rate": 0.0003102199028325095, "loss": 1.4392, "step": 3496 }, { "epoch": 0.44, "learning_rate": 0.00031012217075623274, "loss": 1.4442, "step": 3497 }, { "epoch": 0.44, "learning_rate": 0.0003100244289267996, "loss": 1.4171, "step": 3498 }, { "epoch": 0.44, "learning_rate": 0.00030992667736006596, "loss": 1.3891, "step": 3499 }, { "epoch": 0.44, "learning_rate": 0.00030982891607188946, "loss": 1.3762, "step": 3500 }, { "epoch": 0.44, "eval_acceptance_rate": 0.6683300733566284, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.0904196564329487, "eval_expected_tokens": 2.613024570541186, "eval_loss": 1.4686557054519653, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.40325614263529647, "eval_runtime": 50.3925, "eval_samples_per_second": 9.664, "eval_steps_per_second": 0.318, "step": 3500 }, { "epoch": 0.44, "learning_rate": 0.00030973114507812903, "loss": 1.4473, "step": 3501 }, { "epoch": 0.44, "learning_rate": 0.00030963336439464523, "loss": 1.4655, "step": 3502 }, { "epoch": 0.44, "learning_rate": 0.00030953557403730044, "loss": 1.457, "step": 3503 }, { "epoch": 0.44, "learning_rate": 0.00030943777402195826, "loss": 1.4489, "step": 3504 }, { "epoch": 0.44, "learning_rate": 0.0003093399643644842, "loss": 1.4614, "step": 3505 }, { "epoch": 0.44, "learning_rate": 0.00030924214508074504, "loss": 1.3185, "step": 3506 }, { "epoch": 0.44, "learning_rate": 0.00030914431618660936, "loss": 1.445, "step": 3507 }, { "epoch": 0.44, "learning_rate": 0.0003090464776979471, "loss": 1.4233, "step": 3508 }, { "epoch": 0.44, "learning_rate": 0.0003089486296306299, "loss": 1.4984, "step": 3509 }, { "epoch": 0.44, "learning_rate": 0.00030885077200053083, "loss": 1.4205, "step": 3510 }, { "epoch": 0.44, "learning_rate": 0.0003087529048235246, "loss": 1.4286, "step": 3511 }, { "epoch": 0.44, "learning_rate": 0.0003086550281154875, "loss": 1.3705, "step": 3512 }, { "epoch": 0.44, "learning_rate": 0.00030855714189229733, "loss": 1.4162, "step": 3513 }, { "epoch": 0.44, "learning_rate": 0.00030845924616983335, "loss": 1.4998, "step": 3514 }, { "epoch": 0.44, "learning_rate": 0.0003083613409639764, "loss": 1.4016, "step": 3515 }, { "epoch": 0.44, "learning_rate": 0.0003082634262906089, "loss": 1.3742, "step": 3516 }, { "epoch": 0.44, "learning_rate": 0.0003081655021656149, "loss": 1.4072, "step": 3517 }, { "epoch": 0.44, "learning_rate": 0.0003080675686048798, "loss": 1.4095, "step": 3518 }, { "epoch": 0.44, "learning_rate": 0.00030796962562429055, "loss": 1.4448, "step": 3519 }, { "epoch": 0.44, "learning_rate": 0.0003078716732397358, "loss": 1.3338, "step": 3520 }, { "epoch": 0.44, "learning_rate": 0.00030777371146710545, "loss": 1.4493, "step": 3521 }, { "epoch": 0.44, "learning_rate": 0.00030767574032229136, "loss": 1.4297, "step": 3522 }, { "epoch": 0.44, "learning_rate": 0.00030757775982118634, "loss": 1.3967, "step": 3523 }, { "epoch": 0.44, "learning_rate": 0.00030747976997968514, "loss": 1.4374, "step": 3524 }, { "epoch": 0.44, "learning_rate": 0.0003073817708136839, "loss": 1.4721, "step": 3525 }, { "epoch": 0.44, "learning_rate": 0.0003072837623390803, "loss": 1.457, "step": 3526 }, { "epoch": 0.44, "learning_rate": 0.0003071857445717734, "loss": 1.4389, "step": 3527 }, { "epoch": 0.44, "learning_rate": 0.00030708771752766396, "loss": 1.3523, "step": 3528 }, { "epoch": 0.44, "learning_rate": 0.00030698968122265416, "loss": 1.4122, "step": 3529 }, { "epoch": 0.44, "learning_rate": 0.0003068916356726475, "loss": 1.4686, "step": 3530 }, { "epoch": 0.44, "learning_rate": 0.0003067935808935494, "loss": 1.4275, "step": 3531 }, { "epoch": 0.44, "learning_rate": 0.0003066955169012664, "loss": 1.4097, "step": 3532 }, { "epoch": 0.44, "learning_rate": 0.0003065974437117066, "loss": 1.3922, "step": 3533 }, { "epoch": 0.44, "learning_rate": 0.00030649936134077984, "loss": 1.4617, "step": 3534 }, { "epoch": 0.44, "learning_rate": 0.0003064012698043971, "loss": 1.404, "step": 3535 }, { "epoch": 0.44, "learning_rate": 0.00030630316911847114, "loss": 1.4192, "step": 3536 }, { "epoch": 0.44, "learning_rate": 0.00030620505929891597, "loss": 1.3752, "step": 3537 }, { "epoch": 0.44, "learning_rate": 0.0003061069403616472, "loss": 1.3919, "step": 3538 }, { "epoch": 0.44, "learning_rate": 0.00030600881232258206, "loss": 1.3312, "step": 3539 }, { "epoch": 0.44, "learning_rate": 0.0003059106751976389, "loss": 1.3795, "step": 3540 }, { "epoch": 0.44, "learning_rate": 0.00030581252900273786, "loss": 1.4785, "step": 3541 }, { "epoch": 0.44, "learning_rate": 0.00030571437375380043, "loss": 1.4118, "step": 3542 }, { "epoch": 0.44, "learning_rate": 0.00030561620946674954, "loss": 1.398, "step": 3543 }, { "epoch": 0.44, "learning_rate": 0.0003055180361575097, "loss": 1.434, "step": 3544 }, { "epoch": 0.44, "learning_rate": 0.00030541985384200683, "loss": 1.4657, "step": 3545 }, { "epoch": 0.44, "learning_rate": 0.0003053216625361681, "loss": 1.3389, "step": 3546 }, { "epoch": 0.44, "learning_rate": 0.0003052234622559226, "loss": 1.4451, "step": 3547 }, { "epoch": 0.44, "learning_rate": 0.0003051252530172003, "loss": 1.3536, "step": 3548 }, { "epoch": 0.44, "learning_rate": 0.00030502703483593316, "loss": 1.415, "step": 3549 }, { "epoch": 0.44, "learning_rate": 0.0003049288077280543, "loss": 1.42, "step": 3550 }, { "epoch": 0.44, "learning_rate": 0.0003048305717094983, "loss": 1.4539, "step": 3551 }, { "epoch": 0.44, "learning_rate": 0.00030473232679620126, "loss": 1.4281, "step": 3552 }, { "epoch": 0.44, "learning_rate": 0.00030463407300410073, "loss": 1.4391, "step": 3553 }, { "epoch": 0.44, "learning_rate": 0.0003045358103491357, "loss": 1.363, "step": 3554 }, { "epoch": 0.44, "learning_rate": 0.0003044375388472464, "loss": 1.3924, "step": 3555 }, { "epoch": 0.44, "learning_rate": 0.00030433925851437477, "loss": 1.4166, "step": 3556 }, { "epoch": 0.44, "learning_rate": 0.0003042409693664641, "loss": 1.4058, "step": 3557 }, { "epoch": 0.44, "learning_rate": 0.00030414267141945896, "loss": 1.392, "step": 3558 }, { "epoch": 0.44, "learning_rate": 0.0003040443646893056, "loss": 1.4083, "step": 3559 }, { "epoch": 0.45, "learning_rate": 0.00030394604919195154, "loss": 1.4035, "step": 3560 }, { "epoch": 0.45, "learning_rate": 0.0003038477249433457, "loss": 1.3476, "step": 3561 }, { "epoch": 0.45, "learning_rate": 0.0003037493919594385, "loss": 1.4413, "step": 3562 }, { "epoch": 0.45, "learning_rate": 0.00030365105025618165, "loss": 1.3947, "step": 3563 }, { "epoch": 0.45, "learning_rate": 0.0003035526998495285, "loss": 1.4451, "step": 3564 }, { "epoch": 0.45, "learning_rate": 0.00030345434075543357, "loss": 1.4085, "step": 3565 }, { "epoch": 0.45, "learning_rate": 0.000303355972989853, "loss": 1.3497, "step": 3566 }, { "epoch": 0.45, "learning_rate": 0.0003032575965687442, "loss": 1.4159, "step": 3567 }, { "epoch": 0.45, "learning_rate": 0.0003031592115080659, "loss": 1.431, "step": 3568 }, { "epoch": 0.45, "learning_rate": 0.00030306081782377855, "loss": 1.3704, "step": 3569 }, { "epoch": 0.45, "learning_rate": 0.0003029624155318436, "loss": 1.3449, "step": 3570 }, { "epoch": 0.45, "learning_rate": 0.00030286400464822426, "loss": 1.4628, "step": 3571 }, { "epoch": 0.45, "learning_rate": 0.0003027655851888849, "loss": 1.3916, "step": 3572 }, { "epoch": 0.45, "learning_rate": 0.0003026671571697913, "loss": 1.4384, "step": 3573 }, { "epoch": 0.45, "learning_rate": 0.0003025687206069108, "loss": 1.4354, "step": 3574 }, { "epoch": 0.45, "learning_rate": 0.0003024702755162119, "loss": 1.4, "step": 3575 }, { "epoch": 0.45, "learning_rate": 0.00030237182191366463, "loss": 1.4095, "step": 3576 }, { "epoch": 0.45, "learning_rate": 0.00030227335981524036, "loss": 1.4393, "step": 3577 }, { "epoch": 0.45, "learning_rate": 0.00030217488923691186, "loss": 1.3108, "step": 3578 }, { "epoch": 0.45, "learning_rate": 0.00030207641019465323, "loss": 1.4779, "step": 3579 }, { "epoch": 0.45, "learning_rate": 0.0003019779227044398, "loss": 1.4296, "step": 3580 }, { "epoch": 0.45, "learning_rate": 0.00030187942678224877, "loss": 1.4421, "step": 3581 }, { "epoch": 0.45, "learning_rate": 0.0003017809224440581, "loss": 1.3802, "step": 3582 }, { "epoch": 0.45, "learning_rate": 0.00030168240970584754, "loss": 1.4234, "step": 3583 }, { "epoch": 0.45, "learning_rate": 0.0003015838885835981, "loss": 1.3353, "step": 3584 }, { "epoch": 0.45, "learning_rate": 0.0003014853590932918, "loss": 1.3808, "step": 3585 }, { "epoch": 0.45, "learning_rate": 0.00030138682125091266, "loss": 1.3844, "step": 3586 }, { "epoch": 0.45, "learning_rate": 0.00030128827507244554, "loss": 1.4615, "step": 3587 }, { "epoch": 0.45, "learning_rate": 0.00030118972057387687, "loss": 1.4131, "step": 3588 }, { "epoch": 0.45, "learning_rate": 0.0003010911577711944, "loss": 1.3856, "step": 3589 }, { "epoch": 0.45, "learning_rate": 0.0003009925866803872, "loss": 1.4125, "step": 3590 }, { "epoch": 0.45, "learning_rate": 0.0003008940073174457, "loss": 1.3886, "step": 3591 }, { "epoch": 0.45, "learning_rate": 0.0003007954196983616, "loss": 1.4117, "step": 3592 }, { "epoch": 0.45, "learning_rate": 0.0003006968238391281, "loss": 1.4447, "step": 3593 }, { "epoch": 0.45, "learning_rate": 0.0003005982197557397, "loss": 1.4354, "step": 3594 }, { "epoch": 0.45, "learning_rate": 0.00030049960746419204, "loss": 1.3971, "step": 3595 }, { "epoch": 0.45, "learning_rate": 0.0003004009869804823, "loss": 1.3506, "step": 3596 }, { "epoch": 0.45, "learning_rate": 0.0003003023583206089, "loss": 1.4353, "step": 3597 }, { "epoch": 0.45, "learning_rate": 0.0003002037215005716, "loss": 1.4701, "step": 3598 }, { "epoch": 0.45, "learning_rate": 0.0003001050765363716, "loss": 1.3748, "step": 3599 }, { "epoch": 0.45, "learning_rate": 0.00030000642344401115, "loss": 1.4012, "step": 3600 }, { "epoch": 0.45, "eval_acceptance_rate": 0.6719171404838562, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1044556319144987, "eval_expected_tokens": 2.630569539893123, "eval_loss": 1.4508709907531738, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4076423849732807, "eval_runtime": 50.2593, "eval_samples_per_second": 9.69, "eval_steps_per_second": 0.318, "step": 3600 }, { "epoch": 0.45, "learning_rate": 0.00029990776223949404, "loss": 1.405, "step": 3601 }, { "epoch": 0.45, "learning_rate": 0.00029980909293882537, "loss": 1.3746, "step": 3602 }, { "epoch": 0.45, "learning_rate": 0.00029971041555801147, "loss": 1.4854, "step": 3603 }, { "epoch": 0.45, "learning_rate": 0.00029961173011306, "loss": 1.3599, "step": 3604 }, { "epoch": 0.45, "learning_rate": 0.0002995130366199799, "loss": 1.3897, "step": 3605 }, { "epoch": 0.45, "learning_rate": 0.00029941433509478153, "loss": 1.3232, "step": 3606 }, { "epoch": 0.45, "learning_rate": 0.00029931562555347645, "loss": 1.3669, "step": 3607 }, { "epoch": 0.45, "learning_rate": 0.00029921690801207757, "loss": 1.3565, "step": 3608 }, { "epoch": 0.45, "learning_rate": 0.000299118182486599, "loss": 1.4488, "step": 3609 }, { "epoch": 0.45, "learning_rate": 0.0002990194489930563, "loss": 1.3892, "step": 3610 }, { "epoch": 0.45, "learning_rate": 0.0002989207075474662, "loss": 1.3456, "step": 3611 }, { "epoch": 0.45, "learning_rate": 0.0002988219581658468, "loss": 1.3764, "step": 3612 }, { "epoch": 0.45, "learning_rate": 0.00029872320086421744, "loss": 1.4506, "step": 3613 }, { "epoch": 0.45, "learning_rate": 0.0002986244356585987, "loss": 1.3036, "step": 3614 }, { "epoch": 0.45, "learning_rate": 0.0002985256625650125, "loss": 1.3922, "step": 3615 }, { "epoch": 0.45, "learning_rate": 0.00029842688159948224, "loss": 1.4611, "step": 3616 }, { "epoch": 0.45, "learning_rate": 0.0002983280927780321, "loss": 1.4393, "step": 3617 }, { "epoch": 0.45, "learning_rate": 0.000298229296116688, "loss": 1.4094, "step": 3618 }, { "epoch": 0.45, "learning_rate": 0.00029813049163147687, "loss": 1.4453, "step": 3619 }, { "epoch": 0.45, "learning_rate": 0.0002980316793384271, "loss": 1.408, "step": 3620 }, { "epoch": 0.45, "learning_rate": 0.00029793285925356825, "loss": 1.3965, "step": 3621 }, { "epoch": 0.45, "learning_rate": 0.00029783403139293103, "loss": 1.4167, "step": 3622 }, { "epoch": 0.45, "learning_rate": 0.0002977351957725476, "loss": 1.4515, "step": 3623 }, { "epoch": 0.45, "learning_rate": 0.0002976363524084512, "loss": 1.3472, "step": 3624 }, { "epoch": 0.45, "learning_rate": 0.00029753750131667666, "loss": 1.4326, "step": 3625 }, { "epoch": 0.45, "learning_rate": 0.00029743864251325963, "loss": 1.3422, "step": 3626 }, { "epoch": 0.45, "learning_rate": 0.0002973397760142372, "loss": 1.4238, "step": 3627 }, { "epoch": 0.45, "learning_rate": 0.00029724090183564793, "loss": 1.4279, "step": 3628 }, { "epoch": 0.45, "learning_rate": 0.0002971420199935312, "loss": 1.4307, "step": 3629 }, { "epoch": 0.45, "learning_rate": 0.00029704313050392785, "loss": 1.4046, "step": 3630 }, { "epoch": 0.45, "learning_rate": 0.0002969442333828801, "loss": 1.4512, "step": 3631 }, { "epoch": 0.45, "learning_rate": 0.0002968453286464312, "loss": 1.3922, "step": 3632 }, { "epoch": 0.45, "learning_rate": 0.00029674641631062564, "loss": 1.445, "step": 3633 }, { "epoch": 0.45, "learning_rate": 0.0002966474963915093, "loss": 1.3641, "step": 3634 }, { "epoch": 0.45, "learning_rate": 0.0002965485689051292, "loss": 1.4179, "step": 3635 }, { "epoch": 0.45, "learning_rate": 0.0002964496338675335, "loss": 1.3523, "step": 3636 }, { "epoch": 0.45, "learning_rate": 0.0002963506912947717, "loss": 1.3943, "step": 3637 }, { "epoch": 0.45, "learning_rate": 0.0002962517412028946, "loss": 1.4373, "step": 3638 }, { "epoch": 0.45, "learning_rate": 0.00029615278360795396, "loss": 1.3625, "step": 3639 }, { "epoch": 0.46, "learning_rate": 0.0002960538185260029, "loss": 1.3087, "step": 3640 }, { "epoch": 0.46, "learning_rate": 0.00029595484597309585, "loss": 1.497, "step": 3641 }, { "epoch": 0.46, "learning_rate": 0.00029585586596528844, "loss": 1.4048, "step": 3642 }, { "epoch": 0.46, "learning_rate": 0.0002957568785186373, "loss": 1.4655, "step": 3643 }, { "epoch": 0.46, "learning_rate": 0.0002956578836492003, "loss": 1.417, "step": 3644 }, { "epoch": 0.46, "learning_rate": 0.0002955588813730369, "loss": 1.3411, "step": 3645 }, { "epoch": 0.46, "learning_rate": 0.00029545987170620725, "loss": 1.4314, "step": 3646 }, { "epoch": 0.46, "learning_rate": 0.000295360854664773, "loss": 1.3258, "step": 3647 }, { "epoch": 0.46, "learning_rate": 0.000295261830264797, "loss": 1.3651, "step": 3648 }, { "epoch": 0.46, "learning_rate": 0.00029516279852234305, "loss": 1.3687, "step": 3649 }, { "epoch": 0.46, "learning_rate": 0.0002950637594534765, "loss": 1.4079, "step": 3650 }, { "epoch": 0.46, "learning_rate": 0.00029496471307426347, "loss": 1.4182, "step": 3651 }, { "epoch": 0.46, "learning_rate": 0.0002948656594007717, "loss": 1.3865, "step": 3652 }, { "epoch": 0.46, "learning_rate": 0.0002947665984490697, "loss": 1.3718, "step": 3653 }, { "epoch": 0.46, "learning_rate": 0.00029466753023522755, "loss": 1.4039, "step": 3654 }, { "epoch": 0.46, "learning_rate": 0.0002945684547753163, "loss": 1.4378, "step": 3655 }, { "epoch": 0.46, "learning_rate": 0.00029446937208540806, "loss": 1.3841, "step": 3656 }, { "epoch": 0.46, "learning_rate": 0.0002943702821815764, "loss": 1.3952, "step": 3657 }, { "epoch": 0.46, "learning_rate": 0.0002942711850798959, "loss": 1.4129, "step": 3658 }, { "epoch": 0.46, "learning_rate": 0.0002941720807964421, "loss": 1.3748, "step": 3659 }, { "epoch": 0.46, "learning_rate": 0.00029407296934729225, "loss": 1.4084, "step": 3660 }, { "epoch": 0.46, "learning_rate": 0.00029397385074852425, "loss": 1.4795, "step": 3661 }, { "epoch": 0.46, "learning_rate": 0.0002938747250162173, "loss": 1.3446, "step": 3662 }, { "epoch": 0.46, "learning_rate": 0.00029377559216645197, "loss": 1.384, "step": 3663 }, { "epoch": 0.46, "learning_rate": 0.00029367645221530967, "loss": 1.3865, "step": 3664 }, { "epoch": 0.46, "learning_rate": 0.0002935773051788732, "loss": 1.3388, "step": 3665 }, { "epoch": 0.46, "learning_rate": 0.0002934781510732264, "loss": 1.4152, "step": 3666 }, { "epoch": 0.46, "learning_rate": 0.0002933789899144543, "loss": 1.4161, "step": 3667 }, { "epoch": 0.46, "learning_rate": 0.00029327982171864286, "loss": 1.4086, "step": 3668 }, { "epoch": 0.46, "learning_rate": 0.0002931806465018797, "loss": 1.3676, "step": 3669 }, { "epoch": 0.46, "learning_rate": 0.000293081464280253, "loss": 1.4119, "step": 3670 }, { "epoch": 0.46, "learning_rate": 0.0002929822750698524, "loss": 1.4043, "step": 3671 }, { "epoch": 0.46, "learning_rate": 0.0002928830788867686, "loss": 1.4195, "step": 3672 }, { "epoch": 0.46, "learning_rate": 0.00029278387574709346, "loss": 1.4003, "step": 3673 }, { "epoch": 0.46, "learning_rate": 0.00029268466566691995, "loss": 1.3544, "step": 3674 }, { "epoch": 0.46, "learning_rate": 0.000292585448662342, "loss": 1.3844, "step": 3675 }, { "epoch": 0.46, "learning_rate": 0.00029248622474945516, "loss": 1.3864, "step": 3676 }, { "epoch": 0.46, "learning_rate": 0.0002923869939443554, "loss": 1.4482, "step": 3677 }, { "epoch": 0.46, "learning_rate": 0.00029228775626314034, "loss": 1.4918, "step": 3678 }, { "epoch": 0.46, "learning_rate": 0.0002921885117219086, "loss": 1.3534, "step": 3679 }, { "epoch": 0.46, "learning_rate": 0.00029208926033675957, "loss": 1.4105, "step": 3680 }, { "epoch": 0.46, "learning_rate": 0.00029199000212379447, "loss": 1.3685, "step": 3681 }, { "epoch": 0.46, "learning_rate": 0.0002918907370991149, "loss": 1.4224, "step": 3682 }, { "epoch": 0.46, "learning_rate": 0.00029179146527882384, "loss": 1.3795, "step": 3683 }, { "epoch": 0.46, "learning_rate": 0.00029169218667902556, "loss": 1.3931, "step": 3684 }, { "epoch": 0.46, "learning_rate": 0.00029159290131582517, "loss": 1.3346, "step": 3685 }, { "epoch": 0.46, "learning_rate": 0.00029149360920532896, "loss": 1.4137, "step": 3686 }, { "epoch": 0.46, "learning_rate": 0.00029139431036364445, "loss": 1.3764, "step": 3687 }, { "epoch": 0.46, "learning_rate": 0.00029129500480687995, "loss": 1.3687, "step": 3688 }, { "epoch": 0.46, "learning_rate": 0.000291195692551145, "loss": 1.4118, "step": 3689 }, { "epoch": 0.46, "learning_rate": 0.0002910963736125505, "loss": 1.3402, "step": 3690 }, { "epoch": 0.46, "learning_rate": 0.00029099704800720805, "loss": 1.4444, "step": 3691 }, { "epoch": 0.46, "learning_rate": 0.00029089771575123045, "loss": 1.4201, "step": 3692 }, { "epoch": 0.46, "learning_rate": 0.0002907983768607317, "loss": 1.4153, "step": 3693 }, { "epoch": 0.46, "learning_rate": 0.00029069903135182674, "loss": 1.3854, "step": 3694 }, { "epoch": 0.46, "learning_rate": 0.0002905996792406315, "loss": 1.3425, "step": 3695 }, { "epoch": 0.46, "learning_rate": 0.00029050032054326333, "loss": 1.3425, "step": 3696 }, { "epoch": 0.46, "learning_rate": 0.0002904009552758403, "loss": 1.3611, "step": 3697 }, { "epoch": 0.46, "learning_rate": 0.0002903015834544817, "loss": 1.3628, "step": 3698 }, { "epoch": 0.46, "learning_rate": 0.0002902022050953078, "loss": 1.3885, "step": 3699 }, { "epoch": 0.46, "learning_rate": 0.0002901028202144401, "loss": 1.4014, "step": 3700 }, { "epoch": 0.46, "eval_acceptance_rate": 0.6740893721580505, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.113012684309476, "eval_expected_tokens": 2.641265855386845, "eval_loss": 1.4432650804519653, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4103164638467113, "eval_runtime": 50.4117, "eval_samples_per_second": 9.66, "eval_steps_per_second": 0.317, "step": 3700 }, { "epoch": 0.46, "learning_rate": 0.00029000342882800094, "loss": 1.3295, "step": 3701 }, { "epoch": 0.46, "learning_rate": 0.00028990403095211376, "loss": 1.4029, "step": 3702 }, { "epoch": 0.46, "learning_rate": 0.00028980462660290336, "loss": 1.4449, "step": 3703 }, { "epoch": 0.46, "learning_rate": 0.00028970521579649517, "loss": 1.3551, "step": 3704 }, { "epoch": 0.46, "learning_rate": 0.0002896057985490157, "loss": 1.4181, "step": 3705 }, { "epoch": 0.46, "learning_rate": 0.0002895063748765929, "loss": 1.4117, "step": 3706 }, { "epoch": 0.46, "learning_rate": 0.0002894069447953555, "loss": 1.3622, "step": 3707 }, { "epoch": 0.46, "learning_rate": 0.00028930750832143306, "loss": 1.403, "step": 3708 }, { "epoch": 0.46, "learning_rate": 0.0002892080654709565, "loss": 1.4154, "step": 3709 }, { "epoch": 0.46, "learning_rate": 0.00028910861626005774, "loss": 1.3602, "step": 3710 }, { "epoch": 0.46, "learning_rate": 0.00028900916070486956, "loss": 1.3576, "step": 3711 }, { "epoch": 0.46, "learning_rate": 0.0002889096988215259, "loss": 1.3572, "step": 3712 }, { "epoch": 0.46, "learning_rate": 0.0002888102306261618, "loss": 1.3876, "step": 3713 }, { "epoch": 0.46, "learning_rate": 0.000288710756134913, "loss": 1.4019, "step": 3714 }, { "epoch": 0.46, "learning_rate": 0.00028861127536391665, "loss": 1.3697, "step": 3715 }, { "epoch": 0.46, "learning_rate": 0.00028851178832931077, "loss": 1.434, "step": 3716 }, { "epoch": 0.46, "learning_rate": 0.00028841229504723414, "loss": 1.3621, "step": 3717 }, { "epoch": 0.46, "learning_rate": 0.00028831279553382706, "loss": 1.4478, "step": 3718 }, { "epoch": 0.46, "learning_rate": 0.0002882132898052305, "loss": 1.4379, "step": 3719 }, { "epoch": 0.47, "learning_rate": 0.00028811377787758634, "loss": 1.3639, "step": 3720 }, { "epoch": 0.47, "learning_rate": 0.00028801425976703786, "loss": 1.4072, "step": 3721 }, { "epoch": 0.47, "learning_rate": 0.00028791473548972897, "loss": 1.3505, "step": 3722 }, { "epoch": 0.47, "learning_rate": 0.00028781520506180485, "loss": 1.3848, "step": 3723 }, { "epoch": 0.47, "learning_rate": 0.00028771566849941134, "loss": 1.3119, "step": 3724 }, { "epoch": 0.47, "learning_rate": 0.00028761612581869577, "loss": 1.3946, "step": 3725 }, { "epoch": 0.47, "learning_rate": 0.00028751657703580605, "loss": 1.4148, "step": 3726 }, { "epoch": 0.47, "learning_rate": 0.00028741702216689103, "loss": 1.3978, "step": 3727 }, { "epoch": 0.47, "learning_rate": 0.00028731746122810105, "loss": 1.4065, "step": 3728 }, { "epoch": 0.47, "learning_rate": 0.00028721789423558693, "loss": 1.3442, "step": 3729 }, { "epoch": 0.47, "learning_rate": 0.0002871183212055006, "loss": 1.3391, "step": 3730 }, { "epoch": 0.47, "learning_rate": 0.00028701874215399526, "loss": 1.3715, "step": 3731 }, { "epoch": 0.47, "learning_rate": 0.00028691915709722467, "loss": 1.3597, "step": 3732 }, { "epoch": 0.47, "learning_rate": 0.0002868195660513437, "loss": 1.3878, "step": 3733 }, { "epoch": 0.47, "learning_rate": 0.0002867199690325084, "loss": 1.4102, "step": 3734 }, { "epoch": 0.47, "learning_rate": 0.00028662036605687553, "loss": 1.3979, "step": 3735 }, { "epoch": 0.47, "learning_rate": 0.00028652075714060294, "loss": 1.3994, "step": 3736 }, { "epoch": 0.47, "learning_rate": 0.00028642114229984934, "loss": 1.4019, "step": 3737 }, { "epoch": 0.47, "learning_rate": 0.00028632152155077465, "loss": 1.3949, "step": 3738 }, { "epoch": 0.47, "learning_rate": 0.0002862218949095394, "loss": 1.3588, "step": 3739 }, { "epoch": 0.47, "learning_rate": 0.00028612226239230535, "loss": 1.3408, "step": 3740 }, { "epoch": 0.47, "learning_rate": 0.0002860226240152351, "loss": 1.384, "step": 3741 }, { "epoch": 0.47, "learning_rate": 0.00028592297979449215, "loss": 1.3888, "step": 3742 }, { "epoch": 0.47, "learning_rate": 0.00028582332974624115, "loss": 1.47, "step": 3743 }, { "epoch": 0.47, "learning_rate": 0.0002857236738866474, "loss": 1.399, "step": 3744 }, { "epoch": 0.47, "learning_rate": 0.00028562401223187746, "loss": 1.3871, "step": 3745 }, { "epoch": 0.47, "learning_rate": 0.00028552434479809855, "loss": 1.4044, "step": 3746 }, { "epoch": 0.47, "learning_rate": 0.00028542467160147896, "loss": 1.4089, "step": 3747 }, { "epoch": 0.47, "learning_rate": 0.000285324992658188, "loss": 1.4529, "step": 3748 }, { "epoch": 0.47, "learning_rate": 0.00028522530798439564, "loss": 1.4162, "step": 3749 }, { "epoch": 0.47, "learning_rate": 0.0002851256175962732, "loss": 1.3434, "step": 3750 }, { "epoch": 0.47, "learning_rate": 0.0002850259215099925, "loss": 1.3563, "step": 3751 }, { "epoch": 0.47, "learning_rate": 0.0002849262197417265, "loss": 1.3979, "step": 3752 }, { "epoch": 0.47, "learning_rate": 0.0002848265123076491, "loss": 1.3524, "step": 3753 }, { "epoch": 0.47, "learning_rate": 0.0002847267992239351, "loss": 1.4298, "step": 3754 }, { "epoch": 0.47, "learning_rate": 0.00028462708050676003, "loss": 1.3472, "step": 3755 }, { "epoch": 0.47, "learning_rate": 0.0002845273561723005, "loss": 1.3807, "step": 3756 }, { "epoch": 0.47, "learning_rate": 0.0002844276262367343, "loss": 1.4081, "step": 3757 }, { "epoch": 0.47, "learning_rate": 0.00028432789071623955, "loss": 1.4117, "step": 3758 }, { "epoch": 0.47, "learning_rate": 0.00028422814962699567, "loss": 1.3912, "step": 3759 }, { "epoch": 0.47, "learning_rate": 0.00028412840298518296, "loss": 1.3316, "step": 3760 }, { "epoch": 0.47, "learning_rate": 0.0002840286508069824, "loss": 1.4609, "step": 3761 }, { "epoch": 0.47, "learning_rate": 0.0002839288931085761, "loss": 1.3927, "step": 3762 }, { "epoch": 0.47, "learning_rate": 0.0002838291299061471, "loss": 1.4019, "step": 3763 }, { "epoch": 0.47, "learning_rate": 0.00028372936121587897, "loss": 1.4309, "step": 3764 }, { "epoch": 0.47, "learning_rate": 0.0002836295870539566, "loss": 1.3604, "step": 3765 }, { "epoch": 0.47, "learning_rate": 0.0002835298074365655, "loss": 1.414, "step": 3766 }, { "epoch": 0.47, "learning_rate": 0.0002834300223798922, "loss": 1.3235, "step": 3767 }, { "epoch": 0.47, "learning_rate": 0.000283330231900124, "loss": 1.4454, "step": 3768 }, { "epoch": 0.47, "learning_rate": 0.0002832304360134492, "loss": 1.3868, "step": 3769 }, { "epoch": 0.47, "learning_rate": 0.00028313063473605694, "loss": 1.3724, "step": 3770 }, { "epoch": 0.47, "learning_rate": 0.00028303082808413705, "loss": 1.3551, "step": 3771 }, { "epoch": 0.47, "learning_rate": 0.0002829310160738806, "loss": 1.353, "step": 3772 }, { "epoch": 0.47, "learning_rate": 0.00028283119872147915, "loss": 1.4272, "step": 3773 }, { "epoch": 0.47, "learning_rate": 0.0002827313760431255, "loss": 1.3894, "step": 3774 }, { "epoch": 0.47, "learning_rate": 0.000282631548055013, "loss": 1.4451, "step": 3775 }, { "epoch": 0.47, "learning_rate": 0.0002825317147733358, "loss": 1.3658, "step": 3776 }, { "epoch": 0.47, "learning_rate": 0.0002824318762142894, "loss": 1.4344, "step": 3777 }, { "epoch": 0.47, "learning_rate": 0.0002823320323940696, "loss": 1.4163, "step": 3778 }, { "epoch": 0.47, "learning_rate": 0.0002822321833288735, "loss": 1.4286, "step": 3779 }, { "epoch": 0.47, "learning_rate": 0.00028213232903489865, "loss": 1.4063, "step": 3780 }, { "epoch": 0.47, "learning_rate": 0.00028203246952834376, "loss": 1.3652, "step": 3781 }, { "epoch": 0.47, "learning_rate": 0.0002819326048254083, "loss": 1.3547, "step": 3782 }, { "epoch": 0.47, "learning_rate": 0.00028183273494229233, "loss": 1.4061, "step": 3783 }, { "epoch": 0.47, "learning_rate": 0.0002817328598951972, "loss": 1.4534, "step": 3784 }, { "epoch": 0.47, "learning_rate": 0.00028163297970032486, "loss": 1.443, "step": 3785 }, { "epoch": 0.47, "learning_rate": 0.00028153309437387795, "loss": 1.4377, "step": 3786 }, { "epoch": 0.47, "learning_rate": 0.00028143320393206024, "loss": 1.4627, "step": 3787 }, { "epoch": 0.47, "learning_rate": 0.00028133330839107606, "loss": 1.4221, "step": 3788 }, { "epoch": 0.47, "learning_rate": 0.00028123340776713086, "loss": 1.3448, "step": 3789 }, { "epoch": 0.47, "learning_rate": 0.0002811335020764306, "loss": 1.3965, "step": 3790 }, { "epoch": 0.47, "learning_rate": 0.00028103359133518224, "loss": 1.3876, "step": 3791 }, { "epoch": 0.47, "learning_rate": 0.00028093367555959365, "loss": 1.453, "step": 3792 }, { "epoch": 0.47, "learning_rate": 0.00028083375476587315, "loss": 1.464, "step": 3793 }, { "epoch": 0.47, "learning_rate": 0.0002807338289702304, "loss": 1.4037, "step": 3794 }, { "epoch": 0.47, "learning_rate": 0.00028063389818887535, "loss": 1.4735, "step": 3795 }, { "epoch": 0.47, "learning_rate": 0.0002805339624380192, "loss": 1.4133, "step": 3796 }, { "epoch": 0.47, "learning_rate": 0.0002804340217338737, "loss": 1.3094, "step": 3797 }, { "epoch": 0.47, "learning_rate": 0.0002803340760926513, "loss": 1.4119, "step": 3798 }, { "epoch": 0.47, "learning_rate": 0.00028023412553056557, "loss": 1.4104, "step": 3799 }, { "epoch": 0.47, "learning_rate": 0.0002801341700638307, "loss": 1.4405, "step": 3800 }, { "epoch": 0.47, "eval_acceptance_rate": 0.6764605641365051, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.122403051252764, "eval_expected_tokens": 2.653003814065955, "eval_loss": 1.4344199895858765, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4132509535164887, "eval_runtime": 50.263, "eval_samples_per_second": 9.689, "eval_steps_per_second": 0.318, "step": 3800 }, { "epoch": 0.48, "learning_rate": 0.00028003420970866175, "loss": 1.4214, "step": 3801 }, { "epoch": 0.48, "learning_rate": 0.0002799342444812745, "loss": 1.3468, "step": 3802 }, { "epoch": 0.48, "learning_rate": 0.00027983427439788535, "loss": 1.4047, "step": 3803 }, { "epoch": 0.48, "learning_rate": 0.0002797342994747119, "loss": 1.4817, "step": 3804 }, { "epoch": 0.48, "learning_rate": 0.00027963431972797214, "loss": 1.4408, "step": 3805 }, { "epoch": 0.48, "learning_rate": 0.00027953433517388524, "loss": 1.4499, "step": 3806 }, { "epoch": 0.48, "learning_rate": 0.00027943434582867067, "loss": 1.3561, "step": 3807 }, { "epoch": 0.48, "learning_rate": 0.00027933435170854906, "loss": 1.4192, "step": 3808 }, { "epoch": 0.48, "learning_rate": 0.00027923435282974166, "loss": 1.375, "step": 3809 }, { "epoch": 0.48, "learning_rate": 0.00027913434920847057, "loss": 1.354, "step": 3810 }, { "epoch": 0.48, "learning_rate": 0.00027903434086095854, "loss": 1.3928, "step": 3811 }, { "epoch": 0.48, "learning_rate": 0.00027893432780342925, "loss": 1.3877, "step": 3812 }, { "epoch": 0.48, "learning_rate": 0.0002788343100521069, "loss": 1.3435, "step": 3813 }, { "epoch": 0.48, "learning_rate": 0.0002787342876232167, "loss": 1.3602, "step": 3814 }, { "epoch": 0.48, "learning_rate": 0.0002786342605329844, "loss": 1.3681, "step": 3815 }, { "epoch": 0.48, "learning_rate": 0.00027853422879763686, "loss": 1.3922, "step": 3816 }, { "epoch": 0.48, "learning_rate": 0.00027843419243340123, "loss": 1.4121, "step": 3817 }, { "epoch": 0.48, "learning_rate": 0.00027833415145650574, "loss": 1.4039, "step": 3818 }, { "epoch": 0.48, "learning_rate": 0.00027823410588317927, "loss": 1.3291, "step": 3819 }, { "epoch": 0.48, "learning_rate": 0.0002781340557296514, "loss": 1.4348, "step": 3820 }, { "epoch": 0.48, "learning_rate": 0.00027803400101215247, "loss": 1.3067, "step": 3821 }, { "epoch": 0.48, "learning_rate": 0.0002779339417469137, "loss": 1.3367, "step": 3822 }, { "epoch": 0.48, "learning_rate": 0.0002778338779501669, "loss": 1.4133, "step": 3823 }, { "epoch": 0.48, "learning_rate": 0.00027773380963814456, "loss": 1.3864, "step": 3824 }, { "epoch": 0.48, "learning_rate": 0.00027763373682708, "loss": 1.4284, "step": 3825 }, { "epoch": 0.48, "learning_rate": 0.00027753365953320735, "loss": 1.4142, "step": 3826 }, { "epoch": 0.48, "learning_rate": 0.0002774335777727613, "loss": 1.4628, "step": 3827 }, { "epoch": 0.48, "learning_rate": 0.0002773334915619775, "loss": 1.3936, "step": 3828 }, { "epoch": 0.48, "learning_rate": 0.00027723340091709194, "loss": 1.3587, "step": 3829 }, { "epoch": 0.48, "learning_rate": 0.0002771333058543416, "loss": 1.3813, "step": 3830 }, { "epoch": 0.48, "learning_rate": 0.00027703320638996424, "loss": 1.4096, "step": 3831 }, { "epoch": 0.48, "learning_rate": 0.0002769331025401982, "loss": 1.4104, "step": 3832 }, { "epoch": 0.48, "learning_rate": 0.00027683299432128245, "loss": 1.3205, "step": 3833 }, { "epoch": 0.48, "learning_rate": 0.000276732881749457, "loss": 1.3994, "step": 3834 }, { "epoch": 0.48, "learning_rate": 0.00027663276484096207, "loss": 1.3883, "step": 3835 }, { "epoch": 0.48, "learning_rate": 0.00027653264361203895, "loss": 1.4586, "step": 3836 }, { "epoch": 0.48, "learning_rate": 0.0002764325180789296, "loss": 1.3417, "step": 3837 }, { "epoch": 0.48, "learning_rate": 0.0002763323882578766, "loss": 1.3693, "step": 3838 }, { "epoch": 0.48, "learning_rate": 0.0002762322541651232, "loss": 1.3468, "step": 3839 }, { "epoch": 0.48, "learning_rate": 0.0002761321158169134, "loss": 1.4428, "step": 3840 }, { "epoch": 0.48, "learning_rate": 0.0002760319732294919, "loss": 1.3874, "step": 3841 }, { "epoch": 0.48, "learning_rate": 0.0002759318264191039, "loss": 1.359, "step": 3842 }, { "epoch": 0.48, "learning_rate": 0.00027583167540199565, "loss": 1.368, "step": 3843 }, { "epoch": 0.48, "learning_rate": 0.0002757315201944139, "loss": 1.3179, "step": 3844 }, { "epoch": 0.48, "learning_rate": 0.0002756313608126058, "loss": 1.3373, "step": 3845 }, { "epoch": 0.48, "learning_rate": 0.0002755311972728197, "loss": 1.3964, "step": 3846 }, { "epoch": 0.48, "learning_rate": 0.00027543102959130415, "loss": 1.3835, "step": 3847 }, { "epoch": 0.48, "learning_rate": 0.0002753308577843088, "loss": 1.421, "step": 3848 }, { "epoch": 0.48, "learning_rate": 0.00027523068186808354, "loss": 1.3489, "step": 3849 }, { "epoch": 0.48, "learning_rate": 0.00027513050185887927, "loss": 1.3906, "step": 3850 }, { "epoch": 0.48, "learning_rate": 0.0002750303177729474, "loss": 1.395, "step": 3851 }, { "epoch": 0.48, "learning_rate": 0.00027493012962653997, "loss": 1.3759, "step": 3852 }, { "epoch": 0.48, "learning_rate": 0.0002748299374359098, "loss": 1.3838, "step": 3853 }, { "epoch": 0.48, "learning_rate": 0.0002747297412173102, "loss": 1.3855, "step": 3854 }, { "epoch": 0.48, "learning_rate": 0.00027462954098699536, "loss": 1.4526, "step": 3855 }, { "epoch": 0.48, "learning_rate": 0.0002745293367612199, "loss": 1.3943, "step": 3856 }, { "epoch": 0.48, "learning_rate": 0.00027442912855623914, "loss": 1.3567, "step": 3857 }, { "epoch": 0.48, "learning_rate": 0.00027432891638830916, "loss": 1.3868, "step": 3858 }, { "epoch": 0.48, "learning_rate": 0.00027422870027368664, "loss": 1.3306, "step": 3859 }, { "epoch": 0.48, "learning_rate": 0.0002741284802286288, "loss": 1.3563, "step": 3860 }, { "epoch": 0.48, "learning_rate": 0.0002740282562693936, "loss": 1.3744, "step": 3861 }, { "epoch": 0.48, "learning_rate": 0.0002739280284122395, "loss": 1.3151, "step": 3862 }, { "epoch": 0.48, "learning_rate": 0.00027382779667342594, "loss": 1.3753, "step": 3863 }, { "epoch": 0.48, "learning_rate": 0.0002737275610692125, "loss": 1.3483, "step": 3864 }, { "epoch": 0.48, "learning_rate": 0.0002736273216158597, "loss": 1.3479, "step": 3865 }, { "epoch": 0.48, "learning_rate": 0.0002735270783296286, "loss": 1.2979, "step": 3866 }, { "epoch": 0.48, "learning_rate": 0.000273426831226781, "loss": 1.47, "step": 3867 }, { "epoch": 0.48, "learning_rate": 0.000273326580323579, "loss": 1.4033, "step": 3868 }, { "epoch": 0.48, "learning_rate": 0.0002732263256362858, "loss": 1.4243, "step": 3869 }, { "epoch": 0.48, "learning_rate": 0.00027312606718116477, "loss": 1.365, "step": 3870 }, { "epoch": 0.48, "learning_rate": 0.00027302580497448007, "loss": 1.3691, "step": 3871 }, { "epoch": 0.48, "learning_rate": 0.00027292553903249657, "loss": 1.3772, "step": 3872 }, { "epoch": 0.48, "learning_rate": 0.0002728252693714796, "loss": 1.3399, "step": 3873 }, { "epoch": 0.48, "learning_rate": 0.00027272499600769495, "loss": 1.4455, "step": 3874 }, { "epoch": 0.48, "learning_rate": 0.00027262471895740946, "loss": 1.3342, "step": 3875 }, { "epoch": 0.48, "learning_rate": 0.0002725244382368902, "loss": 1.3585, "step": 3876 }, { "epoch": 0.48, "learning_rate": 0.0002724241538624049, "loss": 1.3865, "step": 3877 }, { "epoch": 0.48, "learning_rate": 0.000272323865850222, "loss": 1.382, "step": 3878 }, { "epoch": 0.48, "learning_rate": 0.00027222357421661044, "loss": 1.3504, "step": 3879 }, { "epoch": 0.48, "learning_rate": 0.0002721232789778396, "loss": 1.4003, "step": 3880 }, { "epoch": 0.49, "learning_rate": 0.0002720229801501798, "loss": 1.2752, "step": 3881 }, { "epoch": 0.49, "learning_rate": 0.00027192267774990174, "loss": 1.3298, "step": 3882 }, { "epoch": 0.49, "learning_rate": 0.00027182237179327657, "loss": 1.3711, "step": 3883 }, { "epoch": 0.49, "learning_rate": 0.0002717220622965762, "loss": 1.3945, "step": 3884 }, { "epoch": 0.49, "learning_rate": 0.00027162174927607315, "loss": 1.3993, "step": 3885 }, { "epoch": 0.49, "learning_rate": 0.0002715214327480403, "loss": 1.386, "step": 3886 }, { "epoch": 0.49, "learning_rate": 0.00027142111272875134, "loss": 1.3962, "step": 3887 }, { "epoch": 0.49, "learning_rate": 0.00027132078923448037, "loss": 1.3514, "step": 3888 }, { "epoch": 0.49, "learning_rate": 0.0002712204622815021, "loss": 1.3909, "step": 3889 }, { "epoch": 0.49, "learning_rate": 0.0002711201318860918, "loss": 1.367, "step": 3890 }, { "epoch": 0.49, "learning_rate": 0.0002710197980645253, "loss": 1.4149, "step": 3891 }, { "epoch": 0.49, "learning_rate": 0.00027091946083307894, "loss": 1.3416, "step": 3892 }, { "epoch": 0.49, "learning_rate": 0.0002708191202080296, "loss": 1.3788, "step": 3893 }, { "epoch": 0.49, "learning_rate": 0.000270718776205655, "loss": 1.3502, "step": 3894 }, { "epoch": 0.49, "learning_rate": 0.00027061842884223297, "loss": 1.3824, "step": 3895 }, { "epoch": 0.49, "learning_rate": 0.0002705180781340421, "loss": 1.328, "step": 3896 }, { "epoch": 0.49, "learning_rate": 0.00027041772409736167, "loss": 1.4114, "step": 3897 }, { "epoch": 0.49, "learning_rate": 0.00027031736674847117, "loss": 1.359, "step": 3898 }, { "epoch": 0.49, "learning_rate": 0.00027021700610365085, "loss": 1.3977, "step": 3899 }, { "epoch": 0.49, "learning_rate": 0.0002701166421791815, "loss": 1.4132, "step": 3900 }, { "epoch": 0.49, "eval_acceptance_rate": 0.6775884032249451, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.126887724013625, "eval_expected_tokens": 2.6586096550170315, "eval_loss": 1.4276988506317139, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4146524137542579, "eval_runtime": 50.2675, "eval_samples_per_second": 9.688, "eval_steps_per_second": 0.318, "step": 3900 }, { "epoch": 0.49, "learning_rate": 0.0002700162749913443, "loss": 1.3959, "step": 3901 }, { "epoch": 0.49, "learning_rate": 0.0002699159045564211, "loss": 1.3755, "step": 3902 }, { "epoch": 0.49, "learning_rate": 0.00026981553089069416, "loss": 1.3608, "step": 3903 }, { "epoch": 0.49, "learning_rate": 0.0002697151540104465, "loss": 1.3584, "step": 3904 }, { "epoch": 0.49, "learning_rate": 0.00026961477393196127, "loss": 1.3417, "step": 3905 }, { "epoch": 0.49, "learning_rate": 0.00026951439067152247, "loss": 1.3558, "step": 3906 }, { "epoch": 0.49, "learning_rate": 0.00026941400424541445, "loss": 1.318, "step": 3907 }, { "epoch": 0.49, "learning_rate": 0.0002693136146699222, "loss": 1.4052, "step": 3908 }, { "epoch": 0.49, "learning_rate": 0.00026921322196133107, "loss": 1.3989, "step": 3909 }, { "epoch": 0.49, "learning_rate": 0.0002691128261359271, "loss": 1.44, "step": 3910 }, { "epoch": 0.49, "learning_rate": 0.0002690124272099966, "loss": 1.4014, "step": 3911 }, { "epoch": 0.49, "learning_rate": 0.00026891202519982655, "loss": 1.434, "step": 3912 }, { "epoch": 0.49, "learning_rate": 0.0002688116201217044, "loss": 1.3894, "step": 3913 }, { "epoch": 0.49, "learning_rate": 0.00026871121199191825, "loss": 1.4423, "step": 3914 }, { "epoch": 0.49, "learning_rate": 0.00026861080082675627, "loss": 1.3401, "step": 3915 }, { "epoch": 0.49, "learning_rate": 0.0002685103866425075, "loss": 1.4265, "step": 3916 }, { "epoch": 0.49, "learning_rate": 0.00026840996945546144, "loss": 1.2612, "step": 3917 }, { "epoch": 0.49, "learning_rate": 0.00026830954928190793, "loss": 1.361, "step": 3918 }, { "epoch": 0.49, "learning_rate": 0.00026820912613813736, "loss": 1.3161, "step": 3919 }, { "epoch": 0.49, "learning_rate": 0.00026810870004044063, "loss": 1.3046, "step": 3920 }, { "epoch": 0.49, "learning_rate": 0.000268008271005109, "loss": 1.3904, "step": 3921 }, { "epoch": 0.49, "learning_rate": 0.00026790783904843447, "loss": 1.3263, "step": 3922 }, { "epoch": 0.49, "learning_rate": 0.0002678074041867091, "loss": 1.3641, "step": 3923 }, { "epoch": 0.49, "learning_rate": 0.0002677069664362259, "loss": 1.4533, "step": 3924 }, { "epoch": 0.49, "learning_rate": 0.000267606525813278, "loss": 1.3736, "step": 3925 }, { "epoch": 0.49, "learning_rate": 0.0002675060823341591, "loss": 1.332, "step": 3926 }, { "epoch": 0.49, "learning_rate": 0.00026740563601516346, "loss": 1.4166, "step": 3927 }, { "epoch": 0.49, "learning_rate": 0.0002673051868725855, "loss": 1.344, "step": 3928 }, { "epoch": 0.49, "learning_rate": 0.0002672047349227206, "loss": 1.3513, "step": 3929 }, { "epoch": 0.49, "learning_rate": 0.00026710428018186403, "loss": 1.3765, "step": 3930 }, { "epoch": 0.49, "learning_rate": 0.00026700382266631206, "loss": 1.3821, "step": 3931 }, { "epoch": 0.49, "learning_rate": 0.00026690336239236094, "loss": 1.4294, "step": 3932 }, { "epoch": 0.49, "learning_rate": 0.0002668028993763076, "loss": 1.3281, "step": 3933 }, { "epoch": 0.49, "learning_rate": 0.0002667024336344495, "loss": 1.3407, "step": 3934 }, { "epoch": 0.49, "learning_rate": 0.00026660196518308416, "loss": 1.3631, "step": 3935 }, { "epoch": 0.49, "learning_rate": 0.00026650149403851007, "loss": 1.3341, "step": 3936 }, { "epoch": 0.49, "learning_rate": 0.00026640102021702585, "loss": 1.3808, "step": 3937 }, { "epoch": 0.49, "learning_rate": 0.0002663005437349304, "loss": 1.409, "step": 3938 }, { "epoch": 0.49, "learning_rate": 0.0002662000646085234, "loss": 1.4055, "step": 3939 }, { "epoch": 0.49, "learning_rate": 0.0002660995828541048, "loss": 1.3795, "step": 3940 }, { "epoch": 0.49, "learning_rate": 0.0002659990984879751, "loss": 1.4487, "step": 3941 }, { "epoch": 0.49, "learning_rate": 0.00026589861152643483, "loss": 1.3369, "step": 3942 }, { "epoch": 0.49, "learning_rate": 0.00026579812198578533, "loss": 1.347, "step": 3943 }, { "epoch": 0.49, "learning_rate": 0.0002656976298823284, "loss": 1.369, "step": 3944 }, { "epoch": 0.49, "learning_rate": 0.00026559713523236584, "loss": 1.3638, "step": 3945 }, { "epoch": 0.49, "learning_rate": 0.0002654966380522003, "loss": 1.3361, "step": 3946 }, { "epoch": 0.49, "learning_rate": 0.00026539613835813457, "loss": 1.4037, "step": 3947 }, { "epoch": 0.49, "learning_rate": 0.00026529563616647207, "loss": 1.4621, "step": 3948 }, { "epoch": 0.49, "learning_rate": 0.0002651951314935163, "loss": 1.3323, "step": 3949 }, { "epoch": 0.49, "learning_rate": 0.00026509462435557154, "loss": 1.3463, "step": 3950 }, { "epoch": 0.49, "learning_rate": 0.0002649941147689422, "loss": 1.3126, "step": 3951 }, { "epoch": 0.49, "learning_rate": 0.0002648936027499331, "loss": 1.3454, "step": 3952 }, { "epoch": 0.49, "learning_rate": 0.0002647930883148498, "loss": 1.3063, "step": 3953 }, { "epoch": 0.49, "learning_rate": 0.0002646925714799977, "loss": 1.3641, "step": 3954 }, { "epoch": 0.49, "learning_rate": 0.0002645920522616829, "loss": 1.4391, "step": 3955 }, { "epoch": 0.49, "learning_rate": 0.00026449153067621207, "loss": 1.3776, "step": 3956 }, { "epoch": 0.49, "learning_rate": 0.00026439100673989184, "loss": 1.4196, "step": 3957 }, { "epoch": 0.49, "learning_rate": 0.0002642904804690295, "loss": 1.3216, "step": 3958 }, { "epoch": 0.49, "learning_rate": 0.00026418995187993275, "loss": 1.3804, "step": 3959 }, { "epoch": 0.49, "learning_rate": 0.0002640894209889094, "loss": 1.3583, "step": 3960 }, { "epoch": 0.5, "learning_rate": 0.0002639888878122678, "loss": 1.3234, "step": 3961 }, { "epoch": 0.5, "learning_rate": 0.0002638883523663169, "loss": 1.4137, "step": 3962 }, { "epoch": 0.5, "learning_rate": 0.00026378781466736555, "loss": 1.3531, "step": 3963 }, { "epoch": 0.5, "learning_rate": 0.00026368727473172334, "loss": 1.4858, "step": 3964 }, { "epoch": 0.5, "learning_rate": 0.00026358673257570003, "loss": 1.3674, "step": 3965 }, { "epoch": 0.5, "learning_rate": 0.0002634861882156058, "loss": 1.3691, "step": 3966 }, { "epoch": 0.5, "learning_rate": 0.00026338564166775125, "loss": 1.3714, "step": 3967 }, { "epoch": 0.5, "learning_rate": 0.00026328509294844715, "loss": 1.3716, "step": 3968 }, { "epoch": 0.5, "learning_rate": 0.0002631845420740049, "loss": 1.311, "step": 3969 }, { "epoch": 0.5, "learning_rate": 0.000263083989060736, "loss": 1.3829, "step": 3970 }, { "epoch": 0.5, "learning_rate": 0.00026298343392495234, "loss": 1.3927, "step": 3971 }, { "epoch": 0.5, "learning_rate": 0.0002628828766829663, "loss": 1.3383, "step": 3972 }, { "epoch": 0.5, "learning_rate": 0.00026278231735109053, "loss": 1.3845, "step": 3973 }, { "epoch": 0.5, "learning_rate": 0.0002626817559456378, "loss": 1.3533, "step": 3974 }, { "epoch": 0.5, "learning_rate": 0.00026258119248292165, "loss": 1.3052, "step": 3975 }, { "epoch": 0.5, "learning_rate": 0.00026248062697925567, "loss": 1.3774, "step": 3976 }, { "epoch": 0.5, "learning_rate": 0.0002623800594509536, "loss": 1.2957, "step": 3977 }, { "epoch": 0.5, "learning_rate": 0.00026227948991433005, "loss": 1.4019, "step": 3978 }, { "epoch": 0.5, "learning_rate": 0.0002621789183856995, "loss": 1.3219, "step": 3979 }, { "epoch": 0.5, "learning_rate": 0.0002620783448813768, "loss": 1.3557, "step": 3980 }, { "epoch": 0.5, "learning_rate": 0.00026197776941767735, "loss": 1.4515, "step": 3981 }, { "epoch": 0.5, "learning_rate": 0.00026187719201091665, "loss": 1.4138, "step": 3982 }, { "epoch": 0.5, "learning_rate": 0.00026177661267741067, "loss": 1.4043, "step": 3983 }, { "epoch": 0.5, "learning_rate": 0.0002616760314334756, "loss": 1.4205, "step": 3984 }, { "epoch": 0.5, "learning_rate": 0.00026157544829542796, "loss": 1.4364, "step": 3985 }, { "epoch": 0.5, "learning_rate": 0.00026147486327958447, "loss": 1.3757, "step": 3986 }, { "epoch": 0.5, "learning_rate": 0.0002613742764022624, "loss": 1.3108, "step": 3987 }, { "epoch": 0.5, "learning_rate": 0.00026127368767977924, "loss": 1.3812, "step": 3988 }, { "epoch": 0.5, "learning_rate": 0.0002611730971284525, "loss": 1.4434, "step": 3989 }, { "epoch": 0.5, "learning_rate": 0.0002610725047646004, "loss": 1.4324, "step": 3990 }, { "epoch": 0.5, "learning_rate": 0.00026097191060454113, "loss": 1.3366, "step": 3991 }, { "epoch": 0.5, "learning_rate": 0.0002608713146645934, "loss": 1.4473, "step": 3992 }, { "epoch": 0.5, "learning_rate": 0.0002607707169610761, "loss": 1.3691, "step": 3993 }, { "epoch": 0.5, "learning_rate": 0.00026067011751030835, "loss": 1.4145, "step": 3994 }, { "epoch": 0.5, "learning_rate": 0.00026056951632860974, "loss": 1.3384, "step": 3995 }, { "epoch": 0.5, "learning_rate": 0.0002604689134322999, "loss": 1.4004, "step": 3996 }, { "epoch": 0.5, "learning_rate": 0.000260368308837699, "loss": 1.4036, "step": 3997 }, { "epoch": 0.5, "learning_rate": 0.00026026770256112723, "loss": 1.3479, "step": 3998 }, { "epoch": 0.5, "learning_rate": 0.0002601670946189052, "loss": 1.4009, "step": 3999 }, { "epoch": 0.5, "learning_rate": 0.0002600664850273538, "loss": 1.3855, "step": 4000 }, { "epoch": 0.5, "eval_acceptance_rate": 0.677869439125061, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1280070508085007, "eval_expected_tokens": 2.6600088135106255, "eval_loss": 1.4214684963226318, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4150022033776564, "eval_runtime": 50.5859, "eval_samples_per_second": 9.627, "eval_steps_per_second": 0.316, "step": 4000 }, { "epoch": 0.5, "learning_rate": 0.00025996587380279414, "loss": 1.3795, "step": 4001 }, { "epoch": 0.5, "learning_rate": 0.0002598652609615475, "loss": 1.2994, "step": 4002 }, { "epoch": 0.5, "learning_rate": 0.00025976464651993573, "loss": 1.4046, "step": 4003 }, { "epoch": 0.5, "learning_rate": 0.00025966403049428055, "loss": 1.4145, "step": 4004 }, { "epoch": 0.5, "learning_rate": 0.0002595634129009042, "loss": 1.4171, "step": 4005 }, { "epoch": 0.5, "learning_rate": 0.0002594627937561291, "loss": 1.4036, "step": 4006 }, { "epoch": 0.5, "learning_rate": 0.00025936217307627793, "loss": 1.3838, "step": 4007 }, { "epoch": 0.5, "learning_rate": 0.0002592615508776735, "loss": 1.3338, "step": 4008 }, { "epoch": 0.5, "learning_rate": 0.0002591609271766391, "loss": 1.4018, "step": 4009 }, { "epoch": 0.5, "learning_rate": 0.0002590603019894981, "loss": 1.416, "step": 4010 }, { "epoch": 0.5, "learning_rate": 0.00025895967533257416, "loss": 1.3459, "step": 4011 }, { "epoch": 0.5, "learning_rate": 0.00025885904722219107, "loss": 1.3547, "step": 4012 }, { "epoch": 0.5, "learning_rate": 0.0002587584176746731, "loss": 1.3917, "step": 4013 }, { "epoch": 0.5, "learning_rate": 0.0002586577867063445, "loss": 1.3283, "step": 4014 }, { "epoch": 0.5, "learning_rate": 0.0002585571543335299, "loss": 1.4547, "step": 4015 }, { "epoch": 0.5, "learning_rate": 0.00025845652057255415, "loss": 1.4542, "step": 4016 }, { "epoch": 0.5, "learning_rate": 0.00025835588543974224, "loss": 1.3718, "step": 4017 }, { "epoch": 0.5, "learning_rate": 0.00025825524895141936, "loss": 1.3844, "step": 4018 }, { "epoch": 0.5, "learning_rate": 0.0002581546111239112, "loss": 1.4649, "step": 4019 }, { "epoch": 0.5, "learning_rate": 0.0002580539719735433, "loss": 1.4495, "step": 4020 }, { "epoch": 0.5, "learning_rate": 0.00025795333151664145, "loss": 1.4478, "step": 4021 }, { "epoch": 0.5, "learning_rate": 0.00025785268976953206, "loss": 1.4029, "step": 4022 }, { "epoch": 0.5, "learning_rate": 0.0002577520467485414, "loss": 1.3728, "step": 4023 }, { "epoch": 0.5, "learning_rate": 0.0002576514024699958, "loss": 1.3726, "step": 4024 }, { "epoch": 0.5, "learning_rate": 0.00025755075695022223, "loss": 1.3234, "step": 4025 }, { "epoch": 0.5, "learning_rate": 0.00025745011020554754, "loss": 1.3025, "step": 4026 }, { "epoch": 0.5, "learning_rate": 0.00025734946225229886, "loss": 1.3934, "step": 4027 }, { "epoch": 0.5, "learning_rate": 0.00025724881310680364, "loss": 1.4505, "step": 4028 }, { "epoch": 0.5, "learning_rate": 0.00025714816278538935, "loss": 1.3009, "step": 4029 }, { "epoch": 0.5, "learning_rate": 0.00025704751130438365, "loss": 1.3077, "step": 4030 }, { "epoch": 0.5, "learning_rate": 0.00025694685868011446, "loss": 1.3974, "step": 4031 }, { "epoch": 0.5, "learning_rate": 0.00025684620492891, "loss": 1.3453, "step": 4032 }, { "epoch": 0.5, "learning_rate": 0.00025674555006709854, "loss": 1.3565, "step": 4033 }, { "epoch": 0.5, "learning_rate": 0.00025664489411100837, "loss": 1.4289, "step": 4034 }, { "epoch": 0.5, "learning_rate": 0.00025654423707696834, "loss": 1.3245, "step": 4035 }, { "epoch": 0.5, "learning_rate": 0.0002564435789813071, "loss": 1.3533, "step": 4036 }, { "epoch": 0.5, "learning_rate": 0.00025634291984035374, "loss": 1.4181, "step": 4037 }, { "epoch": 0.5, "learning_rate": 0.00025624225967043734, "loss": 1.3802, "step": 4038 }, { "epoch": 0.5, "learning_rate": 0.00025614159848788733, "loss": 1.3618, "step": 4039 }, { "epoch": 0.51, "learning_rate": 0.0002560409363090331, "loss": 1.3625, "step": 4040 }, { "epoch": 0.51, "learning_rate": 0.00025594027315020435, "loss": 1.4054, "step": 4041 }, { "epoch": 0.51, "learning_rate": 0.00025583960902773086, "loss": 1.3562, "step": 4042 }, { "epoch": 0.51, "learning_rate": 0.00025573894395794256, "loss": 1.3291, "step": 4043 }, { "epoch": 0.51, "learning_rate": 0.0002556382779571697, "loss": 1.3842, "step": 4044 }, { "epoch": 0.51, "learning_rate": 0.00025553761104174245, "loss": 1.3725, "step": 4045 }, { "epoch": 0.51, "learning_rate": 0.00025543694322799115, "loss": 1.3703, "step": 4046 }, { "epoch": 0.51, "learning_rate": 0.0002553362745322466, "loss": 1.3488, "step": 4047 }, { "epoch": 0.51, "learning_rate": 0.00025523560497083924, "loss": 1.367, "step": 4048 }, { "epoch": 0.51, "learning_rate": 0.0002551349345601001, "loss": 1.3708, "step": 4049 }, { "epoch": 0.51, "learning_rate": 0.0002550342633163601, "loss": 1.3678, "step": 4050 }, { "epoch": 0.51, "learning_rate": 0.0002549335912559504, "loss": 1.3847, "step": 4051 }, { "epoch": 0.51, "learning_rate": 0.0002548329183952021, "loss": 1.3623, "step": 4052 }, { "epoch": 0.51, "learning_rate": 0.0002547322447504468, "loss": 1.4298, "step": 4053 }, { "epoch": 0.51, "learning_rate": 0.0002546315703380159, "loss": 1.3766, "step": 4054 }, { "epoch": 0.51, "learning_rate": 0.000254530895174241, "loss": 1.3631, "step": 4055 }, { "epoch": 0.51, "learning_rate": 0.00025443021927545396, "loss": 1.3499, "step": 4056 }, { "epoch": 0.51, "learning_rate": 0.0002543295426579866, "loss": 1.4169, "step": 4057 }, { "epoch": 0.51, "learning_rate": 0.00025422886533817075, "loss": 1.421, "step": 4058 }, { "epoch": 0.51, "learning_rate": 0.0002541281873323388, "loss": 1.4266, "step": 4059 }, { "epoch": 0.51, "learning_rate": 0.0002540275086568228, "loss": 1.4258, "step": 4060 }, { "epoch": 0.51, "learning_rate": 0.0002539268293279552, "loss": 1.2565, "step": 4061 }, { "epoch": 0.51, "learning_rate": 0.00025382614936206825, "loss": 1.3786, "step": 4062 }, { "epoch": 0.51, "learning_rate": 0.0002537254687754946, "loss": 1.3646, "step": 4063 }, { "epoch": 0.51, "learning_rate": 0.0002536247875845669, "loss": 1.3227, "step": 4064 }, { "epoch": 0.51, "learning_rate": 0.0002535241058056178, "loss": 1.3525, "step": 4065 }, { "epoch": 0.51, "learning_rate": 0.0002534234234549802, "loss": 1.3333, "step": 4066 }, { "epoch": 0.51, "learning_rate": 0.00025332274054898704, "loss": 1.363, "step": 4067 }, { "epoch": 0.51, "learning_rate": 0.00025322205710397123, "loss": 1.3984, "step": 4068 }, { "epoch": 0.51, "learning_rate": 0.000253121373136266, "loss": 1.3927, "step": 4069 }, { "epoch": 0.51, "learning_rate": 0.00025302068866220437, "loss": 1.3652, "step": 4070 }, { "epoch": 0.51, "learning_rate": 0.00025292000369811973, "loss": 1.3715, "step": 4071 }, { "epoch": 0.51, "learning_rate": 0.00025281931826034537, "loss": 1.3815, "step": 4072 }, { "epoch": 0.51, "learning_rate": 0.00025271863236521485, "loss": 1.3604, "step": 4073 }, { "epoch": 0.51, "learning_rate": 0.00025261794602906147, "loss": 1.3698, "step": 4074 }, { "epoch": 0.51, "learning_rate": 0.0002525172592682189, "loss": 1.3089, "step": 4075 }, { "epoch": 0.51, "learning_rate": 0.0002524165720990208, "loss": 1.385, "step": 4076 }, { "epoch": 0.51, "learning_rate": 0.00025231588453780067, "loss": 1.3105, "step": 4077 }, { "epoch": 0.51, "learning_rate": 0.00025221519660089265, "loss": 1.3412, "step": 4078 }, { "epoch": 0.51, "learning_rate": 0.0002521145083046303, "loss": 1.2905, "step": 4079 }, { "epoch": 0.51, "learning_rate": 0.00025201381966534746, "loss": 1.2681, "step": 4080 }, { "epoch": 0.51, "learning_rate": 0.00025191313069937833, "loss": 1.3124, "step": 4081 }, { "epoch": 0.51, "learning_rate": 0.00025181244142305666, "loss": 1.2687, "step": 4082 }, { "epoch": 0.51, "learning_rate": 0.0002517117518527166, "loss": 1.4148, "step": 4083 }, { "epoch": 0.51, "learning_rate": 0.0002516110620046923, "loss": 1.4226, "step": 4084 }, { "epoch": 0.51, "learning_rate": 0.00025151037189531775, "loss": 1.3439, "step": 4085 }, { "epoch": 0.51, "learning_rate": 0.0002514096815409272, "loss": 1.3282, "step": 4086 }, { "epoch": 0.51, "learning_rate": 0.0002513089909578549, "loss": 1.2593, "step": 4087 }, { "epoch": 0.51, "learning_rate": 0.0002512083001624351, "loss": 1.3367, "step": 4088 }, { "epoch": 0.51, "learning_rate": 0.0002511076091710021, "loss": 1.3608, "step": 4089 }, { "epoch": 0.51, "learning_rate": 0.0002510069179998901, "loss": 1.3747, "step": 4090 }, { "epoch": 0.51, "learning_rate": 0.0002509062266654337, "loss": 1.3648, "step": 4091 }, { "epoch": 0.51, "learning_rate": 0.000250805535183967, "loss": 1.3259, "step": 4092 }, { "epoch": 0.51, "learning_rate": 0.0002507048435718246, "loss": 1.3788, "step": 4093 }, { "epoch": 0.51, "learning_rate": 0.00025060415184534085, "loss": 1.337, "step": 4094 }, { "epoch": 0.51, "learning_rate": 0.0002505034600208503, "loss": 1.3743, "step": 4095 }, { "epoch": 0.51, "learning_rate": 0.00025040276811468714, "loss": 1.3675, "step": 4096 }, { "epoch": 0.51, "learning_rate": 0.0002503020761431861, "loss": 1.2906, "step": 4097 }, { "epoch": 0.51, "learning_rate": 0.0002502013841226816, "loss": 1.3114, "step": 4098 }, { "epoch": 0.51, "learning_rate": 0.00025010069206950805, "loss": 1.3516, "step": 4099 }, { "epoch": 0.51, "learning_rate": 0.00025, "loss": 1.3546, "step": 4100 }, { "epoch": 0.51, "eval_acceptance_rate": 0.6809515953063965, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.140330955467779, "eval_expected_tokens": 2.6754136943347238, "eval_loss": 1.4133321046829224, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.418853423583681, "eval_runtime": 50.2895, "eval_samples_per_second": 9.684, "eval_steps_per_second": 0.318, "step": 4100 }, { "epoch": 0.51, "learning_rate": 0.000249899307930492, "loss": 1.315, "step": 4101 }, { "epoch": 0.51, "learning_rate": 0.0002497986158773185, "loss": 1.3576, "step": 4102 }, { "epoch": 0.51, "learning_rate": 0.0002496979238568139, "loss": 1.3262, "step": 4103 }, { "epoch": 0.51, "learning_rate": 0.0002495972318853129, "loss": 1.3444, "step": 4104 }, { "epoch": 0.51, "learning_rate": 0.00024949653997914976, "loss": 1.34, "step": 4105 }, { "epoch": 0.51, "learning_rate": 0.00024939584815465916, "loss": 1.3295, "step": 4106 }, { "epoch": 0.51, "learning_rate": 0.00024929515642817544, "loss": 1.3788, "step": 4107 }, { "epoch": 0.51, "learning_rate": 0.000249194464816033, "loss": 1.3437, "step": 4108 }, { "epoch": 0.51, "learning_rate": 0.00024909377333456634, "loss": 1.3908, "step": 4109 }, { "epoch": 0.51, "learning_rate": 0.0002489930820001099, "loss": 1.3945, "step": 4110 }, { "epoch": 0.51, "learning_rate": 0.0002488923908289979, "loss": 1.3647, "step": 4111 }, { "epoch": 0.51, "learning_rate": 0.00024879169983756496, "loss": 1.3953, "step": 4112 }, { "epoch": 0.51, "learning_rate": 0.00024869100904214507, "loss": 1.3249, "step": 4113 }, { "epoch": 0.51, "learning_rate": 0.00024859031845907284, "loss": 1.4143, "step": 4114 }, { "epoch": 0.51, "learning_rate": 0.0002484896281046823, "loss": 1.3374, "step": 4115 }, { "epoch": 0.51, "learning_rate": 0.00024838893799530783, "loss": 1.2919, "step": 4116 }, { "epoch": 0.51, "learning_rate": 0.0002482882481472834, "loss": 1.3849, "step": 4117 }, { "epoch": 0.51, "learning_rate": 0.0002481875585769433, "loss": 1.4174, "step": 4118 }, { "epoch": 0.51, "learning_rate": 0.00024808686930062173, "loss": 1.3565, "step": 4119 }, { "epoch": 0.52, "learning_rate": 0.00024798618033465255, "loss": 1.3107, "step": 4120 }, { "epoch": 0.52, "learning_rate": 0.0002478854916953698, "loss": 1.409, "step": 4121 }, { "epoch": 0.52, "learning_rate": 0.00024778480339910747, "loss": 1.3067, "step": 4122 }, { "epoch": 0.52, "learning_rate": 0.0002476841154621993, "loss": 1.3756, "step": 4123 }, { "epoch": 0.52, "learning_rate": 0.00024758342790097925, "loss": 1.3676, "step": 4124 }, { "epoch": 0.52, "learning_rate": 0.00024748274073178113, "loss": 1.3303, "step": 4125 }, { "epoch": 0.52, "learning_rate": 0.00024738205397093865, "loss": 1.4201, "step": 4126 }, { "epoch": 0.52, "learning_rate": 0.0002472813676347852, "loss": 1.3625, "step": 4127 }, { "epoch": 0.52, "learning_rate": 0.0002471806817396546, "loss": 1.3896, "step": 4128 }, { "epoch": 0.52, "learning_rate": 0.0002470799963018803, "loss": 1.3434, "step": 4129 }, { "epoch": 0.52, "learning_rate": 0.00024697931133779564, "loss": 1.4102, "step": 4130 }, { "epoch": 0.52, "learning_rate": 0.0002468786268637341, "loss": 1.4096, "step": 4131 }, { "epoch": 0.52, "learning_rate": 0.0002467779428960288, "loss": 1.3306, "step": 4132 }, { "epoch": 0.52, "learning_rate": 0.000246677259451013, "loss": 1.3465, "step": 4133 }, { "epoch": 0.52, "learning_rate": 0.0002465765765450198, "loss": 1.3444, "step": 4134 }, { "epoch": 0.52, "learning_rate": 0.00024647589419438224, "loss": 1.3453, "step": 4135 }, { "epoch": 0.52, "learning_rate": 0.00024637521241543313, "loss": 1.3859, "step": 4136 }, { "epoch": 0.52, "learning_rate": 0.00024627453122450535, "loss": 1.3458, "step": 4137 }, { "epoch": 0.52, "learning_rate": 0.00024617385063793176, "loss": 1.3839, "step": 4138 }, { "epoch": 0.52, "learning_rate": 0.0002460731706720449, "loss": 1.3429, "step": 4139 }, { "epoch": 0.52, "learning_rate": 0.0002459724913431772, "loss": 1.3406, "step": 4140 }, { "epoch": 0.52, "learning_rate": 0.00024587181266766124, "loss": 1.4206, "step": 4141 }, { "epoch": 0.52, "learning_rate": 0.0002457711346618292, "loss": 1.3062, "step": 4142 }, { "epoch": 0.52, "learning_rate": 0.00024567045734201346, "loss": 1.3071, "step": 4143 }, { "epoch": 0.52, "learning_rate": 0.0002455697807245461, "loss": 1.3298, "step": 4144 }, { "epoch": 0.52, "learning_rate": 0.0002454691048257591, "loss": 1.3404, "step": 4145 }, { "epoch": 0.52, "learning_rate": 0.00024536842966198415, "loss": 1.358, "step": 4146 }, { "epoch": 0.52, "learning_rate": 0.0002452677552495532, "loss": 1.3403, "step": 4147 }, { "epoch": 0.52, "learning_rate": 0.0002451670816047979, "loss": 1.3495, "step": 4148 }, { "epoch": 0.52, "learning_rate": 0.0002450664087440497, "loss": 1.3243, "step": 4149 }, { "epoch": 0.52, "learning_rate": 0.00024496573668363996, "loss": 1.3682, "step": 4150 }, { "epoch": 0.52, "learning_rate": 0.0002448650654399, "loss": 1.3529, "step": 4151 }, { "epoch": 0.52, "learning_rate": 0.00024476439502916077, "loss": 1.371, "step": 4152 }, { "epoch": 0.52, "learning_rate": 0.00024466372546775347, "loss": 1.3546, "step": 4153 }, { "epoch": 0.52, "learning_rate": 0.00024456305677200886, "loss": 1.2424, "step": 4154 }, { "epoch": 0.52, "learning_rate": 0.0002444623889582576, "loss": 1.353, "step": 4155 }, { "epoch": 0.52, "learning_rate": 0.0002443617220428304, "loss": 1.356, "step": 4156 }, { "epoch": 0.52, "learning_rate": 0.0002442610560420574, "loss": 1.3334, "step": 4157 }, { "epoch": 0.52, "learning_rate": 0.0002441603909722692, "loss": 1.3371, "step": 4158 }, { "epoch": 0.52, "learning_rate": 0.0002440597268497957, "loss": 1.4049, "step": 4159 }, { "epoch": 0.52, "learning_rate": 0.000243959063690967, "loss": 1.3711, "step": 4160 }, { "epoch": 0.52, "learning_rate": 0.00024385840151211274, "loss": 1.347, "step": 4161 }, { "epoch": 0.52, "learning_rate": 0.00024375774032956265, "loss": 1.3934, "step": 4162 }, { "epoch": 0.52, "learning_rate": 0.0002436570801596463, "loss": 1.3248, "step": 4163 }, { "epoch": 0.52, "learning_rate": 0.00024355642101869294, "loss": 1.3817, "step": 4164 }, { "epoch": 0.52, "learning_rate": 0.00024345576292303175, "loss": 1.3288, "step": 4165 }, { "epoch": 0.52, "learning_rate": 0.0002433551058889917, "loss": 1.38, "step": 4166 }, { "epoch": 0.52, "learning_rate": 0.00024325444993290152, "loss": 1.3084, "step": 4167 }, { "epoch": 0.52, "learning_rate": 0.00024315379507109, "loss": 1.3438, "step": 4168 }, { "epoch": 0.52, "learning_rate": 0.00024305314131988555, "loss": 1.3089, "step": 4169 }, { "epoch": 0.52, "learning_rate": 0.00024295248869561642, "loss": 1.3673, "step": 4170 }, { "epoch": 0.52, "learning_rate": 0.00024285183721461077, "loss": 1.3851, "step": 4171 }, { "epoch": 0.52, "learning_rate": 0.00024275118689319637, "loss": 1.3023, "step": 4172 }, { "epoch": 0.52, "learning_rate": 0.00024265053774770115, "loss": 1.3078, "step": 4173 }, { "epoch": 0.52, "learning_rate": 0.0002425498897944525, "loss": 1.2727, "step": 4174 }, { "epoch": 0.52, "learning_rate": 0.00024244924304977784, "loss": 1.3661, "step": 4175 }, { "epoch": 0.52, "learning_rate": 0.00024234859753000419, "loss": 1.3158, "step": 4176 }, { "epoch": 0.52, "learning_rate": 0.00024224795325145863, "loss": 1.327, "step": 4177 }, { "epoch": 0.52, "learning_rate": 0.00024214731023046793, "loss": 1.2984, "step": 4178 }, { "epoch": 0.52, "learning_rate": 0.00024204666848335859, "loss": 1.3136, "step": 4179 }, { "epoch": 0.52, "learning_rate": 0.00024194602802645682, "loss": 1.3105, "step": 4180 }, { "epoch": 0.52, "learning_rate": 0.0002418453888760888, "loss": 1.4323, "step": 4181 }, { "epoch": 0.52, "learning_rate": 0.00024174475104858062, "loss": 1.3565, "step": 4182 }, { "epoch": 0.52, "learning_rate": 0.00024164411456025777, "loss": 1.3685, "step": 4183 }, { "epoch": 0.52, "learning_rate": 0.0002415434794274459, "loss": 1.3473, "step": 4184 }, { "epoch": 0.52, "learning_rate": 0.00024144284566647014, "loss": 1.3202, "step": 4185 }, { "epoch": 0.52, "learning_rate": 0.0002413422132936555, "loss": 1.3843, "step": 4186 }, { "epoch": 0.52, "learning_rate": 0.00024124158232532693, "loss": 1.3381, "step": 4187 }, { "epoch": 0.52, "learning_rate": 0.00024114095277780895, "loss": 1.3634, "step": 4188 }, { "epoch": 0.52, "learning_rate": 0.00024104032466742593, "loss": 1.3396, "step": 4189 }, { "epoch": 0.52, "learning_rate": 0.000240939698010502, "loss": 1.3639, "step": 4190 }, { "epoch": 0.52, "learning_rate": 0.0002408390728233609, "loss": 1.4015, "step": 4191 }, { "epoch": 0.52, "learning_rate": 0.00024073844912232657, "loss": 1.3789, "step": 4192 }, { "epoch": 0.52, "learning_rate": 0.00024063782692372216, "loss": 1.3678, "step": 4193 }, { "epoch": 0.52, "learning_rate": 0.000240537206243871, "loss": 1.3208, "step": 4194 }, { "epoch": 0.52, "learning_rate": 0.00024043658709909586, "loss": 1.3431, "step": 4195 }, { "epoch": 0.52, "learning_rate": 0.00024033596950571943, "loss": 1.3237, "step": 4196 }, { "epoch": 0.52, "learning_rate": 0.0002402353534800643, "loss": 1.2851, "step": 4197 }, { "epoch": 0.52, "learning_rate": 0.0002401347390384525, "loss": 1.3322, "step": 4198 }, { "epoch": 0.52, "learning_rate": 0.00024003412619720595, "loss": 1.3579, "step": 4199 }, { "epoch": 0.53, "learning_rate": 0.0002399335149726463, "loss": 1.3318, "step": 4200 }, { "epoch": 0.53, "eval_acceptance_rate": 0.6815897822380066, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1428937727591153, "eval_expected_tokens": 2.678617215948894, "eval_loss": 1.4086946249008179, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4196543039872236, "eval_runtime": 50.3154, "eval_samples_per_second": 9.679, "eval_steps_per_second": 0.318, "step": 4200 }, { "epoch": 0.53, "learning_rate": 0.00023983290538109482, "loss": 1.3526, "step": 4201 }, { "epoch": 0.53, "learning_rate": 0.00023973229743887278, "loss": 1.3469, "step": 4202 }, { "epoch": 0.53, "learning_rate": 0.00023963169116230105, "loss": 1.3436, "step": 4203 }, { "epoch": 0.53, "learning_rate": 0.00023953108656770016, "loss": 1.3578, "step": 4204 }, { "epoch": 0.53, "learning_rate": 0.00023943048367139035, "loss": 1.2584, "step": 4205 }, { "epoch": 0.53, "learning_rate": 0.00023932988248969163, "loss": 1.2744, "step": 4206 }, { "epoch": 0.53, "learning_rate": 0.00023922928303892396, "loss": 1.3313, "step": 4207 }, { "epoch": 0.53, "learning_rate": 0.00023912868533540663, "loss": 1.3616, "step": 4208 }, { "epoch": 0.53, "learning_rate": 0.00023902808939545896, "loss": 1.4208, "step": 4209 }, { "epoch": 0.53, "learning_rate": 0.00023892749523539975, "loss": 1.3529, "step": 4210 }, { "epoch": 0.53, "learning_rate": 0.00023882690287154755, "loss": 1.4034, "step": 4211 }, { "epoch": 0.53, "learning_rate": 0.00023872631232022082, "loss": 1.3635, "step": 4212 }, { "epoch": 0.53, "learning_rate": 0.0002386257235977376, "loss": 1.3213, "step": 4213 }, { "epoch": 0.53, "learning_rate": 0.00023852513672041554, "loss": 1.3675, "step": 4214 }, { "epoch": 0.53, "learning_rate": 0.00023842455170457216, "loss": 1.3142, "step": 4215 }, { "epoch": 0.53, "learning_rate": 0.0002383239685665244, "loss": 1.3258, "step": 4216 }, { "epoch": 0.53, "learning_rate": 0.00023822338732258937, "loss": 1.3503, "step": 4217 }, { "epoch": 0.53, "learning_rate": 0.00023812280798908336, "loss": 1.4314, "step": 4218 }, { "epoch": 0.53, "learning_rate": 0.00023802223058232275, "loss": 1.3736, "step": 4219 }, { "epoch": 0.53, "learning_rate": 0.00023792165511862326, "loss": 1.3211, "step": 4220 }, { "epoch": 0.53, "learning_rate": 0.00023782108161430056, "loss": 1.3444, "step": 4221 }, { "epoch": 0.53, "learning_rate": 0.00023772051008566999, "loss": 1.3129, "step": 4222 }, { "epoch": 0.53, "learning_rate": 0.00023761994054904642, "loss": 1.3718, "step": 4223 }, { "epoch": 0.53, "learning_rate": 0.00023751937302074445, "loss": 1.3677, "step": 4224 }, { "epoch": 0.53, "learning_rate": 0.0002374188075170783, "loss": 1.3573, "step": 4225 }, { "epoch": 0.53, "learning_rate": 0.0002373182440543622, "loss": 1.3651, "step": 4226 }, { "epoch": 0.53, "learning_rate": 0.00023721768264890953, "loss": 1.3366, "step": 4227 }, { "epoch": 0.53, "learning_rate": 0.00023711712331703374, "loss": 1.3019, "step": 4228 }, { "epoch": 0.53, "learning_rate": 0.00023701656607504773, "loss": 1.308, "step": 4229 }, { "epoch": 0.53, "learning_rate": 0.00023691601093926405, "loss": 1.3761, "step": 4230 }, { "epoch": 0.53, "learning_rate": 0.00023681545792599509, "loss": 1.3273, "step": 4231 }, { "epoch": 0.53, "learning_rate": 0.00023671490705155286, "loss": 1.4003, "step": 4232 }, { "epoch": 0.53, "learning_rate": 0.00023661435833224878, "loss": 1.338, "step": 4233 }, { "epoch": 0.53, "learning_rate": 0.00023651381178439423, "loss": 1.3657, "step": 4234 }, { "epoch": 0.53, "learning_rate": 0.00023641326742429998, "loss": 1.3231, "step": 4235 }, { "epoch": 0.53, "learning_rate": 0.00023631272526827672, "loss": 1.321, "step": 4236 }, { "epoch": 0.53, "learning_rate": 0.00023621218533263446, "loss": 1.3652, "step": 4237 }, { "epoch": 0.53, "learning_rate": 0.00023611164763368318, "loss": 1.3388, "step": 4238 }, { "epoch": 0.53, "learning_rate": 0.00023601111218773222, "loss": 1.3562, "step": 4239 }, { "epoch": 0.53, "learning_rate": 0.0002359105790110906, "loss": 1.3822, "step": 4240 }, { "epoch": 0.53, "learning_rate": 0.00023581004812006732, "loss": 1.3925, "step": 4241 }, { "epoch": 0.53, "learning_rate": 0.00023570951953097053, "loss": 1.2817, "step": 4242 }, { "epoch": 0.53, "learning_rate": 0.0002356089932601082, "loss": 1.3405, "step": 4243 }, { "epoch": 0.53, "learning_rate": 0.00023550846932378803, "loss": 1.3083, "step": 4244 }, { "epoch": 0.53, "learning_rate": 0.00023540794773831706, "loss": 1.3551, "step": 4245 }, { "epoch": 0.53, "learning_rate": 0.00023530742852000233, "loss": 1.381, "step": 4246 }, { "epoch": 0.53, "learning_rate": 0.00023520691168515026, "loss": 1.4362, "step": 4247 }, { "epoch": 0.53, "learning_rate": 0.00023510639725006692, "loss": 1.3748, "step": 4248 }, { "epoch": 0.53, "learning_rate": 0.00023500588523105788, "loss": 1.3155, "step": 4249 }, { "epoch": 0.53, "learning_rate": 0.00023490537564442845, "loss": 1.3782, "step": 4250 }, { "epoch": 0.53, "learning_rate": 0.00023480486850648371, "loss": 1.3432, "step": 4251 }, { "epoch": 0.53, "learning_rate": 0.000234704363833528, "loss": 1.3367, "step": 4252 }, { "epoch": 0.53, "learning_rate": 0.00023460386164186547, "loss": 1.3511, "step": 4253 }, { "epoch": 0.53, "learning_rate": 0.0002345033619477998, "loss": 1.3021, "step": 4254 }, { "epoch": 0.53, "learning_rate": 0.0002344028647676342, "loss": 1.3406, "step": 4255 }, { "epoch": 0.53, "learning_rate": 0.00023430237011767165, "loss": 1.4312, "step": 4256 }, { "epoch": 0.53, "learning_rate": 0.00023420187801421468, "loss": 1.3431, "step": 4257 }, { "epoch": 0.53, "learning_rate": 0.0002341013884735652, "loss": 1.3308, "step": 4258 }, { "epoch": 0.53, "learning_rate": 0.00023400090151202502, "loss": 1.3275, "step": 4259 }, { "epoch": 0.53, "learning_rate": 0.00023390041714589513, "loss": 1.3145, "step": 4260 }, { "epoch": 0.53, "learning_rate": 0.0002337999353914766, "loss": 1.3234, "step": 4261 }, { "epoch": 0.53, "learning_rate": 0.00023369945626506962, "loss": 1.3579, "step": 4262 }, { "epoch": 0.53, "learning_rate": 0.0002335989797829743, "loss": 1.3959, "step": 4263 }, { "epoch": 0.53, "learning_rate": 0.00023349850596148997, "loss": 1.2674, "step": 4264 }, { "epoch": 0.53, "learning_rate": 0.0002333980348169158, "loss": 1.3428, "step": 4265 }, { "epoch": 0.53, "learning_rate": 0.0002332975663655506, "loss": 1.2806, "step": 4266 }, { "epoch": 0.53, "learning_rate": 0.00023319710062369247, "loss": 1.282, "step": 4267 }, { "epoch": 0.53, "learning_rate": 0.00023309663760763912, "loss": 1.3102, "step": 4268 }, { "epoch": 0.53, "learning_rate": 0.00023299617733368806, "loss": 1.2706, "step": 4269 }, { "epoch": 0.53, "learning_rate": 0.00023289571981813592, "loss": 1.3568, "step": 4270 }, { "epoch": 0.53, "learning_rate": 0.00023279526507727943, "loss": 1.2733, "step": 4271 }, { "epoch": 0.53, "learning_rate": 0.00023269481312741453, "loss": 1.2867, "step": 4272 }, { "epoch": 0.53, "learning_rate": 0.00023259436398483669, "loss": 1.3692, "step": 4273 }, { "epoch": 0.53, "learning_rate": 0.0002324939176658409, "loss": 1.3709, "step": 4274 }, { "epoch": 0.53, "learning_rate": 0.00023239347418672202, "loss": 1.2886, "step": 4275 }, { "epoch": 0.53, "learning_rate": 0.00023229303356377416, "loss": 1.4196, "step": 4276 }, { "epoch": 0.53, "learning_rate": 0.00023219259581329092, "loss": 1.3165, "step": 4277 }, { "epoch": 0.53, "learning_rate": 0.00023209216095156565, "loss": 1.3226, "step": 4278 }, { "epoch": 0.53, "learning_rate": 0.00023199172899489098, "loss": 1.3046, "step": 4279 }, { "epoch": 0.54, "learning_rate": 0.00023189129995955944, "loss": 1.338, "step": 4280 }, { "epoch": 0.54, "learning_rate": 0.00023179087386186265, "loss": 1.3948, "step": 4281 }, { "epoch": 0.54, "learning_rate": 0.00023169045071809213, "loss": 1.3197, "step": 4282 }, { "epoch": 0.54, "learning_rate": 0.00023159003054453857, "loss": 1.349, "step": 4283 }, { "epoch": 0.54, "learning_rate": 0.00023148961335749246, "loss": 1.3604, "step": 4284 }, { "epoch": 0.54, "learning_rate": 0.00023138919917324376, "loss": 1.2755, "step": 4285 }, { "epoch": 0.54, "learning_rate": 0.0002312887880080818, "loss": 1.3616, "step": 4286 }, { "epoch": 0.54, "learning_rate": 0.0002311883798782956, "loss": 1.2721, "step": 4287 }, { "epoch": 0.54, "learning_rate": 0.00023108797480017357, "loss": 1.314, "step": 4288 }, { "epoch": 0.54, "learning_rate": 0.0002309875727900035, "loss": 1.3177, "step": 4289 }, { "epoch": 0.54, "learning_rate": 0.00023088717386407295, "loss": 1.319, "step": 4290 }, { "epoch": 0.54, "learning_rate": 0.000230786778038669, "loss": 1.3444, "step": 4291 }, { "epoch": 0.54, "learning_rate": 0.00023068638533007784, "loss": 1.3132, "step": 4292 }, { "epoch": 0.54, "learning_rate": 0.00023058599575458561, "loss": 1.3579, "step": 4293 }, { "epoch": 0.54, "learning_rate": 0.00023048560932847757, "loss": 1.3173, "step": 4294 }, { "epoch": 0.54, "learning_rate": 0.0002303852260680388, "loss": 1.3156, "step": 4295 }, { "epoch": 0.54, "learning_rate": 0.00023028484598955354, "loss": 1.3079, "step": 4296 }, { "epoch": 0.54, "learning_rate": 0.00023018446910930585, "loss": 1.2757, "step": 4297 }, { "epoch": 0.54, "learning_rate": 0.00023008409544357893, "loss": 1.3121, "step": 4298 }, { "epoch": 0.54, "learning_rate": 0.00022998372500865566, "loss": 1.3181, "step": 4299 }, { "epoch": 0.54, "learning_rate": 0.00022988335782081853, "loss": 1.3937, "step": 4300 }, { "epoch": 0.54, "eval_acceptance_rate": 0.6837840676307678, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1517345641445598, "eval_expected_tokens": 2.6896682051806997, "eval_loss": 1.4035065174102783, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4224170512951749, "eval_runtime": 50.3193, "eval_samples_per_second": 9.678, "eval_steps_per_second": 0.318, "step": 4300 }, { "epoch": 0.54, "learning_rate": 0.0002297829938963492, "loss": 1.3212, "step": 4301 }, { "epoch": 0.54, "learning_rate": 0.0002296826332515289, "loss": 1.322, "step": 4302 }, { "epoch": 0.54, "learning_rate": 0.00022958227590263842, "loss": 1.3844, "step": 4303 }, { "epoch": 0.54, "learning_rate": 0.00022948192186595786, "loss": 1.2846, "step": 4304 }, { "epoch": 0.54, "learning_rate": 0.00022938157115776704, "loss": 1.2835, "step": 4305 }, { "epoch": 0.54, "learning_rate": 0.00022928122379434505, "loss": 1.285, "step": 4306 }, { "epoch": 0.54, "learning_rate": 0.00022918087979197045, "loss": 1.3608, "step": 4307 }, { "epoch": 0.54, "learning_rate": 0.00022908053916692117, "loss": 1.2873, "step": 4308 }, { "epoch": 0.54, "learning_rate": 0.00022898020193547476, "loss": 1.3466, "step": 4309 }, { "epoch": 0.54, "learning_rate": 0.00022887986811390826, "loss": 1.3551, "step": 4310 }, { "epoch": 0.54, "learning_rate": 0.0002287795377184979, "loss": 1.3789, "step": 4311 }, { "epoch": 0.54, "learning_rate": 0.00022867921076551967, "loss": 1.3734, "step": 4312 }, { "epoch": 0.54, "learning_rate": 0.00022857888727124872, "loss": 1.3102, "step": 4313 }, { "epoch": 0.54, "learning_rate": 0.0002284785672519597, "loss": 1.2659, "step": 4314 }, { "epoch": 0.54, "learning_rate": 0.00022837825072392686, "loss": 1.3454, "step": 4315 }, { "epoch": 0.54, "learning_rate": 0.00022827793770342384, "loss": 1.3813, "step": 4316 }, { "epoch": 0.54, "learning_rate": 0.0002281776282067235, "loss": 1.3565, "step": 4317 }, { "epoch": 0.54, "learning_rate": 0.00022807732225009836, "loss": 1.3126, "step": 4318 }, { "epoch": 0.54, "learning_rate": 0.00022797701984982017, "loss": 1.3014, "step": 4319 }, { "epoch": 0.54, "learning_rate": 0.00022787672102216042, "loss": 1.377, "step": 4320 }, { "epoch": 0.54, "learning_rate": 0.00022777642578338965, "loss": 1.3336, "step": 4321 }, { "epoch": 0.54, "learning_rate": 0.00022767613414977807, "loss": 1.3215, "step": 4322 }, { "epoch": 0.54, "learning_rate": 0.00022757584613759506, "loss": 1.2716, "step": 4323 }, { "epoch": 0.54, "learning_rate": 0.00022747556176310978, "loss": 1.331, "step": 4324 }, { "epoch": 0.54, "learning_rate": 0.00022737528104259055, "loss": 1.3531, "step": 4325 }, { "epoch": 0.54, "learning_rate": 0.0002272750039923051, "loss": 1.354, "step": 4326 }, { "epoch": 0.54, "learning_rate": 0.00022717473062852054, "loss": 1.2836, "step": 4327 }, { "epoch": 0.54, "learning_rate": 0.00022707446096750344, "loss": 1.3522, "step": 4328 }, { "epoch": 0.54, "learning_rate": 0.00022697419502551994, "loss": 1.3189, "step": 4329 }, { "epoch": 0.54, "learning_rate": 0.00022687393281883524, "loss": 1.3394, "step": 4330 }, { "epoch": 0.54, "learning_rate": 0.00022677367436371428, "loss": 1.3419, "step": 4331 }, { "epoch": 0.54, "learning_rate": 0.00022667341967642105, "loss": 1.3714, "step": 4332 }, { "epoch": 0.54, "learning_rate": 0.00022657316877321907, "loss": 1.2611, "step": 4333 }, { "epoch": 0.54, "learning_rate": 0.00022647292167037142, "loss": 1.4318, "step": 4334 }, { "epoch": 0.54, "learning_rate": 0.0002263726783841404, "loss": 1.3473, "step": 4335 }, { "epoch": 0.54, "learning_rate": 0.0002262724389307876, "loss": 1.2997, "step": 4336 }, { "epoch": 0.54, "learning_rate": 0.00022617220332657418, "loss": 1.2888, "step": 4337 }, { "epoch": 0.54, "learning_rate": 0.00022607197158776044, "loss": 1.3453, "step": 4338 }, { "epoch": 0.54, "learning_rate": 0.00022597174373060647, "loss": 1.2858, "step": 4339 }, { "epoch": 0.54, "learning_rate": 0.00022587151977137123, "loss": 1.3918, "step": 4340 }, { "epoch": 0.54, "learning_rate": 0.00022577129972631345, "loss": 1.2789, "step": 4341 }, { "epoch": 0.54, "learning_rate": 0.00022567108361169086, "loss": 1.3583, "step": 4342 }, { "epoch": 0.54, "learning_rate": 0.00022557087144376085, "loss": 1.3785, "step": 4343 }, { "epoch": 0.54, "learning_rate": 0.0002254706632387802, "loss": 1.2797, "step": 4344 }, { "epoch": 0.54, "learning_rate": 0.00022537045901300474, "loss": 1.3086, "step": 4345 }, { "epoch": 0.54, "learning_rate": 0.00022527025878268985, "loss": 1.3793, "step": 4346 }, { "epoch": 0.54, "learning_rate": 0.00022517006256409032, "loss": 1.3296, "step": 4347 }, { "epoch": 0.54, "learning_rate": 0.00022506987037346004, "loss": 1.3437, "step": 4348 }, { "epoch": 0.54, "learning_rate": 0.0002249696822270526, "loss": 1.4014, "step": 4349 }, { "epoch": 0.54, "learning_rate": 0.00022486949814112076, "loss": 1.3585, "step": 4350 }, { "epoch": 0.54, "learning_rate": 0.00022476931813191652, "loss": 1.3095, "step": 4351 }, { "epoch": 0.54, "learning_rate": 0.0002246691422156913, "loss": 1.3638, "step": 4352 }, { "epoch": 0.54, "learning_rate": 0.0002245689704086958, "loss": 1.3492, "step": 4353 }, { "epoch": 0.54, "learning_rate": 0.00022446880272718037, "loss": 1.3447, "step": 4354 }, { "epoch": 0.54, "learning_rate": 0.0002243686391873942, "loss": 1.3818, "step": 4355 }, { "epoch": 0.54, "learning_rate": 0.0002242684798055862, "loss": 1.375, "step": 4356 }, { "epoch": 0.54, "learning_rate": 0.0002241683245980044, "loss": 1.3655, "step": 4357 }, { "epoch": 0.54, "learning_rate": 0.00022406817358089609, "loss": 1.2757, "step": 4358 }, { "epoch": 0.54, "learning_rate": 0.00022396802677050818, "loss": 1.379, "step": 4359 }, { "epoch": 0.55, "learning_rate": 0.00022386788418308668, "loss": 1.3153, "step": 4360 }, { "epoch": 0.55, "learning_rate": 0.00022376774583487686, "loss": 1.3819, "step": 4361 }, { "epoch": 0.55, "learning_rate": 0.00022366761174212347, "loss": 1.337, "step": 4362 }, { "epoch": 0.55, "learning_rate": 0.00022356748192107042, "loss": 1.3737, "step": 4363 }, { "epoch": 0.55, "learning_rate": 0.0002234673563879611, "loss": 1.3432, "step": 4364 }, { "epoch": 0.55, "learning_rate": 0.000223367235159038, "loss": 1.4075, "step": 4365 }, { "epoch": 0.55, "learning_rate": 0.00022326711825054312, "loss": 1.3116, "step": 4366 }, { "epoch": 0.55, "learning_rate": 0.00022316700567871756, "loss": 1.3065, "step": 4367 }, { "epoch": 0.55, "learning_rate": 0.0002230668974598018, "loss": 1.3081, "step": 4368 }, { "epoch": 0.55, "learning_rate": 0.00022296679361003577, "loss": 1.3471, "step": 4369 }, { "epoch": 0.55, "learning_rate": 0.00022286669414565846, "loss": 1.3582, "step": 4370 }, { "epoch": 0.55, "learning_rate": 0.00022276659908290817, "loss": 1.3281, "step": 4371 }, { "epoch": 0.55, "learning_rate": 0.00022266650843802256, "loss": 1.2976, "step": 4372 }, { "epoch": 0.55, "learning_rate": 0.00022256642222723868, "loss": 1.3821, "step": 4373 }, { "epoch": 0.55, "learning_rate": 0.00022246634046679264, "loss": 1.343, "step": 4374 }, { "epoch": 0.55, "learning_rate": 0.00022236626317292005, "loss": 1.2895, "step": 4375 }, { "epoch": 0.55, "learning_rate": 0.00022226619036185558, "loss": 1.3254, "step": 4376 }, { "epoch": 0.55, "learning_rate": 0.00022216612204983317, "loss": 1.3005, "step": 4377 }, { "epoch": 0.55, "learning_rate": 0.00022206605825308629, "loss": 1.3035, "step": 4378 }, { "epoch": 0.55, "learning_rate": 0.00022196599898784752, "loss": 1.3245, "step": 4379 }, { "epoch": 0.55, "learning_rate": 0.00022186594427034864, "loss": 1.3221, "step": 4380 }, { "epoch": 0.55, "learning_rate": 0.0002217658941168208, "loss": 1.3386, "step": 4381 }, { "epoch": 0.55, "learning_rate": 0.00022166584854349424, "loss": 1.3489, "step": 4382 }, { "epoch": 0.55, "learning_rate": 0.0002215658075665988, "loss": 1.3908, "step": 4383 }, { "epoch": 0.55, "learning_rate": 0.00022146577120236315, "loss": 1.3939, "step": 4384 }, { "epoch": 0.55, "learning_rate": 0.0002213657394670156, "loss": 1.2905, "step": 4385 }, { "epoch": 0.55, "learning_rate": 0.00022126571237678338, "loss": 1.3553, "step": 4386 }, { "epoch": 0.55, "learning_rate": 0.0002211656899478931, "loss": 1.3339, "step": 4387 }, { "epoch": 0.55, "learning_rate": 0.0002210656721965708, "loss": 1.3384, "step": 4388 }, { "epoch": 0.55, "learning_rate": 0.0002209656591390415, "loss": 1.3359, "step": 4389 }, { "epoch": 0.55, "learning_rate": 0.00022086565079152947, "loss": 1.3225, "step": 4390 }, { "epoch": 0.55, "learning_rate": 0.0002207656471702584, "loss": 1.3325, "step": 4391 }, { "epoch": 0.55, "learning_rate": 0.00022066564829145097, "loss": 1.3157, "step": 4392 }, { "epoch": 0.55, "learning_rate": 0.00022056565417132934, "loss": 1.3664, "step": 4393 }, { "epoch": 0.55, "learning_rate": 0.00022046566482611485, "loss": 1.2648, "step": 4394 }, { "epoch": 0.55, "learning_rate": 0.00022036568027202793, "loss": 1.3842, "step": 4395 }, { "epoch": 0.55, "learning_rate": 0.00022026570052528817, "loss": 1.3756, "step": 4396 }, { "epoch": 0.55, "learning_rate": 0.00022016572560211464, "loss": 1.359, "step": 4397 }, { "epoch": 0.55, "learning_rate": 0.0002200657555187256, "loss": 1.2939, "step": 4398 }, { "epoch": 0.55, "learning_rate": 0.00021996579029133823, "loss": 1.3113, "step": 4399 }, { "epoch": 0.55, "learning_rate": 0.00021986582993616926, "loss": 1.3621, "step": 4400 }, { "epoch": 0.55, "eval_acceptance_rate": 0.684598982334137, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.155029340820147, "eval_expected_tokens": 2.6937866760251836, "eval_loss": 1.3975615501403809, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4234466690062959, "eval_runtime": 50.3436, "eval_samples_per_second": 9.674, "eval_steps_per_second": 0.318, "step": 4400 }, { "epoch": 0.55, "learning_rate": 0.0002197658744694344, "loss": 1.3892, "step": 4401 }, { "epoch": 0.55, "learning_rate": 0.0002196659239073487, "loss": 1.2643, "step": 4402 }, { "epoch": 0.55, "learning_rate": 0.00021956597826612638, "loss": 1.3405, "step": 4403 }, { "epoch": 0.55, "learning_rate": 0.00021946603756198084, "loss": 1.32, "step": 4404 }, { "epoch": 0.55, "learning_rate": 0.00021936610181112466, "loss": 1.3643, "step": 4405 }, { "epoch": 0.55, "learning_rate": 0.00021926617102976972, "loss": 1.3506, "step": 4406 }, { "epoch": 0.55, "learning_rate": 0.00021916624523412684, "loss": 1.374, "step": 4407 }, { "epoch": 0.55, "learning_rate": 0.00021906632444040642, "loss": 1.2734, "step": 4408 }, { "epoch": 0.55, "learning_rate": 0.0002189664086648178, "loss": 1.3031, "step": 4409 }, { "epoch": 0.55, "learning_rate": 0.0002188664979235695, "loss": 1.3606, "step": 4410 }, { "epoch": 0.55, "learning_rate": 0.00021876659223286923, "loss": 1.3511, "step": 4411 }, { "epoch": 0.55, "learning_rate": 0.00021866669160892392, "loss": 1.3579, "step": 4412 }, { "epoch": 0.55, "learning_rate": 0.00021856679606793982, "loss": 1.3932, "step": 4413 }, { "epoch": 0.55, "learning_rate": 0.00021846690562612206, "loss": 1.4123, "step": 4414 }, { "epoch": 0.55, "learning_rate": 0.00021836702029967524, "loss": 1.4357, "step": 4415 }, { "epoch": 0.55, "learning_rate": 0.00021826714010480286, "loss": 1.3807, "step": 4416 }, { "epoch": 0.55, "learning_rate": 0.00021816726505770768, "loss": 1.3813, "step": 4417 }, { "epoch": 0.55, "learning_rate": 0.0002180673951745918, "loss": 1.3429, "step": 4418 }, { "epoch": 0.55, "learning_rate": 0.0002179675304716563, "loss": 1.3621, "step": 4419 }, { "epoch": 0.55, "learning_rate": 0.00021786767096510138, "loss": 1.3971, "step": 4420 }, { "epoch": 0.55, "learning_rate": 0.00021776781667112648, "loss": 1.4299, "step": 4421 }, { "epoch": 0.55, "learning_rate": 0.00021766796760593036, "loss": 1.3536, "step": 4422 }, { "epoch": 0.55, "learning_rate": 0.00021756812378571067, "loss": 1.4171, "step": 4423 }, { "epoch": 0.55, "learning_rate": 0.0002174682852266642, "loss": 1.3716, "step": 4424 }, { "epoch": 0.55, "learning_rate": 0.00021736845194498717, "loss": 1.3568, "step": 4425 }, { "epoch": 0.55, "learning_rate": 0.0002172686239568745, "loss": 1.2887, "step": 4426 }, { "epoch": 0.55, "learning_rate": 0.00021716880127852078, "loss": 1.411, "step": 4427 }, { "epoch": 0.55, "learning_rate": 0.00021706898392611946, "loss": 1.2801, "step": 4428 }, { "epoch": 0.55, "learning_rate": 0.00021696917191586302, "loss": 1.2784, "step": 4429 }, { "epoch": 0.55, "learning_rate": 0.00021686936526394318, "loss": 1.2755, "step": 4430 }, { "epoch": 0.55, "learning_rate": 0.0002167695639865508, "loss": 1.2915, "step": 4431 }, { "epoch": 0.55, "learning_rate": 0.00021666976809987603, "loss": 1.3198, "step": 4432 }, { "epoch": 0.55, "learning_rate": 0.00021656997762010787, "loss": 1.3209, "step": 4433 }, { "epoch": 0.55, "learning_rate": 0.0002164701925634346, "loss": 1.3119, "step": 4434 }, { "epoch": 0.55, "learning_rate": 0.00021637041294604353, "loss": 1.354, "step": 4435 }, { "epoch": 0.55, "learning_rate": 0.00021627063878412107, "loss": 1.3389, "step": 4436 }, { "epoch": 0.55, "learning_rate": 0.00021617087009385296, "loss": 1.3209, "step": 4437 }, { "epoch": 0.55, "learning_rate": 0.00021607110689142393, "loss": 1.3342, "step": 4438 }, { "epoch": 0.55, "learning_rate": 0.00021597134919301765, "loss": 1.3785, "step": 4439 }, { "epoch": 0.56, "learning_rate": 0.00021587159701481716, "loss": 1.3565, "step": 4440 }, { "epoch": 0.56, "learning_rate": 0.00021577185037300434, "loss": 1.3749, "step": 4441 }, { "epoch": 0.56, "learning_rate": 0.0002156721092837605, "loss": 1.3564, "step": 4442 }, { "epoch": 0.56, "learning_rate": 0.00021557237376326576, "loss": 1.3942, "step": 4443 }, { "epoch": 0.56, "learning_rate": 0.0002154726438276995, "loss": 1.4006, "step": 4444 }, { "epoch": 0.56, "learning_rate": 0.00021537291949324006, "loss": 1.3348, "step": 4445 }, { "epoch": 0.56, "learning_rate": 0.00021527320077606495, "loss": 1.3499, "step": 4446 }, { "epoch": 0.56, "learning_rate": 0.0002151734876923509, "loss": 1.4214, "step": 4447 }, { "epoch": 0.56, "learning_rate": 0.00021507378025827352, "loss": 1.3444, "step": 4448 }, { "epoch": 0.56, "learning_rate": 0.00021497407849000756, "loss": 1.3147, "step": 4449 }, { "epoch": 0.56, "learning_rate": 0.0002148743824037269, "loss": 1.3432, "step": 4450 }, { "epoch": 0.56, "learning_rate": 0.00021477469201560434, "loss": 1.3038, "step": 4451 }, { "epoch": 0.56, "learning_rate": 0.00021467500734181203, "loss": 1.2642, "step": 4452 }, { "epoch": 0.56, "learning_rate": 0.0002145753283985211, "loss": 1.3697, "step": 4453 }, { "epoch": 0.56, "learning_rate": 0.00021447565520190157, "loss": 1.4059, "step": 4454 }, { "epoch": 0.56, "learning_rate": 0.00021437598776812263, "loss": 1.3875, "step": 4455 }, { "epoch": 0.56, "learning_rate": 0.00021427632611335258, "loss": 1.3006, "step": 4456 }, { "epoch": 0.56, "learning_rate": 0.0002141766702537589, "loss": 1.363, "step": 4457 }, { "epoch": 0.56, "learning_rate": 0.00021407702020550786, "loss": 1.3549, "step": 4458 }, { "epoch": 0.56, "learning_rate": 0.00021397737598476496, "loss": 1.3475, "step": 4459 }, { "epoch": 0.56, "learning_rate": 0.00021387773760769474, "loss": 1.2944, "step": 4460 }, { "epoch": 0.56, "learning_rate": 0.00021377810509046064, "loss": 1.2933, "step": 4461 }, { "epoch": 0.56, "learning_rate": 0.00021367847844922538, "loss": 1.3581, "step": 4462 }, { "epoch": 0.56, "learning_rate": 0.00021357885770015067, "loss": 1.3695, "step": 4463 }, { "epoch": 0.56, "learning_rate": 0.00021347924285939712, "loss": 1.3953, "step": 4464 }, { "epoch": 0.56, "learning_rate": 0.00021337963394312454, "loss": 1.3781, "step": 4465 }, { "epoch": 0.56, "learning_rate": 0.0002132800309674916, "loss": 1.3599, "step": 4466 }, { "epoch": 0.56, "learning_rate": 0.00021318043394865632, "loss": 1.3835, "step": 4467 }, { "epoch": 0.56, "learning_rate": 0.0002130808429027754, "loss": 1.3729, "step": 4468 }, { "epoch": 0.56, "learning_rate": 0.0002129812578460048, "loss": 1.3029, "step": 4469 }, { "epoch": 0.56, "learning_rate": 0.00021288167879449934, "loss": 1.3432, "step": 4470 }, { "epoch": 0.56, "learning_rate": 0.00021278210576441308, "loss": 1.2925, "step": 4471 }, { "epoch": 0.56, "learning_rate": 0.000212682538771899, "loss": 1.3259, "step": 4472 }, { "epoch": 0.56, "learning_rate": 0.000212582977833109, "loss": 1.2917, "step": 4473 }, { "epoch": 0.56, "learning_rate": 0.00021248342296419407, "loss": 1.3578, "step": 4474 }, { "epoch": 0.56, "learning_rate": 0.00021238387418130424, "loss": 1.3129, "step": 4475 }, { "epoch": 0.56, "learning_rate": 0.00021228433150058862, "loss": 1.2773, "step": 4476 }, { "epoch": 0.56, "learning_rate": 0.0002121847949381952, "loss": 1.2327, "step": 4477 }, { "epoch": 0.56, "learning_rate": 0.00021208526451027104, "loss": 1.3491, "step": 4478 }, { "epoch": 0.56, "learning_rate": 0.00021198574023296223, "loss": 1.3443, "step": 4479 }, { "epoch": 0.56, "learning_rate": 0.00021188622212241364, "loss": 1.3473, "step": 4480 }, { "epoch": 0.56, "learning_rate": 0.00021178671019476955, "loss": 1.3961, "step": 4481 }, { "epoch": 0.56, "learning_rate": 0.00021168720446617298, "loss": 1.3387, "step": 4482 }, { "epoch": 0.56, "learning_rate": 0.00021158770495276585, "loss": 1.311, "step": 4483 }, { "epoch": 0.56, "learning_rate": 0.00021148821167068938, "loss": 1.2769, "step": 4484 }, { "epoch": 0.56, "learning_rate": 0.00021138872463608333, "loss": 1.2904, "step": 4485 }, { "epoch": 0.56, "learning_rate": 0.00021128924386508704, "loss": 1.3014, "step": 4486 }, { "epoch": 0.56, "learning_rate": 0.00021118976937383827, "loss": 1.3441, "step": 4487 }, { "epoch": 0.56, "learning_rate": 0.00021109030117847413, "loss": 1.3162, "step": 4488 }, { "epoch": 0.56, "learning_rate": 0.00021099083929513048, "loss": 1.2634, "step": 4489 }, { "epoch": 0.56, "learning_rate": 0.00021089138373994224, "loss": 1.3307, "step": 4490 }, { "epoch": 0.56, "learning_rate": 0.0002107919345290435, "loss": 1.3367, "step": 4491 }, { "epoch": 0.56, "learning_rate": 0.000210692491678567, "loss": 1.355, "step": 4492 }, { "epoch": 0.56, "learning_rate": 0.00021059305520464462, "loss": 1.2304, "step": 4493 }, { "epoch": 0.56, "learning_rate": 0.00021049362512340713, "loss": 1.3265, "step": 4494 }, { "epoch": 0.56, "learning_rate": 0.00021039420145098426, "loss": 1.3551, "step": 4495 }, { "epoch": 0.56, "learning_rate": 0.0002102947842035049, "loss": 1.3505, "step": 4496 }, { "epoch": 0.56, "learning_rate": 0.00021019537339709668, "loss": 1.3312, "step": 4497 }, { "epoch": 0.56, "learning_rate": 0.00021009596904788625, "loss": 1.2682, "step": 4498 }, { "epoch": 0.56, "learning_rate": 0.00020999657117199916, "loss": 1.2546, "step": 4499 }, { "epoch": 0.56, "learning_rate": 0.0002098971797855599, "loss": 1.367, "step": 4500 }, { "epoch": 0.56, "eval_acceptance_rate": 0.6862798929214478, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1618451180680287, "eval_expected_tokens": 2.702306397585036, "eval_loss": 1.3925145864486694, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.425576599396259, "eval_runtime": 50.2982, "eval_samples_per_second": 9.682, "eval_steps_per_second": 0.318, "step": 4500 }, { "epoch": 0.56, "learning_rate": 0.0002097977949046922, "loss": 1.3368, "step": 4501 }, { "epoch": 0.56, "learning_rate": 0.00020969841654551832, "loss": 1.3558, "step": 4502 }, { "epoch": 0.56, "learning_rate": 0.00020959904472415974, "loss": 1.332, "step": 4503 }, { "epoch": 0.56, "learning_rate": 0.00020949967945673676, "loss": 1.3424, "step": 4504 }, { "epoch": 0.56, "learning_rate": 0.0002094003207593685, "loss": 1.388, "step": 4505 }, { "epoch": 0.56, "learning_rate": 0.00020930096864817333, "loss": 1.4562, "step": 4506 }, { "epoch": 0.56, "learning_rate": 0.00020920162313926835, "loss": 1.3568, "step": 4507 }, { "epoch": 0.56, "learning_rate": 0.00020910228424876956, "loss": 1.3311, "step": 4508 }, { "epoch": 0.56, "learning_rate": 0.00020900295199279204, "loss": 1.3159, "step": 4509 }, { "epoch": 0.56, "learning_rate": 0.0002089036263874495, "loss": 1.366, "step": 4510 }, { "epoch": 0.56, "learning_rate": 0.00020880430744885502, "loss": 1.2929, "step": 4511 }, { "epoch": 0.56, "learning_rate": 0.00020870499519312015, "loss": 1.3959, "step": 4512 }, { "epoch": 0.56, "learning_rate": 0.00020860568963635567, "loss": 1.2898, "step": 4513 }, { "epoch": 0.56, "learning_rate": 0.00020850639079467105, "loss": 1.3472, "step": 4514 }, { "epoch": 0.56, "learning_rate": 0.0002084070986841748, "loss": 1.3067, "step": 4515 }, { "epoch": 0.56, "learning_rate": 0.00020830781332097445, "loss": 1.3199, "step": 4516 }, { "epoch": 0.56, "learning_rate": 0.0002082085347211762, "loss": 1.3606, "step": 4517 }, { "epoch": 0.56, "learning_rate": 0.0002081092629008852, "loss": 1.3201, "step": 4518 }, { "epoch": 0.56, "learning_rate": 0.00020800999787620554, "loss": 1.3442, "step": 4519 }, { "epoch": 0.56, "learning_rate": 0.00020791073966324036, "loss": 1.3198, "step": 4520 }, { "epoch": 0.57, "learning_rate": 0.00020781148827809145, "loss": 1.395, "step": 4521 }, { "epoch": 0.57, "learning_rate": 0.0002077122437368597, "loss": 1.3447, "step": 4522 }, { "epoch": 0.57, "learning_rate": 0.0002076130060556446, "loss": 1.307, "step": 4523 }, { "epoch": 0.57, "learning_rate": 0.00020751377525054488, "loss": 1.3066, "step": 4524 }, { "epoch": 0.57, "learning_rate": 0.0002074145513376579, "loss": 1.3325, "step": 4525 }, { "epoch": 0.57, "learning_rate": 0.0002073153343330801, "loss": 1.2933, "step": 4526 }, { "epoch": 0.57, "learning_rate": 0.00020721612425290652, "loss": 1.2771, "step": 4527 }, { "epoch": 0.57, "learning_rate": 0.00020711692111323143, "loss": 1.3522, "step": 4528 }, { "epoch": 0.57, "learning_rate": 0.00020701772493014758, "loss": 1.3168, "step": 4529 }, { "epoch": 0.57, "learning_rate": 0.00020691853571974705, "loss": 1.2965, "step": 4530 }, { "epoch": 0.57, "learning_rate": 0.00020681935349812036, "loss": 1.3764, "step": 4531 }, { "epoch": 0.57, "learning_rate": 0.00020672017828135718, "loss": 1.3038, "step": 4532 }, { "epoch": 0.57, "learning_rate": 0.00020662101008554583, "loss": 1.3165, "step": 4533 }, { "epoch": 0.57, "learning_rate": 0.0002065218489267736, "loss": 1.289, "step": 4534 }, { "epoch": 0.57, "learning_rate": 0.0002064226948211268, "loss": 1.3455, "step": 4535 }, { "epoch": 0.57, "learning_rate": 0.0002063235477846903, "loss": 1.3187, "step": 4536 }, { "epoch": 0.57, "learning_rate": 0.0002062244078335481, "loss": 1.308, "step": 4537 }, { "epoch": 0.57, "learning_rate": 0.00020612527498378274, "loss": 1.2314, "step": 4538 }, { "epoch": 0.57, "learning_rate": 0.0002060261492514758, "loss": 1.3514, "step": 4539 }, { "epoch": 0.57, "learning_rate": 0.00020592703065270776, "loss": 1.3442, "step": 4540 }, { "epoch": 0.57, "learning_rate": 0.0002058279192035579, "loss": 1.3606, "step": 4541 }, { "epoch": 0.57, "learning_rate": 0.0002057288149201042, "loss": 1.3332, "step": 4542 }, { "epoch": 0.57, "learning_rate": 0.00020562971781842366, "loss": 1.3633, "step": 4543 }, { "epoch": 0.57, "learning_rate": 0.00020553062791459192, "loss": 1.3101, "step": 4544 }, { "epoch": 0.57, "learning_rate": 0.00020543154522468377, "loss": 1.267, "step": 4545 }, { "epoch": 0.57, "learning_rate": 0.00020533246976477243, "loss": 1.2824, "step": 4546 }, { "epoch": 0.57, "learning_rate": 0.00020523340155093034, "loss": 1.2657, "step": 4547 }, { "epoch": 0.57, "learning_rate": 0.0002051343405992284, "loss": 1.2327, "step": 4548 }, { "epoch": 0.57, "learning_rate": 0.0002050352869257365, "loss": 1.3435, "step": 4549 }, { "epoch": 0.57, "learning_rate": 0.00020493624054652358, "loss": 1.269, "step": 4550 }, { "epoch": 0.57, "learning_rate": 0.000204837201477657, "loss": 1.3262, "step": 4551 }, { "epoch": 0.57, "learning_rate": 0.00020473816973520307, "loss": 1.3554, "step": 4552 }, { "epoch": 0.57, "learning_rate": 0.00020463914533522704, "loss": 1.2874, "step": 4553 }, { "epoch": 0.57, "learning_rate": 0.00020454012829379276, "loss": 1.3579, "step": 4554 }, { "epoch": 0.57, "learning_rate": 0.00020444111862696313, "loss": 1.3729, "step": 4555 }, { "epoch": 0.57, "learning_rate": 0.00020434211635079973, "loss": 1.323, "step": 4556 }, { "epoch": 0.57, "learning_rate": 0.00020424312148136284, "loss": 1.3364, "step": 4557 }, { "epoch": 0.57, "learning_rate": 0.00020414413403471165, "loss": 1.352, "step": 4558 }, { "epoch": 0.57, "learning_rate": 0.0002040451540269041, "loss": 1.3332, "step": 4559 }, { "epoch": 0.57, "learning_rate": 0.00020394618147399714, "loss": 1.2533, "step": 4560 }, { "epoch": 0.57, "learning_rate": 0.00020384721639204613, "loss": 1.2829, "step": 4561 }, { "epoch": 0.57, "learning_rate": 0.0002037482587971055, "loss": 1.2985, "step": 4562 }, { "epoch": 0.57, "learning_rate": 0.00020364930870522837, "loss": 1.288, "step": 4563 }, { "epoch": 0.57, "learning_rate": 0.00020355036613246654, "loss": 1.3206, "step": 4564 }, { "epoch": 0.57, "learning_rate": 0.00020345143109487082, "loss": 1.3051, "step": 4565 }, { "epoch": 0.57, "learning_rate": 0.00020335250360849073, "loss": 1.2774, "step": 4566 }, { "epoch": 0.57, "learning_rate": 0.0002032535836893744, "loss": 1.3058, "step": 4567 }, { "epoch": 0.57, "learning_rate": 0.0002031546713535688, "loss": 1.2798, "step": 4568 }, { "epoch": 0.57, "learning_rate": 0.0002030557666171199, "loss": 1.3169, "step": 4569 }, { "epoch": 0.57, "learning_rate": 0.00020295686949607216, "loss": 1.2828, "step": 4570 }, { "epoch": 0.57, "learning_rate": 0.00020285798000646888, "loss": 1.3229, "step": 4571 }, { "epoch": 0.57, "learning_rate": 0.00020275909816435216, "loss": 1.3001, "step": 4572 }, { "epoch": 0.57, "learning_rate": 0.00020266022398576273, "loss": 1.2749, "step": 4573 }, { "epoch": 0.57, "learning_rate": 0.00020256135748674038, "loss": 1.3005, "step": 4574 }, { "epoch": 0.57, "learning_rate": 0.00020246249868332336, "loss": 1.2591, "step": 4575 }, { "epoch": 0.57, "learning_rate": 0.0002023636475915488, "loss": 1.3359, "step": 4576 }, { "epoch": 0.57, "learning_rate": 0.00020226480422745247, "loss": 1.2689, "step": 4577 }, { "epoch": 0.57, "learning_rate": 0.00020216596860706898, "loss": 1.3766, "step": 4578 }, { "epoch": 0.57, "learning_rate": 0.00020206714074643182, "loss": 1.2662, "step": 4579 }, { "epoch": 0.57, "learning_rate": 0.00020196832066157288, "loss": 1.2561, "step": 4580 }, { "epoch": 0.57, "learning_rate": 0.0002018695083685232, "loss": 1.3101, "step": 4581 }, { "epoch": 0.57, "learning_rate": 0.00020177070388331214, "loss": 1.247, "step": 4582 }, { "epoch": 0.57, "learning_rate": 0.00020167190722196796, "loss": 1.2604, "step": 4583 }, { "epoch": 0.57, "learning_rate": 0.00020157311840051785, "loss": 1.3399, "step": 4584 }, { "epoch": 0.57, "learning_rate": 0.0002014743374349875, "loss": 1.3194, "step": 4585 }, { "epoch": 0.57, "learning_rate": 0.00020137556434140136, "loss": 1.3036, "step": 4586 }, { "epoch": 0.57, "learning_rate": 0.00020127679913578268, "loss": 1.3082, "step": 4587 }, { "epoch": 0.57, "learning_rate": 0.00020117804183415324, "loss": 1.3052, "step": 4588 }, { "epoch": 0.57, "learning_rate": 0.00020107929245253385, "loss": 1.327, "step": 4589 }, { "epoch": 0.57, "learning_rate": 0.00020098055100694373, "loss": 1.345, "step": 4590 }, { "epoch": 0.57, "learning_rate": 0.00020088181751340105, "loss": 1.34, "step": 4591 }, { "epoch": 0.57, "learning_rate": 0.0002007830919879225, "loss": 1.308, "step": 4592 }, { "epoch": 0.57, "learning_rate": 0.0002006843744465235, "loss": 1.3409, "step": 4593 }, { "epoch": 0.57, "learning_rate": 0.00020058566490521845, "loss": 1.3372, "step": 4594 }, { "epoch": 0.57, "learning_rate": 0.00020048696338002014, "loss": 1.292, "step": 4595 }, { "epoch": 0.57, "learning_rate": 0.00020038826988694007, "loss": 1.3287, "step": 4596 }, { "epoch": 0.57, "learning_rate": 0.00020028958444198863, "loss": 1.2909, "step": 4597 }, { "epoch": 0.57, "learning_rate": 0.00020019090706117464, "loss": 1.2948, "step": 4598 }, { "epoch": 0.57, "learning_rate": 0.00020009223776050594, "loss": 1.2956, "step": 4599 }, { "epoch": 0.57, "learning_rate": 0.00019999357655598894, "loss": 1.3146, "step": 4600 }, { "epoch": 0.57, "eval_acceptance_rate": 0.6875178813934326, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1668819362484597, "eval_expected_tokens": 2.7086024203105743, "eval_loss": 1.3872885704040527, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.42715060507764363, "eval_runtime": 50.3283, "eval_samples_per_second": 9.676, "eval_steps_per_second": 0.318, "step": 4600 }, { "epoch": 0.58, "learning_rate": 0.00019989492346362853, "loss": 1.3062, "step": 4601 }, { "epoch": 0.58, "learning_rate": 0.00019979627849942843, "loss": 1.3641, "step": 4602 }, { "epoch": 0.58, "learning_rate": 0.0001996976416793911, "loss": 1.268, "step": 4603 }, { "epoch": 0.58, "learning_rate": 0.00019959901301951777, "loss": 1.2564, "step": 4604 }, { "epoch": 0.58, "learning_rate": 0.000199500392535808, "loss": 1.2666, "step": 4605 }, { "epoch": 0.58, "learning_rate": 0.0001994017802442604, "loss": 1.2902, "step": 4606 }, { "epoch": 0.58, "learning_rate": 0.00019930317616087195, "loss": 1.2744, "step": 4607 }, { "epoch": 0.58, "learning_rate": 0.0001992045803016384, "loss": 1.2987, "step": 4608 }, { "epoch": 0.58, "learning_rate": 0.00019910599268255436, "loss": 1.296, "step": 4609 }, { "epoch": 0.58, "learning_rate": 0.00019900741331961287, "loss": 1.2622, "step": 4610 }, { "epoch": 0.58, "learning_rate": 0.00019890884222880563, "loss": 1.2668, "step": 4611 }, { "epoch": 0.58, "learning_rate": 0.00019881027942612306, "loss": 1.3835, "step": 4612 }, { "epoch": 0.58, "learning_rate": 0.00019871172492755447, "loss": 1.3382, "step": 4613 }, { "epoch": 0.58, "learning_rate": 0.00019861317874908738, "loss": 1.3465, "step": 4614 }, { "epoch": 0.58, "learning_rate": 0.0001985146409067082, "loss": 1.3083, "step": 4615 }, { "epoch": 0.58, "learning_rate": 0.00019841611141640204, "loss": 1.282, "step": 4616 }, { "epoch": 0.58, "learning_rate": 0.00019831759029415242, "loss": 1.3011, "step": 4617 }, { "epoch": 0.58, "learning_rate": 0.00019821907755594183, "loss": 1.4285, "step": 4618 }, { "epoch": 0.58, "learning_rate": 0.00019812057321775127, "loss": 1.2734, "step": 4619 }, { "epoch": 0.58, "learning_rate": 0.0001980220772955602, "loss": 1.2672, "step": 4620 }, { "epoch": 0.58, "learning_rate": 0.00019792358980534692, "loss": 1.2974, "step": 4621 }, { "epoch": 0.58, "learning_rate": 0.00019782511076308815, "loss": 1.3266, "step": 4622 }, { "epoch": 0.58, "learning_rate": 0.00019772664018475965, "loss": 1.3354, "step": 4623 }, { "epoch": 0.58, "learning_rate": 0.00019762817808633538, "loss": 1.404, "step": 4624 }, { "epoch": 0.58, "learning_rate": 0.00019752972448378815, "loss": 1.3435, "step": 4625 }, { "epoch": 0.58, "learning_rate": 0.0001974312793930893, "loss": 1.2898, "step": 4626 }, { "epoch": 0.58, "learning_rate": 0.0001973328428302087, "loss": 1.2785, "step": 4627 }, { "epoch": 0.58, "learning_rate": 0.00019723441481111513, "loss": 1.3804, "step": 4628 }, { "epoch": 0.58, "learning_rate": 0.00019713599535177578, "loss": 1.3297, "step": 4629 }, { "epoch": 0.58, "learning_rate": 0.0001970375844681564, "loss": 1.3895, "step": 4630 }, { "epoch": 0.58, "learning_rate": 0.00019693918217622157, "loss": 1.3205, "step": 4631 }, { "epoch": 0.58, "learning_rate": 0.0001968407884919341, "loss": 1.3495, "step": 4632 }, { "epoch": 0.58, "learning_rate": 0.00019674240343125587, "loss": 1.3163, "step": 4633 }, { "epoch": 0.58, "learning_rate": 0.00019664402701014702, "loss": 1.3059, "step": 4634 }, { "epoch": 0.58, "learning_rate": 0.00019654565924456647, "loss": 1.3336, "step": 4635 }, { "epoch": 0.58, "learning_rate": 0.00019644730015047155, "loss": 1.3323, "step": 4636 }, { "epoch": 0.58, "learning_rate": 0.0001963489497438183, "loss": 1.301, "step": 4637 }, { "epoch": 0.58, "learning_rate": 0.00019625060804056156, "loss": 1.3323, "step": 4638 }, { "epoch": 0.58, "learning_rate": 0.00019615227505665438, "loss": 1.3223, "step": 4639 }, { "epoch": 0.58, "learning_rate": 0.0001960539508080485, "loss": 1.3018, "step": 4640 }, { "epoch": 0.58, "learning_rate": 0.00019595563531069442, "loss": 1.321, "step": 4641 }, { "epoch": 0.58, "learning_rate": 0.00019585732858054102, "loss": 1.3093, "step": 4642 }, { "epoch": 0.58, "learning_rate": 0.00019575903063353593, "loss": 1.307, "step": 4643 }, { "epoch": 0.58, "learning_rate": 0.00019566074148562527, "loss": 1.3697, "step": 4644 }, { "epoch": 0.58, "learning_rate": 0.00019556246115275363, "loss": 1.304, "step": 4645 }, { "epoch": 0.58, "learning_rate": 0.00019546418965086444, "loss": 1.2655, "step": 4646 }, { "epoch": 0.58, "learning_rate": 0.00019536592699589926, "loss": 1.3223, "step": 4647 }, { "epoch": 0.58, "learning_rate": 0.00019526767320379875, "loss": 1.3183, "step": 4648 }, { "epoch": 0.58, "learning_rate": 0.00019516942829050172, "loss": 1.3755, "step": 4649 }, { "epoch": 0.58, "learning_rate": 0.0001950711922719458, "loss": 1.3249, "step": 4650 }, { "epoch": 0.58, "learning_rate": 0.0001949729651640669, "loss": 1.342, "step": 4651 }, { "epoch": 0.58, "learning_rate": 0.00019487474698279974, "loss": 1.3369, "step": 4652 }, { "epoch": 0.58, "learning_rate": 0.0001947765377440775, "loss": 1.3116, "step": 4653 }, { "epoch": 0.58, "learning_rate": 0.00019467833746383197, "loss": 1.3086, "step": 4654 }, { "epoch": 0.58, "learning_rate": 0.00019458014615799326, "loss": 1.2822, "step": 4655 }, { "epoch": 0.58, "learning_rate": 0.00019448196384249037, "loss": 1.2925, "step": 4656 }, { "epoch": 0.58, "learning_rate": 0.00019438379053325045, "loss": 1.3252, "step": 4657 }, { "epoch": 0.58, "learning_rate": 0.00019428562624619958, "loss": 1.3186, "step": 4658 }, { "epoch": 0.58, "learning_rate": 0.0001941874709972622, "loss": 1.277, "step": 4659 }, { "epoch": 0.58, "learning_rate": 0.00019408932480236119, "loss": 1.2606, "step": 4660 }, { "epoch": 0.58, "learning_rate": 0.000193991187677418, "loss": 1.3119, "step": 4661 }, { "epoch": 0.58, "learning_rate": 0.00019389305963835276, "loss": 1.3362, "step": 4662 }, { "epoch": 0.58, "learning_rate": 0.00019379494070108412, "loss": 1.3502, "step": 4663 }, { "epoch": 0.58, "learning_rate": 0.00019369683088152893, "loss": 1.3431, "step": 4664 }, { "epoch": 0.58, "learning_rate": 0.00019359873019560297, "loss": 1.3368, "step": 4665 }, { "epoch": 0.58, "learning_rate": 0.0001935006386592202, "loss": 1.3283, "step": 4666 }, { "epoch": 0.58, "learning_rate": 0.0001934025562882934, "loss": 1.3357, "step": 4667 }, { "epoch": 0.58, "learning_rate": 0.00019330448309873367, "loss": 1.3555, "step": 4668 }, { "epoch": 0.58, "learning_rate": 0.00019320641910645068, "loss": 1.2675, "step": 4669 }, { "epoch": 0.58, "learning_rate": 0.00019310836432735252, "loss": 1.3321, "step": 4670 }, { "epoch": 0.58, "learning_rate": 0.00019301031877734588, "loss": 1.2684, "step": 4671 }, { "epoch": 0.58, "learning_rate": 0.00019291228247233605, "loss": 1.3676, "step": 4672 }, { "epoch": 0.58, "learning_rate": 0.00019281425542822664, "loss": 1.3516, "step": 4673 }, { "epoch": 0.58, "learning_rate": 0.00019271623766091976, "loss": 1.2869, "step": 4674 }, { "epoch": 0.58, "learning_rate": 0.00019261822918631617, "loss": 1.3647, "step": 4675 }, { "epoch": 0.58, "learning_rate": 0.00019252023002031487, "loss": 1.3619, "step": 4676 }, { "epoch": 0.58, "learning_rate": 0.00019242224017881367, "loss": 1.2986, "step": 4677 }, { "epoch": 0.58, "learning_rate": 0.00019232425967770874, "loss": 1.3344, "step": 4678 }, { "epoch": 0.58, "learning_rate": 0.00019222628853289458, "loss": 1.3226, "step": 4679 }, { "epoch": 0.58, "learning_rate": 0.0001921283267602643, "loss": 1.38, "step": 4680 }, { "epoch": 0.59, "learning_rate": 0.00019203037437570943, "loss": 1.359, "step": 4681 }, { "epoch": 0.59, "learning_rate": 0.00019193243139512027, "loss": 1.3105, "step": 4682 }, { "epoch": 0.59, "learning_rate": 0.00019183449783438512, "loss": 1.3541, "step": 4683 }, { "epoch": 0.59, "learning_rate": 0.00019173657370939113, "loss": 1.2465, "step": 4684 }, { "epoch": 0.59, "learning_rate": 0.00019163865903602372, "loss": 1.3832, "step": 4685 }, { "epoch": 0.59, "learning_rate": 0.00019154075383016672, "loss": 1.3771, "step": 4686 }, { "epoch": 0.59, "learning_rate": 0.00019144285810770268, "loss": 1.3547, "step": 4687 }, { "epoch": 0.59, "learning_rate": 0.0001913449718845125, "loss": 1.3433, "step": 4688 }, { "epoch": 0.59, "learning_rate": 0.00019124709517647536, "loss": 1.334, "step": 4689 }, { "epoch": 0.59, "learning_rate": 0.00019114922799946923, "loss": 1.3037, "step": 4690 }, { "epoch": 0.59, "learning_rate": 0.0001910513703693701, "loss": 1.2943, "step": 4691 }, { "epoch": 0.59, "learning_rate": 0.00019095352230205293, "loss": 1.354, "step": 4692 }, { "epoch": 0.59, "learning_rate": 0.00019085568381339062, "loss": 1.3477, "step": 4693 }, { "epoch": 0.59, "learning_rate": 0.000190757854919255, "loss": 1.2716, "step": 4694 }, { "epoch": 0.59, "learning_rate": 0.00019066003563551585, "loss": 1.3371, "step": 4695 }, { "epoch": 0.59, "learning_rate": 0.00019056222597804167, "loss": 1.3472, "step": 4696 }, { "epoch": 0.59, "learning_rate": 0.0001904644259626996, "loss": 1.3529, "step": 4697 }, { "epoch": 0.59, "learning_rate": 0.00019036663560535483, "loss": 1.3694, "step": 4698 }, { "epoch": 0.59, "learning_rate": 0.00019026885492187106, "loss": 1.3259, "step": 4699 }, { "epoch": 0.59, "learning_rate": 0.00019017108392811063, "loss": 1.3108, "step": 4700 }, { "epoch": 0.59, "eval_acceptance_rate": 0.6886455416679382, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.171482466138646, "eval_expected_tokens": 2.7143530826733073, "eval_loss": 1.3813830614089966, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4285882706683269, "eval_runtime": 50.2819, "eval_samples_per_second": 9.685, "eval_steps_per_second": 0.318, "step": 4700 }, { "epoch": 0.59, "learning_rate": 0.00019007332263993402, "loss": 1.3099, "step": 4701 }, { "epoch": 0.59, "learning_rate": 0.00018997557107320042, "loss": 1.3646, "step": 4702 }, { "epoch": 0.59, "learning_rate": 0.00018987782924376735, "loss": 1.3541, "step": 4703 }, { "epoch": 0.59, "learning_rate": 0.00018978009716749062, "loss": 1.3066, "step": 4704 }, { "epoch": 0.59, "learning_rate": 0.00018968237486022453, "loss": 1.3052, "step": 4705 }, { "epoch": 0.59, "learning_rate": 0.0001895846623378218, "loss": 1.2919, "step": 4706 }, { "epoch": 0.59, "learning_rate": 0.00018948695961613373, "loss": 1.2676, "step": 4707 }, { "epoch": 0.59, "learning_rate": 0.0001893892667110098, "loss": 1.2887, "step": 4708 }, { "epoch": 0.59, "learning_rate": 0.00018929158363829797, "loss": 1.2639, "step": 4709 }, { "epoch": 0.59, "learning_rate": 0.00018919391041384448, "loss": 1.2867, "step": 4710 }, { "epoch": 0.59, "learning_rate": 0.00018909624705349437, "loss": 1.3138, "step": 4711 }, { "epoch": 0.59, "learning_rate": 0.00018899859357309062, "loss": 1.3221, "step": 4712 }, { "epoch": 0.59, "learning_rate": 0.00018890094998847494, "loss": 1.2948, "step": 4713 }, { "epoch": 0.59, "learning_rate": 0.00018880331631548715, "loss": 1.2913, "step": 4714 }, { "epoch": 0.59, "learning_rate": 0.0001887056925699656, "loss": 1.254, "step": 4715 }, { "epoch": 0.59, "learning_rate": 0.00018860807876774723, "loss": 1.3972, "step": 4716 }, { "epoch": 0.59, "learning_rate": 0.00018851047492466703, "loss": 1.3494, "step": 4717 }, { "epoch": 0.59, "learning_rate": 0.00018841288105655852, "loss": 1.3131, "step": 4718 }, { "epoch": 0.59, "learning_rate": 0.00018831529717925366, "loss": 1.3121, "step": 4719 }, { "epoch": 0.59, "learning_rate": 0.0001882177233085826, "loss": 1.2995, "step": 4720 }, { "epoch": 0.59, "learning_rate": 0.00018812015946037414, "loss": 1.2796, "step": 4721 }, { "epoch": 0.59, "learning_rate": 0.0001880226056504553, "loss": 1.3231, "step": 4722 }, { "epoch": 0.59, "learning_rate": 0.00018792506189465146, "loss": 1.2222, "step": 4723 }, { "epoch": 0.59, "learning_rate": 0.00018782752820878634, "loss": 1.3288, "step": 4724 }, { "epoch": 0.59, "learning_rate": 0.00018773000460868205, "loss": 1.3779, "step": 4725 }, { "epoch": 0.59, "learning_rate": 0.0001876324911101593, "loss": 1.3303, "step": 4726 }, { "epoch": 0.59, "learning_rate": 0.00018753498772903671, "loss": 1.2568, "step": 4727 }, { "epoch": 0.59, "learning_rate": 0.00018743749448113167, "loss": 1.2372, "step": 4728 }, { "epoch": 0.59, "learning_rate": 0.0001873400113822597, "loss": 1.2941, "step": 4729 }, { "epoch": 0.59, "learning_rate": 0.00018724253844823457, "loss": 1.3154, "step": 4730 }, { "epoch": 0.59, "learning_rate": 0.00018714507569486883, "loss": 1.2829, "step": 4731 }, { "epoch": 0.59, "learning_rate": 0.00018704762313797302, "loss": 1.3293, "step": 4732 }, { "epoch": 0.59, "learning_rate": 0.00018695018079335608, "loss": 1.3688, "step": 4733 }, { "epoch": 0.59, "learning_rate": 0.00018685274867682536, "loss": 1.2523, "step": 4734 }, { "epoch": 0.59, "learning_rate": 0.00018675532680418642, "loss": 1.2759, "step": 4735 }, { "epoch": 0.59, "learning_rate": 0.0001866579151912434, "loss": 1.3274, "step": 4736 }, { "epoch": 0.59, "learning_rate": 0.0001865605138537986, "loss": 1.3513, "step": 4737 }, { "epoch": 0.59, "learning_rate": 0.0001864631228076527, "loss": 1.2783, "step": 4738 }, { "epoch": 0.59, "learning_rate": 0.00018636574206860464, "loss": 1.2749, "step": 4739 }, { "epoch": 0.59, "learning_rate": 0.00018626837165245164, "loss": 1.2861, "step": 4740 }, { "epoch": 0.59, "learning_rate": 0.00018617101157498967, "loss": 1.2846, "step": 4741 }, { "epoch": 0.59, "learning_rate": 0.00018607366185201246, "loss": 1.2753, "step": 4742 }, { "epoch": 0.59, "learning_rate": 0.00018597632249931238, "loss": 1.2862, "step": 4743 }, { "epoch": 0.59, "learning_rate": 0.00018587899353268006, "loss": 1.3574, "step": 4744 }, { "epoch": 0.59, "learning_rate": 0.00018578167496790426, "loss": 1.2956, "step": 4745 }, { "epoch": 0.59, "learning_rate": 0.00018568436682077244, "loss": 1.2917, "step": 4746 }, { "epoch": 0.59, "learning_rate": 0.00018558706910707013, "loss": 1.2494, "step": 4747 }, { "epoch": 0.59, "learning_rate": 0.0001854897818425811, "loss": 1.2727, "step": 4748 }, { "epoch": 0.59, "learning_rate": 0.00018539250504308753, "loss": 1.3104, "step": 4749 }, { "epoch": 0.59, "learning_rate": 0.0001852952387243698, "loss": 1.305, "step": 4750 }, { "epoch": 0.59, "learning_rate": 0.0001851979829022069, "loss": 1.2821, "step": 4751 }, { "epoch": 0.59, "learning_rate": 0.00018510073759237569, "loss": 1.2353, "step": 4752 }, { "epoch": 0.59, "learning_rate": 0.00018500350281065167, "loss": 1.3448, "step": 4753 }, { "epoch": 0.59, "learning_rate": 0.0001849062785728084, "loss": 1.2386, "step": 4754 }, { "epoch": 0.59, "learning_rate": 0.00018480906489461775, "loss": 1.2829, "step": 4755 }, { "epoch": 0.59, "learning_rate": 0.00018471186179185012, "loss": 1.2984, "step": 4756 }, { "epoch": 0.59, "learning_rate": 0.00018461466928027393, "loss": 1.2481, "step": 4757 }, { "epoch": 0.59, "learning_rate": 0.00018451748737565596, "loss": 1.3251, "step": 4758 }, { "epoch": 0.59, "learning_rate": 0.00018442031609376125, "loss": 1.2886, "step": 4759 }, { "epoch": 0.59, "learning_rate": 0.00018432315545035326, "loss": 1.3234, "step": 4760 }, { "epoch": 0.6, "learning_rate": 0.0001842260054611936, "loss": 1.3469, "step": 4761 }, { "epoch": 0.6, "learning_rate": 0.00018412886614204202, "loss": 1.3357, "step": 4762 }, { "epoch": 0.6, "learning_rate": 0.00018403173750865685, "loss": 1.3143, "step": 4763 }, { "epoch": 0.6, "learning_rate": 0.0001839346195767943, "loss": 1.3523, "step": 4764 }, { "epoch": 0.6, "learning_rate": 0.0001838375123622093, "loss": 1.3227, "step": 4765 }, { "epoch": 0.6, "learning_rate": 0.00018374041588065476, "loss": 1.2894, "step": 4766 }, { "epoch": 0.6, "learning_rate": 0.00018364333014788187, "loss": 1.303, "step": 4767 }, { "epoch": 0.6, "learning_rate": 0.00018354625517963999, "loss": 1.2751, "step": 4768 }, { "epoch": 0.6, "learning_rate": 0.0001834491909916769, "loss": 1.264, "step": 4769 }, { "epoch": 0.6, "learning_rate": 0.00018335213759973867, "loss": 1.2827, "step": 4770 }, { "epoch": 0.6, "learning_rate": 0.00018325509501956945, "loss": 1.2892, "step": 4771 }, { "epoch": 0.6, "learning_rate": 0.00018315806326691175, "loss": 1.2494, "step": 4772 }, { "epoch": 0.6, "learning_rate": 0.0001830610423575062, "loss": 1.3185, "step": 4773 }, { "epoch": 0.6, "learning_rate": 0.00018296403230709185, "loss": 1.3415, "step": 4774 }, { "epoch": 0.6, "learning_rate": 0.0001828670331314058, "loss": 1.2819, "step": 4775 }, { "epoch": 0.6, "learning_rate": 0.00018277004484618358, "loss": 1.3159, "step": 4776 }, { "epoch": 0.6, "learning_rate": 0.0001826730674671588, "loss": 1.3308, "step": 4777 }, { "epoch": 0.6, "learning_rate": 0.0001825761010100634, "loss": 1.2439, "step": 4778 }, { "epoch": 0.6, "learning_rate": 0.00018247914549062726, "loss": 1.3004, "step": 4779 }, { "epoch": 0.6, "learning_rate": 0.0001823822009245791, "loss": 1.2382, "step": 4780 }, { "epoch": 0.6, "learning_rate": 0.00018228526732764528, "loss": 1.3137, "step": 4781 }, { "epoch": 0.6, "learning_rate": 0.00018218834471555063, "loss": 1.3365, "step": 4782 }, { "epoch": 0.6, "learning_rate": 0.0001820914331040181, "loss": 1.2643, "step": 4783 }, { "epoch": 0.6, "learning_rate": 0.00018199453250876892, "loss": 1.323, "step": 4784 }, { "epoch": 0.6, "learning_rate": 0.00018189764294552268, "loss": 1.3212, "step": 4785 }, { "epoch": 0.6, "learning_rate": 0.00018180076442999687, "loss": 1.2701, "step": 4786 }, { "epoch": 0.6, "learning_rate": 0.00018170389697790744, "loss": 1.2736, "step": 4787 }, { "epoch": 0.6, "learning_rate": 0.0001816070406049684, "loss": 1.3025, "step": 4788 }, { "epoch": 0.6, "learning_rate": 0.00018151019532689188, "loss": 1.294, "step": 4789 }, { "epoch": 0.6, "learning_rate": 0.00018141336115938856, "loss": 1.2619, "step": 4790 }, { "epoch": 0.6, "learning_rate": 0.00018131653811816707, "loss": 1.2763, "step": 4791 }, { "epoch": 0.6, "learning_rate": 0.00018121972621893418, "loss": 1.2453, "step": 4792 }, { "epoch": 0.6, "learning_rate": 0.000181122925477395, "loss": 1.3528, "step": 4793 }, { "epoch": 0.6, "learning_rate": 0.0001810261359092526, "loss": 1.3077, "step": 4794 }, { "epoch": 0.6, "learning_rate": 0.00018092935753020873, "loss": 1.2814, "step": 4795 }, { "epoch": 0.6, "learning_rate": 0.00018083259035596273, "loss": 1.283, "step": 4796 }, { "epoch": 0.6, "learning_rate": 0.00018073583440221254, "loss": 1.2598, "step": 4797 }, { "epoch": 0.6, "learning_rate": 0.00018063908968465398, "loss": 1.2628, "step": 4798 }, { "epoch": 0.6, "learning_rate": 0.00018054235621898124, "loss": 1.3126, "step": 4799 }, { "epoch": 0.6, "learning_rate": 0.00018044563402088684, "loss": 1.3234, "step": 4800 }, { "epoch": 0.6, "eval_acceptance_rate": 0.690747857093811, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1800914412615855, "eval_expected_tokens": 2.725114301576982, "eval_loss": 1.3777357339859009, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4312785753942455, "eval_runtime": 50.3246, "eval_samples_per_second": 9.677, "eval_steps_per_second": 0.318, "step": 4800 }, { "epoch": 0.6, "learning_rate": 0.00018034892310606114, "loss": 1.3143, "step": 4801 }, { "epoch": 0.6, "learning_rate": 0.0001802522234901927, "loss": 1.2608, "step": 4802 }, { "epoch": 0.6, "learning_rate": 0.00018015553518896854, "loss": 1.2855, "step": 4803 }, { "epoch": 0.6, "learning_rate": 0.00018005885821807346, "loss": 1.2748, "step": 4804 }, { "epoch": 0.6, "learning_rate": 0.0001799621925931908, "loss": 1.3326, "step": 4805 }, { "epoch": 0.6, "learning_rate": 0.00017986553833000186, "loss": 1.29, "step": 4806 }, { "epoch": 0.6, "learning_rate": 0.00017976889544418605, "loss": 1.2276, "step": 4807 }, { "epoch": 0.6, "learning_rate": 0.00017967226395142088, "loss": 1.216, "step": 4808 }, { "epoch": 0.6, "learning_rate": 0.00017957564386738236, "loss": 1.3573, "step": 4809 }, { "epoch": 0.6, "learning_rate": 0.0001794790352077444, "loss": 1.249, "step": 4810 }, { "epoch": 0.6, "learning_rate": 0.00017938243798817894, "loss": 1.2854, "step": 4811 }, { "epoch": 0.6, "learning_rate": 0.00017928585222435627, "loss": 1.3183, "step": 4812 }, { "epoch": 0.6, "learning_rate": 0.00017918927793194472, "loss": 1.2925, "step": 4813 }, { "epoch": 0.6, "learning_rate": 0.0001790927151266109, "loss": 1.3384, "step": 4814 }, { "epoch": 0.6, "learning_rate": 0.00017899616382401935, "loss": 1.3221, "step": 4815 }, { "epoch": 0.6, "learning_rate": 0.0001788996240398329, "loss": 1.2771, "step": 4816 }, { "epoch": 0.6, "learning_rate": 0.00017880309578971237, "loss": 1.3533, "step": 4817 }, { "epoch": 0.6, "learning_rate": 0.0001787065790893168, "loss": 1.2062, "step": 4818 }, { "epoch": 0.6, "learning_rate": 0.00017861007395430356, "loss": 1.2696, "step": 4819 }, { "epoch": 0.6, "learning_rate": 0.00017851358040032772, "loss": 1.319, "step": 4820 }, { "epoch": 0.6, "learning_rate": 0.00017841709844304272, "loss": 1.2787, "step": 4821 }, { "epoch": 0.6, "learning_rate": 0.00017832062809810013, "loss": 1.2345, "step": 4822 }, { "epoch": 0.6, "learning_rate": 0.00017822416938114943, "loss": 1.2949, "step": 4823 }, { "epoch": 0.6, "learning_rate": 0.00017812772230783863, "loss": 1.3432, "step": 4824 }, { "epoch": 0.6, "learning_rate": 0.0001780312868938135, "loss": 1.3184, "step": 4825 }, { "epoch": 0.6, "learning_rate": 0.00017793486315471796, "loss": 1.3009, "step": 4826 }, { "epoch": 0.6, "learning_rate": 0.00017783845110619413, "loss": 1.2581, "step": 4827 }, { "epoch": 0.6, "learning_rate": 0.00017774205076388205, "loss": 1.2875, "step": 4828 }, { "epoch": 0.6, "learning_rate": 0.00017764566214342032, "loss": 1.254, "step": 4829 }, { "epoch": 0.6, "learning_rate": 0.00017754928526044506, "loss": 1.2822, "step": 4830 }, { "epoch": 0.6, "learning_rate": 0.00017745292013059087, "loss": 1.2488, "step": 4831 }, { "epoch": 0.6, "learning_rate": 0.00017735656676949027, "loss": 1.2732, "step": 4832 }, { "epoch": 0.6, "learning_rate": 0.00017726022519277385, "loss": 1.2933, "step": 4833 }, { "epoch": 0.6, "learning_rate": 0.00017716389541607048, "loss": 1.3021, "step": 4834 }, { "epoch": 0.6, "learning_rate": 0.00017706757745500703, "loss": 1.3027, "step": 4835 }, { "epoch": 0.6, "learning_rate": 0.00017697127132520828, "loss": 1.3885, "step": 4836 }, { "epoch": 0.6, "learning_rate": 0.00017687497704229737, "loss": 1.3214, "step": 4837 }, { "epoch": 0.6, "learning_rate": 0.0001767786946218952, "loss": 1.2435, "step": 4838 }, { "epoch": 0.6, "learning_rate": 0.00017668242407962122, "loss": 1.3088, "step": 4839 }, { "epoch": 0.6, "learning_rate": 0.00017658616543109236, "loss": 1.2781, "step": 4840 }, { "epoch": 0.61, "learning_rate": 0.00017648991869192405, "loss": 1.2782, "step": 4841 }, { "epoch": 0.61, "learning_rate": 0.00017639368387772965, "loss": 1.3132, "step": 4842 }, { "epoch": 0.61, "learning_rate": 0.0001762974610041205, "loss": 1.2897, "step": 4843 }, { "epoch": 0.61, "learning_rate": 0.0001762012500867063, "loss": 1.2759, "step": 4844 }, { "epoch": 0.61, "learning_rate": 0.00017610505114109447, "loss": 1.3569, "step": 4845 }, { "epoch": 0.61, "learning_rate": 0.00017600886418289057, "loss": 1.2474, "step": 4846 }, { "epoch": 0.61, "learning_rate": 0.00017591268922769842, "loss": 1.2596, "step": 4847 }, { "epoch": 0.61, "learning_rate": 0.00017581652629111954, "loss": 1.2801, "step": 4848 }, { "epoch": 0.61, "learning_rate": 0.0001757203753887539, "loss": 1.2443, "step": 4849 }, { "epoch": 0.61, "learning_rate": 0.0001756242365361993, "loss": 1.2247, "step": 4850 }, { "epoch": 0.61, "learning_rate": 0.0001755281097490515, "loss": 1.2438, "step": 4851 }, { "epoch": 0.61, "learning_rate": 0.00017543199504290442, "loss": 1.27, "step": 4852 }, { "epoch": 0.61, "learning_rate": 0.00017533589243335, "loss": 1.2994, "step": 4853 }, { "epoch": 0.61, "learning_rate": 0.00017523980193597836, "loss": 1.2767, "step": 4854 }, { "epoch": 0.61, "learning_rate": 0.00017514372356637736, "loss": 1.2906, "step": 4855 }, { "epoch": 0.61, "learning_rate": 0.00017504765734013323, "loss": 1.2605, "step": 4856 }, { "epoch": 0.61, "learning_rate": 0.00017495160327282976, "loss": 1.2824, "step": 4857 }, { "epoch": 0.61, "learning_rate": 0.0001748555613800494, "loss": 1.3237, "step": 4858 }, { "epoch": 0.61, "learning_rate": 0.00017475953167737208, "loss": 1.256, "step": 4859 }, { "epoch": 0.61, "learning_rate": 0.00017466351418037607, "loss": 1.2943, "step": 4860 }, { "epoch": 0.61, "learning_rate": 0.0001745675089046374, "loss": 1.2836, "step": 4861 }, { "epoch": 0.61, "learning_rate": 0.00017447151586573033, "loss": 1.342, "step": 4862 }, { "epoch": 0.61, "learning_rate": 0.00017437553507922718, "loss": 1.2958, "step": 4863 }, { "epoch": 0.61, "learning_rate": 0.0001742795665606981, "loss": 1.3189, "step": 4864 }, { "epoch": 0.61, "learning_rate": 0.00017418361032571128, "loss": 1.2391, "step": 4865 }, { "epoch": 0.61, "learning_rate": 0.000174087666389833, "loss": 1.2246, "step": 4866 }, { "epoch": 0.61, "learning_rate": 0.00017399173476862742, "loss": 1.2694, "step": 4867 }, { "epoch": 0.61, "learning_rate": 0.00017389581547765692, "loss": 1.2718, "step": 4868 }, { "epoch": 0.61, "learning_rate": 0.00017379990853248173, "loss": 1.3619, "step": 4869 }, { "epoch": 0.61, "learning_rate": 0.0001737040139486601, "loss": 1.3241, "step": 4870 }, { "epoch": 0.61, "learning_rate": 0.0001736081317417481, "loss": 1.2551, "step": 4871 }, { "epoch": 0.61, "learning_rate": 0.00017351226192730007, "loss": 1.2959, "step": 4872 }, { "epoch": 0.61, "learning_rate": 0.00017341640452086838, "loss": 1.287, "step": 4873 }, { "epoch": 0.61, "learning_rate": 0.000173320559538003, "loss": 1.2557, "step": 4874 }, { "epoch": 0.61, "learning_rate": 0.00017322472699425235, "loss": 1.2831, "step": 4875 }, { "epoch": 0.61, "learning_rate": 0.00017312890690516242, "loss": 1.2991, "step": 4876 }, { "epoch": 0.61, "learning_rate": 0.00017303309928627737, "loss": 1.2702, "step": 4877 }, { "epoch": 0.61, "learning_rate": 0.0001729373041531394, "loss": 1.2902, "step": 4878 }, { "epoch": 0.61, "learning_rate": 0.00017284152152128869, "loss": 1.2951, "step": 4879 }, { "epoch": 0.61, "learning_rate": 0.00017274575140626317, "loss": 1.2655, "step": 4880 }, { "epoch": 0.61, "learning_rate": 0.000172649993823599, "loss": 1.2867, "step": 4881 }, { "epoch": 0.61, "learning_rate": 0.00017255424878883006, "loss": 1.1743, "step": 4882 }, { "epoch": 0.61, "learning_rate": 0.0001724585163174885, "loss": 1.2736, "step": 4883 }, { "epoch": 0.61, "learning_rate": 0.00017236279642510413, "loss": 1.3134, "step": 4884 }, { "epoch": 0.61, "learning_rate": 0.00017226708912720495, "loss": 1.3153, "step": 4885 }, { "epoch": 0.61, "learning_rate": 0.00017217139443931673, "loss": 1.3553, "step": 4886 }, { "epoch": 0.61, "learning_rate": 0.00017207571237696324, "loss": 1.3095, "step": 4887 }, { "epoch": 0.61, "learning_rate": 0.00017198004295566644, "loss": 1.2777, "step": 4888 }, { "epoch": 0.61, "learning_rate": 0.0001718843861909459, "loss": 1.2918, "step": 4889 }, { "epoch": 0.61, "learning_rate": 0.00017178874209831927, "loss": 1.2469, "step": 4890 }, { "epoch": 0.61, "learning_rate": 0.00017169311069330223, "loss": 1.2825, "step": 4891 }, { "epoch": 0.61, "learning_rate": 0.00017159749199140818, "loss": 1.2486, "step": 4892 }, { "epoch": 0.61, "learning_rate": 0.00017150188600814875, "loss": 1.2507, "step": 4893 }, { "epoch": 0.61, "learning_rate": 0.00017140629275903338, "loss": 1.2135, "step": 4894 }, { "epoch": 0.61, "learning_rate": 0.00017131071225956932, "loss": 1.3805, "step": 4895 }, { "epoch": 0.61, "learning_rate": 0.00017121514452526194, "loss": 1.2477, "step": 4896 }, { "epoch": 0.61, "learning_rate": 0.00017111958957161433, "loss": 1.2767, "step": 4897 }, { "epoch": 0.61, "learning_rate": 0.0001710240474141278, "loss": 1.2488, "step": 4898 }, { "epoch": 0.61, "learning_rate": 0.00017092851806830133, "loss": 1.3296, "step": 4899 }, { "epoch": 0.61, "learning_rate": 0.00017083300154963193, "loss": 1.2612, "step": 4900 }, { "epoch": 0.61, "eval_acceptance_rate": 0.6911183595657349, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.181612993032552, "eval_expected_tokens": 2.72701624129069, "eval_loss": 1.3746060132980347, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4317540603226725, "eval_runtime": 50.3943, "eval_samples_per_second": 9.664, "eval_steps_per_second": 0.317, "step": 4900 }, { "epoch": 0.61, "learning_rate": 0.00017073749787361448, "loss": 1.2795, "step": 4901 }, { "epoch": 0.61, "learning_rate": 0.00017064200705574174, "loss": 1.3344, "step": 4902 }, { "epoch": 0.61, "learning_rate": 0.00017054652911150465, "loss": 1.2632, "step": 4903 }, { "epoch": 0.61, "learning_rate": 0.00017045106405639174, "loss": 1.3283, "step": 4904 }, { "epoch": 0.61, "learning_rate": 0.00017035561190588953, "loss": 1.2985, "step": 4905 }, { "epoch": 0.61, "learning_rate": 0.0001702601726754825, "loss": 1.2585, "step": 4906 }, { "epoch": 0.61, "learning_rate": 0.0001701647463806531, "loss": 1.3427, "step": 4907 }, { "epoch": 0.61, "learning_rate": 0.00017006933303688153, "loss": 1.3268, "step": 4908 }, { "epoch": 0.61, "learning_rate": 0.00016997393265964607, "loss": 1.262, "step": 4909 }, { "epoch": 0.61, "learning_rate": 0.00016987854526442266, "loss": 1.3061, "step": 4910 }, { "epoch": 0.61, "learning_rate": 0.0001697831708666852, "loss": 1.2961, "step": 4911 }, { "epoch": 0.61, "learning_rate": 0.00016968780948190573, "loss": 1.2968, "step": 4912 }, { "epoch": 0.61, "learning_rate": 0.00016959246112555393, "loss": 1.2597, "step": 4913 }, { "epoch": 0.61, "learning_rate": 0.0001694971258130974, "loss": 1.2068, "step": 4914 }, { "epoch": 0.61, "learning_rate": 0.00016940180356000167, "loss": 1.2954, "step": 4915 }, { "epoch": 0.61, "learning_rate": 0.00016930649438173006, "loss": 1.2662, "step": 4916 }, { "epoch": 0.61, "learning_rate": 0.000169211198293744, "loss": 1.2399, "step": 4917 }, { "epoch": 0.61, "learning_rate": 0.0001691159153115025, "loss": 1.2848, "step": 4918 }, { "epoch": 0.61, "learning_rate": 0.0001690206454504627, "loss": 1.3138, "step": 4919 }, { "epoch": 0.61, "learning_rate": 0.00016892538872607935, "loss": 1.2614, "step": 4920 }, { "epoch": 0.62, "learning_rate": 0.00016883014515380523, "loss": 1.235, "step": 4921 }, { "epoch": 0.62, "learning_rate": 0.0001687349147490912, "loss": 1.3041, "step": 4922 }, { "epoch": 0.62, "learning_rate": 0.00016863969752738554, "loss": 1.1859, "step": 4923 }, { "epoch": 0.62, "learning_rate": 0.0001685444935041346, "loss": 1.2679, "step": 4924 }, { "epoch": 0.62, "learning_rate": 0.00016844930269478272, "loss": 1.2309, "step": 4925 }, { "epoch": 0.62, "learning_rate": 0.0001683541251147718, "loss": 1.251, "step": 4926 }, { "epoch": 0.62, "learning_rate": 0.0001682589607795419, "loss": 1.2816, "step": 4927 }, { "epoch": 0.62, "learning_rate": 0.00016816380970453084, "loss": 1.272, "step": 4928 }, { "epoch": 0.62, "learning_rate": 0.00016806867190517416, "loss": 1.3332, "step": 4929 }, { "epoch": 0.62, "learning_rate": 0.00016797354739690528, "loss": 1.3116, "step": 4930 }, { "epoch": 0.62, "learning_rate": 0.00016787843619515554, "loss": 1.2483, "step": 4931 }, { "epoch": 0.62, "learning_rate": 0.00016778333831535418, "loss": 1.3212, "step": 4932 }, { "epoch": 0.62, "learning_rate": 0.00016768825377292817, "loss": 1.2849, "step": 4933 }, { "epoch": 0.62, "learning_rate": 0.00016759318258330232, "loss": 1.2798, "step": 4934 }, { "epoch": 0.62, "learning_rate": 0.00016749812476189935, "loss": 1.2485, "step": 4935 }, { "epoch": 0.62, "learning_rate": 0.00016740308032413956, "loss": 1.2435, "step": 4936 }, { "epoch": 0.62, "learning_rate": 0.0001673080492854415, "loss": 1.3025, "step": 4937 }, { "epoch": 0.62, "learning_rate": 0.0001672130316612213, "loss": 1.3001, "step": 4938 }, { "epoch": 0.62, "learning_rate": 0.0001671180274668928, "loss": 1.3242, "step": 4939 }, { "epoch": 0.62, "learning_rate": 0.00016702303671786795, "loss": 1.2521, "step": 4940 }, { "epoch": 0.62, "learning_rate": 0.0001669280594295562, "loss": 1.2658, "step": 4941 }, { "epoch": 0.62, "learning_rate": 0.00016683309561736515, "loss": 1.2473, "step": 4942 }, { "epoch": 0.62, "learning_rate": 0.00016673814529669995, "loss": 1.3224, "step": 4943 }, { "epoch": 0.62, "learning_rate": 0.00016664320848296376, "loss": 1.3324, "step": 4944 }, { "epoch": 0.62, "learning_rate": 0.00016654828519155728, "loss": 1.3029, "step": 4945 }, { "epoch": 0.62, "learning_rate": 0.00016645337543787927, "loss": 1.317, "step": 4946 }, { "epoch": 0.62, "learning_rate": 0.0001663584792373263, "loss": 1.2481, "step": 4947 }, { "epoch": 0.62, "learning_rate": 0.00016626359660529256, "loss": 1.298, "step": 4948 }, { "epoch": 0.62, "learning_rate": 0.0001661687275571701, "loss": 1.3218, "step": 4949 }, { "epoch": 0.62, "learning_rate": 0.00016607387210834887, "loss": 1.3294, "step": 4950 }, { "epoch": 0.62, "learning_rate": 0.0001659790302742164, "loss": 1.2839, "step": 4951 }, { "epoch": 0.62, "learning_rate": 0.00016588420207015825, "loss": 1.2938, "step": 4952 }, { "epoch": 0.62, "learning_rate": 0.0001657893875115578, "loss": 1.2775, "step": 4953 }, { "epoch": 0.62, "learning_rate": 0.00016569458661379587, "loss": 1.2465, "step": 4954 }, { "epoch": 0.62, "learning_rate": 0.00016559979939225129, "loss": 1.2281, "step": 4955 }, { "epoch": 0.62, "learning_rate": 0.00016550502586230078, "loss": 1.3133, "step": 4956 }, { "epoch": 0.62, "learning_rate": 0.00016541026603931873, "loss": 1.2721, "step": 4957 }, { "epoch": 0.62, "learning_rate": 0.00016531551993867715, "loss": 1.2214, "step": 4958 }, { "epoch": 0.62, "learning_rate": 0.00016522078757574615, "loss": 1.2908, "step": 4959 }, { "epoch": 0.62, "learning_rate": 0.00016512606896589322, "loss": 1.2615, "step": 4960 }, { "epoch": 0.62, "learning_rate": 0.000165031364124484, "loss": 1.2376, "step": 4961 }, { "epoch": 0.62, "learning_rate": 0.00016493667306688166, "loss": 1.2322, "step": 4962 }, { "epoch": 0.62, "learning_rate": 0.00016484199580844727, "loss": 1.3, "step": 4963 }, { "epoch": 0.62, "learning_rate": 0.00016474733236453948, "loss": 1.2274, "step": 4964 }, { "epoch": 0.62, "learning_rate": 0.00016465268275051478, "loss": 1.3081, "step": 4965 }, { "epoch": 0.62, "learning_rate": 0.00016455804698172765, "loss": 1.2953, "step": 4966 }, { "epoch": 0.62, "learning_rate": 0.00016446342507353, "loss": 1.23, "step": 4967 }, { "epoch": 0.62, "learning_rate": 0.0001643688170412716, "loss": 1.3289, "step": 4968 }, { "epoch": 0.62, "learning_rate": 0.00016427422290030003, "loss": 1.2651, "step": 4969 }, { "epoch": 0.62, "learning_rate": 0.0001641796426659604, "loss": 1.3068, "step": 4970 }, { "epoch": 0.62, "learning_rate": 0.00016408507635359598, "loss": 1.2286, "step": 4971 }, { "epoch": 0.62, "learning_rate": 0.0001639905239785474, "loss": 1.3028, "step": 4972 }, { "epoch": 0.62, "learning_rate": 0.00016389598555615326, "loss": 1.2317, "step": 4973 }, { "epoch": 0.62, "learning_rate": 0.00016380146110174965, "loss": 1.2764, "step": 4974 }, { "epoch": 0.62, "learning_rate": 0.00016370695063067054, "loss": 1.27, "step": 4975 }, { "epoch": 0.62, "learning_rate": 0.00016361245415824784, "loss": 1.3038, "step": 4976 }, { "epoch": 0.62, "learning_rate": 0.0001635179716998108, "loss": 1.2113, "step": 4977 }, { "epoch": 0.62, "learning_rate": 0.0001634235032706867, "loss": 1.2589, "step": 4978 }, { "epoch": 0.62, "learning_rate": 0.00016332904888620035, "loss": 1.2494, "step": 4979 }, { "epoch": 0.62, "learning_rate": 0.00016323460856167428, "loss": 1.245, "step": 4980 }, { "epoch": 0.62, "learning_rate": 0.00016314018231242895, "loss": 1.2688, "step": 4981 }, { "epoch": 0.62, "learning_rate": 0.00016304577015378242, "loss": 1.2629, "step": 4982 }, { "epoch": 0.62, "learning_rate": 0.00016295137210105032, "loss": 1.2396, "step": 4983 }, { "epoch": 0.62, "learning_rate": 0.00016285698816954626, "loss": 1.2995, "step": 4984 }, { "epoch": 0.62, "learning_rate": 0.00016276261837458122, "loss": 1.2262, "step": 4985 }, { "epoch": 0.62, "learning_rate": 0.0001626682627314643, "loss": 1.2945, "step": 4986 }, { "epoch": 0.62, "learning_rate": 0.00016257392125550196, "loss": 1.2532, "step": 4987 }, { "epoch": 0.62, "learning_rate": 0.0001624795939619986, "loss": 1.303, "step": 4988 }, { "epoch": 0.62, "learning_rate": 0.00016238528086625609, "loss": 1.2531, "step": 4989 }, { "epoch": 0.62, "learning_rate": 0.00016229098198357407, "loss": 1.2191, "step": 4990 }, { "epoch": 0.62, "learning_rate": 0.00016219669732925012, "loss": 1.2375, "step": 4991 }, { "epoch": 0.62, "learning_rate": 0.00016210242691857922, "loss": 1.2712, "step": 4992 }, { "epoch": 0.62, "learning_rate": 0.00016200817076685408, "loss": 1.3246, "step": 4993 }, { "epoch": 0.62, "learning_rate": 0.00016191392888936524, "loss": 1.274, "step": 4994 }, { "epoch": 0.62, "learning_rate": 0.00016181970130140066, "loss": 1.2544, "step": 4995 }, { "epoch": 0.62, "learning_rate": 0.00016172548801824633, "loss": 1.2441, "step": 4996 }, { "epoch": 0.62, "learning_rate": 0.00016163128905518576, "loss": 1.2496, "step": 4997 }, { "epoch": 0.62, "learning_rate": 0.00016153710442750002, "loss": 1.2153, "step": 4998 }, { "epoch": 0.62, "learning_rate": 0.00016144293415046795, "loss": 1.2338, "step": 4999 }, { "epoch": 0.62, "learning_rate": 0.00016134877823936609, "loss": 1.1862, "step": 5000 }, { "epoch": 0.62, "eval_acceptance_rate": 0.6927386522293091, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.188282427979586, "eval_expected_tokens": 2.735353034974483, "eval_loss": 1.371904730796814, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.43383825874362064, "eval_runtime": 50.3551, "eval_samples_per_second": 9.671, "eval_steps_per_second": 0.318, "step": 5000 }, { "epoch": 0.63, "learning_rate": 0.0001612546367094687, "loss": 1.2596, "step": 5001 }, { "epoch": 0.63, "learning_rate": 0.00016116050957604755, "loss": 1.2652, "step": 5002 }, { "epoch": 0.63, "learning_rate": 0.00016106639685437225, "loss": 1.3298, "step": 5003 }, { "epoch": 0.63, "learning_rate": 0.00016097229855970984, "loss": 1.3333, "step": 5004 }, { "epoch": 0.63, "learning_rate": 0.00016087821470732528, "loss": 1.3282, "step": 5005 }, { "epoch": 0.63, "learning_rate": 0.00016078414531248104, "loss": 1.2415, "step": 5006 }, { "epoch": 0.63, "learning_rate": 0.00016069009039043728, "loss": 1.2564, "step": 5007 }, { "epoch": 0.63, "learning_rate": 0.00016059604995645173, "loss": 1.3183, "step": 5008 }, { "epoch": 0.63, "learning_rate": 0.00016050202402577988, "loss": 1.2381, "step": 5009 }, { "epoch": 0.63, "learning_rate": 0.00016040801261367493, "loss": 1.2671, "step": 5010 }, { "epoch": 0.63, "learning_rate": 0.00016031401573538757, "loss": 1.2791, "step": 5011 }, { "epoch": 0.63, "learning_rate": 0.0001602200334061661, "loss": 1.2331, "step": 5012 }, { "epoch": 0.63, "learning_rate": 0.0001601260656412567, "loss": 1.29, "step": 5013 }, { "epoch": 0.63, "learning_rate": 0.00016003211245590283, "loss": 1.2151, "step": 5014 }, { "epoch": 0.63, "learning_rate": 0.000159938173865346, "loss": 1.2309, "step": 5015 }, { "epoch": 0.63, "learning_rate": 0.00015984424988482507, "loss": 1.2818, "step": 5016 }, { "epoch": 0.63, "learning_rate": 0.00015975034052957655, "loss": 1.2384, "step": 5017 }, { "epoch": 0.63, "learning_rate": 0.00015965644581483475, "loss": 1.2455, "step": 5018 }, { "epoch": 0.63, "learning_rate": 0.0001595625657558313, "loss": 1.2085, "step": 5019 }, { "epoch": 0.63, "learning_rate": 0.00015946870036779583, "loss": 1.3364, "step": 5020 }, { "epoch": 0.63, "learning_rate": 0.0001593748496659553, "loss": 1.2755, "step": 5021 }, { "epoch": 0.63, "learning_rate": 0.00015928101366553444, "loss": 1.2503, "step": 5022 }, { "epoch": 0.63, "learning_rate": 0.00015918719238175544, "loss": 1.2801, "step": 5023 }, { "epoch": 0.63, "learning_rate": 0.00015909338582983823, "loss": 1.3145, "step": 5024 }, { "epoch": 0.63, "learning_rate": 0.00015899959402500047, "loss": 1.2547, "step": 5025 }, { "epoch": 0.63, "learning_rate": 0.00015890581698245715, "loss": 1.2741, "step": 5026 }, { "epoch": 0.63, "learning_rate": 0.000158812054717421, "loss": 1.1973, "step": 5027 }, { "epoch": 0.63, "learning_rate": 0.0001587183072451024, "loss": 1.2778, "step": 5028 }, { "epoch": 0.63, "learning_rate": 0.00015862457458070913, "loss": 1.2263, "step": 5029 }, { "epoch": 0.63, "learning_rate": 0.00015853085673944695, "loss": 1.3099, "step": 5030 }, { "epoch": 0.63, "learning_rate": 0.00015843715373651884, "loss": 1.2869, "step": 5031 }, { "epoch": 0.63, "learning_rate": 0.00015834346558712563, "loss": 1.2512, "step": 5032 }, { "epoch": 0.63, "learning_rate": 0.00015824979230646543, "loss": 1.3189, "step": 5033 }, { "epoch": 0.63, "learning_rate": 0.00015815613390973422, "loss": 1.3137, "step": 5034 }, { "epoch": 0.63, "learning_rate": 0.00015806249041212561, "loss": 1.2259, "step": 5035 }, { "epoch": 0.63, "learning_rate": 0.00015796886182883053, "loss": 1.2417, "step": 5036 }, { "epoch": 0.63, "learning_rate": 0.00015787524817503768, "loss": 1.2557, "step": 5037 }, { "epoch": 0.63, "learning_rate": 0.00015778164946593325, "loss": 1.2746, "step": 5038 }, { "epoch": 0.63, "learning_rate": 0.00015768806571670095, "loss": 1.2712, "step": 5039 }, { "epoch": 0.63, "learning_rate": 0.00015759449694252226, "loss": 1.2259, "step": 5040 }, { "epoch": 0.63, "learning_rate": 0.0001575009431585762, "loss": 1.2111, "step": 5041 }, { "epoch": 0.63, "learning_rate": 0.00015740740438003912, "loss": 1.2923, "step": 5042 }, { "epoch": 0.63, "learning_rate": 0.00015731388062208517, "loss": 1.3053, "step": 5043 }, { "epoch": 0.63, "learning_rate": 0.00015722037189988593, "loss": 1.2813, "step": 5044 }, { "epoch": 0.63, "learning_rate": 0.0001571268782286107, "loss": 1.2213, "step": 5045 }, { "epoch": 0.63, "learning_rate": 0.0001570333996234262, "loss": 1.3615, "step": 5046 }, { "epoch": 0.63, "learning_rate": 0.00015693993609949673, "loss": 1.3166, "step": 5047 }, { "epoch": 0.63, "learning_rate": 0.00015684648767198413, "loss": 1.2291, "step": 5048 }, { "epoch": 0.63, "learning_rate": 0.00015675305435604775, "loss": 1.2591, "step": 5049 }, { "epoch": 0.63, "learning_rate": 0.00015665963616684475, "loss": 1.2437, "step": 5050 }, { "epoch": 0.63, "learning_rate": 0.00015656623311952959, "loss": 1.275, "step": 5051 }, { "epoch": 0.63, "learning_rate": 0.0001564728452292542, "loss": 1.2675, "step": 5052 }, { "epoch": 0.63, "learning_rate": 0.0001563794725111682, "loss": 1.3026, "step": 5053 }, { "epoch": 0.63, "learning_rate": 0.0001562861149804189, "loss": 1.2903, "step": 5054 }, { "epoch": 0.63, "learning_rate": 0.00015619277265215077, "loss": 1.3031, "step": 5055 }, { "epoch": 0.63, "learning_rate": 0.00015609944554150618, "loss": 1.2917, "step": 5056 }, { "epoch": 0.63, "learning_rate": 0.00015600613366362475, "loss": 1.2263, "step": 5057 }, { "epoch": 0.63, "learning_rate": 0.00015591283703364373, "loss": 1.3315, "step": 5058 }, { "epoch": 0.63, "learning_rate": 0.000155819555666698, "loss": 1.278, "step": 5059 }, { "epoch": 0.63, "learning_rate": 0.0001557262895779199, "loss": 1.2565, "step": 5060 }, { "epoch": 0.63, "learning_rate": 0.00015563303878243917, "loss": 1.3172, "step": 5061 }, { "epoch": 0.63, "learning_rate": 0.00015553980329538326, "loss": 1.3238, "step": 5062 }, { "epoch": 0.63, "learning_rate": 0.0001554465831318769, "loss": 1.2242, "step": 5063 }, { "epoch": 0.63, "learning_rate": 0.00015535337830704265, "loss": 1.2648, "step": 5064 }, { "epoch": 0.63, "learning_rate": 0.00015526018883600036, "loss": 1.2118, "step": 5065 }, { "epoch": 0.63, "learning_rate": 0.00015516701473386737, "loss": 1.2166, "step": 5066 }, { "epoch": 0.63, "learning_rate": 0.00015507385601575864, "loss": 1.298, "step": 5067 }, { "epoch": 0.63, "learning_rate": 0.00015498071269678655, "loss": 1.2684, "step": 5068 }, { "epoch": 0.63, "learning_rate": 0.00015488758479206117, "loss": 1.3127, "step": 5069 }, { "epoch": 0.63, "learning_rate": 0.00015479447231668978, "loss": 1.2813, "step": 5070 }, { "epoch": 0.63, "learning_rate": 0.00015470137528577732, "loss": 1.2606, "step": 5071 }, { "epoch": 0.63, "learning_rate": 0.00015460829371442624, "loss": 1.2665, "step": 5072 }, { "epoch": 0.63, "learning_rate": 0.00015451522761773635, "loss": 1.2612, "step": 5073 }, { "epoch": 0.63, "learning_rate": 0.00015442217701080516, "loss": 1.2569, "step": 5074 }, { "epoch": 0.63, "learning_rate": 0.00015432914190872756, "loss": 1.2608, "step": 5075 }, { "epoch": 0.63, "learning_rate": 0.0001542361223265959, "loss": 1.2544, "step": 5076 }, { "epoch": 0.63, "learning_rate": 0.00015414311827949994, "loss": 1.3139, "step": 5077 }, { "epoch": 0.63, "learning_rate": 0.00015405012978252701, "loss": 1.2663, "step": 5078 }, { "epoch": 0.63, "learning_rate": 0.0001539571568507621, "loss": 1.3288, "step": 5079 }, { "epoch": 0.64, "learning_rate": 0.00015386419949928732, "loss": 1.3328, "step": 5080 }, { "epoch": 0.64, "learning_rate": 0.0001537712577431825, "loss": 1.3206, "step": 5081 }, { "epoch": 0.64, "learning_rate": 0.00015367833159752486, "loss": 1.2574, "step": 5082 }, { "epoch": 0.64, "learning_rate": 0.00015358542107738897, "loss": 1.2643, "step": 5083 }, { "epoch": 0.64, "learning_rate": 0.00015349252619784712, "loss": 1.2738, "step": 5084 }, { "epoch": 0.64, "learning_rate": 0.00015339964697396897, "loss": 1.2706, "step": 5085 }, { "epoch": 0.64, "learning_rate": 0.0001533067834208215, "loss": 1.2381, "step": 5086 }, { "epoch": 0.64, "learning_rate": 0.0001532139355534693, "loss": 1.2985, "step": 5087 }, { "epoch": 0.64, "learning_rate": 0.00015312110338697426, "loss": 1.2798, "step": 5088 }, { "epoch": 0.64, "learning_rate": 0.000153028286936396, "loss": 1.2689, "step": 5089 }, { "epoch": 0.64, "learning_rate": 0.00015293548621679126, "loss": 1.2418, "step": 5090 }, { "epoch": 0.64, "learning_rate": 0.00015284270124321452, "loss": 1.26, "step": 5091 }, { "epoch": 0.64, "learning_rate": 0.00015274993203071748, "loss": 1.3073, "step": 5092 }, { "epoch": 0.64, "learning_rate": 0.00015265717859434935, "loss": 1.2555, "step": 5093 }, { "epoch": 0.64, "learning_rate": 0.0001525644409491569, "loss": 1.2516, "step": 5094 }, { "epoch": 0.64, "learning_rate": 0.00015247171911018427, "loss": 1.2594, "step": 5095 }, { "epoch": 0.64, "learning_rate": 0.0001523790130924728, "loss": 1.2876, "step": 5096 }, { "epoch": 0.64, "learning_rate": 0.00015228632291106173, "loss": 1.3098, "step": 5097 }, { "epoch": 0.64, "learning_rate": 0.00015219364858098721, "loss": 1.3219, "step": 5098 }, { "epoch": 0.64, "learning_rate": 0.0001521009901172833, "loss": 1.2991, "step": 5099 }, { "epoch": 0.64, "learning_rate": 0.00015200834753498126, "loss": 1.2071, "step": 5100 }, { "epoch": 0.64, "eval_acceptance_rate": 0.6933574080467224, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.190835944227234, "eval_expected_tokens": 2.7385449302840423, "eval_loss": 1.3687366247177124, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.43463623257101064, "eval_runtime": 50.2473, "eval_samples_per_second": 9.692, "eval_steps_per_second": 0.318, "step": 5100 }, { "epoch": 0.64, "learning_rate": 0.0001519157208491097, "loss": 1.2504, "step": 5101 }, { "epoch": 0.64, "learning_rate": 0.00015182311007469463, "loss": 1.2444, "step": 5102 }, { "epoch": 0.64, "learning_rate": 0.0001517305152267598, "loss": 1.2649, "step": 5103 }, { "epoch": 0.64, "learning_rate": 0.00015163793632032605, "loss": 1.2257, "step": 5104 }, { "epoch": 0.64, "learning_rate": 0.00015154537337041174, "loss": 1.2409, "step": 5105 }, { "epoch": 0.64, "learning_rate": 0.00015145282639203268, "loss": 1.3066, "step": 5106 }, { "epoch": 0.64, "learning_rate": 0.00015136029540020195, "loss": 1.2654, "step": 5107 }, { "epoch": 0.64, "learning_rate": 0.0001512677804099303, "loss": 1.3527, "step": 5108 }, { "epoch": 0.64, "learning_rate": 0.00015117528143622562, "loss": 1.2365, "step": 5109 }, { "epoch": 0.64, "learning_rate": 0.00015108279849409333, "loss": 1.324, "step": 5110 }, { "epoch": 0.64, "learning_rate": 0.00015099033159853622, "loss": 1.2963, "step": 5111 }, { "epoch": 0.64, "learning_rate": 0.0001508978807645544, "loss": 1.3835, "step": 5112 }, { "epoch": 0.64, "learning_rate": 0.0001508054460071457, "loss": 1.2685, "step": 5113 }, { "epoch": 0.64, "learning_rate": 0.00015071302734130488, "loss": 1.2889, "step": 5114 }, { "epoch": 0.64, "learning_rate": 0.00015062062478202434, "loss": 1.2674, "step": 5115 }, { "epoch": 0.64, "learning_rate": 0.0001505282383442939, "loss": 1.3226, "step": 5116 }, { "epoch": 0.64, "learning_rate": 0.0001504358680431006, "loss": 1.2549, "step": 5117 }, { "epoch": 0.64, "learning_rate": 0.00015034351389342905, "loss": 1.281, "step": 5118 }, { "epoch": 0.64, "learning_rate": 0.00015025117591026117, "loss": 1.3429, "step": 5119 }, { "epoch": 0.64, "learning_rate": 0.00015015885410857616, "loss": 1.2835, "step": 5120 }, { "epoch": 0.64, "learning_rate": 0.00015006654850335073, "loss": 1.2489, "step": 5121 }, { "epoch": 0.64, "learning_rate": 0.00014997425910955874, "loss": 1.2676, "step": 5122 }, { "epoch": 0.64, "learning_rate": 0.0001498819859421719, "loss": 1.3511, "step": 5123 }, { "epoch": 0.64, "learning_rate": 0.00014978972901615877, "loss": 1.3011, "step": 5124 }, { "epoch": 0.64, "learning_rate": 0.00014969748834648555, "loss": 1.2309, "step": 5125 }, { "epoch": 0.64, "learning_rate": 0.0001496052639481157, "loss": 1.2445, "step": 5126 }, { "epoch": 0.64, "learning_rate": 0.00014951305583600999, "loss": 1.255, "step": 5127 }, { "epoch": 0.64, "learning_rate": 0.00014942086402512672, "loss": 1.2633, "step": 5128 }, { "epoch": 0.64, "learning_rate": 0.00014932868853042154, "loss": 1.3479, "step": 5129 }, { "epoch": 0.64, "learning_rate": 0.00014923652936684722, "loss": 1.2937, "step": 5130 }, { "epoch": 0.64, "learning_rate": 0.00014914438654935414, "loss": 1.3313, "step": 5131 }, { "epoch": 0.64, "learning_rate": 0.0001490522600928898, "loss": 1.2706, "step": 5132 }, { "epoch": 0.64, "learning_rate": 0.00014896015001239932, "loss": 1.2711, "step": 5133 }, { "epoch": 0.64, "learning_rate": 0.00014886805632282485, "loss": 1.2861, "step": 5134 }, { "epoch": 0.64, "learning_rate": 0.00014877597903910616, "loss": 1.2298, "step": 5135 }, { "epoch": 0.64, "learning_rate": 0.0001486839181761802, "loss": 1.2609, "step": 5136 }, { "epoch": 0.64, "learning_rate": 0.0001485918737489812, "loss": 1.2945, "step": 5137 }, { "epoch": 0.64, "learning_rate": 0.00014849984577244103, "loss": 1.2969, "step": 5138 }, { "epoch": 0.64, "learning_rate": 0.00014840783426148845, "loss": 1.2586, "step": 5139 }, { "epoch": 0.64, "learning_rate": 0.00014831583923105, "loss": 1.3265, "step": 5140 }, { "epoch": 0.64, "learning_rate": 0.00014822386069604916, "loss": 1.2577, "step": 5141 }, { "epoch": 0.64, "learning_rate": 0.00014813189867140687, "loss": 1.211, "step": 5142 }, { "epoch": 0.64, "learning_rate": 0.00014803995317204152, "loss": 1.3239, "step": 5143 }, { "epoch": 0.64, "learning_rate": 0.00014794802421286878, "loss": 1.3329, "step": 5144 }, { "epoch": 0.64, "learning_rate": 0.00014785611180880142, "loss": 1.3117, "step": 5145 }, { "epoch": 0.64, "learning_rate": 0.00014776421597474982, "loss": 1.2522, "step": 5146 }, { "epoch": 0.64, "learning_rate": 0.00014767233672562135, "loss": 1.289, "step": 5147 }, { "epoch": 0.64, "learning_rate": 0.00014758047407632108, "loss": 1.259, "step": 5148 }, { "epoch": 0.64, "learning_rate": 0.00014748862804175107, "loss": 1.2756, "step": 5149 }, { "epoch": 0.64, "learning_rate": 0.00014739679863681084, "loss": 1.2429, "step": 5150 }, { "epoch": 0.64, "learning_rate": 0.00014730498587639708, "loss": 1.2738, "step": 5151 }, { "epoch": 0.64, "learning_rate": 0.00014721318977540393, "loss": 1.2647, "step": 5152 }, { "epoch": 0.64, "learning_rate": 0.0001471214103487228, "loss": 1.2692, "step": 5153 }, { "epoch": 0.64, "learning_rate": 0.00014702964761124237, "loss": 1.2575, "step": 5154 }, { "epoch": 0.64, "learning_rate": 0.0001469379015778485, "loss": 1.2895, "step": 5155 }, { "epoch": 0.64, "learning_rate": 0.0001468461722634244, "loss": 1.2801, "step": 5156 }, { "epoch": 0.64, "learning_rate": 0.00014675445968285078, "loss": 1.2816, "step": 5157 }, { "epoch": 0.64, "learning_rate": 0.00014666276385100538, "loss": 1.3272, "step": 5158 }, { "epoch": 0.64, "learning_rate": 0.00014657108478276337, "loss": 1.2522, "step": 5159 }, { "epoch": 0.65, "learning_rate": 0.00014647942249299705, "loss": 1.2763, "step": 5160 }, { "epoch": 0.65, "learning_rate": 0.00014638777699657607, "loss": 1.2648, "step": 5161 }, { "epoch": 0.65, "learning_rate": 0.00014629614830836743, "loss": 1.2879, "step": 5162 }, { "epoch": 0.65, "learning_rate": 0.0001462045364432354, "loss": 1.2649, "step": 5163 }, { "epoch": 0.65, "learning_rate": 0.00014611294141604137, "loss": 1.2916, "step": 5164 }, { "epoch": 0.65, "learning_rate": 0.00014602136324164418, "loss": 1.3159, "step": 5165 }, { "epoch": 0.65, "learning_rate": 0.00014592980193489974, "loss": 1.3049, "step": 5166 }, { "epoch": 0.65, "learning_rate": 0.00014583825751066134, "loss": 1.2189, "step": 5167 }, { "epoch": 0.65, "learning_rate": 0.00014574672998377968, "loss": 1.2236, "step": 5168 }, { "epoch": 0.65, "learning_rate": 0.00014565521936910251, "loss": 1.2921, "step": 5169 }, { "epoch": 0.65, "learning_rate": 0.00014556372568147482, "loss": 1.2536, "step": 5170 }, { "epoch": 0.65, "learning_rate": 0.00014547224893573884, "loss": 1.2343, "step": 5171 }, { "epoch": 0.65, "learning_rate": 0.00014538078914673442, "loss": 1.2404, "step": 5172 }, { "epoch": 0.65, "learning_rate": 0.00014528934632929803, "loss": 1.2342, "step": 5173 }, { "epoch": 0.65, "learning_rate": 0.00014519792049826408, "loss": 1.2418, "step": 5174 }, { "epoch": 0.65, "learning_rate": 0.00014510651166846366, "loss": 1.2879, "step": 5175 }, { "epoch": 0.65, "learning_rate": 0.00014501511985472529, "loss": 1.2808, "step": 5176 }, { "epoch": 0.65, "learning_rate": 0.00014492374507187495, "loss": 1.2776, "step": 5177 }, { "epoch": 0.65, "learning_rate": 0.00014483238733473553, "loss": 1.2467, "step": 5178 }, { "epoch": 0.65, "learning_rate": 0.00014474104665812727, "loss": 1.24, "step": 5179 }, { "epoch": 0.65, "learning_rate": 0.00014464972305686777, "loss": 1.3209, "step": 5180 }, { "epoch": 0.65, "learning_rate": 0.00014455841654577162, "loss": 1.2067, "step": 5181 }, { "epoch": 0.65, "learning_rate": 0.00014446712713965092, "loss": 1.2768, "step": 5182 }, { "epoch": 0.65, "learning_rate": 0.00014437585485331482, "loss": 1.3021, "step": 5183 }, { "epoch": 0.65, "learning_rate": 0.00014428459970156966, "loss": 1.1921, "step": 5184 }, { "epoch": 0.65, "learning_rate": 0.00014419336169921892, "loss": 1.2683, "step": 5185 }, { "epoch": 0.65, "learning_rate": 0.00014410214086106365, "loss": 1.2362, "step": 5186 }, { "epoch": 0.65, "learning_rate": 0.00014401093720190194, "loss": 1.2124, "step": 5187 }, { "epoch": 0.65, "learning_rate": 0.00014391975073652896, "loss": 1.2342, "step": 5188 }, { "epoch": 0.65, "learning_rate": 0.00014382858147973717, "loss": 1.2304, "step": 5189 }, { "epoch": 0.65, "learning_rate": 0.00014373742944631622, "loss": 1.227, "step": 5190 }, { "epoch": 0.65, "learning_rate": 0.0001436462946510531, "loss": 1.2253, "step": 5191 }, { "epoch": 0.65, "learning_rate": 0.00014355517710873183, "loss": 1.2656, "step": 5192 }, { "epoch": 0.65, "learning_rate": 0.00014346407683413378, "loss": 1.2484, "step": 5193 }, { "epoch": 0.65, "learning_rate": 0.00014337299384203746, "loss": 1.2869, "step": 5194 }, { "epoch": 0.65, "learning_rate": 0.00014328192814721846, "loss": 1.3095, "step": 5195 }, { "epoch": 0.65, "learning_rate": 0.00014319087976444965, "loss": 1.2502, "step": 5196 }, { "epoch": 0.65, "learning_rate": 0.00014309984870850123, "loss": 1.2524, "step": 5197 }, { "epoch": 0.65, "learning_rate": 0.00014300883499414037, "loss": 1.222, "step": 5198 }, { "epoch": 0.65, "learning_rate": 0.0001429178386361316, "loss": 1.2748, "step": 5199 }, { "epoch": 0.65, "learning_rate": 0.00014282685964923643, "loss": 1.2539, "step": 5200 }, { "epoch": 0.65, "eval_acceptance_rate": 0.6941763162612915, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1942210790496017, "eval_expected_tokens": 2.742776348812002, "eval_loss": 1.3666422367095947, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4356940872030005, "eval_runtime": 50.3082, "eval_samples_per_second": 9.68, "eval_steps_per_second": 0.318, "step": 5200 }, { "epoch": 0.65, "learning_rate": 0.00014273589804821384, "loss": 1.2196, "step": 5201 }, { "epoch": 0.65, "learning_rate": 0.00014264495384781983, "loss": 1.2404, "step": 5202 }, { "epoch": 0.65, "learning_rate": 0.00014255402706280735, "loss": 1.2035, "step": 5203 }, { "epoch": 0.65, "learning_rate": 0.00014246311770792702, "loss": 1.3182, "step": 5204 }, { "epoch": 0.65, "learning_rate": 0.00014237222579792616, "loss": 1.3092, "step": 5205 }, { "epoch": 0.65, "learning_rate": 0.00014228135134754965, "loss": 1.3042, "step": 5206 }, { "epoch": 0.65, "learning_rate": 0.00014219049437153925, "loss": 1.2263, "step": 5207 }, { "epoch": 0.65, "learning_rate": 0.000142099654884634, "loss": 1.2183, "step": 5208 }, { "epoch": 0.65, "learning_rate": 0.00014200883290156995, "loss": 1.3377, "step": 5209 }, { "epoch": 0.65, "learning_rate": 0.00014191802843708062, "loss": 1.299, "step": 5210 }, { "epoch": 0.65, "learning_rate": 0.0001418272415058966, "loss": 1.2861, "step": 5211 }, { "epoch": 0.65, "learning_rate": 0.00014173647212274538, "loss": 1.2808, "step": 5212 }, { "epoch": 0.65, "learning_rate": 0.00014164572030235184, "loss": 1.2756, "step": 5213 }, { "epoch": 0.65, "learning_rate": 0.00014155498605943794, "loss": 1.2186, "step": 5214 }, { "epoch": 0.65, "learning_rate": 0.00014146426940872276, "loss": 1.365, "step": 5215 }, { "epoch": 0.65, "learning_rate": 0.00014137357036492253, "loss": 1.3248, "step": 5216 }, { "epoch": 0.65, "learning_rate": 0.00014128288894275088, "loss": 1.2887, "step": 5217 }, { "epoch": 0.65, "learning_rate": 0.00014119222515691815, "loss": 1.2601, "step": 5218 }, { "epoch": 0.65, "learning_rate": 0.0001411015790221321, "loss": 1.3298, "step": 5219 }, { "epoch": 0.65, "learning_rate": 0.00014101095055309745, "loss": 1.2357, "step": 5220 }, { "epoch": 0.65, "learning_rate": 0.00014092033976451635, "loss": 1.3178, "step": 5221 }, { "epoch": 0.65, "learning_rate": 0.00014082974667108767, "loss": 1.2434, "step": 5222 }, { "epoch": 0.65, "learning_rate": 0.00014073917128750785, "loss": 1.2761, "step": 5223 }, { "epoch": 0.65, "learning_rate": 0.0001406486136284701, "loss": 1.2414, "step": 5224 }, { "epoch": 0.65, "learning_rate": 0.00014055807370866487, "loss": 1.3092, "step": 5225 }, { "epoch": 0.65, "learning_rate": 0.00014046755154277987, "loss": 1.3276, "step": 5226 }, { "epoch": 0.65, "learning_rate": 0.00014037704714549976, "loss": 1.2426, "step": 5227 }, { "epoch": 0.65, "learning_rate": 0.00014028656053150627, "loss": 1.2719, "step": 5228 }, { "epoch": 0.65, "learning_rate": 0.0001401960917154785, "loss": 1.2325, "step": 5229 }, { "epoch": 0.65, "learning_rate": 0.00014010564071209236, "loss": 1.2447, "step": 5230 }, { "epoch": 0.65, "learning_rate": 0.0001400152075360212, "loss": 1.2313, "step": 5231 }, { "epoch": 0.65, "learning_rate": 0.00013992479220193522, "loss": 1.2597, "step": 5232 }, { "epoch": 0.65, "learning_rate": 0.00013983439472450176, "loss": 1.275, "step": 5233 }, { "epoch": 0.65, "learning_rate": 0.00013974401511838524, "loss": 1.2649, "step": 5234 }, { "epoch": 0.65, "learning_rate": 0.00013965365339824732, "loss": 1.2312, "step": 5235 }, { "epoch": 0.65, "learning_rate": 0.0001395633095787468, "loss": 1.2698, "step": 5236 }, { "epoch": 0.65, "learning_rate": 0.0001394729836745394, "loss": 1.2629, "step": 5237 }, { "epoch": 0.65, "learning_rate": 0.00013938267570027792, "loss": 1.2623, "step": 5238 }, { "epoch": 0.65, "learning_rate": 0.00013929238567061237, "loss": 1.2708, "step": 5239 }, { "epoch": 0.66, "learning_rate": 0.0001392021136001897, "loss": 1.2144, "step": 5240 }, { "epoch": 0.66, "learning_rate": 0.00013911185950365415, "loss": 1.2315, "step": 5241 }, { "epoch": 0.66, "learning_rate": 0.00013902162339564706, "loss": 1.2811, "step": 5242 }, { "epoch": 0.66, "learning_rate": 0.0001389314052908066, "loss": 1.1954, "step": 5243 }, { "epoch": 0.66, "learning_rate": 0.00013884120520376818, "loss": 1.2565, "step": 5244 }, { "epoch": 0.66, "learning_rate": 0.0001387510231491642, "loss": 1.2954, "step": 5245 }, { "epoch": 0.66, "learning_rate": 0.00013866085914162436, "loss": 1.2268, "step": 5246 }, { "epoch": 0.66, "learning_rate": 0.00013857071319577508, "loss": 1.2816, "step": 5247 }, { "epoch": 0.66, "learning_rate": 0.00013848058532624024, "loss": 1.315, "step": 5248 }, { "epoch": 0.66, "learning_rate": 0.0001383904755476404, "loss": 1.3146, "step": 5249 }, { "epoch": 0.66, "learning_rate": 0.00013830038387459354, "loss": 1.3023, "step": 5250 }, { "epoch": 0.66, "learning_rate": 0.0001382103103217145, "loss": 1.2929, "step": 5251 }, { "epoch": 0.66, "learning_rate": 0.00013812025490361516, "loss": 1.2911, "step": 5252 }, { "epoch": 0.66, "learning_rate": 0.00013803021763490443, "loss": 1.2539, "step": 5253 }, { "epoch": 0.66, "learning_rate": 0.0001379401985301885, "loss": 1.2381, "step": 5254 }, { "epoch": 0.66, "learning_rate": 0.00013785019760407052, "loss": 1.369, "step": 5255 }, { "epoch": 0.66, "learning_rate": 0.00013776021487115059, "loss": 1.2237, "step": 5256 }, { "epoch": 0.66, "learning_rate": 0.0001376702503460259, "loss": 1.3222, "step": 5257 }, { "epoch": 0.66, "learning_rate": 0.0001375803040432907, "loss": 1.2767, "step": 5258 }, { "epoch": 0.66, "learning_rate": 0.00013749037597753617, "loss": 1.2061, "step": 5259 }, { "epoch": 0.66, "learning_rate": 0.0001374004661633508, "loss": 1.3275, "step": 5260 }, { "epoch": 0.66, "learning_rate": 0.00013731057461531998, "loss": 1.2289, "step": 5261 }, { "epoch": 0.66, "learning_rate": 0.0001372207013480261, "loss": 1.1904, "step": 5262 }, { "epoch": 0.66, "learning_rate": 0.00013713084637604855, "loss": 1.261, "step": 5263 }, { "epoch": 0.66, "learning_rate": 0.00013704100971396378, "loss": 1.2424, "step": 5264 }, { "epoch": 0.66, "learning_rate": 0.00013695119137634542, "loss": 1.2646, "step": 5265 }, { "epoch": 0.66, "learning_rate": 0.0001368613913777639, "loss": 1.2943, "step": 5266 }, { "epoch": 0.66, "learning_rate": 0.00013677160973278685, "loss": 1.2864, "step": 5267 }, { "epoch": 0.66, "learning_rate": 0.00013668184645597886, "loss": 1.2758, "step": 5268 }, { "epoch": 0.66, "learning_rate": 0.0001365921015619014, "loss": 1.2026, "step": 5269 }, { "epoch": 0.66, "learning_rate": 0.00013650237506511331, "loss": 1.3286, "step": 5270 }, { "epoch": 0.66, "learning_rate": 0.00013641266698017013, "loss": 1.2646, "step": 5271 }, { "epoch": 0.66, "learning_rate": 0.00013632297732162437, "loss": 1.3416, "step": 5272 }, { "epoch": 0.66, "learning_rate": 0.00013623330610402595, "loss": 1.2261, "step": 5273 }, { "epoch": 0.66, "learning_rate": 0.00013614365334192128, "loss": 1.2282, "step": 5274 }, { "epoch": 0.66, "learning_rate": 0.00013605401904985426, "loss": 1.3078, "step": 5275 }, { "epoch": 0.66, "learning_rate": 0.0001359644032423655, "loss": 1.2081, "step": 5276 }, { "epoch": 0.66, "learning_rate": 0.00013587480593399266, "loss": 1.2355, "step": 5277 }, { "epoch": 0.66, "learning_rate": 0.00013578522713927028, "loss": 1.2158, "step": 5278 }, { "epoch": 0.66, "learning_rate": 0.0001356956668727302, "loss": 1.2546, "step": 5279 }, { "epoch": 0.66, "learning_rate": 0.0001356061251489012, "loss": 1.2133, "step": 5280 }, { "epoch": 0.66, "learning_rate": 0.00013551660198230875, "loss": 1.2516, "step": 5281 }, { "epoch": 0.66, "learning_rate": 0.0001354270973874756, "loss": 1.2945, "step": 5282 }, { "epoch": 0.66, "learning_rate": 0.00013533761137892136, "loss": 1.2589, "step": 5283 }, { "epoch": 0.66, "learning_rate": 0.00013524814397116252, "loss": 1.2387, "step": 5284 }, { "epoch": 0.66, "learning_rate": 0.00013515869517871288, "loss": 1.248, "step": 5285 }, { "epoch": 0.66, "learning_rate": 0.00013506926501608303, "loss": 1.2582, "step": 5286 }, { "epoch": 0.66, "learning_rate": 0.00013497985349778045, "loss": 1.2403, "step": 5287 }, { "epoch": 0.66, "learning_rate": 0.00013489046063830973, "loss": 1.2851, "step": 5288 }, { "epoch": 0.66, "learning_rate": 0.00013480108645217223, "loss": 1.2067, "step": 5289 }, { "epoch": 0.66, "learning_rate": 0.0001347117309538667, "loss": 1.1782, "step": 5290 }, { "epoch": 0.66, "learning_rate": 0.00013462239415788836, "loss": 1.2102, "step": 5291 }, { "epoch": 0.66, "learning_rate": 0.0001345330760787298, "loss": 1.2114, "step": 5292 }, { "epoch": 0.66, "learning_rate": 0.0001344437767308803, "loss": 1.2761, "step": 5293 }, { "epoch": 0.66, "learning_rate": 0.0001343544961288262, "loss": 1.283, "step": 5294 }, { "epoch": 0.66, "learning_rate": 0.0001342652342870509, "loss": 1.259, "step": 5295 }, { "epoch": 0.66, "learning_rate": 0.00013417599122003462, "loss": 1.3199, "step": 5296 }, { "epoch": 0.66, "learning_rate": 0.00013408676694225447, "loss": 1.2528, "step": 5297 }, { "epoch": 0.66, "learning_rate": 0.0001339975614681847, "loss": 1.3021, "step": 5298 }, { "epoch": 0.66, "learning_rate": 0.0001339083748122965, "loss": 1.2445, "step": 5299 }, { "epoch": 0.66, "learning_rate": 0.00013381920698905787, "loss": 1.2614, "step": 5300 }, { "epoch": 0.66, "eval_acceptance_rate": 0.6949754357337952, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1975305920359163, "eval_expected_tokens": 2.7469132400448952, "eval_loss": 1.3633508682250977, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.43672831001122375, "eval_runtime": 50.347, "eval_samples_per_second": 9.673, "eval_steps_per_second": 0.318, "step": 5300 }, { "epoch": 0.66, "learning_rate": 0.00013373005801293382, "loss": 1.2314, "step": 5301 }, { "epoch": 0.66, "learning_rate": 0.00013364092789838627, "loss": 1.3197, "step": 5302 }, { "epoch": 0.66, "learning_rate": 0.0001335518166598741, "loss": 1.2589, "step": 5303 }, { "epoch": 0.66, "learning_rate": 0.0001334627243118531, "loss": 1.2836, "step": 5304 }, { "epoch": 0.66, "learning_rate": 0.00013337365086877622, "loss": 1.2455, "step": 5305 }, { "epoch": 0.66, "learning_rate": 0.0001332845963450931, "loss": 1.2777, "step": 5306 }, { "epoch": 0.66, "learning_rate": 0.00013319556075525027, "loss": 1.2589, "step": 5307 }, { "epoch": 0.66, "learning_rate": 0.00013310654411369125, "loss": 1.3145, "step": 5308 }, { "epoch": 0.66, "learning_rate": 0.0001330175464348567, "loss": 1.2724, "step": 5309 }, { "epoch": 0.66, "learning_rate": 0.00013292856773318385, "loss": 1.3196, "step": 5310 }, { "epoch": 0.66, "learning_rate": 0.00013283960802310717, "loss": 1.2149, "step": 5311 }, { "epoch": 0.66, "learning_rate": 0.0001327506673190579, "loss": 1.2229, "step": 5312 }, { "epoch": 0.66, "learning_rate": 0.00013266174563546403, "loss": 1.221, "step": 5313 }, { "epoch": 0.66, "learning_rate": 0.00013257284298675087, "loss": 1.2546, "step": 5314 }, { "epoch": 0.66, "learning_rate": 0.00013248395938734032, "loss": 1.2371, "step": 5315 }, { "epoch": 0.66, "learning_rate": 0.00013239509485165113, "loss": 1.2452, "step": 5316 }, { "epoch": 0.66, "learning_rate": 0.00013230624939409934, "loss": 1.1812, "step": 5317 }, { "epoch": 0.66, "learning_rate": 0.00013221742302909752, "loss": 1.2611, "step": 5318 }, { "epoch": 0.66, "learning_rate": 0.00013212861577105545, "loss": 1.2059, "step": 5319 }, { "epoch": 0.67, "learning_rate": 0.0001320398276343795, "loss": 1.2985, "step": 5320 }, { "epoch": 0.67, "learning_rate": 0.00013195105863347313, "loss": 1.2491, "step": 5321 }, { "epoch": 0.67, "learning_rate": 0.00013186230878273653, "loss": 1.3121, "step": 5322 }, { "epoch": 0.67, "learning_rate": 0.00013177357809656702, "loss": 1.2428, "step": 5323 }, { "epoch": 0.67, "learning_rate": 0.00013168486658935878, "loss": 1.2198, "step": 5324 }, { "epoch": 0.67, "learning_rate": 0.00013159617427550268, "loss": 1.3047, "step": 5325 }, { "epoch": 0.67, "learning_rate": 0.00013150750116938664, "loss": 1.2919, "step": 5326 }, { "epoch": 0.67, "learning_rate": 0.00013141884728539542, "loss": 1.2189, "step": 5327 }, { "epoch": 0.67, "learning_rate": 0.0001313302126379105, "loss": 1.2982, "step": 5328 }, { "epoch": 0.67, "learning_rate": 0.0001312415972413105, "loss": 1.2753, "step": 5329 }, { "epoch": 0.67, "learning_rate": 0.00013115300110997098, "loss": 1.25, "step": 5330 }, { "epoch": 0.67, "learning_rate": 0.00013106442425826404, "loss": 1.2422, "step": 5331 }, { "epoch": 0.67, "learning_rate": 0.00013097586670055884, "loss": 1.2557, "step": 5332 }, { "epoch": 0.67, "learning_rate": 0.00013088732845122127, "loss": 1.2679, "step": 5333 }, { "epoch": 0.67, "learning_rate": 0.00013079880952461442, "loss": 1.2866, "step": 5334 }, { "epoch": 0.67, "learning_rate": 0.00013071030993509788, "loss": 1.2533, "step": 5335 }, { "epoch": 0.67, "learning_rate": 0.0001306218296970284, "loss": 1.232, "step": 5336 }, { "epoch": 0.67, "learning_rate": 0.00013053336882475937, "loss": 1.247, "step": 5337 }, { "epoch": 0.67, "learning_rate": 0.00013044492733264103, "loss": 1.2586, "step": 5338 }, { "epoch": 0.67, "learning_rate": 0.00013035650523502071, "loss": 1.253, "step": 5339 }, { "epoch": 0.67, "learning_rate": 0.0001302681025462424, "loss": 1.2844, "step": 5340 }, { "epoch": 0.67, "learning_rate": 0.00013017971928064683, "loss": 1.2733, "step": 5341 }, { "epoch": 0.67, "learning_rate": 0.000130091355452572, "loss": 1.2002, "step": 5342 }, { "epoch": 0.67, "learning_rate": 0.00013000301107635227, "loss": 1.3036, "step": 5343 }, { "epoch": 0.67, "learning_rate": 0.00012991468616631923, "loss": 1.2918, "step": 5344 }, { "epoch": 0.67, "learning_rate": 0.00012982638073680112, "loss": 1.2661, "step": 5345 }, { "epoch": 0.67, "learning_rate": 0.00012973809480212298, "loss": 1.3257, "step": 5346 }, { "epoch": 0.67, "learning_rate": 0.00012964982837660672, "loss": 1.3027, "step": 5347 }, { "epoch": 0.67, "learning_rate": 0.00012956158147457115, "loss": 1.2716, "step": 5348 }, { "epoch": 0.67, "learning_rate": 0.000129473354110332, "loss": 1.2584, "step": 5349 }, { "epoch": 0.67, "learning_rate": 0.0001293851462982017, "loss": 1.2602, "step": 5350 }, { "epoch": 0.67, "learning_rate": 0.0001292969580524894, "loss": 1.2746, "step": 5351 }, { "epoch": 0.67, "learning_rate": 0.0001292087893875012, "loss": 1.3243, "step": 5352 }, { "epoch": 0.67, "learning_rate": 0.0001291206403175402, "loss": 1.2177, "step": 5353 }, { "epoch": 0.67, "learning_rate": 0.0001290325108569059, "loss": 1.2765, "step": 5354 }, { "epoch": 0.67, "learning_rate": 0.0001289444010198951, "loss": 1.259, "step": 5355 }, { "epoch": 0.67, "learning_rate": 0.00012885631082080106, "loss": 1.2166, "step": 5356 }, { "epoch": 0.67, "learning_rate": 0.0001287682402739139, "loss": 1.2169, "step": 5357 }, { "epoch": 0.67, "learning_rate": 0.00012868018939352077, "loss": 1.2712, "step": 5358 }, { "epoch": 0.67, "learning_rate": 0.00012859215819390545, "loss": 1.235, "step": 5359 }, { "epoch": 0.67, "learning_rate": 0.00012850414668934848, "loss": 1.277, "step": 5360 }, { "epoch": 0.67, "learning_rate": 0.0001284161548941274, "loss": 1.3063, "step": 5361 }, { "epoch": 0.67, "learning_rate": 0.0001283281828225163, "loss": 1.24, "step": 5362 }, { "epoch": 0.67, "learning_rate": 0.0001282402304887864, "loss": 1.2343, "step": 5363 }, { "epoch": 0.67, "learning_rate": 0.00012815229790720544, "loss": 1.2221, "step": 5364 }, { "epoch": 0.67, "learning_rate": 0.000128064385092038, "loss": 1.2918, "step": 5365 }, { "epoch": 0.67, "learning_rate": 0.00012797649205754548, "loss": 1.2669, "step": 5366 }, { "epoch": 0.67, "learning_rate": 0.0001278886188179861, "loss": 1.2981, "step": 5367 }, { "epoch": 0.67, "learning_rate": 0.00012780076538761504, "loss": 1.2691, "step": 5368 }, { "epoch": 0.67, "learning_rate": 0.00012771293178068387, "loss": 1.2379, "step": 5369 }, { "epoch": 0.67, "learning_rate": 0.00012762511801144133, "loss": 1.2318, "step": 5370 }, { "epoch": 0.67, "learning_rate": 0.0001275373240941326, "loss": 1.2916, "step": 5371 }, { "epoch": 0.67, "learning_rate": 0.00012744955004299983, "loss": 1.2918, "step": 5372 }, { "epoch": 0.67, "learning_rate": 0.00012736179587228197, "loss": 1.2291, "step": 5373 }, { "epoch": 0.67, "learning_rate": 0.00012727406159621478, "loss": 1.2079, "step": 5374 }, { "epoch": 0.67, "learning_rate": 0.0001271863472290307, "loss": 1.2352, "step": 5375 }, { "epoch": 0.67, "learning_rate": 0.0001270986527849589, "loss": 1.2491, "step": 5376 }, { "epoch": 0.67, "learning_rate": 0.00012701097827822527, "loss": 1.2594, "step": 5377 }, { "epoch": 0.67, "learning_rate": 0.00012692332372305276, "loss": 1.2189, "step": 5378 }, { "epoch": 0.67, "learning_rate": 0.00012683568913366073, "loss": 1.2935, "step": 5379 }, { "epoch": 0.67, "learning_rate": 0.00012674807452426565, "loss": 1.2621, "step": 5380 }, { "epoch": 0.67, "learning_rate": 0.00012666047990908048, "loss": 1.2613, "step": 5381 }, { "epoch": 0.67, "learning_rate": 0.00012657290530231486, "loss": 1.2111, "step": 5382 }, { "epoch": 0.67, "learning_rate": 0.00012648535071817558, "loss": 1.3492, "step": 5383 }, { "epoch": 0.67, "learning_rate": 0.00012639781617086588, "loss": 1.3076, "step": 5384 }, { "epoch": 0.67, "learning_rate": 0.00012631030167458568, "loss": 1.2107, "step": 5385 }, { "epoch": 0.67, "learning_rate": 0.00012622280724353197, "loss": 1.2223, "step": 5386 }, { "epoch": 0.67, "learning_rate": 0.0001261353328918981, "loss": 1.2585, "step": 5387 }, { "epoch": 0.67, "learning_rate": 0.00012604787863387457, "loss": 1.2451, "step": 5388 }, { "epoch": 0.67, "learning_rate": 0.00012596044448364834, "loss": 1.2895, "step": 5389 }, { "epoch": 0.67, "learning_rate": 0.00012587303045540313, "loss": 1.1836, "step": 5390 }, { "epoch": 0.67, "learning_rate": 0.00012578563656331937, "loss": 1.2922, "step": 5391 }, { "epoch": 0.67, "learning_rate": 0.0001256982628215744, "loss": 1.2831, "step": 5392 }, { "epoch": 0.67, "learning_rate": 0.00012561090924434226, "loss": 1.247, "step": 5393 }, { "epoch": 0.67, "learning_rate": 0.0001255235758457936, "loss": 1.2351, "step": 5394 }, { "epoch": 0.67, "learning_rate": 0.00012543626264009577, "loss": 1.2329, "step": 5395 }, { "epoch": 0.67, "learning_rate": 0.0001253489696414129, "loss": 1.1971, "step": 5396 }, { "epoch": 0.67, "learning_rate": 0.000125261696863906, "loss": 1.1669, "step": 5397 }, { "epoch": 0.67, "learning_rate": 0.0001251744443217325, "loss": 1.3141, "step": 5398 }, { "epoch": 0.67, "learning_rate": 0.00012508721202904687, "loss": 1.2414, "step": 5399 }, { "epoch": 0.68, "learning_rate": 0.00012500000000000006, "loss": 1.2406, "step": 5400 }, { "epoch": 0.68, "eval_acceptance_rate": 0.6954663395881653, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.1995666760506447, "eval_expected_tokens": 2.7494583450633057, "eval_loss": 1.3609733581542969, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4373645862658264, "eval_runtime": 50.3236, "eval_samples_per_second": 9.677, "eval_steps_per_second": 0.318, "step": 5400 }, { "epoch": 0.68, "learning_rate": 0.0001249128082487397, "loss": 1.2425, "step": 5401 }, { "epoch": 0.68, "learning_rate": 0.0001248256367894104, "loss": 1.2652, "step": 5402 }, { "epoch": 0.68, "learning_rate": 0.00012473848563615328, "loss": 1.2091, "step": 5403 }, { "epoch": 0.68, "learning_rate": 0.00012465135480310604, "loss": 1.2768, "step": 5404 }, { "epoch": 0.68, "learning_rate": 0.0001245642443044035, "loss": 1.2557, "step": 5405 }, { "epoch": 0.68, "learning_rate": 0.00012447715415417668, "loss": 1.2059, "step": 5406 }, { "epoch": 0.68, "learning_rate": 0.00012439008436655375, "loss": 1.2577, "step": 5407 }, { "epoch": 0.68, "learning_rate": 0.00012430303495565928, "loss": 1.2564, "step": 5408 }, { "epoch": 0.68, "learning_rate": 0.0001242160059356146, "loss": 1.2752, "step": 5409 }, { "epoch": 0.68, "learning_rate": 0.00012412899732053772, "loss": 1.2429, "step": 5410 }, { "epoch": 0.68, "learning_rate": 0.00012404200912454345, "loss": 1.2747, "step": 5411 }, { "epoch": 0.68, "learning_rate": 0.00012395504136174327, "loss": 1.2358, "step": 5412 }, { "epoch": 0.68, "learning_rate": 0.00012386809404624523, "loss": 1.2075, "step": 5413 }, { "epoch": 0.68, "learning_rate": 0.0001237811671921541, "loss": 1.3196, "step": 5414 }, { "epoch": 0.68, "learning_rate": 0.00012369426081357138, "loss": 1.2446, "step": 5415 }, { "epoch": 0.68, "learning_rate": 0.0001236073749245951, "loss": 1.2655, "step": 5416 }, { "epoch": 0.68, "learning_rate": 0.0001235205095393202, "loss": 1.2175, "step": 5417 }, { "epoch": 0.68, "learning_rate": 0.00012343366467183832, "loss": 1.2197, "step": 5418 }, { "epoch": 0.68, "learning_rate": 0.00012334684033623745, "loss": 1.2517, "step": 5419 }, { "epoch": 0.68, "learning_rate": 0.00012326003654660248, "loss": 1.2885, "step": 5420 }, { "epoch": 0.68, "learning_rate": 0.00012317325331701482, "loss": 1.2192, "step": 5421 }, { "epoch": 0.68, "learning_rate": 0.00012308649066155284, "loss": 1.2635, "step": 5422 }, { "epoch": 0.68, "learning_rate": 0.0001229997485942912, "loss": 1.2113, "step": 5423 }, { "epoch": 0.68, "learning_rate": 0.00012291302712930157, "loss": 1.2721, "step": 5424 }, { "epoch": 0.68, "learning_rate": 0.00012282632628065197, "loss": 1.2099, "step": 5425 }, { "epoch": 0.68, "learning_rate": 0.00012273964606240718, "loss": 1.2442, "step": 5426 }, { "epoch": 0.68, "learning_rate": 0.00012265298648862883, "loss": 1.2167, "step": 5427 }, { "epoch": 0.68, "learning_rate": 0.00012256634757337493, "loss": 1.2252, "step": 5428 }, { "epoch": 0.68, "learning_rate": 0.00012247972933070015, "loss": 1.2951, "step": 5429 }, { "epoch": 0.68, "learning_rate": 0.00012239313177465612, "loss": 1.2269, "step": 5430 }, { "epoch": 0.68, "learning_rate": 0.00012230655491929067, "loss": 1.2827, "step": 5431 }, { "epoch": 0.68, "learning_rate": 0.0001222199987786487, "loss": 1.2265, "step": 5432 }, { "epoch": 0.68, "learning_rate": 0.00012213346336677143, "loss": 1.2365, "step": 5433 }, { "epoch": 0.68, "learning_rate": 0.00012204694869769686, "loss": 1.2858, "step": 5434 }, { "epoch": 0.68, "learning_rate": 0.0001219604547854595, "loss": 1.2396, "step": 5435 }, { "epoch": 0.68, "learning_rate": 0.00012187398164409067, "loss": 1.1959, "step": 5436 }, { "epoch": 0.68, "learning_rate": 0.00012178752928761832, "loss": 1.2959, "step": 5437 }, { "epoch": 0.68, "learning_rate": 0.00012170109773006694, "loss": 1.3101, "step": 5438 }, { "epoch": 0.68, "learning_rate": 0.00012161468698545755, "loss": 1.2404, "step": 5439 }, { "epoch": 0.68, "learning_rate": 0.00012152829706780786, "loss": 1.2212, "step": 5440 }, { "epoch": 0.68, "learning_rate": 0.0001214419279911324, "loss": 1.2454, "step": 5441 }, { "epoch": 0.68, "learning_rate": 0.00012135557976944203, "loss": 1.1867, "step": 5442 }, { "epoch": 0.68, "learning_rate": 0.00012126925241674447, "loss": 1.2382, "step": 5443 }, { "epoch": 0.68, "learning_rate": 0.0001211829459470439, "loss": 1.2572, "step": 5444 }, { "epoch": 0.68, "learning_rate": 0.00012109666037434103, "loss": 1.3005, "step": 5445 }, { "epoch": 0.68, "learning_rate": 0.0001210103957126335, "loss": 1.2348, "step": 5446 }, { "epoch": 0.68, "learning_rate": 0.00012092415197591519, "loss": 1.2248, "step": 5447 }, { "epoch": 0.68, "learning_rate": 0.0001208379291781769, "loss": 1.2427, "step": 5448 }, { "epoch": 0.68, "learning_rate": 0.00012075172733340586, "loss": 1.2629, "step": 5449 }, { "epoch": 0.68, "learning_rate": 0.00012066554645558578, "loss": 1.2275, "step": 5450 }, { "epoch": 0.68, "learning_rate": 0.00012057938655869735, "loss": 1.2865, "step": 5451 }, { "epoch": 0.68, "learning_rate": 0.00012049324765671748, "loss": 1.2865, "step": 5452 }, { "epoch": 0.68, "learning_rate": 0.00012040712976361979, "loss": 1.2396, "step": 5453 }, { "epoch": 0.68, "learning_rate": 0.00012032103289337467, "loss": 1.2217, "step": 5454 }, { "epoch": 0.68, "learning_rate": 0.00012023495705994878, "loss": 1.2787, "step": 5455 }, { "epoch": 0.68, "learning_rate": 0.00012014890227730569, "loss": 1.331, "step": 5456 }, { "epoch": 0.68, "learning_rate": 0.00012006286855940538, "loss": 1.2291, "step": 5457 }, { "epoch": 0.68, "learning_rate": 0.00011997685592020436, "loss": 1.2716, "step": 5458 }, { "epoch": 0.68, "learning_rate": 0.00011989086437365576, "loss": 1.1993, "step": 5459 }, { "epoch": 0.68, "learning_rate": 0.00011980489393370938, "loss": 1.2923, "step": 5460 }, { "epoch": 0.68, "learning_rate": 0.00011971894461431165, "loss": 1.2427, "step": 5461 }, { "epoch": 0.68, "learning_rate": 0.00011963301642940541, "loss": 1.1985, "step": 5462 }, { "epoch": 0.68, "learning_rate": 0.00011954710939293007, "loss": 1.1967, "step": 5463 }, { "epoch": 0.68, "learning_rate": 0.00011946122351882171, "loss": 1.2074, "step": 5464 }, { "epoch": 0.68, "learning_rate": 0.00011937535882101281, "loss": 1.1977, "step": 5465 }, { "epoch": 0.68, "learning_rate": 0.00011928951531343263, "loss": 1.2502, "step": 5466 }, { "epoch": 0.68, "learning_rate": 0.00011920369301000702, "loss": 1.2666, "step": 5467 }, { "epoch": 0.68, "learning_rate": 0.00011911789192465816, "loss": 1.2054, "step": 5468 }, { "epoch": 0.68, "learning_rate": 0.00011903211207130491, "loss": 1.2109, "step": 5469 }, { "epoch": 0.68, "learning_rate": 0.00011894635346386255, "loss": 1.2148, "step": 5470 }, { "epoch": 0.68, "learning_rate": 0.00011886061611624327, "loss": 1.2576, "step": 5471 }, { "epoch": 0.68, "learning_rate": 0.00011877490004235535, "loss": 1.1891, "step": 5472 }, { "epoch": 0.68, "learning_rate": 0.00011868920525610408, "loss": 1.2789, "step": 5473 }, { "epoch": 0.68, "learning_rate": 0.00011860353177139094, "loss": 1.2483, "step": 5474 }, { "epoch": 0.68, "learning_rate": 0.00011851787960211397, "loss": 1.3303, "step": 5475 }, { "epoch": 0.68, "learning_rate": 0.00011843224876216807, "loss": 1.266, "step": 5476 }, { "epoch": 0.68, "learning_rate": 0.0001183466392654444, "loss": 1.3123, "step": 5477 }, { "epoch": 0.68, "learning_rate": 0.0001182610511258306, "loss": 1.2711, "step": 5478 }, { "epoch": 0.68, "learning_rate": 0.00011817548435721117, "loss": 1.2074, "step": 5479 }, { "epoch": 0.69, "learning_rate": 0.00011808993897346678, "loss": 1.2472, "step": 5480 }, { "epoch": 0.69, "learning_rate": 0.00011800441498847497, "loss": 1.2397, "step": 5481 }, { "epoch": 0.69, "learning_rate": 0.00011791891241610952, "loss": 1.2305, "step": 5482 }, { "epoch": 0.69, "learning_rate": 0.00011783343127024091, "loss": 1.2371, "step": 5483 }, { "epoch": 0.69, "learning_rate": 0.00011774797156473594, "loss": 1.2021, "step": 5484 }, { "epoch": 0.69, "learning_rate": 0.00011766253331345819, "loss": 1.2609, "step": 5485 }, { "epoch": 0.69, "learning_rate": 0.00011757711653026776, "loss": 1.303, "step": 5486 }, { "epoch": 0.69, "learning_rate": 0.00011749172122902102, "loss": 1.2144, "step": 5487 }, { "epoch": 0.69, "learning_rate": 0.00011740634742357104, "loss": 1.2919, "step": 5488 }, { "epoch": 0.69, "learning_rate": 0.00011732099512776733, "loss": 1.3052, "step": 5489 }, { "epoch": 0.69, "learning_rate": 0.00011723566435545585, "loss": 1.2274, "step": 5490 }, { "epoch": 0.69, "learning_rate": 0.00011715035512047925, "loss": 1.257, "step": 5491 }, { "epoch": 0.69, "learning_rate": 0.00011706506743667666, "loss": 1.2236, "step": 5492 }, { "epoch": 0.69, "learning_rate": 0.00011697980131788358, "loss": 1.2505, "step": 5493 }, { "epoch": 0.69, "learning_rate": 0.00011689455677793196, "loss": 1.2405, "step": 5494 }, { "epoch": 0.69, "learning_rate": 0.00011680933383065057, "loss": 1.2527, "step": 5495 }, { "epoch": 0.69, "learning_rate": 0.00011672413248986435, "loss": 1.2638, "step": 5496 }, { "epoch": 0.69, "learning_rate": 0.0001166389527693948, "loss": 1.164, "step": 5497 }, { "epoch": 0.69, "learning_rate": 0.00011655379468306013, "loss": 1.2172, "step": 5498 }, { "epoch": 0.69, "learning_rate": 0.0001164686582446747, "loss": 1.3232, "step": 5499 }, { "epoch": 0.69, "learning_rate": 0.00011638354346804972, "loss": 1.2502, "step": 5500 }, { "epoch": 0.69, "eval_acceptance_rate": 0.6961297988891602, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2023221244758378, "eval_expected_tokens": 2.7529026555947973, "eval_loss": 1.359587550163269, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4382256638986993, "eval_runtime": 50.2655, "eval_samples_per_second": 9.689, "eval_steps_per_second": 0.318, "step": 5500 }, { "epoch": 0.69, "learning_rate": 0.00011629845036699263, "loss": 1.2496, "step": 5501 }, { "epoch": 0.69, "learning_rate": 0.0001162133789553074, "loss": 1.2755, "step": 5502 }, { "epoch": 0.69, "learning_rate": 0.00011612832924679446, "loss": 1.2772, "step": 5503 }, { "epoch": 0.69, "learning_rate": 0.00011604330125525078, "loss": 1.2005, "step": 5504 }, { "epoch": 0.69, "learning_rate": 0.00011595829499446998, "loss": 1.2152, "step": 5505 }, { "epoch": 0.69, "learning_rate": 0.00011587331047824185, "loss": 1.1928, "step": 5506 }, { "epoch": 0.69, "learning_rate": 0.00011578834772035274, "loss": 1.219, "step": 5507 }, { "epoch": 0.69, "learning_rate": 0.00011570340673458552, "loss": 1.2584, "step": 5508 }, { "epoch": 0.69, "learning_rate": 0.00011561848753471943, "loss": 1.2342, "step": 5509 }, { "epoch": 0.69, "learning_rate": 0.00011553359013453035, "loss": 1.2184, "step": 5510 }, { "epoch": 0.69, "learning_rate": 0.0001154487145477906, "loss": 1.2174, "step": 5511 }, { "epoch": 0.69, "learning_rate": 0.00011536386078826883, "loss": 1.2736, "step": 5512 }, { "epoch": 0.69, "learning_rate": 0.0001152790288697302, "loss": 1.2779, "step": 5513 }, { "epoch": 0.69, "learning_rate": 0.00011519421880593626, "loss": 1.2305, "step": 5514 }, { "epoch": 0.69, "learning_rate": 0.00011510943061064524, "loss": 1.2374, "step": 5515 }, { "epoch": 0.69, "learning_rate": 0.00011502466429761152, "loss": 1.237, "step": 5516 }, { "epoch": 0.69, "learning_rate": 0.00011493991988058625, "loss": 1.2475, "step": 5517 }, { "epoch": 0.69, "learning_rate": 0.00011485519737331681, "loss": 1.252, "step": 5518 }, { "epoch": 0.69, "learning_rate": 0.00011477049678954693, "loss": 1.3265, "step": 5519 }, { "epoch": 0.69, "learning_rate": 0.00011468581814301718, "loss": 1.201, "step": 5520 }, { "epoch": 0.69, "learning_rate": 0.00011460116144746419, "loss": 1.2198, "step": 5521 }, { "epoch": 0.69, "learning_rate": 0.00011451652671662108, "loss": 1.198, "step": 5522 }, { "epoch": 0.69, "learning_rate": 0.00011443191396421768, "loss": 1.1875, "step": 5523 }, { "epoch": 0.69, "learning_rate": 0.00011434732320397986, "loss": 1.2727, "step": 5524 }, { "epoch": 0.69, "learning_rate": 0.00011426275444963033, "loss": 1.2355, "step": 5525 }, { "epoch": 0.69, "learning_rate": 0.00011417820771488796, "loss": 1.2489, "step": 5526 }, { "epoch": 0.69, "learning_rate": 0.00011409368301346807, "loss": 1.1994, "step": 5527 }, { "epoch": 0.69, "learning_rate": 0.00011400918035908236, "loss": 1.2722, "step": 5528 }, { "epoch": 0.69, "learning_rate": 0.00011392469976543917, "loss": 1.2973, "step": 5529 }, { "epoch": 0.69, "learning_rate": 0.00011384024124624323, "loss": 1.2288, "step": 5530 }, { "epoch": 0.69, "learning_rate": 0.00011375580481519546, "loss": 1.2471, "step": 5531 }, { "epoch": 0.69, "learning_rate": 0.00011367139048599334, "loss": 1.1979, "step": 5532 }, { "epoch": 0.69, "learning_rate": 0.0001135869982723308, "loss": 1.2385, "step": 5533 }, { "epoch": 0.69, "learning_rate": 0.00011350262818789799, "loss": 1.305, "step": 5534 }, { "epoch": 0.69, "learning_rate": 0.00011341828024638176, "loss": 1.2755, "step": 5535 }, { "epoch": 0.69, "learning_rate": 0.0001133339544614653, "loss": 1.2681, "step": 5536 }, { "epoch": 0.69, "learning_rate": 0.00011324965084682804, "loss": 1.2594, "step": 5537 }, { "epoch": 0.69, "learning_rate": 0.00011316536941614589, "loss": 1.24, "step": 5538 }, { "epoch": 0.69, "learning_rate": 0.00011308111018309108, "loss": 1.2578, "step": 5539 }, { "epoch": 0.69, "learning_rate": 0.00011299687316133256, "loss": 1.2368, "step": 5540 }, { "epoch": 0.69, "learning_rate": 0.00011291265836453523, "loss": 1.2606, "step": 5541 }, { "epoch": 0.69, "learning_rate": 0.00011282846580636081, "loss": 1.2063, "step": 5542 }, { "epoch": 0.69, "learning_rate": 0.00011274429550046702, "loss": 1.2164, "step": 5543 }, { "epoch": 0.69, "learning_rate": 0.00011266014746050837, "loss": 1.2455, "step": 5544 }, { "epoch": 0.69, "learning_rate": 0.0001125760217001354, "loss": 1.2573, "step": 5545 }, { "epoch": 0.69, "learning_rate": 0.00011249191823299523, "loss": 1.2261, "step": 5546 }, { "epoch": 0.69, "learning_rate": 0.00011240783707273122, "loss": 1.2575, "step": 5547 }, { "epoch": 0.69, "learning_rate": 0.0001123237782329833, "loss": 1.2597, "step": 5548 }, { "epoch": 0.69, "learning_rate": 0.00011223974172738774, "loss": 1.2892, "step": 5549 }, { "epoch": 0.69, "learning_rate": 0.0001121557275695771, "loss": 1.2659, "step": 5550 }, { "epoch": 0.69, "learning_rate": 0.0001120717357731803, "loss": 1.2715, "step": 5551 }, { "epoch": 0.69, "learning_rate": 0.00011198776635182272, "loss": 1.2066, "step": 5552 }, { "epoch": 0.69, "learning_rate": 0.00011190381931912596, "loss": 1.1889, "step": 5553 }, { "epoch": 0.69, "learning_rate": 0.0001118198946887082, "loss": 1.2019, "step": 5554 }, { "epoch": 0.69, "learning_rate": 0.00011173599247418397, "loss": 1.2218, "step": 5555 }, { "epoch": 0.69, "learning_rate": 0.000111652112689164, "loss": 1.2124, "step": 5556 }, { "epoch": 0.69, "learning_rate": 0.00011156825534725545, "loss": 1.2378, "step": 5557 }, { "epoch": 0.69, "learning_rate": 0.00011148442046206178, "loss": 1.2615, "step": 5558 }, { "epoch": 0.69, "learning_rate": 0.00011140060804718302, "loss": 1.1778, "step": 5559 }, { "epoch": 0.69, "learning_rate": 0.00011131681811621528, "loss": 1.2784, "step": 5560 }, { "epoch": 0.7, "learning_rate": 0.00011123305068275127, "loss": 1.2533, "step": 5561 }, { "epoch": 0.7, "learning_rate": 0.00011114930576037991, "loss": 1.234, "step": 5562 }, { "epoch": 0.7, "learning_rate": 0.00011106558336268639, "loss": 1.2413, "step": 5563 }, { "epoch": 0.7, "learning_rate": 0.00011098188350325248, "loss": 1.2395, "step": 5564 }, { "epoch": 0.7, "learning_rate": 0.00011089820619565614, "loss": 1.2087, "step": 5565 }, { "epoch": 0.7, "learning_rate": 0.00011081455145347152, "loss": 1.2549, "step": 5566 }, { "epoch": 0.7, "learning_rate": 0.00011073091929026958, "loss": 1.2582, "step": 5567 }, { "epoch": 0.7, "learning_rate": 0.00011064730971961703, "loss": 1.2266, "step": 5568 }, { "epoch": 0.7, "learning_rate": 0.00011056372275507747, "loss": 1.227, "step": 5569 }, { "epoch": 0.7, "learning_rate": 0.00011048015841021044, "loss": 1.2456, "step": 5570 }, { "epoch": 0.7, "learning_rate": 0.00011039661669857193, "loss": 1.2369, "step": 5571 }, { "epoch": 0.7, "learning_rate": 0.0001103130976337142, "loss": 1.2156, "step": 5572 }, { "epoch": 0.7, "learning_rate": 0.00011022960122918599, "loss": 1.2776, "step": 5573 }, { "epoch": 0.7, "learning_rate": 0.00011014612749853233, "loss": 1.2776, "step": 5574 }, { "epoch": 0.7, "learning_rate": 0.0001100626764552945, "loss": 1.2607, "step": 5575 }, { "epoch": 0.7, "learning_rate": 0.00010997924811301008, "loss": 1.2787, "step": 5576 }, { "epoch": 0.7, "learning_rate": 0.000109895842485213, "loss": 1.264, "step": 5577 }, { "epoch": 0.7, "learning_rate": 0.00010981245958543343, "loss": 1.2743, "step": 5578 }, { "epoch": 0.7, "learning_rate": 0.00010972909942719802, "loss": 1.2275, "step": 5579 }, { "epoch": 0.7, "learning_rate": 0.00010964576202402979, "loss": 1.2187, "step": 5580 }, { "epoch": 0.7, "learning_rate": 0.00010956244738944773, "loss": 1.219, "step": 5581 }, { "epoch": 0.7, "learning_rate": 0.0001094791555369674, "loss": 1.1919, "step": 5582 }, { "epoch": 0.7, "learning_rate": 0.00010939588648010054, "loss": 1.2952, "step": 5583 }, { "epoch": 0.7, "learning_rate": 0.00010931264023235534, "loss": 1.2233, "step": 5584 }, { "epoch": 0.7, "learning_rate": 0.00010922941680723609, "loss": 1.2872, "step": 5585 }, { "epoch": 0.7, "learning_rate": 0.00010914621621824359, "loss": 1.2404, "step": 5586 }, { "epoch": 0.7, "learning_rate": 0.00010906303847887472, "loss": 1.1954, "step": 5587 }, { "epoch": 0.7, "learning_rate": 0.0001089798836026229, "loss": 1.2226, "step": 5588 }, { "epoch": 0.7, "learning_rate": 0.00010889675160297766, "loss": 1.2087, "step": 5589 }, { "epoch": 0.7, "learning_rate": 0.00010881364249342481, "loss": 1.2211, "step": 5590 }, { "epoch": 0.7, "learning_rate": 0.00010873055628744647, "loss": 1.2215, "step": 5591 }, { "epoch": 0.7, "learning_rate": 0.00010864749299852108, "loss": 1.2509, "step": 5592 }, { "epoch": 0.7, "learning_rate": 0.00010856445264012354, "loss": 1.2334, "step": 5593 }, { "epoch": 0.7, "learning_rate": 0.00010848143522572468, "loss": 1.2527, "step": 5594 }, { "epoch": 0.7, "learning_rate": 0.00010839844076879185, "loss": 1.2344, "step": 5595 }, { "epoch": 0.7, "learning_rate": 0.00010831546928278854, "loss": 1.2094, "step": 5596 }, { "epoch": 0.7, "learning_rate": 0.00010823252078117448, "loss": 1.2183, "step": 5597 }, { "epoch": 0.7, "learning_rate": 0.00010814959527740592, "loss": 1.2549, "step": 5598 }, { "epoch": 0.7, "learning_rate": 0.00010806669278493527, "loss": 1.2471, "step": 5599 }, { "epoch": 0.7, "learning_rate": 0.00010798381331721108, "loss": 1.3212, "step": 5600 }, { "epoch": 0.7, "eval_acceptance_rate": 0.6965970993041992, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2042654292296877, "eval_expected_tokens": 2.7553317865371096, "eval_loss": 1.357659935951233, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4388329466342774, "eval_runtime": 50.3138, "eval_samples_per_second": 9.679, "eval_steps_per_second": 0.318, "step": 5600 }, { "epoch": 0.7, "learning_rate": 0.00010790095688767826, "loss": 1.2666, "step": 5601 }, { "epoch": 0.7, "learning_rate": 0.00010781812350977785, "loss": 1.2386, "step": 5602 }, { "epoch": 0.7, "learning_rate": 0.00010773531319694746, "loss": 1.2288, "step": 5603 }, { "epoch": 0.7, "learning_rate": 0.00010765252596262059, "loss": 1.3163, "step": 5604 }, { "epoch": 0.7, "learning_rate": 0.00010756976182022737, "loss": 1.25, "step": 5605 }, { "epoch": 0.7, "learning_rate": 0.00010748702078319391, "loss": 1.3188, "step": 5606 }, { "epoch": 0.7, "learning_rate": 0.00010740430286494252, "loss": 1.19, "step": 5607 }, { "epoch": 0.7, "learning_rate": 0.0001073216080788921, "loss": 1.228, "step": 5608 }, { "epoch": 0.7, "learning_rate": 0.0001072389364384575, "loss": 1.1912, "step": 5609 }, { "epoch": 0.7, "learning_rate": 0.00010715628795704977, "loss": 1.233, "step": 5610 }, { "epoch": 0.7, "learning_rate": 0.00010707366264807658, "loss": 1.2595, "step": 5611 }, { "epoch": 0.7, "learning_rate": 0.00010699106052494137, "loss": 1.3521, "step": 5612 }, { "epoch": 0.7, "learning_rate": 0.00010690848160104424, "loss": 1.2214, "step": 5613 }, { "epoch": 0.7, "learning_rate": 0.00010682592588978123, "loss": 1.2864, "step": 5614 }, { "epoch": 0.7, "learning_rate": 0.00010674339340454476, "loss": 1.2406, "step": 5615 }, { "epoch": 0.7, "learning_rate": 0.0001066608841587233, "loss": 1.2354, "step": 5616 }, { "epoch": 0.7, "learning_rate": 0.0001065783981657018, "loss": 1.3086, "step": 5617 }, { "epoch": 0.7, "learning_rate": 0.00010649593543886141, "loss": 1.2402, "step": 5618 }, { "epoch": 0.7, "learning_rate": 0.00010641349599157935, "loss": 1.2538, "step": 5619 }, { "epoch": 0.7, "learning_rate": 0.00010633107983722912, "loss": 1.2307, "step": 5620 }, { "epoch": 0.7, "learning_rate": 0.00010624868698918044, "loss": 1.2864, "step": 5621 }, { "epoch": 0.7, "learning_rate": 0.00010616631746079921, "loss": 1.2614, "step": 5622 }, { "epoch": 0.7, "learning_rate": 0.00010608397126544769, "loss": 1.2015, "step": 5623 }, { "epoch": 0.7, "learning_rate": 0.00010600164841648435, "loss": 1.2616, "step": 5624 }, { "epoch": 0.7, "learning_rate": 0.00010591934892726371, "loss": 1.2279, "step": 5625 }, { "epoch": 0.7, "learning_rate": 0.00010583707281113655, "loss": 1.2582, "step": 5626 }, { "epoch": 0.7, "learning_rate": 0.00010575482008144984, "loss": 1.2121, "step": 5627 }, { "epoch": 0.7, "learning_rate": 0.00010567259075154695, "loss": 1.2475, "step": 5628 }, { "epoch": 0.7, "learning_rate": 0.00010559038483476716, "loss": 1.2172, "step": 5629 }, { "epoch": 0.7, "learning_rate": 0.00010550820234444627, "loss": 1.2478, "step": 5630 }, { "epoch": 0.7, "learning_rate": 0.00010542604329391604, "loss": 1.2679, "step": 5631 }, { "epoch": 0.7, "learning_rate": 0.00010534390769650437, "loss": 1.2938, "step": 5632 }, { "epoch": 0.7, "learning_rate": 0.00010526179556553573, "loss": 1.2457, "step": 5633 }, { "epoch": 0.7, "learning_rate": 0.00010517970691433035, "loss": 1.2334, "step": 5634 }, { "epoch": 0.7, "learning_rate": 0.00010509764175620485, "loss": 1.2586, "step": 5635 }, { "epoch": 0.7, "learning_rate": 0.00010501560010447205, "loss": 1.2819, "step": 5636 }, { "epoch": 0.7, "learning_rate": 0.00010493358197244104, "loss": 1.2854, "step": 5637 }, { "epoch": 0.7, "learning_rate": 0.00010485158737341693, "loss": 1.2304, "step": 5638 }, { "epoch": 0.7, "learning_rate": 0.00010476961632070106, "loss": 1.2245, "step": 5639 }, { "epoch": 0.7, "learning_rate": 0.00010468766882759092, "loss": 1.2064, "step": 5640 }, { "epoch": 0.71, "learning_rate": 0.00010460574490738017, "loss": 1.299, "step": 5641 }, { "epoch": 0.71, "learning_rate": 0.00010452384457335879, "loss": 1.3421, "step": 5642 }, { "epoch": 0.71, "learning_rate": 0.0001044419678388129, "loss": 1.2771, "step": 5643 }, { "epoch": 0.71, "learning_rate": 0.0001043601147170247, "loss": 1.2911, "step": 5644 }, { "epoch": 0.71, "learning_rate": 0.0001042782852212725, "loss": 1.2679, "step": 5645 }, { "epoch": 0.71, "learning_rate": 0.00010419647936483087, "loss": 1.2453, "step": 5646 }, { "epoch": 0.71, "learning_rate": 0.00010411469716097066, "loss": 1.2292, "step": 5647 }, { "epoch": 0.71, "learning_rate": 0.00010403293862295862, "loss": 1.2237, "step": 5648 }, { "epoch": 0.71, "learning_rate": 0.00010395120376405798, "loss": 1.272, "step": 5649 }, { "epoch": 0.71, "learning_rate": 0.00010386949259752784, "loss": 1.2908, "step": 5650 }, { "epoch": 0.71, "learning_rate": 0.00010378780513662354, "loss": 1.3649, "step": 5651 }, { "epoch": 0.71, "learning_rate": 0.00010370614139459671, "loss": 1.2505, "step": 5652 }, { "epoch": 0.71, "learning_rate": 0.000103624501384695, "loss": 1.3267, "step": 5653 }, { "epoch": 0.71, "learning_rate": 0.00010354288512016216, "loss": 1.2323, "step": 5654 }, { "epoch": 0.71, "learning_rate": 0.00010346129261423828, "loss": 1.2917, "step": 5655 }, { "epoch": 0.71, "learning_rate": 0.00010337972388015937, "loss": 1.2422, "step": 5656 }, { "epoch": 0.71, "learning_rate": 0.00010329817893115784, "loss": 1.2683, "step": 5657 }, { "epoch": 0.71, "learning_rate": 0.00010321665778046203, "loss": 1.2423, "step": 5658 }, { "epoch": 0.71, "learning_rate": 0.00010313516044129645, "loss": 1.2574, "step": 5659 }, { "epoch": 0.71, "learning_rate": 0.00010305368692688174, "loss": 1.2267, "step": 5660 }, { "epoch": 0.71, "learning_rate": 0.00010297223725043478, "loss": 1.2602, "step": 5661 }, { "epoch": 0.71, "learning_rate": 0.00010289081142516865, "loss": 1.2885, "step": 5662 }, { "epoch": 0.71, "learning_rate": 0.00010280940946429229, "loss": 1.3759, "step": 5663 }, { "epoch": 0.71, "learning_rate": 0.00010272803138101095, "loss": 1.2292, "step": 5664 }, { "epoch": 0.71, "learning_rate": 0.00010264667718852596, "loss": 1.2093, "step": 5665 }, { "epoch": 0.71, "learning_rate": 0.00010256534690003472, "loss": 1.2647, "step": 5666 }, { "epoch": 0.71, "learning_rate": 0.00010248404052873083, "loss": 1.2606, "step": 5667 }, { "epoch": 0.71, "learning_rate": 0.00010240275808780416, "loss": 1.2525, "step": 5668 }, { "epoch": 0.71, "learning_rate": 0.00010232149959044043, "loss": 1.2298, "step": 5669 }, { "epoch": 0.71, "learning_rate": 0.00010224026504982159, "loss": 1.2872, "step": 5670 }, { "epoch": 0.71, "learning_rate": 0.00010215905447912555, "loss": 1.2143, "step": 5671 }, { "epoch": 0.71, "learning_rate": 0.0001020778678915267, "loss": 1.2202, "step": 5672 }, { "epoch": 0.71, "learning_rate": 0.00010199670530019511, "loss": 1.2372, "step": 5673 }, { "epoch": 0.71, "learning_rate": 0.00010191556671829739, "loss": 1.2998, "step": 5674 }, { "epoch": 0.71, "learning_rate": 0.00010183445215899583, "loss": 1.2675, "step": 5675 }, { "epoch": 0.71, "learning_rate": 0.00010175336163544907, "loss": 1.2113, "step": 5676 }, { "epoch": 0.71, "learning_rate": 0.00010167229516081189, "loss": 1.263, "step": 5677 }, { "epoch": 0.71, "learning_rate": 0.000101591252748235, "loss": 1.2437, "step": 5678 }, { "epoch": 0.71, "learning_rate": 0.00010151023441086521, "loss": 1.1665, "step": 5679 }, { "epoch": 0.71, "learning_rate": 0.00010142924016184569, "loss": 1.2236, "step": 5680 }, { "epoch": 0.71, "learning_rate": 0.00010134827001431532, "loss": 1.2654, "step": 5681 }, { "epoch": 0.71, "learning_rate": 0.00010126732398140943, "loss": 1.2418, "step": 5682 }, { "epoch": 0.71, "learning_rate": 0.00010118640207625917, "loss": 1.2952, "step": 5683 }, { "epoch": 0.71, "learning_rate": 0.00010110550431199187, "loss": 1.3169, "step": 5684 }, { "epoch": 0.71, "learning_rate": 0.00010102463070173101, "loss": 1.2959, "step": 5685 }, { "epoch": 0.71, "learning_rate": 0.00010094378125859602, "loss": 1.2333, "step": 5686 }, { "epoch": 0.71, "learning_rate": 0.00010086295599570261, "loss": 1.2392, "step": 5687 }, { "epoch": 0.71, "learning_rate": 0.00010078215492616232, "loss": 1.2259, "step": 5688 }, { "epoch": 0.71, "learning_rate": 0.00010070137806308297, "loss": 1.3016, "step": 5689 }, { "epoch": 0.71, "learning_rate": 0.00010062062541956818, "loss": 1.2282, "step": 5690 }, { "epoch": 0.71, "learning_rate": 0.00010053989700871799, "loss": 1.1827, "step": 5691 }, { "epoch": 0.71, "learning_rate": 0.00010045919284362842, "loss": 1.2539, "step": 5692 }, { "epoch": 0.71, "learning_rate": 0.0001003785129373914, "loss": 1.2783, "step": 5693 }, { "epoch": 0.71, "learning_rate": 0.00010029785730309496, "loss": 1.2485, "step": 5694 }, { "epoch": 0.71, "learning_rate": 0.00010021722595382324, "loss": 1.1993, "step": 5695 }, { "epoch": 0.71, "learning_rate": 0.00010013661890265655, "loss": 1.2238, "step": 5696 }, { "epoch": 0.71, "learning_rate": 0.00010005603616267098, "loss": 1.2413, "step": 5697 }, { "epoch": 0.71, "learning_rate": 9.997547774693905e-05, "loss": 1.3096, "step": 5698 }, { "epoch": 0.71, "learning_rate": 9.989494366852902e-05, "loss": 1.2072, "step": 5699 }, { "epoch": 0.71, "learning_rate": 9.981443394050524e-05, "loss": 1.218, "step": 5700 }, { "epoch": 0.71, "eval_acceptance_rate": 0.6972362399101257, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2069267401605286, "eval_expected_tokens": 2.758658425200661, "eval_loss": 1.3570197820663452, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.43966460630016524, "eval_runtime": 50.2555, "eval_samples_per_second": 9.69, "eval_steps_per_second": 0.318, "step": 5700 }, { "epoch": 0.71, "learning_rate": 9.973394857592838e-05, "loss": 1.2145, "step": 5701 }, { "epoch": 0.71, "learning_rate": 9.965348758785483e-05, "loss": 1.1761, "step": 5702 }, { "epoch": 0.71, "learning_rate": 9.957305098933708e-05, "loss": 1.2454, "step": 5703 }, { "epoch": 0.71, "learning_rate": 9.949263879342393e-05, "loss": 1.2479, "step": 5704 }, { "epoch": 0.71, "learning_rate": 9.941225101315985e-05, "loss": 1.2337, "step": 5705 }, { "epoch": 0.71, "learning_rate": 9.933188766158569e-05, "loss": 1.2497, "step": 5706 }, { "epoch": 0.71, "learning_rate": 9.925154875173809e-05, "loss": 1.2598, "step": 5707 }, { "epoch": 0.71, "learning_rate": 9.917123429664982e-05, "loss": 1.2767, "step": 5708 }, { "epoch": 0.71, "learning_rate": 9.909094430934962e-05, "loss": 1.2222, "step": 5709 }, { "epoch": 0.71, "learning_rate": 9.901067880286235e-05, "loss": 1.2521, "step": 5710 }, { "epoch": 0.71, "learning_rate": 9.893043779020897e-05, "loss": 1.2648, "step": 5711 }, { "epoch": 0.71, "learning_rate": 9.885022128440629e-05, "loss": 1.2544, "step": 5712 }, { "epoch": 0.71, "learning_rate": 9.877002929846721e-05, "loss": 1.2959, "step": 5713 }, { "epoch": 0.71, "learning_rate": 9.868986184540066e-05, "loss": 1.2408, "step": 5714 }, { "epoch": 0.71, "learning_rate": 9.860971893821147e-05, "loss": 1.2431, "step": 5715 }, { "epoch": 0.71, "learning_rate": 9.852960058990073e-05, "loss": 1.2449, "step": 5716 }, { "epoch": 0.71, "learning_rate": 9.844950681346548e-05, "loss": 1.211, "step": 5717 }, { "epoch": 0.71, "learning_rate": 9.836943762189865e-05, "loss": 1.2167, "step": 5718 }, { "epoch": 0.71, "learning_rate": 9.828939302818927e-05, "loss": 1.1972, "step": 5719 }, { "epoch": 0.71, "learning_rate": 9.820937304532221e-05, "loss": 1.2018, "step": 5720 }, { "epoch": 0.72, "learning_rate": 9.812937768627872e-05, "loss": 1.2427, "step": 5721 }, { "epoch": 0.72, "learning_rate": 9.804940696403561e-05, "loss": 1.2293, "step": 5722 }, { "epoch": 0.72, "learning_rate": 9.796946089156613e-05, "loss": 1.2111, "step": 5723 }, { "epoch": 0.72, "learning_rate": 9.788953948183923e-05, "loss": 1.2172, "step": 5724 }, { "epoch": 0.72, "learning_rate": 9.780964274781984e-05, "loss": 1.3151, "step": 5725 }, { "epoch": 0.72, "learning_rate": 9.772977070246913e-05, "loss": 1.2744, "step": 5726 }, { "epoch": 0.72, "learning_rate": 9.764992335874412e-05, "loss": 1.2234, "step": 5727 }, { "epoch": 0.72, "learning_rate": 9.757010072959768e-05, "loss": 1.2494, "step": 5728 }, { "epoch": 0.72, "learning_rate": 9.749030282797902e-05, "loss": 1.2364, "step": 5729 }, { "epoch": 0.72, "learning_rate": 9.741052966683295e-05, "loss": 1.2199, "step": 5730 }, { "epoch": 0.72, "learning_rate": 9.733078125910063e-05, "loss": 1.2393, "step": 5731 }, { "epoch": 0.72, "learning_rate": 9.725105761771897e-05, "loss": 1.2495, "step": 5732 }, { "epoch": 0.72, "learning_rate": 9.717135875562085e-05, "loss": 1.1767, "step": 5733 }, { "epoch": 0.72, "learning_rate": 9.709168468573523e-05, "loss": 1.2524, "step": 5734 }, { "epoch": 0.72, "learning_rate": 9.7012035420987e-05, "loss": 1.2041, "step": 5735 }, { "epoch": 0.72, "learning_rate": 9.69324109742972e-05, "loss": 1.1967, "step": 5736 }, { "epoch": 0.72, "learning_rate": 9.685281135858256e-05, "loss": 1.2853, "step": 5737 }, { "epoch": 0.72, "learning_rate": 9.677323658675594e-05, "loss": 1.2296, "step": 5738 }, { "epoch": 0.72, "learning_rate": 9.669368667172604e-05, "loss": 1.1416, "step": 5739 }, { "epoch": 0.72, "learning_rate": 9.66141616263978e-05, "loss": 1.1852, "step": 5740 }, { "epoch": 0.72, "learning_rate": 9.653466146367182e-05, "loss": 1.1931, "step": 5741 }, { "epoch": 0.72, "learning_rate": 9.645518619644491e-05, "loss": 1.2777, "step": 5742 }, { "epoch": 0.72, "learning_rate": 9.63757358376097e-05, "loss": 1.2453, "step": 5743 }, { "epoch": 0.72, "learning_rate": 9.629631040005469e-05, "loss": 1.2618, "step": 5744 }, { "epoch": 0.72, "learning_rate": 9.621690989666465e-05, "loss": 1.2261, "step": 5745 }, { "epoch": 0.72, "learning_rate": 9.613753434031999e-05, "loss": 1.2477, "step": 5746 }, { "epoch": 0.72, "learning_rate": 9.605818374389715e-05, "loss": 1.2341, "step": 5747 }, { "epoch": 0.72, "learning_rate": 9.597885812026871e-05, "loss": 1.232, "step": 5748 }, { "epoch": 0.72, "learning_rate": 9.589955748230291e-05, "loss": 1.2629, "step": 5749 }, { "epoch": 0.72, "learning_rate": 9.582028184286423e-05, "loss": 1.1998, "step": 5750 }, { "epoch": 0.72, "learning_rate": 9.574103121481287e-05, "loss": 1.2177, "step": 5751 }, { "epoch": 0.72, "learning_rate": 9.566180561100505e-05, "loss": 1.2484, "step": 5752 }, { "epoch": 0.72, "learning_rate": 9.55826050442928e-05, "loss": 1.2203, "step": 5753 }, { "epoch": 0.72, "learning_rate": 9.550342952752438e-05, "loss": 1.2467, "step": 5754 }, { "epoch": 0.72, "learning_rate": 9.542427907354387e-05, "loss": 1.206, "step": 5755 }, { "epoch": 0.72, "learning_rate": 9.534515369519111e-05, "loss": 1.2854, "step": 5756 }, { "epoch": 0.72, "learning_rate": 9.526605340530209e-05, "loss": 1.2404, "step": 5757 }, { "epoch": 0.72, "learning_rate": 9.518697821670857e-05, "loss": 1.2578, "step": 5758 }, { "epoch": 0.72, "learning_rate": 9.510792814223826e-05, "loss": 1.2124, "step": 5759 }, { "epoch": 0.72, "learning_rate": 9.502890319471491e-05, "loss": 1.2734, "step": 5760 }, { "epoch": 0.72, "learning_rate": 9.494990338695822e-05, "loss": 1.2228, "step": 5761 }, { "epoch": 0.72, "learning_rate": 9.487092873178364e-05, "loss": 1.2615, "step": 5762 }, { "epoch": 0.72, "learning_rate": 9.479197924200261e-05, "loss": 1.1814, "step": 5763 }, { "epoch": 0.72, "learning_rate": 9.471305493042242e-05, "loss": 1.2085, "step": 5764 }, { "epoch": 0.72, "learning_rate": 9.463415580984653e-05, "loss": 1.218, "step": 5765 }, { "epoch": 0.72, "learning_rate": 9.455528189307391e-05, "loss": 1.2809, "step": 5766 }, { "epoch": 0.72, "learning_rate": 9.447643319289992e-05, "loss": 1.2612, "step": 5767 }, { "epoch": 0.72, "learning_rate": 9.439760972211545e-05, "loss": 1.2699, "step": 5768 }, { "epoch": 0.72, "learning_rate": 9.431881149350735e-05, "loss": 1.2771, "step": 5769 }, { "epoch": 0.72, "learning_rate": 9.42400385198586e-05, "loss": 1.2989, "step": 5770 }, { "epoch": 0.72, "learning_rate": 9.416129081394787e-05, "loss": 1.2731, "step": 5771 }, { "epoch": 0.72, "learning_rate": 9.408256838854967e-05, "loss": 1.2628, "step": 5772 }, { "epoch": 0.72, "learning_rate": 9.400387125643478e-05, "loss": 1.2589, "step": 5773 }, { "epoch": 0.72, "learning_rate": 9.392519943036937e-05, "loss": 1.2094, "step": 5774 }, { "epoch": 0.72, "learning_rate": 9.384655292311601e-05, "loss": 1.1375, "step": 5775 }, { "epoch": 0.72, "learning_rate": 9.376793174743276e-05, "loss": 1.2263, "step": 5776 }, { "epoch": 0.72, "learning_rate": 9.368933591607378e-05, "loss": 1.184, "step": 5777 }, { "epoch": 0.72, "learning_rate": 9.361076544178895e-05, "loss": 1.2468, "step": 5778 }, { "epoch": 0.72, "learning_rate": 9.353222033732428e-05, "loss": 1.2003, "step": 5779 }, { "epoch": 0.72, "learning_rate": 9.345370061542158e-05, "loss": 1.229, "step": 5780 }, { "epoch": 0.72, "learning_rate": 9.337520628881843e-05, "loss": 1.2447, "step": 5781 }, { "epoch": 0.72, "learning_rate": 9.329673737024835e-05, "loss": 1.2458, "step": 5782 }, { "epoch": 0.72, "learning_rate": 9.321829387244065e-05, "loss": 1.288, "step": 5783 }, { "epoch": 0.72, "learning_rate": 9.313987580812083e-05, "loss": 1.2673, "step": 5784 }, { "epoch": 0.72, "learning_rate": 9.306148319000988e-05, "loss": 1.2265, "step": 5785 }, { "epoch": 0.72, "learning_rate": 9.298311603082494e-05, "loss": 1.2071, "step": 5786 }, { "epoch": 0.72, "learning_rate": 9.290477434327885e-05, "loss": 1.2352, "step": 5787 }, { "epoch": 0.72, "learning_rate": 9.282645814008031e-05, "loss": 1.1921, "step": 5788 }, { "epoch": 0.72, "learning_rate": 9.274816743393411e-05, "loss": 1.3091, "step": 5789 }, { "epoch": 0.72, "learning_rate": 9.266990223754068e-05, "loss": 1.2586, "step": 5790 }, { "epoch": 0.72, "learning_rate": 9.259166256359627e-05, "loss": 1.1962, "step": 5791 }, { "epoch": 0.72, "learning_rate": 9.251344842479331e-05, "loss": 1.2128, "step": 5792 }, { "epoch": 0.72, "learning_rate": 9.243525983381964e-05, "loss": 1.2168, "step": 5793 }, { "epoch": 0.72, "learning_rate": 9.235709680335943e-05, "loss": 1.2377, "step": 5794 }, { "epoch": 0.72, "learning_rate": 9.227895934609234e-05, "loss": 1.2762, "step": 5795 }, { "epoch": 0.72, "learning_rate": 9.220084747469401e-05, "loss": 1.2334, "step": 5796 }, { "epoch": 0.72, "learning_rate": 9.212276120183588e-05, "loss": 1.297, "step": 5797 }, { "epoch": 0.72, "learning_rate": 9.204470054018532e-05, "loss": 1.2477, "step": 5798 }, { "epoch": 0.72, "learning_rate": 9.196666550240562e-05, "loss": 1.3186, "step": 5799 }, { "epoch": 0.72, "learning_rate": 9.18886561011557e-05, "loss": 1.1925, "step": 5800 }, { "epoch": 0.72, "eval_acceptance_rate": 0.6971514225006104, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2065733439339135, "eval_expected_tokens": 2.7582166799173917, "eval_loss": 1.3542160987854004, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4395541699793479, "eval_runtime": 50.4048, "eval_samples_per_second": 9.662, "eval_steps_per_second": 0.317, "step": 5800 }, { "epoch": 0.73, "learning_rate": 9.181067234909047e-05, "loss": 1.2082, "step": 5801 }, { "epoch": 0.73, "learning_rate": 9.173271425886056e-05, "loss": 1.2439, "step": 5802 }, { "epoch": 0.73, "learning_rate": 9.165478184311248e-05, "loss": 1.189, "step": 5803 }, { "epoch": 0.73, "learning_rate": 9.157687511448868e-05, "loss": 1.2598, "step": 5804 }, { "epoch": 0.73, "learning_rate": 9.149899408562745e-05, "loss": 1.2374, "step": 5805 }, { "epoch": 0.73, "learning_rate": 9.142113876916272e-05, "loss": 1.2975, "step": 5806 }, { "epoch": 0.73, "learning_rate": 9.134330917772441e-05, "loss": 1.1802, "step": 5807 }, { "epoch": 0.73, "learning_rate": 9.12655053239381e-05, "loss": 1.2103, "step": 5808 }, { "epoch": 0.73, "learning_rate": 9.118772722042546e-05, "loss": 1.2323, "step": 5809 }, { "epoch": 0.73, "learning_rate": 9.110997487980369e-05, "loss": 1.1844, "step": 5810 }, { "epoch": 0.73, "learning_rate": 9.103224831468609e-05, "loss": 1.1858, "step": 5811 }, { "epoch": 0.73, "learning_rate": 9.095454753768161e-05, "loss": 1.2728, "step": 5812 }, { "epoch": 0.73, "learning_rate": 9.087687256139493e-05, "loss": 1.2913, "step": 5813 }, { "epoch": 0.73, "learning_rate": 9.079922339842683e-05, "loss": 1.2513, "step": 5814 }, { "epoch": 0.73, "learning_rate": 9.072160006137367e-05, "loss": 1.2469, "step": 5815 }, { "epoch": 0.73, "learning_rate": 9.064400256282756e-05, "loss": 1.1882, "step": 5816 }, { "epoch": 0.73, "learning_rate": 9.056643091537679e-05, "loss": 1.2599, "step": 5817 }, { "epoch": 0.73, "learning_rate": 9.048888513160494e-05, "loss": 1.1872, "step": 5818 }, { "epoch": 0.73, "learning_rate": 9.041136522409191e-05, "loss": 1.2172, "step": 5819 }, { "epoch": 0.73, "learning_rate": 9.033387120541306e-05, "loss": 1.25, "step": 5820 }, { "epoch": 0.73, "learning_rate": 9.025640308813962e-05, "loss": 1.2638, "step": 5821 }, { "epoch": 0.73, "learning_rate": 9.017896088483859e-05, "loss": 1.3234, "step": 5822 }, { "epoch": 0.73, "learning_rate": 9.010154460807288e-05, "loss": 1.2718, "step": 5823 }, { "epoch": 0.73, "learning_rate": 9.002415427040125e-05, "loss": 1.1929, "step": 5824 }, { "epoch": 0.73, "learning_rate": 8.994678988437802e-05, "loss": 1.2625, "step": 5825 }, { "epoch": 0.73, "learning_rate": 8.986945146255346e-05, "loss": 1.204, "step": 5826 }, { "epoch": 0.73, "learning_rate": 8.979213901747354e-05, "loss": 1.2196, "step": 5827 }, { "epoch": 0.73, "learning_rate": 8.971485256167999e-05, "loss": 1.2744, "step": 5828 }, { "epoch": 0.73, "learning_rate": 8.963759210771053e-05, "loss": 1.2832, "step": 5829 }, { "epoch": 0.73, "learning_rate": 8.956035766809852e-05, "loss": 1.1948, "step": 5830 }, { "epoch": 0.73, "learning_rate": 8.948314925537313e-05, "loss": 1.2515, "step": 5831 }, { "epoch": 0.73, "learning_rate": 8.940596688205909e-05, "loss": 1.2133, "step": 5832 }, { "epoch": 0.73, "learning_rate": 8.932881056067737e-05, "loss": 1.2207, "step": 5833 }, { "epoch": 0.73, "learning_rate": 8.92516803037443e-05, "loss": 1.1901, "step": 5834 }, { "epoch": 0.73, "learning_rate": 8.917457612377209e-05, "loss": 1.2082, "step": 5835 }, { "epoch": 0.73, "learning_rate": 8.909749803326886e-05, "loss": 1.2681, "step": 5836 }, { "epoch": 0.73, "learning_rate": 8.902044604473828e-05, "loss": 1.1828, "step": 5837 }, { "epoch": 0.73, "learning_rate": 8.894342017068005e-05, "loss": 1.2812, "step": 5838 }, { "epoch": 0.73, "learning_rate": 8.886642042358941e-05, "loss": 1.262, "step": 5839 }, { "epoch": 0.73, "learning_rate": 8.878944681595741e-05, "loss": 1.2849, "step": 5840 }, { "epoch": 0.73, "learning_rate": 8.871249936027081e-05, "loss": 1.2827, "step": 5841 }, { "epoch": 0.73, "learning_rate": 8.863557806901232e-05, "loss": 1.2878, "step": 5842 }, { "epoch": 0.73, "learning_rate": 8.855868295466033e-05, "loss": 1.3095, "step": 5843 }, { "epoch": 0.73, "learning_rate": 8.848181402968888e-05, "loss": 1.1815, "step": 5844 }, { "epoch": 0.73, "learning_rate": 8.840497130656783e-05, "loss": 1.2245, "step": 5845 }, { "epoch": 0.73, "learning_rate": 8.832815479776274e-05, "loss": 1.228, "step": 5846 }, { "epoch": 0.73, "learning_rate": 8.825136451573493e-05, "loss": 1.3624, "step": 5847 }, { "epoch": 0.73, "learning_rate": 8.817460047294154e-05, "loss": 1.2606, "step": 5848 }, { "epoch": 0.73, "learning_rate": 8.80978626818355e-05, "loss": 1.2153, "step": 5849 }, { "epoch": 0.73, "learning_rate": 8.802115115486533e-05, "loss": 1.2924, "step": 5850 }, { "epoch": 0.73, "learning_rate": 8.794446590447536e-05, "loss": 1.2435, "step": 5851 }, { "epoch": 0.73, "learning_rate": 8.786780694310551e-05, "loss": 1.2598, "step": 5852 }, { "epoch": 0.73, "learning_rate": 8.779117428319175e-05, "loss": 1.2644, "step": 5853 }, { "epoch": 0.73, "learning_rate": 8.771456793716548e-05, "loss": 1.2403, "step": 5854 }, { "epoch": 0.73, "learning_rate": 8.763798791745412e-05, "loss": 1.2576, "step": 5855 }, { "epoch": 0.73, "learning_rate": 8.756143423648052e-05, "loss": 1.2722, "step": 5856 }, { "epoch": 0.73, "learning_rate": 8.748490690666337e-05, "loss": 1.218, "step": 5857 }, { "epoch": 0.73, "learning_rate": 8.740840594041721e-05, "loss": 1.2438, "step": 5858 }, { "epoch": 0.73, "learning_rate": 8.733193135015219e-05, "loss": 1.2104, "step": 5859 }, { "epoch": 0.73, "learning_rate": 8.725548314827405e-05, "loss": 1.2281, "step": 5860 }, { "epoch": 0.73, "learning_rate": 8.717906134718462e-05, "loss": 1.2537, "step": 5861 }, { "epoch": 0.73, "learning_rate": 8.710266595928101e-05, "loss": 1.2941, "step": 5862 }, { "epoch": 0.73, "learning_rate": 8.702629699695644e-05, "loss": 1.3141, "step": 5863 }, { "epoch": 0.73, "learning_rate": 8.694995447259955e-05, "loss": 1.3072, "step": 5864 }, { "epoch": 0.73, "learning_rate": 8.687363839859483e-05, "loss": 1.2061, "step": 5865 }, { "epoch": 0.73, "learning_rate": 8.679734878732235e-05, "loss": 1.2252, "step": 5866 }, { "epoch": 0.73, "learning_rate": 8.672108565115805e-05, "loss": 1.3151, "step": 5867 }, { "epoch": 0.73, "learning_rate": 8.664484900247364e-05, "loss": 1.2182, "step": 5868 }, { "epoch": 0.73, "learning_rate": 8.656863885363628e-05, "loss": 1.2485, "step": 5869 }, { "epoch": 0.73, "learning_rate": 8.649245521700897e-05, "loss": 1.2224, "step": 5870 }, { "epoch": 0.73, "learning_rate": 8.641629810495041e-05, "loss": 1.2421, "step": 5871 }, { "epoch": 0.73, "learning_rate": 8.634016752981485e-05, "loss": 1.2471, "step": 5872 }, { "epoch": 0.73, "learning_rate": 8.626406350395252e-05, "loss": 1.3208, "step": 5873 }, { "epoch": 0.73, "learning_rate": 8.61879860397092e-05, "loss": 1.3085, "step": 5874 }, { "epoch": 0.73, "learning_rate": 8.611193514942634e-05, "loss": 1.1941, "step": 5875 }, { "epoch": 0.73, "learning_rate": 8.603591084544102e-05, "loss": 1.3094, "step": 5876 }, { "epoch": 0.73, "learning_rate": 8.595991314008606e-05, "loss": 1.2884, "step": 5877 }, { "epoch": 0.73, "learning_rate": 8.588394204569009e-05, "loss": 1.2729, "step": 5878 }, { "epoch": 0.73, "learning_rate": 8.580799757457716e-05, "loss": 1.2877, "step": 5879 }, { "epoch": 0.73, "learning_rate": 8.573207973906736e-05, "loss": 1.2837, "step": 5880 }, { "epoch": 0.74, "learning_rate": 8.565618855147602e-05, "loss": 1.2191, "step": 5881 }, { "epoch": 0.74, "learning_rate": 8.558032402411465e-05, "loss": 1.2232, "step": 5882 }, { "epoch": 0.74, "learning_rate": 8.550448616928999e-05, "loss": 1.1791, "step": 5883 }, { "epoch": 0.74, "learning_rate": 8.542867499930468e-05, "loss": 1.2079, "step": 5884 }, { "epoch": 0.74, "learning_rate": 8.535289052645689e-05, "loss": 1.3094, "step": 5885 }, { "epoch": 0.74, "learning_rate": 8.527713276304064e-05, "loss": 1.2433, "step": 5886 }, { "epoch": 0.74, "learning_rate": 8.520140172134557e-05, "loss": 1.1663, "step": 5887 }, { "epoch": 0.74, "learning_rate": 8.512569741365692e-05, "loss": 1.2157, "step": 5888 }, { "epoch": 0.74, "learning_rate": 8.505001985225558e-05, "loss": 1.2711, "step": 5889 }, { "epoch": 0.74, "learning_rate": 8.497436904941813e-05, "loss": 1.2263, "step": 5890 }, { "epoch": 0.74, "learning_rate": 8.489874501741676e-05, "loss": 1.2614, "step": 5891 }, { "epoch": 0.74, "learning_rate": 8.482314776851946e-05, "loss": 1.2068, "step": 5892 }, { "epoch": 0.74, "learning_rate": 8.474757731498986e-05, "loss": 1.2429, "step": 5893 }, { "epoch": 0.74, "learning_rate": 8.467203366908707e-05, "loss": 1.2711, "step": 5894 }, { "epoch": 0.74, "learning_rate": 8.459651684306596e-05, "loss": 1.3036, "step": 5895 }, { "epoch": 0.74, "learning_rate": 8.4521026849177e-05, "loss": 1.2123, "step": 5896 }, { "epoch": 0.74, "learning_rate": 8.444556369966646e-05, "loss": 1.2409, "step": 5897 }, { "epoch": 0.74, "learning_rate": 8.437012740677603e-05, "loss": 1.2519, "step": 5898 }, { "epoch": 0.74, "learning_rate": 8.429471798274329e-05, "loss": 1.2606, "step": 5899 }, { "epoch": 0.74, "learning_rate": 8.421933543980126e-05, "loss": 1.1997, "step": 5900 }, { "epoch": 0.74, "eval_acceptance_rate": 0.6975709199905396, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2083218774053, "eval_expected_tokens": 2.7604023467566248, "eval_loss": 1.3529020547866821, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44010058668915625, "eval_runtime": 50.2156, "eval_samples_per_second": 9.698, "eval_steps_per_second": 0.319, "step": 5900 }, { "epoch": 0.74, "learning_rate": 8.414397979017857e-05, "loss": 1.2157, "step": 5901 }, { "epoch": 0.74, "learning_rate": 8.406865104609976e-05, "loss": 1.3042, "step": 5902 }, { "epoch": 0.74, "learning_rate": 8.399334921978477e-05, "loss": 1.1917, "step": 5903 }, { "epoch": 0.74, "learning_rate": 8.391807432344913e-05, "loss": 1.2145, "step": 5904 }, { "epoch": 0.74, "learning_rate": 8.384282636930426e-05, "loss": 1.2115, "step": 5905 }, { "epoch": 0.74, "learning_rate": 8.376760536955693e-05, "loss": 1.2195, "step": 5906 }, { "epoch": 0.74, "learning_rate": 8.369241133640981e-05, "loss": 1.2509, "step": 5907 }, { "epoch": 0.74, "learning_rate": 8.361724428206094e-05, "loss": 1.2406, "step": 5908 }, { "epoch": 0.74, "learning_rate": 8.354210421870412e-05, "loss": 1.2629, "step": 5909 }, { "epoch": 0.74, "learning_rate": 8.346699115852865e-05, "loss": 1.2628, "step": 5910 }, { "epoch": 0.74, "learning_rate": 8.339190511371964e-05, "loss": 1.2916, "step": 5911 }, { "epoch": 0.74, "learning_rate": 8.331684609645779e-05, "loss": 1.2004, "step": 5912 }, { "epoch": 0.74, "learning_rate": 8.324181411891926e-05, "loss": 1.2364, "step": 5913 }, { "epoch": 0.74, "learning_rate": 8.316680919327593e-05, "loss": 1.252, "step": 5914 }, { "epoch": 0.74, "learning_rate": 8.309183133169524e-05, "loss": 1.2627, "step": 5915 }, { "epoch": 0.74, "learning_rate": 8.301688054634021e-05, "loss": 1.3308, "step": 5916 }, { "epoch": 0.74, "learning_rate": 8.294195684936959e-05, "loss": 1.2171, "step": 5917 }, { "epoch": 0.74, "learning_rate": 8.286706025293778e-05, "loss": 1.2809, "step": 5918 }, { "epoch": 0.74, "learning_rate": 8.279219076919459e-05, "loss": 1.167, "step": 5919 }, { "epoch": 0.74, "learning_rate": 8.271734841028553e-05, "loss": 1.2766, "step": 5920 }, { "epoch": 0.74, "learning_rate": 8.264253318835158e-05, "loss": 1.1853, "step": 5921 }, { "epoch": 0.74, "learning_rate": 8.256774511552962e-05, "loss": 1.2574, "step": 5922 }, { "epoch": 0.74, "learning_rate": 8.24929842039518e-05, "loss": 1.2382, "step": 5923 }, { "epoch": 0.74, "learning_rate": 8.241825046574613e-05, "loss": 1.2183, "step": 5924 }, { "epoch": 0.74, "learning_rate": 8.234354391303605e-05, "loss": 1.2703, "step": 5925 }, { "epoch": 0.74, "learning_rate": 8.226886455794053e-05, "loss": 1.2642, "step": 5926 }, { "epoch": 0.74, "learning_rate": 8.219421241257435e-05, "loss": 1.2641, "step": 5927 }, { "epoch": 0.74, "learning_rate": 8.211958748904766e-05, "loss": 1.2124, "step": 5928 }, { "epoch": 0.74, "learning_rate": 8.204498979946639e-05, "loss": 1.2383, "step": 5929 }, { "epoch": 0.74, "learning_rate": 8.19704193559318e-05, "loss": 1.2644, "step": 5930 }, { "epoch": 0.74, "learning_rate": 8.189587617054106e-05, "loss": 1.2328, "step": 5931 }, { "epoch": 0.74, "learning_rate": 8.182136025538667e-05, "loss": 1.2452, "step": 5932 }, { "epoch": 0.74, "learning_rate": 8.174687162255672e-05, "loss": 1.1843, "step": 5933 }, { "epoch": 0.74, "learning_rate": 8.167241028413489e-05, "loss": 1.2551, "step": 5934 }, { "epoch": 0.74, "learning_rate": 8.159797625220058e-05, "loss": 1.1935, "step": 5935 }, { "epoch": 0.74, "learning_rate": 8.152356953882856e-05, "loss": 1.1916, "step": 5936 }, { "epoch": 0.74, "learning_rate": 8.144919015608935e-05, "loss": 1.1352, "step": 5937 }, { "epoch": 0.74, "learning_rate": 8.13748381160489e-05, "loss": 1.1618, "step": 5938 }, { "epoch": 0.74, "learning_rate": 8.130051343076878e-05, "loss": 1.1909, "step": 5939 }, { "epoch": 0.74, "learning_rate": 8.122621611230599e-05, "loss": 1.1945, "step": 5940 }, { "epoch": 0.74, "learning_rate": 8.115194617271332e-05, "loss": 1.2804, "step": 5941 }, { "epoch": 0.74, "learning_rate": 8.107770362403908e-05, "loss": 1.2974, "step": 5942 }, { "epoch": 0.74, "learning_rate": 8.1003488478327e-05, "loss": 1.2417, "step": 5943 }, { "epoch": 0.74, "learning_rate": 8.092930074761642e-05, "loss": 1.1867, "step": 5944 }, { "epoch": 0.74, "learning_rate": 8.085514044394216e-05, "loss": 1.1906, "step": 5945 }, { "epoch": 0.74, "learning_rate": 8.078100757933485e-05, "loss": 1.2095, "step": 5946 }, { "epoch": 0.74, "learning_rate": 8.07069021658203e-05, "loss": 1.2411, "step": 5947 }, { "epoch": 0.74, "learning_rate": 8.063282421542026e-05, "loss": 1.2774, "step": 5948 }, { "epoch": 0.74, "learning_rate": 8.055877374015175e-05, "loss": 1.2234, "step": 5949 }, { "epoch": 0.74, "learning_rate": 8.048475075202727e-05, "loss": 1.2457, "step": 5950 }, { "epoch": 0.74, "learning_rate": 8.041075526305519e-05, "loss": 1.1641, "step": 5951 }, { "epoch": 0.74, "learning_rate": 8.033678728523918e-05, "loss": 1.2393, "step": 5952 }, { "epoch": 0.74, "learning_rate": 8.026284683057838e-05, "loss": 1.2824, "step": 5953 }, { "epoch": 0.74, "learning_rate": 8.018893391106774e-05, "loss": 1.2353, "step": 5954 }, { "epoch": 0.74, "learning_rate": 8.011504853869742e-05, "loss": 1.179, "step": 5955 }, { "epoch": 0.74, "learning_rate": 8.004119072545346e-05, "loss": 1.2303, "step": 5956 }, { "epoch": 0.74, "learning_rate": 7.996736048331715e-05, "loss": 1.2198, "step": 5957 }, { "epoch": 0.74, "learning_rate": 7.989355782426538e-05, "loss": 1.2391, "step": 5958 }, { "epoch": 0.74, "learning_rate": 7.981978276027054e-05, "loss": 1.1921, "step": 5959 }, { "epoch": 0.74, "learning_rate": 7.974603530330068e-05, "loss": 1.2271, "step": 5960 }, { "epoch": 0.75, "learning_rate": 7.96723154653193e-05, "loss": 1.2286, "step": 5961 }, { "epoch": 0.75, "learning_rate": 7.959862325828537e-05, "loss": 1.273, "step": 5962 }, { "epoch": 0.75, "learning_rate": 7.952495869415341e-05, "loss": 1.202, "step": 5963 }, { "epoch": 0.75, "learning_rate": 7.945132178487344e-05, "loss": 1.2155, "step": 5964 }, { "epoch": 0.75, "learning_rate": 7.937771254239093e-05, "loss": 1.2305, "step": 5965 }, { "epoch": 0.75, "learning_rate": 7.9304130978647e-05, "loss": 1.2868, "step": 5966 }, { "epoch": 0.75, "learning_rate": 7.923057710557834e-05, "loss": 1.2763, "step": 5967 }, { "epoch": 0.75, "learning_rate": 7.915705093511694e-05, "loss": 1.2503, "step": 5968 }, { "epoch": 0.75, "learning_rate": 7.908355247919036e-05, "loss": 1.2684, "step": 5969 }, { "epoch": 0.75, "learning_rate": 7.90100817497216e-05, "loss": 1.242, "step": 5970 }, { "epoch": 0.75, "learning_rate": 7.893663875862944e-05, "loss": 1.2218, "step": 5971 }, { "epoch": 0.75, "learning_rate": 7.886322351782782e-05, "loss": 1.2494, "step": 5972 }, { "epoch": 0.75, "learning_rate": 7.878983603922646e-05, "loss": 1.245, "step": 5973 }, { "epoch": 0.75, "learning_rate": 7.871647633473036e-05, "loss": 1.2034, "step": 5974 }, { "epoch": 0.75, "learning_rate": 7.864314441624005e-05, "loss": 1.2729, "step": 5975 }, { "epoch": 0.75, "learning_rate": 7.856984029565173e-05, "loss": 1.2642, "step": 5976 }, { "epoch": 0.75, "learning_rate": 7.84965639848569e-05, "loss": 1.2085, "step": 5977 }, { "epoch": 0.75, "learning_rate": 7.842331549574256e-05, "loss": 1.207, "step": 5978 }, { "epoch": 0.75, "learning_rate": 7.835009484019129e-05, "loss": 1.23, "step": 5979 }, { "epoch": 0.75, "learning_rate": 7.827690203008117e-05, "loss": 1.249, "step": 5980 }, { "epoch": 0.75, "learning_rate": 7.820373707728568e-05, "loss": 1.248, "step": 5981 }, { "epoch": 0.75, "learning_rate": 7.813059999367378e-05, "loss": 1.2288, "step": 5982 }, { "epoch": 0.75, "learning_rate": 7.80574907911099e-05, "loss": 1.2371, "step": 5983 }, { "epoch": 0.75, "learning_rate": 7.7984409481454e-05, "loss": 1.2716, "step": 5984 }, { "epoch": 0.75, "learning_rate": 7.791135607656147e-05, "loss": 1.2615, "step": 5985 }, { "epoch": 0.75, "learning_rate": 7.783833058828335e-05, "loss": 1.2293, "step": 5986 }, { "epoch": 0.75, "learning_rate": 7.776533302846589e-05, "loss": 1.2217, "step": 5987 }, { "epoch": 0.75, "learning_rate": 7.769236340895092e-05, "loss": 1.3196, "step": 5988 }, { "epoch": 0.75, "learning_rate": 7.761942174157568e-05, "loss": 1.2306, "step": 5989 }, { "epoch": 0.75, "learning_rate": 7.754650803817309e-05, "loss": 1.1687, "step": 5990 }, { "epoch": 0.75, "learning_rate": 7.747362231057123e-05, "loss": 1.2924, "step": 5991 }, { "epoch": 0.75, "learning_rate": 7.740076457059391e-05, "loss": 1.1896, "step": 5992 }, { "epoch": 0.75, "learning_rate": 7.732793483006023e-05, "loss": 1.2519, "step": 5993 }, { "epoch": 0.75, "learning_rate": 7.725513310078473e-05, "loss": 1.2335, "step": 5994 }, { "epoch": 0.75, "learning_rate": 7.718235939457757e-05, "loss": 1.2459, "step": 5995 }, { "epoch": 0.75, "learning_rate": 7.710961372324423e-05, "loss": 1.2375, "step": 5996 }, { "epoch": 0.75, "learning_rate": 7.703689609858563e-05, "loss": 1.2334, "step": 5997 }, { "epoch": 0.75, "learning_rate": 7.696420653239832e-05, "loss": 1.2092, "step": 5998 }, { "epoch": 0.75, "learning_rate": 7.6891545036474e-05, "loss": 1.2774, "step": 5999 }, { "epoch": 0.75, "learning_rate": 7.681891162260016e-05, "loss": 1.2089, "step": 6000 }, { "epoch": 0.75, "eval_acceptance_rate": 0.6980666518211365, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2103903503392166, "eval_expected_tokens": 2.7629879379240205, "eval_loss": 1.3523883819580078, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.440746984481005, "eval_runtime": 50.231, "eval_samples_per_second": 9.695, "eval_steps_per_second": 0.319, "step": 6000 }, { "epoch": 0.75, "learning_rate": 7.674630630255946e-05, "loss": 1.1778, "step": 6001 }, { "epoch": 0.75, "learning_rate": 7.667372908813011e-05, "loss": 1.2444, "step": 6002 }, { "epoch": 0.75, "learning_rate": 7.660117999108568e-05, "loss": 1.2168, "step": 6003 }, { "epoch": 0.75, "learning_rate": 7.652865902319533e-05, "loss": 1.2506, "step": 6004 }, { "epoch": 0.75, "learning_rate": 7.645616619622364e-05, "loss": 1.2012, "step": 6005 }, { "epoch": 0.75, "learning_rate": 7.638370152193053e-05, "loss": 1.2466, "step": 6006 }, { "epoch": 0.75, "learning_rate": 7.631126501207131e-05, "loss": 1.204, "step": 6007 }, { "epoch": 0.75, "learning_rate": 7.623885667839687e-05, "loss": 1.2629, "step": 6008 }, { "epoch": 0.75, "learning_rate": 7.616647653265332e-05, "loss": 1.2095, "step": 6009 }, { "epoch": 0.75, "learning_rate": 7.609412458658243e-05, "loss": 1.2112, "step": 6010 }, { "epoch": 0.75, "learning_rate": 7.602180085192142e-05, "loss": 1.196, "step": 6011 }, { "epoch": 0.75, "learning_rate": 7.594950534040268e-05, "loss": 1.1725, "step": 6012 }, { "epoch": 0.75, "learning_rate": 7.587723806375418e-05, "loss": 1.2206, "step": 6013 }, { "epoch": 0.75, "learning_rate": 7.580499903369917e-05, "loss": 1.2183, "step": 6014 }, { "epoch": 0.75, "learning_rate": 7.573278826195664e-05, "loss": 1.2375, "step": 6015 }, { "epoch": 0.75, "learning_rate": 7.566060576024058e-05, "loss": 1.2145, "step": 6016 }, { "epoch": 0.75, "learning_rate": 7.558845154026081e-05, "loss": 1.2835, "step": 6017 }, { "epoch": 0.75, "learning_rate": 7.551632561372224e-05, "loss": 1.225, "step": 6018 }, { "epoch": 0.75, "learning_rate": 7.544422799232523e-05, "loss": 1.2026, "step": 6019 }, { "epoch": 0.75, "learning_rate": 7.537215868776579e-05, "loss": 1.2782, "step": 6020 }, { "epoch": 0.75, "learning_rate": 7.530011771173507e-05, "loss": 1.1867, "step": 6021 }, { "epoch": 0.75, "learning_rate": 7.522810507591962e-05, "loss": 1.2737, "step": 6022 }, { "epoch": 0.75, "learning_rate": 7.515612079200171e-05, "loss": 1.2829, "step": 6023 }, { "epoch": 0.75, "learning_rate": 7.508416487165862e-05, "loss": 1.1974, "step": 6024 }, { "epoch": 0.75, "learning_rate": 7.501223732656337e-05, "loss": 1.2258, "step": 6025 }, { "epoch": 0.75, "learning_rate": 7.494033816838408e-05, "loss": 1.228, "step": 6026 }, { "epoch": 0.75, "learning_rate": 7.486846740878445e-05, "loss": 1.1775, "step": 6027 }, { "epoch": 0.75, "learning_rate": 7.479662505942342e-05, "loss": 1.3153, "step": 6028 }, { "epoch": 0.75, "learning_rate": 7.47248111319555e-05, "loss": 1.2357, "step": 6029 }, { "epoch": 0.75, "learning_rate": 7.465302563803062e-05, "loss": 1.163, "step": 6030 }, { "epoch": 0.75, "learning_rate": 7.458126858929387e-05, "loss": 1.1715, "step": 6031 }, { "epoch": 0.75, "learning_rate": 7.450953999738582e-05, "loss": 1.2204, "step": 6032 }, { "epoch": 0.75, "learning_rate": 7.443783987394245e-05, "loss": 1.2415, "step": 6033 }, { "epoch": 0.75, "learning_rate": 7.43661682305952e-05, "loss": 1.2069, "step": 6034 }, { "epoch": 0.75, "learning_rate": 7.429452507897069e-05, "loss": 1.2128, "step": 6035 }, { "epoch": 0.75, "learning_rate": 7.422291043069116e-05, "loss": 1.2254, "step": 6036 }, { "epoch": 0.75, "learning_rate": 7.415132429737406e-05, "loss": 1.189, "step": 6037 }, { "epoch": 0.75, "learning_rate": 7.407976669063216e-05, "loss": 1.2115, "step": 6038 }, { "epoch": 0.75, "learning_rate": 7.400823762207387e-05, "loss": 1.2863, "step": 6039 }, { "epoch": 0.76, "learning_rate": 7.39367371033027e-05, "loss": 1.2471, "step": 6040 }, { "epoch": 0.76, "learning_rate": 7.386526514591758e-05, "loss": 1.252, "step": 6041 }, { "epoch": 0.76, "learning_rate": 7.379382176151298e-05, "loss": 1.2356, "step": 6042 }, { "epoch": 0.76, "learning_rate": 7.372240696167845e-05, "loss": 1.2081, "step": 6043 }, { "epoch": 0.76, "learning_rate": 7.365102075799923e-05, "loss": 1.2906, "step": 6044 }, { "epoch": 0.76, "learning_rate": 7.357966316205573e-05, "loss": 1.2452, "step": 6045 }, { "epoch": 0.76, "learning_rate": 7.350833418542363e-05, "loss": 1.2068, "step": 6046 }, { "epoch": 0.76, "learning_rate": 7.34370338396741e-05, "loss": 1.211, "step": 6047 }, { "epoch": 0.76, "learning_rate": 7.336576213637369e-05, "loss": 1.2509, "step": 6048 }, { "epoch": 0.76, "learning_rate": 7.329451908708432e-05, "loss": 1.2822, "step": 6049 }, { "epoch": 0.76, "learning_rate": 7.322330470336314e-05, "loss": 1.2543, "step": 6050 }, { "epoch": 0.76, "learning_rate": 7.315211899676271e-05, "loss": 1.2605, "step": 6051 }, { "epoch": 0.76, "learning_rate": 7.308096197883093e-05, "loss": 1.2478, "step": 6052 }, { "epoch": 0.76, "learning_rate": 7.3009833661111e-05, "loss": 1.2055, "step": 6053 }, { "epoch": 0.76, "learning_rate": 7.293873405514157e-05, "loss": 1.2008, "step": 6054 }, { "epoch": 0.76, "learning_rate": 7.286766317245667e-05, "loss": 1.2294, "step": 6055 }, { "epoch": 0.76, "learning_rate": 7.27966210245855e-05, "loss": 1.2099, "step": 6056 }, { "epoch": 0.76, "learning_rate": 7.27256076230527e-05, "loss": 1.2675, "step": 6057 }, { "epoch": 0.76, "learning_rate": 7.265462297937811e-05, "loss": 1.2544, "step": 6058 }, { "epoch": 0.76, "learning_rate": 7.258366710507719e-05, "loss": 1.2, "step": 6059 }, { "epoch": 0.76, "learning_rate": 7.251274001166044e-05, "loss": 1.2157, "step": 6060 }, { "epoch": 0.76, "learning_rate": 7.244184171063395e-05, "loss": 1.2432, "step": 6061 }, { "epoch": 0.76, "learning_rate": 7.237097221349892e-05, "loss": 1.1925, "step": 6062 }, { "epoch": 0.76, "learning_rate": 7.230013153175189e-05, "loss": 1.1633, "step": 6063 }, { "epoch": 0.76, "learning_rate": 7.222931967688496e-05, "loss": 1.2163, "step": 6064 }, { "epoch": 0.76, "learning_rate": 7.21585366603853e-05, "loss": 1.245, "step": 6065 }, { "epoch": 0.76, "learning_rate": 7.20877824937354e-05, "loss": 1.2246, "step": 6066 }, { "epoch": 0.76, "learning_rate": 7.201705718841339e-05, "loss": 1.2694, "step": 6067 }, { "epoch": 0.76, "learning_rate": 7.194636075589226e-05, "loss": 1.2377, "step": 6068 }, { "epoch": 0.76, "learning_rate": 7.187569320764073e-05, "loss": 1.2544, "step": 6069 }, { "epoch": 0.76, "learning_rate": 7.180505455512257e-05, "loss": 1.2231, "step": 6070 }, { "epoch": 0.76, "learning_rate": 7.173444480979696e-05, "loss": 1.2155, "step": 6071 }, { "epoch": 0.76, "learning_rate": 7.166386398311828e-05, "loss": 1.2262, "step": 6072 }, { "epoch": 0.76, "learning_rate": 7.159331208653641e-05, "loss": 1.1711, "step": 6073 }, { "epoch": 0.76, "learning_rate": 7.152278913149649e-05, "loss": 1.2588, "step": 6074 }, { "epoch": 0.76, "learning_rate": 7.145229512943885e-05, "loss": 1.2895, "step": 6075 }, { "epoch": 0.76, "learning_rate": 7.138183009179921e-05, "loss": 1.2484, "step": 6076 }, { "epoch": 0.76, "learning_rate": 7.131139403000847e-05, "loss": 1.2843, "step": 6077 }, { "epoch": 0.76, "learning_rate": 7.124098695549308e-05, "loss": 1.1501, "step": 6078 }, { "epoch": 0.76, "learning_rate": 7.117060887967447e-05, "loss": 1.2187, "step": 6079 }, { "epoch": 0.76, "learning_rate": 7.110025981396976e-05, "loss": 1.2179, "step": 6080 }, { "epoch": 0.76, "learning_rate": 7.102993976979097e-05, "loss": 1.2386, "step": 6081 }, { "epoch": 0.76, "learning_rate": 7.095964875854552e-05, "loss": 1.2403, "step": 6082 }, { "epoch": 0.76, "learning_rate": 7.088938679163639e-05, "loss": 1.2865, "step": 6083 }, { "epoch": 0.76, "learning_rate": 7.08191538804615e-05, "loss": 1.236, "step": 6084 }, { "epoch": 0.76, "learning_rate": 7.074895003641416e-05, "loss": 1.2092, "step": 6085 }, { "epoch": 0.76, "learning_rate": 7.06787752708831e-05, "loss": 1.1665, "step": 6086 }, { "epoch": 0.76, "learning_rate": 7.060862959525214e-05, "loss": 1.2347, "step": 6087 }, { "epoch": 0.76, "learning_rate": 7.053851302090059e-05, "loss": 1.2531, "step": 6088 }, { "epoch": 0.76, "learning_rate": 7.046842555920283e-05, "loss": 1.2448, "step": 6089 }, { "epoch": 0.76, "learning_rate": 7.039836722152864e-05, "loss": 1.2678, "step": 6090 }, { "epoch": 0.76, "learning_rate": 7.032833801924296e-05, "loss": 1.2563, "step": 6091 }, { "epoch": 0.76, "learning_rate": 7.025833796370615e-05, "loss": 1.2344, "step": 6092 }, { "epoch": 0.76, "learning_rate": 7.018836706627387e-05, "loss": 1.1814, "step": 6093 }, { "epoch": 0.76, "learning_rate": 7.011842533829687e-05, "loss": 1.2616, "step": 6094 }, { "epoch": 0.76, "learning_rate": 7.004851279112128e-05, "loss": 1.247, "step": 6095 }, { "epoch": 0.76, "learning_rate": 6.997862943608846e-05, "loss": 1.2197, "step": 6096 }, { "epoch": 0.76, "learning_rate": 6.990877528453496e-05, "loss": 1.2301, "step": 6097 }, { "epoch": 0.76, "learning_rate": 6.983895034779277e-05, "loss": 1.2885, "step": 6098 }, { "epoch": 0.76, "learning_rate": 6.976915463718914e-05, "loss": 1.2226, "step": 6099 }, { "epoch": 0.76, "learning_rate": 6.96993881640464e-05, "loss": 1.2594, "step": 6100 }, { "epoch": 0.76, "eval_acceptance_rate": 0.6981348991394043, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2106753019184384, "eval_expected_tokens": 2.763344127398048, "eval_loss": 1.3517311811447144, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.440836031849512, "eval_runtime": 50.216, "eval_samples_per_second": 9.698, "eval_steps_per_second": 0.319, "step": 6100 }, { "epoch": 0.76, "learning_rate": 6.962965093968221e-05, "loss": 1.1875, "step": 6101 }, { "epoch": 0.76, "learning_rate": 6.955994297540946e-05, "loss": 1.2519, "step": 6102 }, { "epoch": 0.76, "learning_rate": 6.949026428253646e-05, "loss": 1.2851, "step": 6103 }, { "epoch": 0.76, "learning_rate": 6.942061487236653e-05, "loss": 1.2294, "step": 6104 }, { "epoch": 0.76, "learning_rate": 6.935099475619847e-05, "loss": 1.3434, "step": 6105 }, { "epoch": 0.76, "learning_rate": 6.928140394532614e-05, "loss": 1.2835, "step": 6106 }, { "epoch": 0.76, "learning_rate": 6.921184245103867e-05, "loss": 1.2148, "step": 6107 }, { "epoch": 0.76, "learning_rate": 6.914231028462065e-05, "loss": 1.2572, "step": 6108 }, { "epoch": 0.76, "learning_rate": 6.907280745735161e-05, "loss": 1.2001, "step": 6109 }, { "epoch": 0.76, "learning_rate": 6.900333398050646e-05, "loss": 1.2206, "step": 6110 }, { "epoch": 0.76, "learning_rate": 6.893388986535542e-05, "loss": 1.2368, "step": 6111 }, { "epoch": 0.76, "learning_rate": 6.886447512316376e-05, "loss": 1.187, "step": 6112 }, { "epoch": 0.76, "learning_rate": 6.879508976519228e-05, "loss": 1.3424, "step": 6113 }, { "epoch": 0.76, "learning_rate": 6.872573380269672e-05, "loss": 1.2204, "step": 6114 }, { "epoch": 0.76, "learning_rate": 6.865640724692814e-05, "loss": 1.1971, "step": 6115 }, { "epoch": 0.76, "learning_rate": 6.858711010913284e-05, "loss": 1.2139, "step": 6116 }, { "epoch": 0.76, "learning_rate": 6.85178424005524e-05, "loss": 1.2234, "step": 6117 }, { "epoch": 0.76, "learning_rate": 6.844860413242366e-05, "loss": 1.1865, "step": 6118 }, { "epoch": 0.76, "learning_rate": 6.837939531597854e-05, "loss": 1.2618, "step": 6119 }, { "epoch": 0.77, "learning_rate": 6.831021596244425e-05, "loss": 1.2186, "step": 6120 }, { "epoch": 0.77, "learning_rate": 6.824106608304323e-05, "loss": 1.2516, "step": 6121 }, { "epoch": 0.77, "learning_rate": 6.817194568899304e-05, "loss": 1.2091, "step": 6122 }, { "epoch": 0.77, "learning_rate": 6.810285479150665e-05, "loss": 1.2355, "step": 6123 }, { "epoch": 0.77, "learning_rate": 6.803379340179223e-05, "loss": 1.1772, "step": 6124 }, { "epoch": 0.77, "learning_rate": 6.796476153105294e-05, "loss": 1.1708, "step": 6125 }, { "epoch": 0.77, "learning_rate": 6.789575919048726e-05, "loss": 1.2063, "step": 6126 }, { "epoch": 0.77, "learning_rate": 6.782678639128905e-05, "loss": 1.2184, "step": 6127 }, { "epoch": 0.77, "learning_rate": 6.775784314464717e-05, "loss": 1.2064, "step": 6128 }, { "epoch": 0.77, "learning_rate": 6.768892946174565e-05, "loss": 1.2122, "step": 6129 }, { "epoch": 0.77, "learning_rate": 6.7620045353764e-05, "loss": 1.1801, "step": 6130 }, { "epoch": 0.77, "learning_rate": 6.75511908318766e-05, "loss": 1.2471, "step": 6131 }, { "epoch": 0.77, "learning_rate": 6.748236590725335e-05, "loss": 1.2031, "step": 6132 }, { "epoch": 0.77, "learning_rate": 6.741357059105907e-05, "loss": 1.2614, "step": 6133 }, { "epoch": 0.77, "learning_rate": 6.734480489445397e-05, "loss": 1.2628, "step": 6134 }, { "epoch": 0.77, "learning_rate": 6.727606882859321e-05, "loss": 1.2127, "step": 6135 }, { "epoch": 0.77, "learning_rate": 6.720736240462746e-05, "loss": 1.1618, "step": 6136 }, { "epoch": 0.77, "learning_rate": 6.713868563370249e-05, "loss": 1.2075, "step": 6137 }, { "epoch": 0.77, "learning_rate": 6.707003852695908e-05, "loss": 1.2203, "step": 6138 }, { "epoch": 0.77, "learning_rate": 6.700142109553342e-05, "loss": 1.3069, "step": 6139 }, { "epoch": 0.77, "learning_rate": 6.69328333505567e-05, "loss": 1.1932, "step": 6140 }, { "epoch": 0.77, "learning_rate": 6.686427530315534e-05, "loss": 1.2549, "step": 6141 }, { "epoch": 0.77, "learning_rate": 6.679574696445107e-05, "loss": 1.2062, "step": 6142 }, { "epoch": 0.77, "learning_rate": 6.672724834556074e-05, "loss": 1.1901, "step": 6143 }, { "epoch": 0.77, "learning_rate": 6.665877945759632e-05, "loss": 1.2261, "step": 6144 }, { "epoch": 0.77, "learning_rate": 6.6590340311665e-05, "loss": 1.1773, "step": 6145 }, { "epoch": 0.77, "learning_rate": 6.652193091886905e-05, "loss": 1.2128, "step": 6146 }, { "epoch": 0.77, "learning_rate": 6.645355129030611e-05, "loss": 1.2084, "step": 6147 }, { "epoch": 0.77, "learning_rate": 6.638520143706878e-05, "loss": 1.2118, "step": 6148 }, { "epoch": 0.77, "learning_rate": 6.631688137024505e-05, "loss": 1.2403, "step": 6149 }, { "epoch": 0.77, "learning_rate": 6.624859110091791e-05, "loss": 1.1999, "step": 6150 }, { "epoch": 0.77, "learning_rate": 6.618033064016545e-05, "loss": 1.2766, "step": 6151 }, { "epoch": 0.77, "learning_rate": 6.611209999906123e-05, "loss": 1.1841, "step": 6152 }, { "epoch": 0.77, "learning_rate": 6.604389918867368e-05, "loss": 1.2116, "step": 6153 }, { "epoch": 0.77, "learning_rate": 6.597572822006642e-05, "loss": 1.2123, "step": 6154 }, { "epoch": 0.77, "learning_rate": 6.590758710429848e-05, "loss": 1.1839, "step": 6155 }, { "epoch": 0.77, "learning_rate": 6.583947585242364e-05, "loss": 1.1941, "step": 6156 }, { "epoch": 0.77, "learning_rate": 6.577139447549133e-05, "loss": 1.234, "step": 6157 }, { "epoch": 0.77, "learning_rate": 6.570334298454566e-05, "loss": 1.2192, "step": 6158 }, { "epoch": 0.77, "learning_rate": 6.563532139062622e-05, "loss": 1.2401, "step": 6159 }, { "epoch": 0.77, "learning_rate": 6.556732970476748e-05, "loss": 1.2426, "step": 6160 }, { "epoch": 0.77, "learning_rate": 6.54993679379993e-05, "loss": 1.235, "step": 6161 }, { "epoch": 0.77, "learning_rate": 6.543143610134669e-05, "loss": 1.1832, "step": 6162 }, { "epoch": 0.77, "learning_rate": 6.53635342058296e-05, "loss": 1.1659, "step": 6163 }, { "epoch": 0.77, "learning_rate": 6.529566226246326e-05, "loss": 1.2291, "step": 6164 }, { "epoch": 0.77, "learning_rate": 6.522782028225788e-05, "loss": 1.226, "step": 6165 }, { "epoch": 0.77, "learning_rate": 6.516000827621918e-05, "loss": 1.2011, "step": 6166 }, { "epoch": 0.77, "learning_rate": 6.509222625534755e-05, "loss": 1.2382, "step": 6167 }, { "epoch": 0.77, "learning_rate": 6.502447423063895e-05, "loss": 1.257, "step": 6168 }, { "epoch": 0.77, "learning_rate": 6.495675221308417e-05, "loss": 1.264, "step": 6169 }, { "epoch": 0.77, "learning_rate": 6.488906021366925e-05, "loss": 1.2399, "step": 6170 }, { "epoch": 0.77, "learning_rate": 6.482139824337522e-05, "loss": 1.1742, "step": 6171 }, { "epoch": 0.77, "learning_rate": 6.475376631317856e-05, "loss": 1.1967, "step": 6172 }, { "epoch": 0.77, "learning_rate": 6.468616443405048e-05, "loss": 1.2362, "step": 6173 }, { "epoch": 0.77, "learning_rate": 6.461859261695771e-05, "loss": 1.2701, "step": 6174 }, { "epoch": 0.77, "learning_rate": 6.455105087286172e-05, "loss": 1.2263, "step": 6175 }, { "epoch": 0.77, "learning_rate": 6.448353921271949e-05, "loss": 1.1591, "step": 6176 }, { "epoch": 0.77, "learning_rate": 6.441605764748276e-05, "loss": 1.2378, "step": 6177 }, { "epoch": 0.77, "learning_rate": 6.434860618809851e-05, "loss": 1.2551, "step": 6178 }, { "epoch": 0.77, "learning_rate": 6.428118484550905e-05, "loss": 1.2618, "step": 6179 }, { "epoch": 0.77, "learning_rate": 6.421379363065141e-05, "loss": 1.2218, "step": 6180 }, { "epoch": 0.77, "learning_rate": 6.414643255445815e-05, "loss": 1.1429, "step": 6181 }, { "epoch": 0.77, "learning_rate": 6.407910162785662e-05, "loss": 1.2311, "step": 6182 }, { "epoch": 0.77, "learning_rate": 6.401180086176947e-05, "loss": 1.2389, "step": 6183 }, { "epoch": 0.77, "learning_rate": 6.394453026711424e-05, "loss": 1.2013, "step": 6184 }, { "epoch": 0.77, "learning_rate": 6.387728985480381e-05, "loss": 1.1803, "step": 6185 }, { "epoch": 0.77, "learning_rate": 6.381007963574617e-05, "loss": 1.2647, "step": 6186 }, { "epoch": 0.77, "learning_rate": 6.37428996208442e-05, "loss": 1.1795, "step": 6187 }, { "epoch": 0.77, "learning_rate": 6.367574982099605e-05, "loss": 1.216, "step": 6188 }, { "epoch": 0.77, "learning_rate": 6.36086302470949e-05, "loss": 1.2534, "step": 6189 }, { "epoch": 0.77, "learning_rate": 6.354154091002892e-05, "loss": 1.1925, "step": 6190 }, { "epoch": 0.77, "learning_rate": 6.347448182068163e-05, "loss": 1.2223, "step": 6191 }, { "epoch": 0.77, "learning_rate": 6.340745298993156e-05, "loss": 1.2785, "step": 6192 }, { "epoch": 0.77, "learning_rate": 6.334045442865219e-05, "loss": 1.2056, "step": 6193 }, { "epoch": 0.77, "learning_rate": 6.327348614771219e-05, "loss": 1.2595, "step": 6194 }, { "epoch": 0.77, "learning_rate": 6.320654815797525e-05, "loss": 1.1834, "step": 6195 }, { "epoch": 0.77, "learning_rate": 6.313964047030032e-05, "loss": 1.2207, "step": 6196 }, { "epoch": 0.77, "learning_rate": 6.307276309554122e-05, "loss": 1.1558, "step": 6197 }, { "epoch": 0.77, "learning_rate": 6.300591604454706e-05, "loss": 1.2992, "step": 6198 }, { "epoch": 0.77, "learning_rate": 6.293909932816185e-05, "loss": 1.2079, "step": 6199 }, { "epoch": 0.78, "learning_rate": 6.28723129572247e-05, "loss": 1.2652, "step": 6200 }, { "epoch": 0.78, "eval_acceptance_rate": 0.6987568140029907, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2132740372751205, "eval_expected_tokens": 2.7665925465939005, "eval_loss": 1.3513942956924438, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4416481366484752, "eval_runtime": 50.2054, "eval_samples_per_second": 9.7, "eval_steps_per_second": 0.319, "step": 6200 }, { "epoch": 0.78, "learning_rate": 6.280555694256998e-05, "loss": 1.19, "step": 6201 }, { "epoch": 0.78, "learning_rate": 6.273883129502694e-05, "loss": 1.2146, "step": 6202 }, { "epoch": 0.78, "learning_rate": 6.267213602541988e-05, "loss": 1.2006, "step": 6203 }, { "epoch": 0.78, "learning_rate": 6.260547114456842e-05, "loss": 1.2549, "step": 6204 }, { "epoch": 0.78, "learning_rate": 6.253883666328689e-05, "loss": 1.2322, "step": 6205 }, { "epoch": 0.78, "learning_rate": 6.24722325923851e-05, "loss": 1.2828, "step": 6206 }, { "epoch": 0.78, "learning_rate": 6.240565894266764e-05, "loss": 1.2435, "step": 6207 }, { "epoch": 0.78, "learning_rate": 6.233911572493422e-05, "loss": 1.2304, "step": 6208 }, { "epoch": 0.78, "learning_rate": 6.227260294997955e-05, "loss": 1.2474, "step": 6209 }, { "epoch": 0.78, "learning_rate": 6.220612062859357e-05, "loss": 1.2375, "step": 6210 }, { "epoch": 0.78, "learning_rate": 6.213966877156126e-05, "loss": 1.2185, "step": 6211 }, { "epoch": 0.78, "learning_rate": 6.207324738966255e-05, "loss": 1.2753, "step": 6212 }, { "epoch": 0.78, "learning_rate": 6.200685649367244e-05, "loss": 1.2917, "step": 6213 }, { "epoch": 0.78, "learning_rate": 6.1940496094361e-05, "loss": 1.27, "step": 6214 }, { "epoch": 0.78, "learning_rate": 6.187416620249333e-05, "loss": 1.1534, "step": 6215 }, { "epoch": 0.78, "learning_rate": 6.180786682882969e-05, "loss": 1.2371, "step": 6216 }, { "epoch": 0.78, "learning_rate": 6.174159798412534e-05, "loss": 1.2585, "step": 6217 }, { "epoch": 0.78, "learning_rate": 6.167535967913055e-05, "loss": 1.2274, "step": 6218 }, { "epoch": 0.78, "learning_rate": 6.160915192459057e-05, "loss": 1.241, "step": 6219 }, { "epoch": 0.78, "learning_rate": 6.154297473124582e-05, "loss": 1.2208, "step": 6220 }, { "epoch": 0.78, "learning_rate": 6.147682810983174e-05, "loss": 1.2834, "step": 6221 }, { "epoch": 0.78, "learning_rate": 6.141071207107873e-05, "loss": 1.2726, "step": 6222 }, { "epoch": 0.78, "learning_rate": 6.134462662571238e-05, "loss": 1.265, "step": 6223 }, { "epoch": 0.78, "learning_rate": 6.12785717844531e-05, "loss": 1.1967, "step": 6224 }, { "epoch": 0.78, "learning_rate": 6.121254755801658e-05, "loss": 1.1801, "step": 6225 }, { "epoch": 0.78, "learning_rate": 6.114655395711338e-05, "loss": 1.1999, "step": 6226 }, { "epoch": 0.78, "learning_rate": 6.108059099244912e-05, "loss": 1.2888, "step": 6227 }, { "epoch": 0.78, "learning_rate": 6.101465867472439e-05, "loss": 1.2541, "step": 6228 }, { "epoch": 0.78, "learning_rate": 6.094875701463495e-05, "loss": 1.1891, "step": 6229 }, { "epoch": 0.78, "learning_rate": 6.088288602287159e-05, "loss": 1.2295, "step": 6230 }, { "epoch": 0.78, "learning_rate": 6.0817045710120006e-05, "loss": 1.2561, "step": 6231 }, { "epoch": 0.78, "learning_rate": 6.075123608706093e-05, "loss": 1.2329, "step": 6232 }, { "epoch": 0.78, "learning_rate": 6.0685457164370155e-05, "loss": 1.1869, "step": 6233 }, { "epoch": 0.78, "learning_rate": 6.0619708952718454e-05, "loss": 1.2272, "step": 6234 }, { "epoch": 0.78, "learning_rate": 6.055399146277166e-05, "loss": 1.216, "step": 6235 }, { "epoch": 0.78, "learning_rate": 6.0488304705190734e-05, "loss": 1.2244, "step": 6236 }, { "epoch": 0.78, "learning_rate": 6.042264869063146e-05, "loss": 1.2245, "step": 6237 }, { "epoch": 0.78, "learning_rate": 6.0357023429744695e-05, "loss": 1.201, "step": 6238 }, { "epoch": 0.78, "learning_rate": 6.029142893317624e-05, "loss": 1.2304, "step": 6239 }, { "epoch": 0.78, "learning_rate": 6.0225865211567146e-05, "loss": 1.2664, "step": 6240 }, { "epoch": 0.78, "learning_rate": 6.016033227555315e-05, "loss": 1.3004, "step": 6241 }, { "epoch": 0.78, "learning_rate": 6.00948301357653e-05, "loss": 1.2403, "step": 6242 }, { "epoch": 0.78, "learning_rate": 6.0029358802829445e-05, "loss": 1.2188, "step": 6243 }, { "epoch": 0.78, "learning_rate": 5.9963918287366444e-05, "loss": 1.2501, "step": 6244 }, { "epoch": 0.78, "learning_rate": 5.989850859999227e-05, "loss": 1.2648, "step": 6245 }, { "epoch": 0.78, "learning_rate": 5.983312975131785e-05, "loss": 1.2137, "step": 6246 }, { "epoch": 0.78, "learning_rate": 5.976778175194897e-05, "loss": 1.2212, "step": 6247 }, { "epoch": 0.78, "learning_rate": 5.9702464612486686e-05, "loss": 1.2308, "step": 6248 }, { "epoch": 0.78, "learning_rate": 5.9637178343526774e-05, "loss": 1.2018, "step": 6249 }, { "epoch": 0.78, "learning_rate": 5.957192295566022e-05, "loss": 1.2312, "step": 6250 }, { "epoch": 0.78, "learning_rate": 5.950669845947287e-05, "loss": 1.252, "step": 6251 }, { "epoch": 0.78, "learning_rate": 5.944150486554556e-05, "loss": 1.1567, "step": 6252 }, { "epoch": 0.78, "learning_rate": 5.937634218445412e-05, "loss": 1.3475, "step": 6253 }, { "epoch": 0.78, "learning_rate": 5.93112104267694e-05, "loss": 1.1733, "step": 6254 }, { "epoch": 0.78, "learning_rate": 5.924610960305732e-05, "loss": 1.2416, "step": 6255 }, { "epoch": 0.78, "learning_rate": 5.918103972387864e-05, "loss": 1.2443, "step": 6256 }, { "epoch": 0.78, "learning_rate": 5.911600079978913e-05, "loss": 1.2223, "step": 6257 }, { "epoch": 0.78, "learning_rate": 5.905099284133952e-05, "loss": 1.2377, "step": 6258 }, { "epoch": 0.78, "learning_rate": 5.898601585907551e-05, "loss": 1.2313, "step": 6259 }, { "epoch": 0.78, "learning_rate": 5.89210698635379e-05, "loss": 1.2026, "step": 6260 }, { "epoch": 0.78, "learning_rate": 5.885615486526241e-05, "loss": 1.224, "step": 6261 }, { "epoch": 0.78, "learning_rate": 5.879127087477967e-05, "loss": 1.2381, "step": 6262 }, { "epoch": 0.78, "learning_rate": 5.872641790261529e-05, "loss": 1.2533, "step": 6263 }, { "epoch": 0.78, "learning_rate": 5.866159595928977e-05, "loss": 1.2661, "step": 6264 }, { "epoch": 0.78, "learning_rate": 5.859680505531886e-05, "loss": 1.2001, "step": 6265 }, { "epoch": 0.78, "learning_rate": 5.853204520121294e-05, "loss": 1.1935, "step": 6266 }, { "epoch": 0.78, "learning_rate": 5.8467316407477626e-05, "loss": 1.2344, "step": 6267 }, { "epoch": 0.78, "learning_rate": 5.84026186846133e-05, "loss": 1.234, "step": 6268 }, { "epoch": 0.78, "learning_rate": 5.833795204311532e-05, "loss": 1.2791, "step": 6269 }, { "epoch": 0.78, "learning_rate": 5.827331649347417e-05, "loss": 1.2905, "step": 6270 }, { "epoch": 0.78, "learning_rate": 5.8208712046175144e-05, "loss": 1.2694, "step": 6271 }, { "epoch": 0.78, "learning_rate": 5.8144138711698436e-05, "loss": 1.2663, "step": 6272 }, { "epoch": 0.78, "learning_rate": 5.807959650051936e-05, "loss": 1.3213, "step": 6273 }, { "epoch": 0.78, "learning_rate": 5.8015085423108155e-05, "loss": 1.2765, "step": 6274 }, { "epoch": 0.78, "learning_rate": 5.7950605489929895e-05, "loss": 1.2925, "step": 6275 }, { "epoch": 0.78, "learning_rate": 5.788615671144465e-05, "loss": 1.2147, "step": 6276 }, { "epoch": 0.78, "learning_rate": 5.782173909810745e-05, "loss": 1.2878, "step": 6277 }, { "epoch": 0.78, "learning_rate": 5.7757352660368216e-05, "loss": 1.3171, "step": 6278 }, { "epoch": 0.78, "learning_rate": 5.7692997408671926e-05, "loss": 1.2214, "step": 6279 }, { "epoch": 0.79, "learning_rate": 5.7628673353458506e-05, "loss": 1.2793, "step": 6280 }, { "epoch": 0.79, "learning_rate": 5.756438050516269e-05, "loss": 1.2632, "step": 6281 }, { "epoch": 0.79, "learning_rate": 5.750011887421419e-05, "loss": 1.3091, "step": 6282 }, { "epoch": 0.79, "learning_rate": 5.7435888471037615e-05, "loss": 1.2311, "step": 6283 }, { "epoch": 0.79, "learning_rate": 5.737168930605272e-05, "loss": 1.2403, "step": 6284 }, { "epoch": 0.79, "learning_rate": 5.730752138967388e-05, "loss": 1.2568, "step": 6285 }, { "epoch": 0.79, "learning_rate": 5.724338473231072e-05, "loss": 1.2235, "step": 6286 }, { "epoch": 0.79, "learning_rate": 5.7179279344367566e-05, "loss": 1.1169, "step": 6287 }, { "epoch": 0.79, "learning_rate": 5.711520523624364e-05, "loss": 1.2097, "step": 6288 }, { "epoch": 0.79, "learning_rate": 5.705116241833339e-05, "loss": 1.2276, "step": 6289 }, { "epoch": 0.79, "learning_rate": 5.6987150901025855e-05, "loss": 1.2177, "step": 6290 }, { "epoch": 0.79, "learning_rate": 5.6923170694705065e-05, "loss": 1.2081, "step": 6291 }, { "epoch": 0.79, "learning_rate": 5.6859221809750226e-05, "loss": 1.2321, "step": 6292 }, { "epoch": 0.79, "learning_rate": 5.679530425653512e-05, "loss": 1.2347, "step": 6293 }, { "epoch": 0.79, "learning_rate": 5.6731418045428694e-05, "loss": 1.242, "step": 6294 }, { "epoch": 0.79, "learning_rate": 5.66675631867947e-05, "loss": 1.2217, "step": 6295 }, { "epoch": 0.79, "learning_rate": 5.660373969099178e-05, "loss": 1.2444, "step": 6296 }, { "epoch": 0.79, "learning_rate": 5.6539947568373466e-05, "loss": 1.2842, "step": 6297 }, { "epoch": 0.79, "learning_rate": 5.647618682928835e-05, "loss": 1.2305, "step": 6298 }, { "epoch": 0.79, "learning_rate": 5.641245748407989e-05, "loss": 1.278, "step": 6299 }, { "epoch": 0.79, "learning_rate": 5.634875954308638e-05, "loss": 1.2661, "step": 6300 }, { "epoch": 0.79, "eval_acceptance_rate": 0.6988047957420349, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2134746888109107, "eval_expected_tokens": 2.7668433610136387, "eval_loss": 1.3499488830566406, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4417108402534096, "eval_runtime": 50.2269, "eval_samples_per_second": 9.696, "eval_steps_per_second": 0.319, "step": 6300 }, { "epoch": 0.79, "learning_rate": 5.6285093016641e-05, "loss": 1.3071, "step": 6301 }, { "epoch": 0.79, "learning_rate": 5.6221457915071885e-05, "loss": 1.2719, "step": 6302 }, { "epoch": 0.79, "learning_rate": 5.615785424870201e-05, "loss": 1.2622, "step": 6303 }, { "epoch": 0.79, "learning_rate": 5.609428202784941e-05, "loss": 1.2844, "step": 6304 }, { "epoch": 0.79, "learning_rate": 5.603074126282692e-05, "loss": 1.2961, "step": 6305 }, { "epoch": 0.79, "learning_rate": 5.596723196394224e-05, "loss": 1.203, "step": 6306 }, { "epoch": 0.79, "learning_rate": 5.590375414149798e-05, "loss": 1.2812, "step": 6307 }, { "epoch": 0.79, "learning_rate": 5.5840307805791605e-05, "loss": 1.2067, "step": 6308 }, { "epoch": 0.79, "learning_rate": 5.577689296711563e-05, "loss": 1.2042, "step": 6309 }, { "epoch": 0.79, "learning_rate": 5.5713509635757274e-05, "loss": 1.2282, "step": 6310 }, { "epoch": 0.79, "learning_rate": 5.565015782199881e-05, "loss": 1.2437, "step": 6311 }, { "epoch": 0.79, "learning_rate": 5.558683753611726e-05, "loss": 1.2775, "step": 6312 }, { "epoch": 0.79, "learning_rate": 5.552354878838453e-05, "loss": 1.2134, "step": 6313 }, { "epoch": 0.79, "learning_rate": 5.5460291589067596e-05, "loss": 1.2047, "step": 6314 }, { "epoch": 0.79, "learning_rate": 5.539706594842814e-05, "loss": 1.2573, "step": 6315 }, { "epoch": 0.79, "learning_rate": 5.533387187672265e-05, "loss": 1.2293, "step": 6316 }, { "epoch": 0.79, "learning_rate": 5.5270709384202785e-05, "loss": 1.2308, "step": 6317 }, { "epoch": 0.79, "learning_rate": 5.5207578481114764e-05, "loss": 1.1628, "step": 6318 }, { "epoch": 0.79, "learning_rate": 5.5144479177699966e-05, "loss": 1.2265, "step": 6319 }, { "epoch": 0.79, "learning_rate": 5.508141148419443e-05, "loss": 1.2522, "step": 6320 }, { "epoch": 0.79, "learning_rate": 5.501837541082916e-05, "loss": 1.2598, "step": 6321 }, { "epoch": 0.79, "learning_rate": 5.495537096782993e-05, "loss": 1.3005, "step": 6322 }, { "epoch": 0.79, "learning_rate": 5.489239816541755e-05, "loss": 1.27, "step": 6323 }, { "epoch": 0.79, "learning_rate": 5.482945701380765e-05, "loss": 1.3011, "step": 6324 }, { "epoch": 0.79, "learning_rate": 5.476654752321064e-05, "loss": 1.2947, "step": 6325 }, { "epoch": 0.79, "learning_rate": 5.470366970383184e-05, "loss": 1.2451, "step": 6326 }, { "epoch": 0.79, "learning_rate": 5.4640823565871355e-05, "loss": 1.1867, "step": 6327 }, { "epoch": 0.79, "learning_rate": 5.4578009119524375e-05, "loss": 1.1963, "step": 6328 }, { "epoch": 0.79, "learning_rate": 5.451522637498066e-05, "loss": 1.2286, "step": 6329 }, { "epoch": 0.79, "learning_rate": 5.4452475342425116e-05, "loss": 1.2108, "step": 6330 }, { "epoch": 0.79, "learning_rate": 5.43897560320373e-05, "loss": 1.2257, "step": 6331 }, { "epoch": 0.79, "learning_rate": 5.4327068453991604e-05, "loss": 1.2106, "step": 6332 }, { "epoch": 0.79, "learning_rate": 5.4264412618457496e-05, "loss": 1.2163, "step": 6333 }, { "epoch": 0.79, "learning_rate": 5.420178853559904e-05, "loss": 1.1483, "step": 6334 }, { "epoch": 0.79, "learning_rate": 5.4139196215575225e-05, "loss": 1.2057, "step": 6335 }, { "epoch": 0.79, "learning_rate": 5.407663566854007e-05, "loss": 1.2523, "step": 6336 }, { "epoch": 0.79, "learning_rate": 5.401410690464214e-05, "loss": 1.2773, "step": 6337 }, { "epoch": 0.79, "learning_rate": 5.395160993402512e-05, "loss": 1.2227, "step": 6338 }, { "epoch": 0.79, "learning_rate": 5.3889144766827356e-05, "loss": 1.2732, "step": 6339 }, { "epoch": 0.79, "learning_rate": 5.382671141318207e-05, "loss": 1.2677, "step": 6340 }, { "epoch": 0.79, "learning_rate": 5.3764309883217315e-05, "loss": 1.1992, "step": 6341 }, { "epoch": 0.79, "learning_rate": 5.370194018705604e-05, "loss": 1.1608, "step": 6342 }, { "epoch": 0.79, "learning_rate": 5.3639602334816104e-05, "loss": 1.2203, "step": 6343 }, { "epoch": 0.79, "learning_rate": 5.357729633660999e-05, "loss": 1.2095, "step": 6344 }, { "epoch": 0.79, "learning_rate": 5.351502220254512e-05, "loss": 1.2074, "step": 6345 }, { "epoch": 0.79, "learning_rate": 5.34527799427238e-05, "loss": 1.2219, "step": 6346 }, { "epoch": 0.79, "learning_rate": 5.3390569567242974e-05, "loss": 1.1903, "step": 6347 }, { "epoch": 0.79, "learning_rate": 5.332839108619467e-05, "loss": 1.2217, "step": 6348 }, { "epoch": 0.79, "learning_rate": 5.326624450966569e-05, "loss": 1.2752, "step": 6349 }, { "epoch": 0.79, "learning_rate": 5.320412984773748e-05, "loss": 1.2047, "step": 6350 }, { "epoch": 0.79, "learning_rate": 5.314204711048648e-05, "loss": 1.1655, "step": 6351 }, { "epoch": 0.79, "learning_rate": 5.30799963079838e-05, "loss": 1.2468, "step": 6352 }, { "epoch": 0.79, "learning_rate": 5.301797745029557e-05, "loss": 1.2233, "step": 6353 }, { "epoch": 0.79, "learning_rate": 5.295599054748257e-05, "loss": 1.2708, "step": 6354 }, { "epoch": 0.79, "learning_rate": 5.289403560960051e-05, "loss": 1.2228, "step": 6355 }, { "epoch": 0.79, "learning_rate": 5.283211264669985e-05, "loss": 1.1775, "step": 6356 }, { "epoch": 0.79, "learning_rate": 5.277022166882581e-05, "loss": 1.1836, "step": 6357 }, { "epoch": 0.79, "learning_rate": 5.2708362686018585e-05, "loss": 1.2853, "step": 6358 }, { "epoch": 0.79, "learning_rate": 5.2646535708313014e-05, "loss": 1.2144, "step": 6359 }, { "epoch": 0.8, "learning_rate": 5.2584740745738765e-05, "loss": 1.2716, "step": 6360 }, { "epoch": 0.8, "learning_rate": 5.2522977808320494e-05, "loss": 1.277, "step": 6361 }, { "epoch": 0.8, "learning_rate": 5.2461246906077396e-05, "loss": 1.2551, "step": 6362 }, { "epoch": 0.8, "learning_rate": 5.239954804902372e-05, "loss": 1.2759, "step": 6363 }, { "epoch": 0.8, "learning_rate": 5.2337881247168314e-05, "loss": 1.2877, "step": 6364 }, { "epoch": 0.8, "learning_rate": 5.227624651051496e-05, "loss": 1.2629, "step": 6365 }, { "epoch": 0.8, "learning_rate": 5.221464384906205e-05, "loss": 1.2572, "step": 6366 }, { "epoch": 0.8, "learning_rate": 5.215307327280303e-05, "loss": 1.2392, "step": 6367 }, { "epoch": 0.8, "learning_rate": 5.209153479172607e-05, "loss": 1.2268, "step": 6368 }, { "epoch": 0.8, "learning_rate": 5.2030028415814e-05, "loss": 1.2358, "step": 6369 }, { "epoch": 0.8, "learning_rate": 5.196855415504456e-05, "loss": 1.2415, "step": 6370 }, { "epoch": 0.8, "learning_rate": 5.1907112019390166e-05, "loss": 1.2203, "step": 6371 }, { "epoch": 0.8, "learning_rate": 5.184570201881819e-05, "loss": 1.2707, "step": 6372 }, { "epoch": 0.8, "learning_rate": 5.178432416329062e-05, "loss": 1.2131, "step": 6373 }, { "epoch": 0.8, "learning_rate": 5.172297846276447e-05, "loss": 1.2193, "step": 6374 }, { "epoch": 0.8, "learning_rate": 5.1661664927191235e-05, "loss": 1.2413, "step": 6375 }, { "epoch": 0.8, "learning_rate": 5.1600383566517334e-05, "loss": 1.2612, "step": 6376 }, { "epoch": 0.8, "learning_rate": 5.153913439068408e-05, "loss": 1.2166, "step": 6377 }, { "epoch": 0.8, "learning_rate": 5.14779174096274e-05, "loss": 1.2536, "step": 6378 }, { "epoch": 0.8, "learning_rate": 5.141673263327798e-05, "loss": 1.275, "step": 6379 }, { "epoch": 0.8, "learning_rate": 5.135558007156146e-05, "loss": 1.2594, "step": 6380 }, { "epoch": 0.8, "learning_rate": 5.129445973439806e-05, "loss": 1.2369, "step": 6381 }, { "epoch": 0.8, "learning_rate": 5.1233371631702966e-05, "loss": 1.1733, "step": 6382 }, { "epoch": 0.8, "learning_rate": 5.1172315773385966e-05, "loss": 1.2758, "step": 6383 }, { "epoch": 0.8, "learning_rate": 5.1111292169351706e-05, "loss": 1.1657, "step": 6384 }, { "epoch": 0.8, "learning_rate": 5.105030082949946e-05, "loss": 1.2177, "step": 6385 }, { "epoch": 0.8, "learning_rate": 5.09893417637235e-05, "loss": 1.2316, "step": 6386 }, { "epoch": 0.8, "learning_rate": 5.0928414981912776e-05, "loss": 1.2191, "step": 6387 }, { "epoch": 0.8, "learning_rate": 5.086752049395094e-05, "loss": 1.2875, "step": 6388 }, { "epoch": 0.8, "learning_rate": 5.0806658309716426e-05, "loss": 1.2482, "step": 6389 }, { "epoch": 0.8, "learning_rate": 5.0745828439082406e-05, "loss": 1.2352, "step": 6390 }, { "epoch": 0.8, "learning_rate": 5.0685030891916785e-05, "loss": 1.2265, "step": 6391 }, { "epoch": 0.8, "learning_rate": 5.062426567808237e-05, "loss": 1.2246, "step": 6392 }, { "epoch": 0.8, "learning_rate": 5.056353280743667e-05, "loss": 1.2374, "step": 6393 }, { "epoch": 0.8, "learning_rate": 5.050283228983188e-05, "loss": 1.2526, "step": 6394 }, { "epoch": 0.8, "learning_rate": 5.044216413511493e-05, "loss": 1.1713, "step": 6395 }, { "epoch": 0.8, "learning_rate": 5.0381528353127536e-05, "loss": 1.1572, "step": 6396 }, { "epoch": 0.8, "learning_rate": 5.032092495370627e-05, "loss": 1.2503, "step": 6397 }, { "epoch": 0.8, "learning_rate": 5.0260353946682215e-05, "loss": 1.2149, "step": 6398 }, { "epoch": 0.8, "learning_rate": 5.0199815341881506e-05, "loss": 1.3013, "step": 6399 }, { "epoch": 0.8, "learning_rate": 5.013930914912476e-05, "loss": 1.2872, "step": 6400 }, { "epoch": 0.8, "eval_acceptance_rate": 0.6992217302322388, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2152191732300643, "eval_expected_tokens": 2.7690239665375804, "eval_loss": 1.3504338264465332, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4422559916343951, "eval_runtime": 50.2619, "eval_samples_per_second": 9.689, "eval_steps_per_second": 0.318, "step": 6400 }, { "epoch": 0.8, "learning_rate": 5.007883537822736e-05, "loss": 1.2148, "step": 6401 }, { "epoch": 0.8, "learning_rate": 5.001839403899966e-05, "loss": 1.1467, "step": 6402 }, { "epoch": 0.8, "learning_rate": 4.9957985141246495e-05, "loss": 1.1882, "step": 6403 }, { "epoch": 0.8, "learning_rate": 4.98976086947675e-05, "loss": 1.1847, "step": 6404 }, { "epoch": 0.8, "learning_rate": 4.983726470935721e-05, "loss": 1.1493, "step": 6405 }, { "epoch": 0.8, "learning_rate": 4.9776953194804574e-05, "loss": 1.1842, "step": 6406 }, { "epoch": 0.8, "learning_rate": 4.971667416089368e-05, "loss": 1.208, "step": 6407 }, { "epoch": 0.8, "learning_rate": 4.965642761740302e-05, "loss": 1.2738, "step": 6408 }, { "epoch": 0.8, "learning_rate": 4.959621357410593e-05, "loss": 1.1624, "step": 6409 }, { "epoch": 0.8, "learning_rate": 4.9536032040770395e-05, "loss": 1.2774, "step": 6410 }, { "epoch": 0.8, "learning_rate": 4.9475883027159304e-05, "loss": 1.1864, "step": 6411 }, { "epoch": 0.8, "learning_rate": 4.9415766543030175e-05, "loss": 1.2962, "step": 6412 }, { "epoch": 0.8, "learning_rate": 4.9355682598135246e-05, "loss": 1.2379, "step": 6413 }, { "epoch": 0.8, "learning_rate": 4.929563120222141e-05, "loss": 1.2548, "step": 6414 }, { "epoch": 0.8, "learning_rate": 4.9235612365030325e-05, "loss": 1.2304, "step": 6415 }, { "epoch": 0.8, "learning_rate": 4.917562609629847e-05, "loss": 1.2368, "step": 6416 }, { "epoch": 0.8, "learning_rate": 4.9115672405756874e-05, "loss": 1.2221, "step": 6417 }, { "epoch": 0.8, "learning_rate": 4.9055751303131447e-05, "loss": 1.1788, "step": 6418 }, { "epoch": 0.8, "learning_rate": 4.899586279814269e-05, "loss": 1.2105, "step": 6419 }, { "epoch": 0.8, "learning_rate": 4.893600690050579e-05, "loss": 1.2134, "step": 6420 }, { "epoch": 0.8, "learning_rate": 4.8876183619930795e-05, "loss": 1.1382, "step": 6421 }, { "epoch": 0.8, "learning_rate": 4.881639296612231e-05, "loss": 1.2928, "step": 6422 }, { "epoch": 0.8, "learning_rate": 4.87566349487798e-05, "loss": 1.2066, "step": 6423 }, { "epoch": 0.8, "learning_rate": 4.8696909577597285e-05, "loss": 1.1812, "step": 6424 }, { "epoch": 0.8, "learning_rate": 4.8637216862263494e-05, "loss": 1.221, "step": 6425 }, { "epoch": 0.8, "learning_rate": 4.8577556812462063e-05, "loss": 1.2011, "step": 6426 }, { "epoch": 0.8, "learning_rate": 4.851792943787109e-05, "loss": 1.248, "step": 6427 }, { "epoch": 0.8, "learning_rate": 4.8458334748163416e-05, "loss": 1.2066, "step": 6428 }, { "epoch": 0.8, "learning_rate": 4.839877275300672e-05, "loss": 1.2229, "step": 6429 }, { "epoch": 0.8, "learning_rate": 4.8339243462063226e-05, "loss": 1.1635, "step": 6430 }, { "epoch": 0.8, "learning_rate": 4.827974688498998e-05, "loss": 1.2041, "step": 6431 }, { "epoch": 0.8, "learning_rate": 4.8220283031438634e-05, "loss": 1.2269, "step": 6432 }, { "epoch": 0.8, "learning_rate": 4.816085191105551e-05, "loss": 1.1741, "step": 6433 }, { "epoch": 0.8, "learning_rate": 4.810145353348161e-05, "loss": 1.2609, "step": 6434 }, { "epoch": 0.8, "learning_rate": 4.804208790835277e-05, "loss": 1.1882, "step": 6435 }, { "epoch": 0.8, "learning_rate": 4.798275504529942e-05, "loss": 1.2053, "step": 6436 }, { "epoch": 0.8, "learning_rate": 4.7923454953946676e-05, "loss": 1.2148, "step": 6437 }, { "epoch": 0.8, "learning_rate": 4.786418764391428e-05, "loss": 1.1448, "step": 6438 }, { "epoch": 0.8, "learning_rate": 4.7804953124816754e-05, "loss": 1.2498, "step": 6439 }, { "epoch": 0.81, "learning_rate": 4.7745751406263163e-05, "loss": 1.1095, "step": 6440 }, { "epoch": 0.81, "learning_rate": 4.768658249785743e-05, "loss": 1.2008, "step": 6441 }, { "epoch": 0.81, "learning_rate": 4.762744640919811e-05, "loss": 1.2425, "step": 6442 }, { "epoch": 0.81, "learning_rate": 4.756834314987835e-05, "loss": 1.2122, "step": 6443 }, { "epoch": 0.81, "learning_rate": 4.750927272948602e-05, "loss": 1.2269, "step": 6444 }, { "epoch": 0.81, "learning_rate": 4.7450235157603576e-05, "loss": 1.2015, "step": 6445 }, { "epoch": 0.81, "learning_rate": 4.739123044380836e-05, "loss": 1.1845, "step": 6446 }, { "epoch": 0.81, "learning_rate": 4.733225859767215e-05, "loss": 1.2685, "step": 6447 }, { "epoch": 0.81, "learning_rate": 4.727331962876158e-05, "loss": 1.2418, "step": 6448 }, { "epoch": 0.81, "learning_rate": 4.721441354663786e-05, "loss": 1.2693, "step": 6449 }, { "epoch": 0.81, "learning_rate": 4.715554036085673e-05, "loss": 1.2131, "step": 6450 }, { "epoch": 0.81, "learning_rate": 4.7096700080968915e-05, "loss": 1.2291, "step": 6451 }, { "epoch": 0.81, "learning_rate": 4.703789271651957e-05, "loss": 1.244, "step": 6452 }, { "epoch": 0.81, "learning_rate": 4.6979118277048426e-05, "loss": 1.1914, "step": 6453 }, { "epoch": 0.81, "learning_rate": 4.692037677209021e-05, "loss": 1.2153, "step": 6454 }, { "epoch": 0.81, "learning_rate": 4.686166821117394e-05, "loss": 1.2429, "step": 6455 }, { "epoch": 0.81, "learning_rate": 4.680299260382356e-05, "loss": 1.2419, "step": 6456 }, { "epoch": 0.81, "learning_rate": 4.6744349959557534e-05, "loss": 1.1785, "step": 6457 }, { "epoch": 0.81, "learning_rate": 4.6685740287889e-05, "loss": 1.2546, "step": 6458 }, { "epoch": 0.81, "learning_rate": 4.6627163598325695e-05, "loss": 1.2351, "step": 6459 }, { "epoch": 0.81, "learning_rate": 4.656861990037009e-05, "loss": 1.2101, "step": 6460 }, { "epoch": 0.81, "learning_rate": 4.651010920351939e-05, "loss": 1.1912, "step": 6461 }, { "epoch": 0.81, "learning_rate": 4.6451631517265225e-05, "loss": 1.1346, "step": 6462 }, { "epoch": 0.81, "learning_rate": 4.639318685109403e-05, "loss": 1.2456, "step": 6463 }, { "epoch": 0.81, "learning_rate": 4.633477521448679e-05, "loss": 1.1861, "step": 6464 }, { "epoch": 0.81, "learning_rate": 4.6276396616919126e-05, "loss": 1.2058, "step": 6465 }, { "epoch": 0.81, "learning_rate": 4.621805106786142e-05, "loss": 1.2325, "step": 6466 }, { "epoch": 0.81, "learning_rate": 4.6159738576778685e-05, "loss": 1.1695, "step": 6467 }, { "epoch": 0.81, "learning_rate": 4.610145915313041e-05, "loss": 1.1582, "step": 6468 }, { "epoch": 0.81, "learning_rate": 4.6043212806370805e-05, "loss": 1.235, "step": 6469 }, { "epoch": 0.81, "learning_rate": 4.5984999545948816e-05, "loss": 1.2518, "step": 6470 }, { "epoch": 0.81, "learning_rate": 4.5926819381307854e-05, "loss": 1.2943, "step": 6471 }, { "epoch": 0.81, "learning_rate": 4.586867232188602e-05, "loss": 1.2361, "step": 6472 }, { "epoch": 0.81, "learning_rate": 4.5810558377116175e-05, "loss": 1.1918, "step": 6473 }, { "epoch": 0.81, "learning_rate": 4.575247755642556e-05, "loss": 1.2269, "step": 6474 }, { "epoch": 0.81, "learning_rate": 4.56944298692363e-05, "loss": 1.2123, "step": 6475 }, { "epoch": 0.81, "learning_rate": 4.5636415324964986e-05, "loss": 1.2964, "step": 6476 }, { "epoch": 0.81, "learning_rate": 4.557843393302286e-05, "loss": 1.1532, "step": 6477 }, { "epoch": 0.81, "learning_rate": 4.552048570281569e-05, "loss": 1.1735, "step": 6478 }, { "epoch": 0.81, "learning_rate": 4.54625706437441e-05, "loss": 1.2126, "step": 6479 }, { "epoch": 0.81, "learning_rate": 4.540468876520323e-05, "loss": 1.274, "step": 6480 }, { "epoch": 0.81, "learning_rate": 4.534684007658277e-05, "loss": 1.2173, "step": 6481 }, { "epoch": 0.81, "learning_rate": 4.528902458726705e-05, "loss": 1.3483, "step": 6482 }, { "epoch": 0.81, "learning_rate": 4.523124230663503e-05, "loss": 1.1766, "step": 6483 }, { "epoch": 0.81, "learning_rate": 4.517349324406023e-05, "loss": 1.196, "step": 6484 }, { "epoch": 0.81, "learning_rate": 4.511577740891087e-05, "loss": 1.2253, "step": 6485 }, { "epoch": 0.81, "learning_rate": 4.505809481054984e-05, "loss": 1.2914, "step": 6486 }, { "epoch": 0.81, "learning_rate": 4.500044545833448e-05, "loss": 1.2401, "step": 6487 }, { "epoch": 0.81, "learning_rate": 4.494282936161681e-05, "loss": 1.2569, "step": 6488 }, { "epoch": 0.81, "learning_rate": 4.488524652974335e-05, "loss": 1.2746, "step": 6489 }, { "epoch": 0.81, "learning_rate": 4.482769697205544e-05, "loss": 1.2497, "step": 6490 }, { "epoch": 0.81, "learning_rate": 4.477018069788882e-05, "loss": 1.2247, "step": 6491 }, { "epoch": 0.81, "learning_rate": 4.471269771657399e-05, "loss": 1.2438, "step": 6492 }, { "epoch": 0.81, "learning_rate": 4.465524803743595e-05, "loss": 1.2681, "step": 6493 }, { "epoch": 0.81, "learning_rate": 4.4597831669794224e-05, "loss": 1.2576, "step": 6494 }, { "epoch": 0.81, "learning_rate": 4.454044862296319e-05, "loss": 1.2198, "step": 6495 }, { "epoch": 0.81, "learning_rate": 4.4483098906251544e-05, "loss": 1.2535, "step": 6496 }, { "epoch": 0.81, "learning_rate": 4.442578252896265e-05, "loss": 1.2402, "step": 6497 }, { "epoch": 0.81, "learning_rate": 4.436849950039465e-05, "loss": 1.17, "step": 6498 }, { "epoch": 0.81, "learning_rate": 4.431124982983997e-05, "loss": 1.2654, "step": 6499 }, { "epoch": 0.81, "learning_rate": 4.425403352658591e-05, "loss": 1.2551, "step": 6500 }, { "epoch": 0.81, "eval_acceptance_rate": 0.6991474628448486, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2149083102951836, "eval_expected_tokens": 2.7686353878689793, "eval_loss": 1.3497247695922852, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4421588469672449, "eval_runtime": 50.3077, "eval_samples_per_second": 9.68, "eval_steps_per_second": 0.318, "step": 6500 }, { "epoch": 0.81, "learning_rate": 4.419685059991418e-05, "loss": 1.2413, "step": 6501 }, { "epoch": 0.81, "learning_rate": 4.413970105910112e-05, "loss": 1.2175, "step": 6502 }, { "epoch": 0.81, "learning_rate": 4.40825849134176e-05, "loss": 1.2191, "step": 6503 }, { "epoch": 0.81, "learning_rate": 4.4025502172129224e-05, "loss": 1.1978, "step": 6504 }, { "epoch": 0.81, "learning_rate": 4.3968452844496076e-05, "loss": 1.2293, "step": 6505 }, { "epoch": 0.81, "learning_rate": 4.3911436939772816e-05, "loss": 1.2526, "step": 6506 }, { "epoch": 0.81, "learning_rate": 4.38544544672087e-05, "loss": 1.2501, "step": 6507 }, { "epoch": 0.81, "learning_rate": 4.379750543604755e-05, "loss": 1.247, "step": 6508 }, { "epoch": 0.81, "learning_rate": 4.3740589855527674e-05, "loss": 1.2139, "step": 6509 }, { "epoch": 0.81, "learning_rate": 4.368370773488214e-05, "loss": 1.2446, "step": 6510 }, { "epoch": 0.81, "learning_rate": 4.362685908333855e-05, "loss": 1.219, "step": 6511 }, { "epoch": 0.81, "learning_rate": 4.357004391011898e-05, "loss": 1.196, "step": 6512 }, { "epoch": 0.81, "learning_rate": 4.35132622244401e-05, "loss": 1.2222, "step": 6513 }, { "epoch": 0.81, "learning_rate": 4.3456514035513094e-05, "loss": 1.2014, "step": 6514 }, { "epoch": 0.81, "learning_rate": 4.33997993525439e-05, "loss": 1.2388, "step": 6515 }, { "epoch": 0.81, "learning_rate": 4.334311818473283e-05, "loss": 1.2115, "step": 6516 }, { "epoch": 0.81, "learning_rate": 4.328647054127491e-05, "loss": 1.1812, "step": 6517 }, { "epoch": 0.81, "learning_rate": 4.3229856431359515e-05, "loss": 1.2273, "step": 6518 }, { "epoch": 0.81, "learning_rate": 4.3173275864170884e-05, "loss": 1.2217, "step": 6519 }, { "epoch": 0.81, "learning_rate": 4.3116728848887567e-05, "loss": 1.2308, "step": 6520 }, { "epoch": 0.82, "learning_rate": 4.306021539468277e-05, "loss": 1.2728, "step": 6521 }, { "epoch": 0.82, "learning_rate": 4.300373551072412e-05, "loss": 1.2524, "step": 6522 }, { "epoch": 0.82, "learning_rate": 4.2947289206174024e-05, "loss": 1.2482, "step": 6523 }, { "epoch": 0.82, "learning_rate": 4.289087649018936e-05, "loss": 1.271, "step": 6524 }, { "epoch": 0.82, "learning_rate": 4.283449737192149e-05, "loss": 1.2584, "step": 6525 }, { "epoch": 0.82, "learning_rate": 4.2778151860516366e-05, "loss": 1.2688, "step": 6526 }, { "epoch": 0.82, "learning_rate": 4.272183996511447e-05, "loss": 1.1879, "step": 6527 }, { "epoch": 0.82, "learning_rate": 4.266556169485078e-05, "loss": 1.2433, "step": 6528 }, { "epoch": 0.82, "learning_rate": 4.2609317058854985e-05, "loss": 1.1833, "step": 6529 }, { "epoch": 0.82, "learning_rate": 4.255310606625123e-05, "loss": 1.3303, "step": 6530 }, { "epoch": 0.82, "learning_rate": 4.2496928726158154e-05, "loss": 1.2142, "step": 6531 }, { "epoch": 0.82, "learning_rate": 4.244078504768897e-05, "loss": 1.2392, "step": 6532 }, { "epoch": 0.82, "learning_rate": 4.2384675039951386e-05, "loss": 1.2974, "step": 6533 }, { "epoch": 0.82, "learning_rate": 4.232859871204783e-05, "loss": 1.2956, "step": 6534 }, { "epoch": 0.82, "learning_rate": 4.2272556073074954e-05, "loss": 1.2275, "step": 6535 }, { "epoch": 0.82, "learning_rate": 4.2216547132124306e-05, "loss": 1.2108, "step": 6536 }, { "epoch": 0.82, "learning_rate": 4.216057189828168e-05, "loss": 1.3057, "step": 6537 }, { "epoch": 0.82, "learning_rate": 4.210463038062748e-05, "loss": 1.2938, "step": 6538 }, { "epoch": 0.82, "learning_rate": 4.204872258823675e-05, "loss": 1.2259, "step": 6539 }, { "epoch": 0.82, "learning_rate": 4.1992848530178955e-05, "loss": 1.246, "step": 6540 }, { "epoch": 0.82, "learning_rate": 4.1937008215518044e-05, "loss": 1.2111, "step": 6541 }, { "epoch": 0.82, "learning_rate": 4.188120165331269e-05, "loss": 1.2171, "step": 6542 }, { "epoch": 0.82, "learning_rate": 4.182542885261581e-05, "loss": 1.2951, "step": 6543 }, { "epoch": 0.82, "learning_rate": 4.176968982247514e-05, "loss": 1.2959, "step": 6544 }, { "epoch": 0.82, "learning_rate": 4.1713984571932737e-05, "loss": 1.2586, "step": 6545 }, { "epoch": 0.82, "learning_rate": 4.1658313110025185e-05, "loss": 1.3047, "step": 6546 }, { "epoch": 0.82, "learning_rate": 4.160267544578367e-05, "loss": 1.2078, "step": 6547 }, { "epoch": 0.82, "learning_rate": 4.154707158823384e-05, "loss": 1.2587, "step": 6548 }, { "epoch": 0.82, "learning_rate": 4.1491501546395974e-05, "loss": 1.2386, "step": 6549 }, { "epoch": 0.82, "learning_rate": 4.1435965329284673e-05, "loss": 1.2832, "step": 6550 }, { "epoch": 0.82, "learning_rate": 4.138046294590922e-05, "loss": 1.2063, "step": 6551 }, { "epoch": 0.82, "learning_rate": 4.132499440527329e-05, "loss": 1.2361, "step": 6552 }, { "epoch": 0.82, "learning_rate": 4.126955971637508e-05, "loss": 1.2818, "step": 6553 }, { "epoch": 0.82, "learning_rate": 4.121415888820737e-05, "loss": 1.2546, "step": 6554 }, { "epoch": 0.82, "learning_rate": 4.115879192975747e-05, "loss": 1.2908, "step": 6555 }, { "epoch": 0.82, "learning_rate": 4.1103458850007106e-05, "loss": 1.2471, "step": 6556 }, { "epoch": 0.82, "learning_rate": 4.104815965793249e-05, "loss": 1.2355, "step": 6557 }, { "epoch": 0.82, "learning_rate": 4.099289436250436e-05, "loss": 1.2531, "step": 6558 }, { "epoch": 0.82, "learning_rate": 4.0937662972688085e-05, "loss": 1.2594, "step": 6559 }, { "epoch": 0.82, "learning_rate": 4.088246549744332e-05, "loss": 1.2326, "step": 6560 }, { "epoch": 0.82, "learning_rate": 4.0827301945724426e-05, "loss": 1.2945, "step": 6561 }, { "epoch": 0.82, "learning_rate": 4.077217232648003e-05, "loss": 1.2025, "step": 6562 }, { "epoch": 0.82, "learning_rate": 4.0717076648653567e-05, "loss": 1.2536, "step": 6563 }, { "epoch": 0.82, "learning_rate": 4.0662014921182655e-05, "loss": 1.2337, "step": 6564 }, { "epoch": 0.82, "learning_rate": 4.06069871529996e-05, "loss": 1.1883, "step": 6565 }, { "epoch": 0.82, "learning_rate": 4.0551993353031e-05, "loss": 1.204, "step": 6566 }, { "epoch": 0.82, "learning_rate": 4.049703353019818e-05, "loss": 1.1781, "step": 6567 }, { "epoch": 0.82, "learning_rate": 4.0442107693416884e-05, "loss": 1.2594, "step": 6568 }, { "epoch": 0.82, "learning_rate": 4.038721585159727e-05, "loss": 1.2087, "step": 6569 }, { "epoch": 0.82, "learning_rate": 4.033235801364402e-05, "loss": 1.2245, "step": 6570 }, { "epoch": 0.82, "learning_rate": 4.027753418845628e-05, "loss": 1.2403, "step": 6571 }, { "epoch": 0.82, "learning_rate": 4.022274438492765e-05, "loss": 1.172, "step": 6572 }, { "epoch": 0.82, "learning_rate": 4.016798861194632e-05, "loss": 1.2193, "step": 6573 }, { "epoch": 0.82, "learning_rate": 4.0113266878394947e-05, "loss": 1.3106, "step": 6574 }, { "epoch": 0.82, "learning_rate": 4.0058579193150535e-05, "loss": 1.2737, "step": 6575 }, { "epoch": 0.82, "learning_rate": 4.0003925565084695e-05, "loss": 1.1973, "step": 6576 }, { "epoch": 0.82, "learning_rate": 3.9949306003063376e-05, "loss": 1.2676, "step": 6577 }, { "epoch": 0.82, "learning_rate": 3.98947205159472e-05, "loss": 1.2158, "step": 6578 }, { "epoch": 0.82, "learning_rate": 3.9840169112591055e-05, "loss": 1.2234, "step": 6579 }, { "epoch": 0.82, "learning_rate": 3.9785651801844485e-05, "loss": 1.2149, "step": 6580 }, { "epoch": 0.82, "learning_rate": 3.973116859255138e-05, "loss": 1.195, "step": 6581 }, { "epoch": 0.82, "learning_rate": 3.9676719493550036e-05, "loss": 1.2516, "step": 6582 }, { "epoch": 0.82, "learning_rate": 3.962230451367349e-05, "loss": 1.3052, "step": 6583 }, { "epoch": 0.82, "learning_rate": 3.9567923661748936e-05, "loss": 1.2098, "step": 6584 }, { "epoch": 0.82, "learning_rate": 3.9513576946598155e-05, "loss": 1.1723, "step": 6585 }, { "epoch": 0.82, "learning_rate": 3.945926437703748e-05, "loss": 1.2084, "step": 6586 }, { "epoch": 0.82, "learning_rate": 3.940498596187753e-05, "loss": 1.1687, "step": 6587 }, { "epoch": 0.82, "learning_rate": 3.935074170992356e-05, "loss": 1.2365, "step": 6588 }, { "epoch": 0.82, "learning_rate": 3.929653162997518e-05, "loss": 1.2223, "step": 6589 }, { "epoch": 0.82, "learning_rate": 3.9242355730826436e-05, "loss": 1.2467, "step": 6590 }, { "epoch": 0.82, "learning_rate": 3.9188214021265796e-05, "loss": 1.2348, "step": 6591 }, { "epoch": 0.82, "learning_rate": 3.9134106510076375e-05, "loss": 1.2422, "step": 6592 }, { "epoch": 0.82, "learning_rate": 3.9080033206035623e-05, "loss": 1.1791, "step": 6593 }, { "epoch": 0.82, "learning_rate": 3.902599411791538e-05, "loss": 1.2283, "step": 6594 }, { "epoch": 0.82, "learning_rate": 3.8971989254482005e-05, "loss": 1.2768, "step": 6595 }, { "epoch": 0.82, "learning_rate": 3.891801862449629e-05, "loss": 1.177, "step": 6596 }, { "epoch": 0.82, "learning_rate": 3.8864082236713423e-05, "loss": 1.2358, "step": 6597 }, { "epoch": 0.82, "learning_rate": 3.881018009988313e-05, "loss": 1.2004, "step": 6598 }, { "epoch": 0.82, "learning_rate": 3.875631222274959e-05, "loss": 1.1858, "step": 6599 }, { "epoch": 0.82, "learning_rate": 3.870247861405135e-05, "loss": 1.1887, "step": 6600 }, { "epoch": 0.82, "eval_acceptance_rate": 0.6994252800941467, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2160714500497645, "eval_expected_tokens": 2.770089312562206, "eval_loss": 1.3499805927276611, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4425223281405514, "eval_runtime": 50.223, "eval_samples_per_second": 9.697, "eval_steps_per_second": 0.319, "step": 6600 }, { "epoch": 0.83, "learning_rate": 3.8648679282521386e-05, "loss": 1.2286, "step": 6601 }, { "epoch": 0.83, "learning_rate": 3.85949142368871e-05, "loss": 1.2688, "step": 6602 }, { "epoch": 0.83, "learning_rate": 3.8541183485870516e-05, "loss": 1.2251, "step": 6603 }, { "epoch": 0.83, "learning_rate": 3.848748703818783e-05, "loss": 1.2399, "step": 6604 }, { "epoch": 0.83, "learning_rate": 3.843382490254993e-05, "loss": 1.1374, "step": 6605 }, { "epoch": 0.83, "learning_rate": 3.8380197087661914e-05, "loss": 1.2026, "step": 6606 }, { "epoch": 0.83, "learning_rate": 3.83266036022234e-05, "loss": 1.222, "step": 6607 }, { "epoch": 0.83, "learning_rate": 3.8273044454928546e-05, "loss": 1.2162, "step": 6608 }, { "epoch": 0.83, "learning_rate": 3.821951965446577e-05, "loss": 1.1684, "step": 6609 }, { "epoch": 0.83, "learning_rate": 3.816602920951792e-05, "loss": 1.2053, "step": 6610 }, { "epoch": 0.83, "learning_rate": 3.811257312876243e-05, "loss": 1.2395, "step": 6611 }, { "epoch": 0.83, "learning_rate": 3.8059151420871104e-05, "loss": 1.186, "step": 6612 }, { "epoch": 0.83, "learning_rate": 3.800576409451006e-05, "loss": 1.2104, "step": 6613 }, { "epoch": 0.83, "learning_rate": 3.795241115833997e-05, "loss": 1.2594, "step": 6614 }, { "epoch": 0.83, "learning_rate": 3.789909262101582e-05, "loss": 1.1807, "step": 6615 }, { "epoch": 0.83, "learning_rate": 3.784580849118702e-05, "loss": 1.25, "step": 6616 }, { "epoch": 0.83, "learning_rate": 3.7792558777497485e-05, "loss": 1.1539, "step": 6617 }, { "epoch": 0.83, "learning_rate": 3.7739343488585596e-05, "loss": 1.2793, "step": 6618 }, { "epoch": 0.83, "learning_rate": 3.7686162633084e-05, "loss": 1.2243, "step": 6619 }, { "epoch": 0.83, "learning_rate": 3.763301621961979e-05, "loss": 1.2905, "step": 6620 }, { "epoch": 0.83, "learning_rate": 3.7579904256814454e-05, "loss": 1.2334, "step": 6621 }, { "epoch": 0.83, "learning_rate": 3.752682675328406e-05, "loss": 1.221, "step": 6622 }, { "epoch": 0.83, "learning_rate": 3.747378371763882e-05, "loss": 1.2273, "step": 6623 }, { "epoch": 0.83, "learning_rate": 3.7420775158483636e-05, "loss": 1.2627, "step": 6624 }, { "epoch": 0.83, "learning_rate": 3.7367801084417615e-05, "loss": 1.2128, "step": 6625 }, { "epoch": 0.83, "learning_rate": 3.731486150403429e-05, "loss": 1.1777, "step": 6626 }, { "epoch": 0.83, "learning_rate": 3.7261956425921725e-05, "loss": 1.2097, "step": 6627 }, { "epoch": 0.83, "learning_rate": 3.7209085858662284e-05, "loss": 1.2084, "step": 6628 }, { "epoch": 0.83, "learning_rate": 3.715624981083265e-05, "loss": 1.1863, "step": 6629 }, { "epoch": 0.83, "learning_rate": 3.710344829100418e-05, "loss": 1.2159, "step": 6630 }, { "epoch": 0.83, "learning_rate": 3.705068130774228e-05, "loss": 1.2383, "step": 6631 }, { "epoch": 0.83, "learning_rate": 3.699794886960706e-05, "loss": 1.2342, "step": 6632 }, { "epoch": 0.83, "learning_rate": 3.694525098515286e-05, "loss": 1.2003, "step": 6633 }, { "epoch": 0.83, "learning_rate": 3.689258766292847e-05, "loss": 1.2361, "step": 6634 }, { "epoch": 0.83, "learning_rate": 3.6839958911476953e-05, "loss": 1.2327, "step": 6635 }, { "epoch": 0.83, "learning_rate": 3.678736473933594e-05, "loss": 1.23, "step": 6636 }, { "epoch": 0.83, "learning_rate": 3.673480515503747e-05, "loss": 1.1735, "step": 6637 }, { "epoch": 0.83, "learning_rate": 3.6682280167107754e-05, "loss": 1.1771, "step": 6638 }, { "epoch": 0.83, "learning_rate": 3.6629789784067557e-05, "loss": 1.2861, "step": 6639 }, { "epoch": 0.83, "learning_rate": 3.6577334014432004e-05, "loss": 1.2272, "step": 6640 }, { "epoch": 0.83, "learning_rate": 3.6524912866710505e-05, "loss": 1.2413, "step": 6641 }, { "epoch": 0.83, "learning_rate": 3.6472526349406984e-05, "loss": 1.2233, "step": 6642 }, { "epoch": 0.83, "learning_rate": 3.6420174471019805e-05, "loss": 1.2114, "step": 6643 }, { "epoch": 0.83, "learning_rate": 3.636785724004152e-05, "loss": 1.2051, "step": 6644 }, { "epoch": 0.83, "learning_rate": 3.631557466495916e-05, "loss": 1.1945, "step": 6645 }, { "epoch": 0.83, "learning_rate": 3.6263326754254034e-05, "loss": 1.2493, "step": 6646 }, { "epoch": 0.83, "learning_rate": 3.6211113516402105e-05, "loss": 1.2071, "step": 6647 }, { "epoch": 0.83, "learning_rate": 3.615893495987335e-05, "loss": 1.2135, "step": 6648 }, { "epoch": 0.83, "learning_rate": 3.610679109313242e-05, "loss": 1.1952, "step": 6649 }, { "epoch": 0.83, "learning_rate": 3.605468192463815e-05, "loss": 1.1631, "step": 6650 }, { "epoch": 0.83, "learning_rate": 3.6002607462843785e-05, "loss": 1.286, "step": 6651 }, { "epoch": 0.83, "learning_rate": 3.595056771619704e-05, "loss": 1.1934, "step": 6652 }, { "epoch": 0.83, "learning_rate": 3.58985626931399e-05, "loss": 1.2141, "step": 6653 }, { "epoch": 0.83, "learning_rate": 3.584659240210863e-05, "loss": 1.2135, "step": 6654 }, { "epoch": 0.83, "learning_rate": 3.579465685153416e-05, "loss": 1.1979, "step": 6655 }, { "epoch": 0.83, "learning_rate": 3.5742756049841396e-05, "loss": 1.1856, "step": 6656 }, { "epoch": 0.83, "learning_rate": 3.569089000544998e-05, "loss": 1.2083, "step": 6657 }, { "epoch": 0.83, "learning_rate": 3.563905872677364e-05, "loss": 1.2304, "step": 6658 }, { "epoch": 0.83, "learning_rate": 3.558726222222056e-05, "loss": 1.2032, "step": 6659 }, { "epoch": 0.83, "learning_rate": 3.553550050019336e-05, "loss": 1.2259, "step": 6660 }, { "epoch": 0.83, "learning_rate": 3.548377356908886e-05, "loss": 1.2157, "step": 6661 }, { "epoch": 0.83, "learning_rate": 3.5432081437298395e-05, "loss": 1.2228, "step": 6662 }, { "epoch": 0.83, "learning_rate": 3.538042411320755e-05, "loss": 1.2401, "step": 6663 }, { "epoch": 0.83, "learning_rate": 3.53288016051963e-05, "loss": 1.2269, "step": 6664 }, { "epoch": 0.83, "learning_rate": 3.527721392163891e-05, "loss": 1.1385, "step": 6665 }, { "epoch": 0.83, "learning_rate": 3.522566107090408e-05, "loss": 1.1825, "step": 6666 }, { "epoch": 0.83, "learning_rate": 3.517414306135491e-05, "loss": 1.2268, "step": 6667 }, { "epoch": 0.83, "learning_rate": 3.512265990134872e-05, "loss": 1.1715, "step": 6668 }, { "epoch": 0.83, "learning_rate": 3.5071211599237234e-05, "loss": 1.2596, "step": 6669 }, { "epoch": 0.83, "learning_rate": 3.5019798163366426e-05, "loss": 1.2074, "step": 6670 }, { "epoch": 0.83, "learning_rate": 3.4968419602076834e-05, "loss": 1.2436, "step": 6671 }, { "epoch": 0.83, "learning_rate": 3.491707592370308e-05, "loss": 1.2701, "step": 6672 }, { "epoch": 0.83, "learning_rate": 3.4865767136574383e-05, "loss": 1.1855, "step": 6673 }, { "epoch": 0.83, "learning_rate": 3.481449324901412e-05, "loss": 1.2362, "step": 6674 }, { "epoch": 0.83, "learning_rate": 3.4763254269339964e-05, "loss": 1.2316, "step": 6675 }, { "epoch": 0.83, "learning_rate": 3.471205020586418e-05, "loss": 1.1171, "step": 6676 }, { "epoch": 0.83, "learning_rate": 3.466088106689313e-05, "loss": 1.2227, "step": 6677 }, { "epoch": 0.83, "learning_rate": 3.460974686072751e-05, "loss": 1.2437, "step": 6678 }, { "epoch": 0.83, "learning_rate": 3.45586475956626e-05, "loss": 1.2163, "step": 6679 }, { "epoch": 0.83, "learning_rate": 3.4507583279987684e-05, "loss": 1.1582, "step": 6680 }, { "epoch": 0.84, "learning_rate": 3.4456553921986625e-05, "loss": 1.2715, "step": 6681 }, { "epoch": 0.84, "learning_rate": 3.440555952993751e-05, "loss": 1.2204, "step": 6682 }, { "epoch": 0.84, "learning_rate": 3.435460011211275e-05, "loss": 1.2565, "step": 6683 }, { "epoch": 0.84, "learning_rate": 3.4303675676779054e-05, "loss": 1.2081, "step": 6684 }, { "epoch": 0.84, "learning_rate": 3.425278623219752e-05, "loss": 1.2288, "step": 6685 }, { "epoch": 0.84, "learning_rate": 3.420193178662365e-05, "loss": 1.2192, "step": 6686 }, { "epoch": 0.84, "learning_rate": 3.415111234830709e-05, "loss": 1.137, "step": 6687 }, { "epoch": 0.84, "learning_rate": 3.410032792549189e-05, "loss": 1.209, "step": 6688 }, { "epoch": 0.84, "learning_rate": 3.404957852641641e-05, "loss": 1.1956, "step": 6689 }, { "epoch": 0.84, "learning_rate": 3.39988641593133e-05, "loss": 1.2324, "step": 6690 }, { "epoch": 0.84, "learning_rate": 3.394818483240958e-05, "loss": 1.2355, "step": 6691 }, { "epoch": 0.84, "learning_rate": 3.389754055392666e-05, "loss": 1.19, "step": 6692 }, { "epoch": 0.84, "learning_rate": 3.384693133208008e-05, "loss": 1.2919, "step": 6693 }, { "epoch": 0.84, "learning_rate": 3.379635717507981e-05, "loss": 1.3052, "step": 6694 }, { "epoch": 0.84, "learning_rate": 3.374581809113003e-05, "loss": 1.2028, "step": 6695 }, { "epoch": 0.84, "learning_rate": 3.369531408842946e-05, "loss": 1.1575, "step": 6696 }, { "epoch": 0.84, "learning_rate": 3.36448451751708e-05, "loss": 1.2402, "step": 6697 }, { "epoch": 0.84, "learning_rate": 3.359441135954139e-05, "loss": 1.2223, "step": 6698 }, { "epoch": 0.84, "learning_rate": 3.354401264972265e-05, "loss": 1.1751, "step": 6699 }, { "epoch": 0.84, "learning_rate": 3.3493649053890325e-05, "loss": 1.2434, "step": 6700 }, { "epoch": 0.84, "eval_acceptance_rate": 0.6996253132820129, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2169093919846943, "eval_expected_tokens": 2.771136739980868, "eval_loss": 1.3498731851577759, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.442784184995217, "eval_runtime": 50.4298, "eval_samples_per_second": 9.657, "eval_steps_per_second": 0.317, "step": 6700 }, { "epoch": 0.84, "learning_rate": 3.344332058021463e-05, "loss": 1.2673, "step": 6701 }, { "epoch": 0.84, "learning_rate": 3.3393027236859886e-05, "loss": 1.191, "step": 6702 }, { "epoch": 0.84, "learning_rate": 3.334276903198474e-05, "loss": 1.2616, "step": 6703 }, { "epoch": 0.84, "learning_rate": 3.329254597374232e-05, "loss": 1.1993, "step": 6704 }, { "epoch": 0.84, "learning_rate": 3.324235807027981e-05, "loss": 1.2426, "step": 6705 }, { "epoch": 0.84, "learning_rate": 3.3192205329738906e-05, "loss": 1.157, "step": 6706 }, { "epoch": 0.84, "learning_rate": 3.3142087760255486e-05, "loss": 1.1854, "step": 6707 }, { "epoch": 0.84, "learning_rate": 3.309200536995968e-05, "loss": 1.1865, "step": 6708 }, { "epoch": 0.84, "learning_rate": 3.304195816697597e-05, "loss": 1.1834, "step": 6709 }, { "epoch": 0.84, "learning_rate": 3.299194615942314e-05, "loss": 1.2289, "step": 6710 }, { "epoch": 0.84, "learning_rate": 3.29419693554143e-05, "loss": 1.2313, "step": 6711 }, { "epoch": 0.84, "learning_rate": 3.28920277630568e-05, "loss": 1.1901, "step": 6712 }, { "epoch": 0.84, "learning_rate": 3.284212139045223e-05, "loss": 1.2276, "step": 6713 }, { "epoch": 0.84, "learning_rate": 3.2792250245696486e-05, "loss": 1.2335, "step": 6714 }, { "epoch": 0.84, "learning_rate": 3.2742414336879855e-05, "loss": 1.1959, "step": 6715 }, { "epoch": 0.84, "learning_rate": 3.2692613672086784e-05, "loss": 1.2418, "step": 6716 }, { "epoch": 0.84, "learning_rate": 3.264284825939612e-05, "loss": 1.1967, "step": 6717 }, { "epoch": 0.84, "learning_rate": 3.259311810688087e-05, "loss": 1.2208, "step": 6718 }, { "epoch": 0.84, "learning_rate": 3.254342322260831e-05, "loss": 1.1999, "step": 6719 }, { "epoch": 0.84, "learning_rate": 3.249376361464021e-05, "loss": 1.2273, "step": 6720 }, { "epoch": 0.84, "learning_rate": 3.244413929103235e-05, "loss": 1.2953, "step": 6721 }, { "epoch": 0.84, "learning_rate": 3.23945502598349e-05, "loss": 1.2027, "step": 6722 }, { "epoch": 0.84, "learning_rate": 3.2344996529092366e-05, "loss": 1.1674, "step": 6723 }, { "epoch": 0.84, "learning_rate": 3.229547810684341e-05, "loss": 1.2033, "step": 6724 }, { "epoch": 0.84, "learning_rate": 3.22459950011211e-05, "loss": 1.1694, "step": 6725 }, { "epoch": 0.84, "learning_rate": 3.219654721995266e-05, "loss": 1.2007, "step": 6726 }, { "epoch": 0.84, "learning_rate": 3.214713477135961e-05, "loss": 1.2755, "step": 6727 }, { "epoch": 0.84, "learning_rate": 3.209775766335771e-05, "loss": 1.2386, "step": 6728 }, { "epoch": 0.84, "learning_rate": 3.2048415903957075e-05, "loss": 1.2038, "step": 6729 }, { "epoch": 0.84, "learning_rate": 3.199910950116208e-05, "loss": 1.2194, "step": 6730 }, { "epoch": 0.84, "learning_rate": 3.194983846297128e-05, "loss": 1.2426, "step": 6731 }, { "epoch": 0.84, "learning_rate": 3.190060279737753e-05, "loss": 1.1748, "step": 6732 }, { "epoch": 0.84, "learning_rate": 3.1851402512367984e-05, "loss": 1.236, "step": 6733 }, { "epoch": 0.84, "learning_rate": 3.1802237615923904e-05, "loss": 1.1863, "step": 6734 }, { "epoch": 0.84, "learning_rate": 3.175310811602106e-05, "loss": 1.2401, "step": 6735 }, { "epoch": 0.84, "learning_rate": 3.1704014020629364e-05, "loss": 1.203, "step": 6736 }, { "epoch": 0.84, "learning_rate": 3.165495533771293e-05, "loss": 1.2057, "step": 6737 }, { "epoch": 0.84, "learning_rate": 3.160593207523019e-05, "loss": 1.2476, "step": 6738 }, { "epoch": 0.84, "learning_rate": 3.15569442411337e-05, "loss": 1.1893, "step": 6739 }, { "epoch": 0.84, "learning_rate": 3.1507991843370525e-05, "loss": 1.1299, "step": 6740 }, { "epoch": 0.84, "learning_rate": 3.1459074889881756e-05, "loss": 1.216, "step": 6741 }, { "epoch": 0.84, "learning_rate": 3.141019338860285e-05, "loss": 1.2313, "step": 6742 }, { "epoch": 0.84, "learning_rate": 3.136134734746349e-05, "loss": 1.2375, "step": 6743 }, { "epoch": 0.84, "learning_rate": 3.131253677438747e-05, "loss": 1.2694, "step": 6744 }, { "epoch": 0.84, "learning_rate": 3.126376167729314e-05, "loss": 1.252, "step": 6745 }, { "epoch": 0.84, "learning_rate": 3.121502206409279e-05, "loss": 1.2219, "step": 6746 }, { "epoch": 0.84, "learning_rate": 3.116631794269306e-05, "loss": 1.2029, "step": 6747 }, { "epoch": 0.84, "learning_rate": 3.111764932099492e-05, "loss": 1.2403, "step": 6748 }, { "epoch": 0.84, "learning_rate": 3.106901620689342e-05, "loss": 1.1698, "step": 6749 }, { "epoch": 0.84, "learning_rate": 3.102041860827804e-05, "loss": 1.1875, "step": 6750 }, { "epoch": 0.84, "learning_rate": 3.097185653303233e-05, "loss": 1.1858, "step": 6751 }, { "epoch": 0.84, "learning_rate": 3.092332998903416e-05, "loss": 1.1928, "step": 6752 }, { "epoch": 0.84, "learning_rate": 3.087483898415552e-05, "loss": 1.2662, "step": 6753 }, { "epoch": 0.84, "learning_rate": 3.0826383526262824e-05, "loss": 1.1976, "step": 6754 }, { "epoch": 0.84, "learning_rate": 3.077796362321669e-05, "loss": 1.2039, "step": 6755 }, { "epoch": 0.84, "learning_rate": 3.072957928287182e-05, "loss": 1.2228, "step": 6756 }, { "epoch": 0.84, "learning_rate": 3.0681230513077254e-05, "loss": 1.239, "step": 6757 }, { "epoch": 0.84, "learning_rate": 3.0632917321676194e-05, "loss": 1.2005, "step": 6758 }, { "epoch": 0.84, "learning_rate": 3.0584639716506185e-05, "loss": 1.2431, "step": 6759 }, { "epoch": 0.84, "learning_rate": 3.053639770539885e-05, "loss": 1.2771, "step": 6760 }, { "epoch": 0.85, "learning_rate": 3.0488191296180245e-05, "loss": 1.1741, "step": 6761 }, { "epoch": 0.85, "learning_rate": 3.044002049667041e-05, "loss": 1.2583, "step": 6762 }, { "epoch": 0.85, "learning_rate": 3.039188531468376e-05, "loss": 1.2299, "step": 6763 }, { "epoch": 0.85, "learning_rate": 3.0343785758028902e-05, "loss": 1.2344, "step": 6764 }, { "epoch": 0.85, "learning_rate": 3.0295721834508684e-05, "loss": 1.2185, "step": 6765 }, { "epoch": 0.85, "learning_rate": 3.0247693551920057e-05, "loss": 1.2177, "step": 6766 }, { "epoch": 0.85, "learning_rate": 3.0199700918054378e-05, "loss": 1.2684, "step": 6767 }, { "epoch": 0.85, "learning_rate": 3.015174394069703e-05, "loss": 1.2231, "step": 6768 }, { "epoch": 0.85, "learning_rate": 3.0103822627627814e-05, "loss": 1.2103, "step": 6769 }, { "epoch": 0.85, "learning_rate": 3.005593698662057e-05, "loss": 1.239, "step": 6770 }, { "epoch": 0.85, "learning_rate": 3.0008087025443416e-05, "loss": 1.1473, "step": 6771 }, { "epoch": 0.85, "learning_rate": 2.9960272751858613e-05, "loss": 1.1867, "step": 6772 }, { "epoch": 0.85, "learning_rate": 2.9912494173622796e-05, "loss": 1.2274, "step": 6773 }, { "epoch": 0.85, "learning_rate": 2.986475129848676e-05, "loss": 1.1362, "step": 6774 }, { "epoch": 0.85, "learning_rate": 2.9817044134195392e-05, "loss": 1.223, "step": 6775 }, { "epoch": 0.85, "learning_rate": 2.9769372688487868e-05, "loss": 1.1858, "step": 6776 }, { "epoch": 0.85, "learning_rate": 2.9721736969097553e-05, "loss": 1.2316, "step": 6777 }, { "epoch": 0.85, "learning_rate": 2.967413698375196e-05, "loss": 1.2016, "step": 6778 }, { "epoch": 0.85, "learning_rate": 2.962657274017294e-05, "loss": 1.1821, "step": 6779 }, { "epoch": 0.85, "learning_rate": 2.9579044246076518e-05, "loss": 1.2198, "step": 6780 }, { "epoch": 0.85, "learning_rate": 2.9531551509172832e-05, "loss": 1.1251, "step": 6781 }, { "epoch": 0.85, "learning_rate": 2.9484094537166274e-05, "loss": 1.2192, "step": 6782 }, { "epoch": 0.85, "learning_rate": 2.943667333775532e-05, "loss": 1.1807, "step": 6783 }, { "epoch": 0.85, "learning_rate": 2.9389287918632877e-05, "loss": 1.1565, "step": 6784 }, { "epoch": 0.85, "learning_rate": 2.9341938287485854e-05, "loss": 1.2475, "step": 6785 }, { "epoch": 0.85, "learning_rate": 2.929462445199546e-05, "loss": 1.212, "step": 6786 }, { "epoch": 0.85, "learning_rate": 2.9247346419837033e-05, "loss": 1.2622, "step": 6787 }, { "epoch": 0.85, "learning_rate": 2.920010419868005e-05, "loss": 1.2214, "step": 6788 }, { "epoch": 0.85, "learning_rate": 2.9152897796188377e-05, "loss": 1.2525, "step": 6789 }, { "epoch": 0.85, "learning_rate": 2.910572722001989e-05, "loss": 1.2342, "step": 6790 }, { "epoch": 0.85, "learning_rate": 2.9058592477826635e-05, "loss": 1.1996, "step": 6791 }, { "epoch": 0.85, "learning_rate": 2.9011493577255077e-05, "loss": 1.248, "step": 6792 }, { "epoch": 0.85, "learning_rate": 2.8964430525945524e-05, "loss": 1.199, "step": 6793 }, { "epoch": 0.85, "learning_rate": 2.891740333153284e-05, "loss": 1.1847, "step": 6794 }, { "epoch": 0.85, "learning_rate": 2.887041200164575e-05, "loss": 1.2113, "step": 6795 }, { "epoch": 0.85, "learning_rate": 2.8823456543907363e-05, "loss": 1.3273, "step": 6796 }, { "epoch": 0.85, "learning_rate": 2.8776536965934797e-05, "loss": 1.2011, "step": 6797 }, { "epoch": 0.85, "learning_rate": 2.872965327533955e-05, "loss": 1.1617, "step": 6798 }, { "epoch": 0.85, "learning_rate": 2.8682805479727236e-05, "loss": 1.2069, "step": 6799 }, { "epoch": 0.85, "learning_rate": 2.863599358669755e-05, "loss": 1.175, "step": 6800 }, { "epoch": 0.85, "eval_acceptance_rate": 0.6997743248939514, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2175338549304895, "eval_expected_tokens": 2.771917318663112, "eval_loss": 1.3499584197998047, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44297932966577797, "eval_runtime": 50.2144, "eval_samples_per_second": 9.698, "eval_steps_per_second": 0.319, "step": 6800 }, { "epoch": 0.85, "learning_rate": 2.858921760384442e-05, "loss": 1.2684, "step": 6801 }, { "epoch": 0.85, "learning_rate": 2.8542477538755967e-05, "loss": 1.2323, "step": 6802 }, { "epoch": 0.85, "learning_rate": 2.8495773399014435e-05, "loss": 1.2357, "step": 6803 }, { "epoch": 0.85, "learning_rate": 2.8449105192196318e-05, "loss": 1.2099, "step": 6804 }, { "epoch": 0.85, "learning_rate": 2.840247292587225e-05, "loss": 1.2158, "step": 6805 }, { "epoch": 0.85, "learning_rate": 2.8355876607607052e-05, "loss": 1.2634, "step": 6806 }, { "epoch": 0.85, "learning_rate": 2.8309316244959554e-05, "loss": 1.2723, "step": 6807 }, { "epoch": 0.85, "learning_rate": 2.8262791845483028e-05, "loss": 1.2563, "step": 6808 }, { "epoch": 0.85, "learning_rate": 2.8216303416724738e-05, "loss": 1.1924, "step": 6809 }, { "epoch": 0.85, "learning_rate": 2.816985096622604e-05, "loss": 1.2653, "step": 6810 }, { "epoch": 0.85, "learning_rate": 2.8123434501522688e-05, "loss": 1.2076, "step": 6811 }, { "epoch": 0.85, "learning_rate": 2.8077054030144378e-05, "loss": 1.1861, "step": 6812 }, { "epoch": 0.85, "learning_rate": 2.803070955961512e-05, "loss": 1.2137, "step": 6813 }, { "epoch": 0.85, "learning_rate": 2.798440109745301e-05, "loss": 1.2612, "step": 6814 }, { "epoch": 0.85, "learning_rate": 2.7938128651170293e-05, "loss": 1.1568, "step": 6815 }, { "epoch": 0.85, "learning_rate": 2.789189222827332e-05, "loss": 1.178, "step": 6816 }, { "epoch": 0.85, "learning_rate": 2.784569183626276e-05, "loss": 1.2005, "step": 6817 }, { "epoch": 0.85, "learning_rate": 2.7799527482633398e-05, "loss": 1.192, "step": 6818 }, { "epoch": 0.85, "learning_rate": 2.7753399174874043e-05, "loss": 1.1887, "step": 6819 }, { "epoch": 0.85, "learning_rate": 2.7707306920467744e-05, "loss": 1.2316, "step": 6820 }, { "epoch": 0.85, "learning_rate": 2.7661250726891683e-05, "loss": 1.1627, "step": 6821 }, { "epoch": 0.85, "learning_rate": 2.7615230601617163e-05, "loss": 1.2137, "step": 6822 }, { "epoch": 0.85, "learning_rate": 2.756924655210974e-05, "loss": 1.2068, "step": 6823 }, { "epoch": 0.85, "learning_rate": 2.7523298585829053e-05, "loss": 1.2366, "step": 6824 }, { "epoch": 0.85, "learning_rate": 2.7477386710228897e-05, "loss": 1.1991, "step": 6825 }, { "epoch": 0.85, "learning_rate": 2.743151093275717e-05, "loss": 1.2068, "step": 6826 }, { "epoch": 0.85, "learning_rate": 2.7385671260855893e-05, "loss": 1.2089, "step": 6827 }, { "epoch": 0.85, "learning_rate": 2.733986770196137e-05, "loss": 1.2272, "step": 6828 }, { "epoch": 0.85, "learning_rate": 2.7294100263503903e-05, "loss": 1.1831, "step": 6829 }, { "epoch": 0.85, "learning_rate": 2.7248368952908055e-05, "loss": 1.1793, "step": 6830 }, { "epoch": 0.85, "learning_rate": 2.720267377759242e-05, "loss": 1.1788, "step": 6831 }, { "epoch": 0.85, "learning_rate": 2.7157014744969732e-05, "loss": 1.2468, "step": 6832 }, { "epoch": 0.85, "learning_rate": 2.7111391862446987e-05, "loss": 1.1819, "step": 6833 }, { "epoch": 0.85, "learning_rate": 2.706580513742521e-05, "loss": 1.2298, "step": 6834 }, { "epoch": 0.85, "learning_rate": 2.7020254577299537e-05, "loss": 1.2095, "step": 6835 }, { "epoch": 0.85, "learning_rate": 2.6974740189459336e-05, "loss": 1.1958, "step": 6836 }, { "epoch": 0.85, "learning_rate": 2.6929261981288005e-05, "loss": 1.1632, "step": 6837 }, { "epoch": 0.85, "learning_rate": 2.6883819960163203e-05, "loss": 1.2367, "step": 6838 }, { "epoch": 0.85, "learning_rate": 2.6838414133456612e-05, "loss": 1.1815, "step": 6839 }, { "epoch": 0.85, "learning_rate": 2.679304450853401e-05, "loss": 1.2203, "step": 6840 }, { "epoch": 0.86, "learning_rate": 2.6747711092755368e-05, "loss": 1.1957, "step": 6841 }, { "epoch": 0.86, "learning_rate": 2.670241389347483e-05, "loss": 1.1952, "step": 6842 }, { "epoch": 0.86, "learning_rate": 2.6657152918040605e-05, "loss": 1.2264, "step": 6843 }, { "epoch": 0.86, "learning_rate": 2.6611928173795042e-05, "loss": 1.1692, "step": 6844 }, { "epoch": 0.86, "learning_rate": 2.6566739668074575e-05, "loss": 1.2061, "step": 6845 }, { "epoch": 0.86, "learning_rate": 2.6521587408209787e-05, "loss": 1.2276, "step": 6846 }, { "epoch": 0.86, "learning_rate": 2.6476471401525313e-05, "loss": 1.1842, "step": 6847 }, { "epoch": 0.86, "learning_rate": 2.6431391655340085e-05, "loss": 1.1925, "step": 6848 }, { "epoch": 0.86, "learning_rate": 2.638634817696703e-05, "loss": 1.2077, "step": 6849 }, { "epoch": 0.86, "learning_rate": 2.6341340973713186e-05, "loss": 1.2279, "step": 6850 }, { "epoch": 0.86, "learning_rate": 2.6296370052879714e-05, "loss": 1.2357, "step": 6851 }, { "epoch": 0.86, "learning_rate": 2.6251435421761865e-05, "loss": 1.1884, "step": 6852 }, { "epoch": 0.86, "learning_rate": 2.620653708764911e-05, "loss": 1.1684, "step": 6853 }, { "epoch": 0.86, "learning_rate": 2.6161675057824878e-05, "loss": 1.215, "step": 6854 }, { "epoch": 0.86, "learning_rate": 2.6116849339566873e-05, "loss": 1.1908, "step": 6855 }, { "epoch": 0.86, "learning_rate": 2.6072059940146776e-05, "loss": 1.1495, "step": 6856 }, { "epoch": 0.86, "learning_rate": 2.602730686683047e-05, "loss": 1.1421, "step": 6857 }, { "epoch": 0.86, "learning_rate": 2.598259012687787e-05, "loss": 1.1262, "step": 6858 }, { "epoch": 0.86, "learning_rate": 2.5937909727543035e-05, "loss": 1.222, "step": 6859 }, { "epoch": 0.86, "learning_rate": 2.5893265676074057e-05, "loss": 1.2113, "step": 6860 }, { "epoch": 0.86, "learning_rate": 2.584865797971328e-05, "loss": 1.2491, "step": 6861 }, { "epoch": 0.86, "learning_rate": 2.5804086645697082e-05, "loss": 1.3238, "step": 6862 }, { "epoch": 0.86, "learning_rate": 2.5759551681255876e-05, "loss": 1.2213, "step": 6863 }, { "epoch": 0.86, "learning_rate": 2.5715053093614243e-05, "loss": 1.1676, "step": 6864 }, { "epoch": 0.86, "learning_rate": 2.5670590889990854e-05, "loss": 1.2037, "step": 6865 }, { "epoch": 0.86, "learning_rate": 2.562616507759838e-05, "loss": 1.2507, "step": 6866 }, { "epoch": 0.86, "learning_rate": 2.558177566364378e-05, "loss": 1.2156, "step": 6867 }, { "epoch": 0.86, "learning_rate": 2.553742265532802e-05, "loss": 1.1753, "step": 6868 }, { "epoch": 0.86, "learning_rate": 2.5493106059846115e-05, "loss": 1.1971, "step": 6869 }, { "epoch": 0.86, "learning_rate": 2.5448825884387183e-05, "loss": 1.2323, "step": 6870 }, { "epoch": 0.86, "learning_rate": 2.5404582136134446e-05, "loss": 1.1746, "step": 6871 }, { "epoch": 0.86, "learning_rate": 2.5360374822265274e-05, "loss": 1.1827, "step": 6872 }, { "epoch": 0.86, "learning_rate": 2.5316203949951016e-05, "loss": 1.1803, "step": 6873 }, { "epoch": 0.86, "learning_rate": 2.5272069526357244e-05, "loss": 1.2314, "step": 6874 }, { "epoch": 0.86, "learning_rate": 2.5227971558643535e-05, "loss": 1.166, "step": 6875 }, { "epoch": 0.86, "learning_rate": 2.5183910053963478e-05, "loss": 1.204, "step": 6876 }, { "epoch": 0.86, "learning_rate": 2.5139885019464958e-05, "loss": 1.1639, "step": 6877 }, { "epoch": 0.86, "learning_rate": 2.5095896462289743e-05, "loss": 1.2271, "step": 6878 }, { "epoch": 0.86, "learning_rate": 2.505194438957373e-05, "loss": 1.219, "step": 6879 }, { "epoch": 0.86, "learning_rate": 2.5008028808446993e-05, "loss": 1.2497, "step": 6880 }, { "epoch": 0.86, "learning_rate": 2.496414972603353e-05, "loss": 1.2043, "step": 6881 }, { "epoch": 0.86, "learning_rate": 2.492030714945162e-05, "loss": 1.2397, "step": 6882 }, { "epoch": 0.86, "learning_rate": 2.4876501085813464e-05, "loss": 1.1871, "step": 6883 }, { "epoch": 0.86, "learning_rate": 2.483273154222537e-05, "loss": 1.2277, "step": 6884 }, { "epoch": 0.86, "learning_rate": 2.478899852578767e-05, "loss": 1.1867, "step": 6885 }, { "epoch": 0.86, "learning_rate": 2.4745302043594875e-05, "loss": 1.2272, "step": 6886 }, { "epoch": 0.86, "learning_rate": 2.4701642102735622e-05, "loss": 1.192, "step": 6887 }, { "epoch": 0.86, "learning_rate": 2.4658018710292467e-05, "loss": 1.2436, "step": 6888 }, { "epoch": 0.86, "learning_rate": 2.461443187334206e-05, "loss": 1.1911, "step": 6889 }, { "epoch": 0.86, "learning_rate": 2.457088159895518e-05, "loss": 1.2424, "step": 6890 }, { "epoch": 0.86, "learning_rate": 2.4527367894196627e-05, "loss": 1.2348, "step": 6891 }, { "epoch": 0.86, "learning_rate": 2.448389076612531e-05, "loss": 1.1881, "step": 6892 }, { "epoch": 0.86, "learning_rate": 2.4440450221794253e-05, "loss": 1.1902, "step": 6893 }, { "epoch": 0.86, "learning_rate": 2.43970462682504e-05, "loss": 1.2192, "step": 6894 }, { "epoch": 0.86, "learning_rate": 2.43536789125349e-05, "loss": 1.2163, "step": 6895 }, { "epoch": 0.86, "learning_rate": 2.431034816168279e-05, "loss": 1.1876, "step": 6896 }, { "epoch": 0.86, "learning_rate": 2.4267054022723445e-05, "loss": 1.2297, "step": 6897 }, { "epoch": 0.86, "learning_rate": 2.4223796502680003e-05, "loss": 1.1359, "step": 6898 }, { "epoch": 0.86, "learning_rate": 2.4180575608569898e-05, "loss": 1.2031, "step": 6899 }, { "epoch": 0.86, "learning_rate": 2.4137391347404474e-05, "loss": 1.2201, "step": 6900 }, { "epoch": 0.86, "eval_acceptance_rate": 0.6997644305229187, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2174923839432465, "eval_expected_tokens": 2.7718654799290583, "eval_loss": 1.3499109745025635, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4429663699822644, "eval_runtime": 50.235, "eval_samples_per_second": 9.694, "eval_steps_per_second": 0.319, "step": 6900 }, { "epoch": 0.86, "learning_rate": 2.4094243726189123e-05, "loss": 1.2284, "step": 6901 }, { "epoch": 0.86, "learning_rate": 2.40511327519235e-05, "loss": 1.2211, "step": 6902 }, { "epoch": 0.86, "learning_rate": 2.4008058431601003e-05, "loss": 1.2164, "step": 6903 }, { "epoch": 0.86, "learning_rate": 2.3965020772209384e-05, "loss": 1.2915, "step": 6904 }, { "epoch": 0.86, "learning_rate": 2.392201978073019e-05, "loss": 1.2117, "step": 6905 }, { "epoch": 0.86, "learning_rate": 2.3879055464139266e-05, "loss": 1.2055, "step": 6906 }, { "epoch": 0.86, "learning_rate": 2.3836127829406308e-05, "loss": 1.1986, "step": 6907 }, { "epoch": 0.86, "learning_rate": 2.379323688349516e-05, "loss": 1.1838, "step": 6908 }, { "epoch": 0.86, "learning_rate": 2.3750382633363597e-05, "loss": 1.1907, "step": 6909 }, { "epoch": 0.86, "learning_rate": 2.3707565085963667e-05, "loss": 1.2129, "step": 6910 }, { "epoch": 0.86, "learning_rate": 2.3664784248241254e-05, "loss": 1.252, "step": 6911 }, { "epoch": 0.86, "learning_rate": 2.3622040127136397e-05, "loss": 1.2369, "step": 6912 }, { "epoch": 0.86, "learning_rate": 2.3579332729583153e-05, "loss": 1.2633, "step": 6913 }, { "epoch": 0.86, "learning_rate": 2.3536662062509595e-05, "loss": 1.242, "step": 6914 }, { "epoch": 0.86, "learning_rate": 2.3494028132837798e-05, "loss": 1.1993, "step": 6915 }, { "epoch": 0.86, "learning_rate": 2.3451430947483976e-05, "loss": 1.2915, "step": 6916 }, { "epoch": 0.86, "learning_rate": 2.340887051335844e-05, "loss": 1.2171, "step": 6917 }, { "epoch": 0.86, "learning_rate": 2.3366346837365326e-05, "loss": 1.2515, "step": 6918 }, { "epoch": 0.86, "learning_rate": 2.3323859926402984e-05, "loss": 1.2633, "step": 6919 }, { "epoch": 0.86, "learning_rate": 2.328140978736365e-05, "loss": 1.2391, "step": 6920 }, { "epoch": 0.87, "learning_rate": 2.3238996427133783e-05, "loss": 1.1695, "step": 6921 }, { "epoch": 0.87, "learning_rate": 2.319661985259369e-05, "loss": 1.1926, "step": 6922 }, { "epoch": 0.87, "learning_rate": 2.31542800706179e-05, "loss": 1.1491, "step": 6923 }, { "epoch": 0.87, "learning_rate": 2.3111977088074803e-05, "loss": 1.2185, "step": 6924 }, { "epoch": 0.87, "learning_rate": 2.306971091182686e-05, "loss": 1.2273, "step": 6925 }, { "epoch": 0.87, "learning_rate": 2.3027481548730634e-05, "loss": 1.2428, "step": 6926 }, { "epoch": 0.87, "learning_rate": 2.2985289005636677e-05, "loss": 1.224, "step": 6927 }, { "epoch": 0.87, "learning_rate": 2.294313328938949e-05, "loss": 1.2209, "step": 6928 }, { "epoch": 0.87, "learning_rate": 2.2901014406827764e-05, "loss": 1.2199, "step": 6929 }, { "epoch": 0.87, "learning_rate": 2.2858932364784036e-05, "loss": 1.1854, "step": 6930 }, { "epoch": 0.87, "learning_rate": 2.281688717008504e-05, "loss": 1.1368, "step": 6931 }, { "epoch": 0.87, "learning_rate": 2.2774878829551398e-05, "loss": 1.1966, "step": 6932 }, { "epoch": 0.87, "learning_rate": 2.2732907349997778e-05, "loss": 1.1733, "step": 6933 }, { "epoch": 0.87, "learning_rate": 2.2690972738232866e-05, "loss": 1.247, "step": 6934 }, { "epoch": 0.87, "learning_rate": 2.264907500105945e-05, "loss": 1.2928, "step": 6935 }, { "epoch": 0.87, "learning_rate": 2.260721414527428e-05, "loss": 1.202, "step": 6936 }, { "epoch": 0.87, "learning_rate": 2.256539017766812e-05, "loss": 1.2492, "step": 6937 }, { "epoch": 0.87, "learning_rate": 2.2523603105025713e-05, "loss": 1.2369, "step": 6938 }, { "epoch": 0.87, "learning_rate": 2.2481852934125886e-05, "loss": 1.2043, "step": 6939 }, { "epoch": 0.87, "learning_rate": 2.2440139671741355e-05, "loss": 1.1933, "step": 6940 }, { "epoch": 0.87, "learning_rate": 2.2398463324639046e-05, "loss": 1.2413, "step": 6941 }, { "epoch": 0.87, "learning_rate": 2.2356823899579797e-05, "loss": 1.189, "step": 6942 }, { "epoch": 0.87, "learning_rate": 2.231522140331843e-05, "loss": 1.2235, "step": 6943 }, { "epoch": 0.87, "learning_rate": 2.2273655842603768e-05, "loss": 1.1993, "step": 6944 }, { "epoch": 0.87, "learning_rate": 2.2232127224178637e-05, "loss": 1.2199, "step": 6945 }, { "epoch": 0.87, "learning_rate": 2.2190635554780014e-05, "loss": 1.2157, "step": 6946 }, { "epoch": 0.87, "learning_rate": 2.214918084113868e-05, "loss": 1.1848, "step": 6947 }, { "epoch": 0.87, "learning_rate": 2.2107763089979582e-05, "loss": 1.2171, "step": 6948 }, { "epoch": 0.87, "learning_rate": 2.2066382308021576e-05, "loss": 1.1798, "step": 6949 }, { "epoch": 0.87, "learning_rate": 2.2025038501977485e-05, "loss": 1.2258, "step": 6950 }, { "epoch": 0.87, "learning_rate": 2.1983731678554302e-05, "loss": 1.2264, "step": 6951 }, { "epoch": 0.87, "learning_rate": 2.194246184445284e-05, "loss": 1.1935, "step": 6952 }, { "epoch": 0.87, "learning_rate": 2.1901229006367986e-05, "loss": 1.2342, "step": 6953 }, { "epoch": 0.87, "learning_rate": 2.186003317098864e-05, "loss": 1.1965, "step": 6954 }, { "epoch": 0.87, "learning_rate": 2.181887434499774e-05, "loss": 1.3153, "step": 6955 }, { "epoch": 0.87, "learning_rate": 2.1777752535072138e-05, "loss": 1.2191, "step": 6956 }, { "epoch": 0.87, "learning_rate": 2.173666774788266e-05, "loss": 1.2642, "step": 6957 }, { "epoch": 0.87, "learning_rate": 2.169561999009423e-05, "loss": 1.1294, "step": 6958 }, { "epoch": 0.87, "learning_rate": 2.1654609268365627e-05, "loss": 1.211, "step": 6959 }, { "epoch": 0.87, "learning_rate": 2.1613635589349755e-05, "loss": 1.209, "step": 6960 }, { "epoch": 0.87, "learning_rate": 2.1572698959693522e-05, "loss": 1.1623, "step": 6961 }, { "epoch": 0.87, "learning_rate": 2.153179938603769e-05, "loss": 1.1818, "step": 6962 }, { "epoch": 0.87, "learning_rate": 2.1490936875017103e-05, "loss": 1.2442, "step": 6963 }, { "epoch": 0.87, "learning_rate": 2.1450111433260507e-05, "loss": 1.2262, "step": 6964 }, { "epoch": 0.87, "learning_rate": 2.1409323067390797e-05, "loss": 1.237, "step": 6965 }, { "epoch": 0.87, "learning_rate": 2.136857178402468e-05, "loss": 1.224, "step": 6966 }, { "epoch": 0.87, "learning_rate": 2.1327857589772987e-05, "loss": 1.2207, "step": 6967 }, { "epoch": 0.87, "learning_rate": 2.1287180491240455e-05, "loss": 1.2133, "step": 6968 }, { "epoch": 0.87, "learning_rate": 2.124654049502575e-05, "loss": 1.2771, "step": 6969 }, { "epoch": 0.87, "learning_rate": 2.1205937607721658e-05, "loss": 1.2537, "step": 6970 }, { "epoch": 0.87, "learning_rate": 2.116537183591488e-05, "loss": 1.1913, "step": 6971 }, { "epoch": 0.87, "learning_rate": 2.112484318618599e-05, "loss": 1.2073, "step": 6972 }, { "epoch": 0.87, "learning_rate": 2.1084351665109745e-05, "loss": 1.2774, "step": 6973 }, { "epoch": 0.87, "learning_rate": 2.104389727925471e-05, "loss": 1.25, "step": 6974 }, { "epoch": 0.87, "learning_rate": 2.1003480035183514e-05, "loss": 1.2336, "step": 6975 }, { "epoch": 0.87, "learning_rate": 2.096309993945278e-05, "loss": 1.2142, "step": 6976 }, { "epoch": 0.87, "learning_rate": 2.0922756998612985e-05, "loss": 1.2815, "step": 6977 }, { "epoch": 0.87, "learning_rate": 2.088245121920862e-05, "loss": 1.1704, "step": 6978 }, { "epoch": 0.87, "learning_rate": 2.084218260777826e-05, "loss": 1.2103, "step": 6979 }, { "epoch": 0.87, "learning_rate": 2.08019511708544e-05, "loss": 1.2318, "step": 6980 }, { "epoch": 0.87, "learning_rate": 2.0761756914963402e-05, "loss": 1.2737, "step": 6981 }, { "epoch": 0.87, "learning_rate": 2.0721599846625723e-05, "loss": 1.2331, "step": 6982 }, { "epoch": 0.87, "learning_rate": 2.0681479972355698e-05, "loss": 1.1675, "step": 6983 }, { "epoch": 0.87, "learning_rate": 2.064139729866163e-05, "loss": 1.2268, "step": 6984 }, { "epoch": 0.87, "learning_rate": 2.0601351832045866e-05, "loss": 1.1484, "step": 6985 }, { "epoch": 0.87, "learning_rate": 2.0561343579004716e-05, "loss": 1.2621, "step": 6986 }, { "epoch": 0.87, "learning_rate": 2.052137254602837e-05, "loss": 1.2876, "step": 6987 }, { "epoch": 0.87, "learning_rate": 2.048143873960101e-05, "loss": 1.2587, "step": 6988 }, { "epoch": 0.87, "learning_rate": 2.0441542166200782e-05, "loss": 1.2047, "step": 6989 }, { "epoch": 0.87, "learning_rate": 2.0401682832299845e-05, "loss": 1.2411, "step": 6990 }, { "epoch": 0.87, "learning_rate": 2.0361860744364192e-05, "loss": 1.235, "step": 6991 }, { "epoch": 0.87, "learning_rate": 2.0322075908853933e-05, "loss": 1.2063, "step": 6992 }, { "epoch": 0.87, "learning_rate": 2.028232833222307e-05, "loss": 1.2043, "step": 6993 }, { "epoch": 0.87, "learning_rate": 2.0242618020919424e-05, "loss": 1.2369, "step": 6994 }, { "epoch": 0.87, "learning_rate": 2.020294498138503e-05, "loss": 1.2513, "step": 6995 }, { "epoch": 0.87, "learning_rate": 2.0163309220055658e-05, "loss": 1.2029, "step": 6996 }, { "epoch": 0.87, "learning_rate": 2.012371074336111e-05, "loss": 1.2333, "step": 6997 }, { "epoch": 0.87, "learning_rate": 2.0084149557725223e-05, "loss": 1.2364, "step": 6998 }, { "epoch": 0.87, "learning_rate": 2.0044625669565582e-05, "loss": 1.1918, "step": 6999 }, { "epoch": 0.88, "learning_rate": 2.0005139085293944e-05, "loss": 1.1579, "step": 7000 }, { "epoch": 0.88, "eval_acceptance_rate": 0.699770450592041, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.217517616178256, "eval_expected_tokens": 2.77189702022282, "eval_loss": 1.3496931791305542, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4429742550557049, "eval_runtime": 50.2042, "eval_samples_per_second": 9.7, "eval_steps_per_second": 0.319, "step": 7000 }, { "epoch": 0.88, "learning_rate": 1.9965689811315913e-05, "loss": 1.2876, "step": 7001 }, { "epoch": 0.88, "learning_rate": 1.9926277854030977e-05, "loss": 1.2918, "step": 7002 }, { "epoch": 0.88, "learning_rate": 1.9886903219832608e-05, "loss": 1.2274, "step": 7003 }, { "epoch": 0.88, "learning_rate": 1.9847565915108308e-05, "loss": 1.2148, "step": 7004 }, { "epoch": 0.88, "learning_rate": 1.9808265946239522e-05, "loss": 1.239, "step": 7005 }, { "epoch": 0.88, "learning_rate": 1.976900331960152e-05, "loss": 1.2326, "step": 7006 }, { "epoch": 0.88, "learning_rate": 1.9729778041563533e-05, "loss": 1.2123, "step": 7007 }, { "epoch": 0.88, "learning_rate": 1.969059011848881e-05, "loss": 1.211, "step": 7008 }, { "epoch": 0.88, "learning_rate": 1.965143955673454e-05, "loss": 1.2245, "step": 7009 }, { "epoch": 0.88, "learning_rate": 1.9612326362651755e-05, "loss": 1.2163, "step": 7010 }, { "epoch": 0.88, "learning_rate": 1.9573250542585573e-05, "loss": 1.1978, "step": 7011 }, { "epoch": 0.88, "learning_rate": 1.95342121028749e-05, "loss": 1.2628, "step": 7012 }, { "epoch": 0.88, "learning_rate": 1.9495211049852608e-05, "loss": 1.1824, "step": 7013 }, { "epoch": 0.88, "learning_rate": 1.945624738984561e-05, "loss": 1.2701, "step": 7014 }, { "epoch": 0.88, "learning_rate": 1.9417321129174658e-05, "loss": 1.2331, "step": 7015 }, { "epoch": 0.88, "learning_rate": 1.9378432274154422e-05, "loss": 1.2306, "step": 7016 }, { "epoch": 0.88, "learning_rate": 1.9339580831093606e-05, "loss": 1.2135, "step": 7017 }, { "epoch": 0.88, "learning_rate": 1.9300766806294722e-05, "loss": 1.2009, "step": 7018 }, { "epoch": 0.88, "learning_rate": 1.926199020605432e-05, "loss": 1.2183, "step": 7019 }, { "epoch": 0.88, "learning_rate": 1.922325103666281e-05, "loss": 1.2561, "step": 7020 }, { "epoch": 0.88, "learning_rate": 1.9184549304404525e-05, "loss": 1.2754, "step": 7021 }, { "epoch": 0.88, "learning_rate": 1.9145885015557723e-05, "loss": 1.2273, "step": 7022 }, { "epoch": 0.88, "learning_rate": 1.9107258176394666e-05, "loss": 1.2761, "step": 7023 }, { "epoch": 0.88, "learning_rate": 1.9068668793181532e-05, "loss": 1.1853, "step": 7024 }, { "epoch": 0.88, "learning_rate": 1.9030116872178316e-05, "loss": 1.2168, "step": 7025 }, { "epoch": 0.88, "learning_rate": 1.8991602419639016e-05, "loss": 1.1604, "step": 7026 }, { "epoch": 0.88, "learning_rate": 1.895312544181152e-05, "loss": 1.2314, "step": 7027 }, { "epoch": 0.88, "learning_rate": 1.8914685944937616e-05, "loss": 1.2236, "step": 7028 }, { "epoch": 0.88, "learning_rate": 1.887628393525309e-05, "loss": 1.231, "step": 7029 }, { "epoch": 0.88, "learning_rate": 1.883791941898766e-05, "loss": 1.1645, "step": 7030 }, { "epoch": 0.88, "learning_rate": 1.879959240236484e-05, "loss": 1.2111, "step": 7031 }, { "epoch": 0.88, "learning_rate": 1.8761302891602132e-05, "loss": 1.219, "step": 7032 }, { "epoch": 0.88, "learning_rate": 1.8723050892910932e-05, "loss": 1.2327, "step": 7033 }, { "epoch": 0.88, "learning_rate": 1.86848364124966e-05, "loss": 1.239, "step": 7034 }, { "epoch": 0.88, "learning_rate": 1.8646659456558358e-05, "loss": 1.1793, "step": 7035 }, { "epoch": 0.88, "learning_rate": 1.8608520031289378e-05, "loss": 1.2237, "step": 7036 }, { "epoch": 0.88, "learning_rate": 1.8570418142876715e-05, "loss": 1.1941, "step": 7037 }, { "epoch": 0.88, "learning_rate": 1.8532353797501317e-05, "loss": 1.2722, "step": 7038 }, { "epoch": 0.88, "learning_rate": 1.849432700133813e-05, "loss": 1.169, "step": 7039 }, { "epoch": 0.88, "learning_rate": 1.8456337760555914e-05, "loss": 1.2468, "step": 7040 }, { "epoch": 0.88, "learning_rate": 1.8418386081317323e-05, "loss": 1.192, "step": 7041 }, { "epoch": 0.88, "learning_rate": 1.8380471969779045e-05, "loss": 1.203, "step": 7042 }, { "epoch": 0.88, "learning_rate": 1.8342595432091546e-05, "loss": 1.2002, "step": 7043 }, { "epoch": 0.88, "learning_rate": 1.8304756474399275e-05, "loss": 1.1862, "step": 7044 }, { "epoch": 0.88, "learning_rate": 1.8266955102840544e-05, "loss": 1.2014, "step": 7045 }, { "epoch": 0.88, "learning_rate": 1.8229191323547582e-05, "loss": 1.1951, "step": 7046 }, { "epoch": 0.88, "learning_rate": 1.819146514264647e-05, "loss": 1.1496, "step": 7047 }, { "epoch": 0.88, "learning_rate": 1.8153776566257273e-05, "loss": 1.2176, "step": 7048 }, { "epoch": 0.88, "learning_rate": 1.8116125600493944e-05, "loss": 1.1769, "step": 7049 }, { "epoch": 0.88, "learning_rate": 1.8078512251464284e-05, "loss": 1.2055, "step": 7050 }, { "epoch": 0.88, "learning_rate": 1.804093652527003e-05, "loss": 1.2216, "step": 7051 }, { "epoch": 0.88, "learning_rate": 1.8003398428006747e-05, "loss": 1.2468, "step": 7052 }, { "epoch": 0.88, "learning_rate": 1.7965897965764034e-05, "loss": 1.2598, "step": 7053 }, { "epoch": 0.88, "learning_rate": 1.7928435144625223e-05, "loss": 1.1989, "step": 7054 }, { "epoch": 0.88, "learning_rate": 1.7891009970667672e-05, "loss": 1.2007, "step": 7055 }, { "epoch": 0.88, "learning_rate": 1.7853622449962582e-05, "loss": 1.1863, "step": 7056 }, { "epoch": 0.88, "learning_rate": 1.781627258857499e-05, "loss": 1.2277, "step": 7057 }, { "epoch": 0.88, "learning_rate": 1.7778960392563936e-05, "loss": 1.2435, "step": 7058 }, { "epoch": 0.88, "learning_rate": 1.774168586798225e-05, "loss": 1.1673, "step": 7059 }, { "epoch": 0.88, "learning_rate": 1.77044490208767e-05, "loss": 1.2034, "step": 7060 }, { "epoch": 0.88, "learning_rate": 1.7667249857287958e-05, "loss": 1.1694, "step": 7061 }, { "epoch": 0.88, "learning_rate": 1.7630088383250475e-05, "loss": 1.2924, "step": 7062 }, { "epoch": 0.88, "learning_rate": 1.7592964604792818e-05, "loss": 1.2117, "step": 7063 }, { "epoch": 0.88, "learning_rate": 1.7555878527937163e-05, "loss": 1.254, "step": 7064 }, { "epoch": 0.88, "learning_rate": 1.7518830158699762e-05, "loss": 1.1837, "step": 7065 }, { "epoch": 0.88, "learning_rate": 1.7481819503090606e-05, "loss": 1.1842, "step": 7066 }, { "epoch": 0.88, "learning_rate": 1.744484656711373e-05, "loss": 1.274, "step": 7067 }, { "epoch": 0.88, "learning_rate": 1.7407911356766977e-05, "loss": 1.2234, "step": 7068 }, { "epoch": 0.88, "learning_rate": 1.7371013878042047e-05, "loss": 1.2019, "step": 7069 }, { "epoch": 0.88, "learning_rate": 1.733415413692449e-05, "loss": 1.2674, "step": 7070 }, { "epoch": 0.88, "learning_rate": 1.729733213939383e-05, "loss": 1.156, "step": 7071 }, { "epoch": 0.88, "learning_rate": 1.7260547891423312e-05, "loss": 1.2381, "step": 7072 }, { "epoch": 0.88, "learning_rate": 1.7223801398980276e-05, "loss": 1.2475, "step": 7073 }, { "epoch": 0.88, "learning_rate": 1.7187092668025817e-05, "loss": 1.264, "step": 7074 }, { "epoch": 0.88, "learning_rate": 1.715042170451486e-05, "loss": 1.2111, "step": 7075 }, { "epoch": 0.88, "learning_rate": 1.7113788514396295e-05, "loss": 1.2164, "step": 7076 }, { "epoch": 0.88, "learning_rate": 1.7077193103612777e-05, "loss": 1.2398, "step": 7077 }, { "epoch": 0.88, "learning_rate": 1.7040635478100974e-05, "loss": 1.2408, "step": 7078 }, { "epoch": 0.88, "learning_rate": 1.7004115643791257e-05, "loss": 1.2526, "step": 7079 }, { "epoch": 0.89, "learning_rate": 1.696763360660808e-05, "loss": 1.2124, "step": 7080 }, { "epoch": 0.89, "learning_rate": 1.693118937246954e-05, "loss": 1.1937, "step": 7081 }, { "epoch": 0.89, "learning_rate": 1.6894782947287716e-05, "loss": 1.2171, "step": 7082 }, { "epoch": 0.89, "learning_rate": 1.6858414336968608e-05, "loss": 1.2847, "step": 7083 }, { "epoch": 0.89, "learning_rate": 1.6822083547411993e-05, "loss": 1.2123, "step": 7084 }, { "epoch": 0.89, "learning_rate": 1.6785790584511433e-05, "loss": 1.2343, "step": 7085 }, { "epoch": 0.89, "learning_rate": 1.6749535454154613e-05, "loss": 1.2234, "step": 7086 }, { "epoch": 0.89, "learning_rate": 1.6713318162222795e-05, "loss": 1.1471, "step": 7087 }, { "epoch": 0.89, "learning_rate": 1.667713871459131e-05, "loss": 1.215, "step": 7088 }, { "epoch": 0.89, "learning_rate": 1.6640997117129268e-05, "loss": 1.1565, "step": 7089 }, { "epoch": 0.89, "learning_rate": 1.6604893375699592e-05, "loss": 1.2706, "step": 7090 }, { "epoch": 0.89, "learning_rate": 1.6568827496159124e-05, "loss": 1.2063, "step": 7091 }, { "epoch": 0.89, "learning_rate": 1.6532799484358568e-05, "loss": 1.2065, "step": 7092 }, { "epoch": 0.89, "learning_rate": 1.6496809346142507e-05, "loss": 1.1911, "step": 7093 }, { "epoch": 0.89, "learning_rate": 1.646085708734929e-05, "loss": 1.2099, "step": 7094 }, { "epoch": 0.89, "learning_rate": 1.6424942713811203e-05, "loss": 1.1981, "step": 7095 }, { "epoch": 0.89, "learning_rate": 1.638906623135436e-05, "loss": 1.2251, "step": 7096 }, { "epoch": 0.89, "learning_rate": 1.635322764579869e-05, "loss": 1.2178, "step": 7097 }, { "epoch": 0.89, "learning_rate": 1.6317426962958014e-05, "loss": 1.169, "step": 7098 }, { "epoch": 0.89, "learning_rate": 1.628166418864005e-05, "loss": 1.1906, "step": 7099 }, { "epoch": 0.89, "learning_rate": 1.624593932864632e-05, "loss": 1.2337, "step": 7100 }, { "epoch": 0.89, "eval_acceptance_rate": 0.699802577495575, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2176522772923706, "eval_expected_tokens": 2.772065346615463, "eval_loss": 1.3496233224868774, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44301633665386575, "eval_runtime": 50.2576, "eval_samples_per_second": 9.69, "eval_steps_per_second": 0.318, "step": 7100 }, { "epoch": 0.89, "learning_rate": 1.6210252388772105e-05, "loss": 1.1817, "step": 7101 }, { "epoch": 0.89, "learning_rate": 1.6174603374806723e-05, "loss": 1.1905, "step": 7102 }, { "epoch": 0.89, "learning_rate": 1.6138992292533183e-05, "loss": 1.2033, "step": 7103 }, { "epoch": 0.89, "learning_rate": 1.610341914772839e-05, "loss": 1.2026, "step": 7104 }, { "epoch": 0.89, "learning_rate": 1.6067883946163154e-05, "loss": 1.2406, "step": 7105 }, { "epoch": 0.89, "learning_rate": 1.603238669360199e-05, "loss": 1.2726, "step": 7106 }, { "epoch": 0.89, "learning_rate": 1.5996927395803466e-05, "loss": 1.2454, "step": 7107 }, { "epoch": 0.89, "learning_rate": 1.596150605851976e-05, "loss": 1.2589, "step": 7108 }, { "epoch": 0.89, "learning_rate": 1.5926122687497073e-05, "loss": 1.2714, "step": 7109 }, { "epoch": 0.89, "learning_rate": 1.5890777288475296e-05, "loss": 1.1887, "step": 7110 }, { "epoch": 0.89, "learning_rate": 1.5855469867188306e-05, "loss": 1.2219, "step": 7111 }, { "epoch": 0.89, "learning_rate": 1.5820200429363774e-05, "loss": 1.1991, "step": 7112 }, { "epoch": 0.89, "learning_rate": 1.5784968980723123e-05, "loss": 1.2478, "step": 7113 }, { "epoch": 0.89, "learning_rate": 1.574977552698176e-05, "loss": 1.2268, "step": 7114 }, { "epoch": 0.89, "learning_rate": 1.5714620073848772e-05, "loss": 1.2343, "step": 7115 }, { "epoch": 0.89, "learning_rate": 1.5679502627027137e-05, "loss": 1.2551, "step": 7116 }, { "epoch": 0.89, "learning_rate": 1.5644423192213736e-05, "loss": 1.1785, "step": 7117 }, { "epoch": 0.89, "learning_rate": 1.5609381775099296e-05, "loss": 1.2447, "step": 7118 }, { "epoch": 0.89, "learning_rate": 1.5574378381368244e-05, "loss": 1.1837, "step": 7119 }, { "epoch": 0.89, "learning_rate": 1.5539413016698924e-05, "loss": 1.1884, "step": 7120 }, { "epoch": 0.89, "learning_rate": 1.5504485686763437e-05, "loss": 1.1688, "step": 7121 }, { "epoch": 0.89, "learning_rate": 1.5469596397227892e-05, "loss": 1.179, "step": 7122 }, { "epoch": 0.89, "learning_rate": 1.5434745153751984e-05, "loss": 1.2439, "step": 7123 }, { "epoch": 0.89, "learning_rate": 1.5399931961989494e-05, "loss": 1.2516, "step": 7124 }, { "epoch": 0.89, "learning_rate": 1.536515682758782e-05, "loss": 1.2247, "step": 7125 }, { "epoch": 0.89, "learning_rate": 1.5330419756188237e-05, "loss": 1.2146, "step": 7126 }, { "epoch": 0.89, "learning_rate": 1.5295720753425945e-05, "loss": 1.174, "step": 7127 }, { "epoch": 0.89, "learning_rate": 1.526105982492987e-05, "loss": 1.2601, "step": 7128 }, { "epoch": 0.89, "learning_rate": 1.5226436976322727e-05, "loss": 1.252, "step": 7129 }, { "epoch": 0.89, "learning_rate": 1.5191852213221196e-05, "loss": 1.2287, "step": 7130 }, { "epoch": 0.89, "learning_rate": 1.5157305541235649e-05, "loss": 1.195, "step": 7131 }, { "epoch": 0.89, "learning_rate": 1.512279696597038e-05, "loss": 1.2148, "step": 7132 }, { "epoch": 0.89, "learning_rate": 1.5088326493023413e-05, "loss": 1.1869, "step": 7133 }, { "epoch": 0.89, "learning_rate": 1.5053894127986606e-05, "loss": 1.2043, "step": 7134 }, { "epoch": 0.89, "learning_rate": 1.5019499876445658e-05, "loss": 1.1585, "step": 7135 }, { "epoch": 0.89, "learning_rate": 1.4985143743980079e-05, "loss": 1.2078, "step": 7136 }, { "epoch": 0.89, "learning_rate": 1.4950825736163298e-05, "loss": 1.202, "step": 7137 }, { "epoch": 0.89, "learning_rate": 1.4916545858562363e-05, "loss": 1.194, "step": 7138 }, { "epoch": 0.89, "learning_rate": 1.4882304116738244e-05, "loss": 1.2108, "step": 7139 }, { "epoch": 0.89, "learning_rate": 1.4848100516245717e-05, "loss": 1.2405, "step": 7140 }, { "epoch": 0.89, "learning_rate": 1.4813935062633399e-05, "loss": 1.1973, "step": 7141 }, { "epoch": 0.89, "learning_rate": 1.4779807761443637e-05, "loss": 1.1691, "step": 7142 }, { "epoch": 0.89, "learning_rate": 1.4745718618212695e-05, "loss": 1.1934, "step": 7143 }, { "epoch": 0.89, "learning_rate": 1.4711667638470566e-05, "loss": 1.2071, "step": 7144 }, { "epoch": 0.89, "learning_rate": 1.467765482774111e-05, "loss": 1.1956, "step": 7145 }, { "epoch": 0.89, "learning_rate": 1.4643680191541887e-05, "loss": 1.2267, "step": 7146 }, { "epoch": 0.89, "learning_rate": 1.46097437353844e-05, "loss": 1.1248, "step": 7147 }, { "epoch": 0.89, "learning_rate": 1.4575845464773863e-05, "loss": 1.1719, "step": 7148 }, { "epoch": 0.89, "learning_rate": 1.45419853852094e-05, "loss": 1.1813, "step": 7149 }, { "epoch": 0.89, "learning_rate": 1.4508163502183786e-05, "loss": 1.1884, "step": 7150 }, { "epoch": 0.89, "learning_rate": 1.4474379821183742e-05, "loss": 1.2356, "step": 7151 }, { "epoch": 0.89, "learning_rate": 1.4440634347689718e-05, "loss": 1.2339, "step": 7152 }, { "epoch": 0.89, "learning_rate": 1.4406927087175948e-05, "loss": 1.2171, "step": 7153 }, { "epoch": 0.89, "learning_rate": 1.4373258045110581e-05, "loss": 1.2507, "step": 7154 }, { "epoch": 0.89, "learning_rate": 1.4339627226955393e-05, "loss": 1.2177, "step": 7155 }, { "epoch": 0.89, "learning_rate": 1.4306034638166126e-05, "loss": 1.2396, "step": 7156 }, { "epoch": 0.89, "learning_rate": 1.427248028419223e-05, "loss": 1.2041, "step": 7157 }, { "epoch": 0.89, "learning_rate": 1.4238964170476931e-05, "loss": 1.2097, "step": 7158 }, { "epoch": 0.89, "learning_rate": 1.4205486302457298e-05, "loss": 1.2274, "step": 7159 }, { "epoch": 0.9, "learning_rate": 1.417204668556421e-05, "loss": 1.2687, "step": 7160 }, { "epoch": 0.9, "learning_rate": 1.4138645325222327e-05, "loss": 1.2564, "step": 7161 }, { "epoch": 0.9, "learning_rate": 1.4105282226850092e-05, "loss": 1.1835, "step": 7162 }, { "epoch": 0.9, "learning_rate": 1.4071957395859758e-05, "loss": 1.2218, "step": 7163 }, { "epoch": 0.9, "learning_rate": 1.403867083765728e-05, "loss": 1.2262, "step": 7164 }, { "epoch": 0.9, "learning_rate": 1.4005422557642533e-05, "loss": 1.2353, "step": 7165 }, { "epoch": 0.9, "learning_rate": 1.3972212561209119e-05, "loss": 1.231, "step": 7166 }, { "epoch": 0.9, "learning_rate": 1.393904085374445e-05, "loss": 1.2767, "step": 7167 }, { "epoch": 0.9, "learning_rate": 1.390590744062975e-05, "loss": 1.2349, "step": 7168 }, { "epoch": 0.9, "learning_rate": 1.3872812327239942e-05, "loss": 1.1968, "step": 7169 }, { "epoch": 0.9, "learning_rate": 1.3839755518943786e-05, "loss": 1.2016, "step": 7170 }, { "epoch": 0.9, "learning_rate": 1.3806737021103882e-05, "loss": 1.1477, "step": 7171 }, { "epoch": 0.9, "learning_rate": 1.37737568390765e-05, "loss": 1.2376, "step": 7172 }, { "epoch": 0.9, "learning_rate": 1.3740814978211836e-05, "loss": 1.1976, "step": 7173 }, { "epoch": 0.9, "learning_rate": 1.3707911443853777e-05, "loss": 1.1583, "step": 7174 }, { "epoch": 0.9, "learning_rate": 1.3675046241339917e-05, "loss": 1.2071, "step": 7175 }, { "epoch": 0.9, "learning_rate": 1.3642219376001879e-05, "loss": 1.2084, "step": 7176 }, { "epoch": 0.9, "learning_rate": 1.3609430853164823e-05, "loss": 1.2032, "step": 7177 }, { "epoch": 0.9, "learning_rate": 1.357668067814774e-05, "loss": 1.187, "step": 7178 }, { "epoch": 0.9, "learning_rate": 1.3543968856263523e-05, "loss": 1.243, "step": 7179 }, { "epoch": 0.9, "learning_rate": 1.35112953928187e-05, "loss": 1.2054, "step": 7180 }, { "epoch": 0.9, "learning_rate": 1.3478660293113675e-05, "loss": 1.2102, "step": 7181 }, { "epoch": 0.9, "learning_rate": 1.3446063562442573e-05, "loss": 1.2126, "step": 7182 }, { "epoch": 0.9, "learning_rate": 1.341350520609333e-05, "loss": 1.2136, "step": 7183 }, { "epoch": 0.9, "learning_rate": 1.3380985229347554e-05, "loss": 1.2246, "step": 7184 }, { "epoch": 0.9, "learning_rate": 1.3348503637480775e-05, "loss": 1.215, "step": 7185 }, { "epoch": 0.9, "learning_rate": 1.3316060435762278e-05, "loss": 1.229, "step": 7186 }, { "epoch": 0.9, "learning_rate": 1.3283655629455021e-05, "loss": 1.226, "step": 7187 }, { "epoch": 0.9, "learning_rate": 1.32512892238158e-05, "loss": 1.1968, "step": 7188 }, { "epoch": 0.9, "learning_rate": 1.3218961224095133e-05, "loss": 1.2252, "step": 7189 }, { "epoch": 0.9, "learning_rate": 1.3186671635537329e-05, "loss": 1.2294, "step": 7190 }, { "epoch": 0.9, "learning_rate": 1.3154420463380528e-05, "loss": 1.2639, "step": 7191 }, { "epoch": 0.9, "learning_rate": 1.3122207712856632e-05, "loss": 1.2474, "step": 7192 }, { "epoch": 0.9, "learning_rate": 1.309003338919118e-05, "loss": 1.1812, "step": 7193 }, { "epoch": 0.9, "learning_rate": 1.305789749760361e-05, "loss": 1.183, "step": 7194 }, { "epoch": 0.9, "learning_rate": 1.302580004330703e-05, "loss": 1.2042, "step": 7195 }, { "epoch": 0.9, "learning_rate": 1.299374103150841e-05, "loss": 1.2096, "step": 7196 }, { "epoch": 0.9, "learning_rate": 1.2961720467408428e-05, "loss": 1.2142, "step": 7197 }, { "epoch": 0.9, "learning_rate": 1.2929738356201537e-05, "loss": 1.2695, "step": 7198 }, { "epoch": 0.9, "learning_rate": 1.289779470307592e-05, "loss": 1.1711, "step": 7199 }, { "epoch": 0.9, "learning_rate": 1.286588951321363e-05, "loss": 1.1878, "step": 7200 }, { "epoch": 0.9, "eval_acceptance_rate": 0.6998010873794556, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2176460311945867, "eval_expected_tokens": 2.7720575389932334, "eval_loss": 1.349700927734375, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44301438474830845, "eval_runtime": 50.185, "eval_samples_per_second": 9.704, "eval_steps_per_second": 0.319, "step": 7200 }, { "epoch": 0.9, "learning_rate": 1.28340227917903e-05, "loss": 1.2263, "step": 7201 }, { "epoch": 0.9, "learning_rate": 1.2802194543975493e-05, "loss": 1.2589, "step": 7202 }, { "epoch": 0.9, "learning_rate": 1.2770404774932381e-05, "loss": 1.2342, "step": 7203 }, { "epoch": 0.9, "learning_rate": 1.2738653489818035e-05, "loss": 1.29, "step": 7204 }, { "epoch": 0.9, "learning_rate": 1.2706940693783225e-05, "loss": 1.1964, "step": 7205 }, { "epoch": 0.9, "learning_rate": 1.2675266391972473e-05, "loss": 1.1995, "step": 7206 }, { "epoch": 0.9, "learning_rate": 1.2643630589524058e-05, "loss": 1.2254, "step": 7207 }, { "epoch": 0.9, "learning_rate": 1.2612033291569985e-05, "loss": 1.2141, "step": 7208 }, { "epoch": 0.9, "learning_rate": 1.2580474503236016e-05, "loss": 1.3064, "step": 7209 }, { "epoch": 0.9, "learning_rate": 1.2548954229641724e-05, "loss": 1.1746, "step": 7210 }, { "epoch": 0.9, "learning_rate": 1.2517472475900432e-05, "loss": 1.2127, "step": 7211 }, { "epoch": 0.9, "learning_rate": 1.2486029247119169e-05, "loss": 1.2148, "step": 7212 }, { "epoch": 0.9, "learning_rate": 1.2454624548398685e-05, "loss": 1.2193, "step": 7213 }, { "epoch": 0.9, "learning_rate": 1.242325838483352e-05, "loss": 1.1583, "step": 7214 }, { "epoch": 0.9, "learning_rate": 1.2391930761512043e-05, "loss": 1.2012, "step": 7215 }, { "epoch": 0.9, "learning_rate": 1.2360641683516193e-05, "loss": 1.1972, "step": 7216 }, { "epoch": 0.9, "learning_rate": 1.2329391155921853e-05, "loss": 1.2049, "step": 7217 }, { "epoch": 0.9, "learning_rate": 1.2298179183798497e-05, "loss": 1.2385, "step": 7218 }, { "epoch": 0.9, "learning_rate": 1.2267005772209406e-05, "loss": 1.2427, "step": 7219 }, { "epoch": 0.9, "learning_rate": 1.2235870926211617e-05, "loss": 1.2433, "step": 7220 }, { "epoch": 0.9, "learning_rate": 1.22047746508559e-05, "loss": 1.1343, "step": 7221 }, { "epoch": 0.9, "learning_rate": 1.2173716951186714e-05, "loss": 1.2099, "step": 7222 }, { "epoch": 0.9, "learning_rate": 1.2142697832242395e-05, "loss": 1.2542, "step": 7223 }, { "epoch": 0.9, "learning_rate": 1.211171729905483e-05, "loss": 1.2318, "step": 7224 }, { "epoch": 0.9, "learning_rate": 1.2080775356649864e-05, "loss": 1.2002, "step": 7225 }, { "epoch": 0.9, "learning_rate": 1.2049872010046925e-05, "loss": 1.2356, "step": 7226 }, { "epoch": 0.9, "learning_rate": 1.2019007264259196e-05, "loss": 1.1789, "step": 7227 }, { "epoch": 0.9, "learning_rate": 1.1988181124293646e-05, "loss": 1.1672, "step": 7228 }, { "epoch": 0.9, "learning_rate": 1.1957393595150939e-05, "loss": 1.2998, "step": 7229 }, { "epoch": 0.9, "learning_rate": 1.1926644681825555e-05, "loss": 1.2166, "step": 7230 }, { "epoch": 0.9, "learning_rate": 1.189593438930564e-05, "loss": 1.1411, "step": 7231 }, { "epoch": 0.9, "learning_rate": 1.1865262722573073e-05, "loss": 1.1858, "step": 7232 }, { "epoch": 0.9, "learning_rate": 1.1834629686603455e-05, "loss": 1.2144, "step": 7233 }, { "epoch": 0.9, "learning_rate": 1.1804035286366171e-05, "loss": 1.2237, "step": 7234 }, { "epoch": 0.9, "learning_rate": 1.1773479526824282e-05, "loss": 1.188, "step": 7235 }, { "epoch": 0.9, "learning_rate": 1.174296241293471e-05, "loss": 1.1782, "step": 7236 }, { "epoch": 0.9, "learning_rate": 1.171248394964794e-05, "loss": 1.2605, "step": 7237 }, { "epoch": 0.9, "learning_rate": 1.1682044141908266e-05, "loss": 1.1758, "step": 7238 }, { "epoch": 0.9, "learning_rate": 1.1651642994653655e-05, "loss": 1.2192, "step": 7239 }, { "epoch": 0.91, "learning_rate": 1.162128051281594e-05, "loss": 1.2075, "step": 7240 }, { "epoch": 0.91, "learning_rate": 1.1590956701320515e-05, "loss": 1.3126, "step": 7241 }, { "epoch": 0.91, "learning_rate": 1.1560671565086667e-05, "loss": 1.2214, "step": 7242 }, { "epoch": 0.91, "learning_rate": 1.1530425109027242e-05, "loss": 1.1555, "step": 7243 }, { "epoch": 0.91, "learning_rate": 1.1500217338048901e-05, "loss": 1.1955, "step": 7244 }, { "epoch": 0.91, "learning_rate": 1.1470048257052085e-05, "loss": 1.2575, "step": 7245 }, { "epoch": 0.91, "learning_rate": 1.1439917870930795e-05, "loss": 1.2088, "step": 7246 }, { "epoch": 0.91, "learning_rate": 1.1409826184572897e-05, "loss": 1.2443, "step": 7247 }, { "epoch": 0.91, "learning_rate": 1.1379773202859905e-05, "loss": 1.1406, "step": 7248 }, { "epoch": 0.91, "learning_rate": 1.134975893066717e-05, "loss": 1.2654, "step": 7249 }, { "epoch": 0.91, "learning_rate": 1.1319783372863602e-05, "loss": 1.2318, "step": 7250 }, { "epoch": 0.91, "learning_rate": 1.1289846534311921e-05, "loss": 1.1779, "step": 7251 }, { "epoch": 0.91, "learning_rate": 1.1259948419868548e-05, "loss": 1.1921, "step": 7252 }, { "epoch": 0.91, "learning_rate": 1.1230089034383579e-05, "loss": 1.1859, "step": 7253 }, { "epoch": 0.91, "learning_rate": 1.1200268382700913e-05, "loss": 1.1602, "step": 7254 }, { "epoch": 0.91, "learning_rate": 1.1170486469658181e-05, "loss": 1.145, "step": 7255 }, { "epoch": 0.91, "learning_rate": 1.1140743300086603e-05, "loss": 1.278, "step": 7256 }, { "epoch": 0.91, "learning_rate": 1.1111038878811207e-05, "loss": 1.2253, "step": 7257 }, { "epoch": 0.91, "learning_rate": 1.108137321065067e-05, "loss": 1.1716, "step": 7258 }, { "epoch": 0.91, "learning_rate": 1.105174630041747e-05, "loss": 1.2068, "step": 7259 }, { "epoch": 0.91, "learning_rate": 1.1022158152917739e-05, "loss": 1.2435, "step": 7260 }, { "epoch": 0.91, "learning_rate": 1.0992608772951357e-05, "loss": 1.1583, "step": 7261 }, { "epoch": 0.91, "learning_rate": 1.096309816531188e-05, "loss": 1.234, "step": 7262 }, { "epoch": 0.91, "learning_rate": 1.0933626334786561e-05, "loss": 1.1694, "step": 7263 }, { "epoch": 0.91, "learning_rate": 1.090419328615641e-05, "loss": 1.2055, "step": 7264 }, { "epoch": 0.91, "learning_rate": 1.0874799024196135e-05, "loss": 1.2275, "step": 7265 }, { "epoch": 0.91, "learning_rate": 1.084544355367409e-05, "loss": 1.2257, "step": 7266 }, { "epoch": 0.91, "learning_rate": 1.0816126879352461e-05, "loss": 1.2532, "step": 7267 }, { "epoch": 0.91, "learning_rate": 1.0786849005986977e-05, "loss": 1.2183, "step": 7268 }, { "epoch": 0.91, "learning_rate": 1.0757609938327278e-05, "loss": 1.1711, "step": 7269 }, { "epoch": 0.91, "learning_rate": 1.0728409681116491e-05, "loss": 1.2359, "step": 7270 }, { "epoch": 0.91, "learning_rate": 1.0699248239091603e-05, "loss": 1.1377, "step": 7271 }, { "epoch": 0.91, "learning_rate": 1.067012561698319e-05, "loss": 1.2466, "step": 7272 }, { "epoch": 0.91, "learning_rate": 1.064104181951564e-05, "loss": 1.1895, "step": 7273 }, { "epoch": 0.91, "learning_rate": 1.0611996851407014e-05, "loss": 1.201, "step": 7274 }, { "epoch": 0.91, "learning_rate": 1.0582990717369013e-05, "loss": 1.2019, "step": 7275 }, { "epoch": 0.91, "learning_rate": 1.0554023422107096e-05, "loss": 1.1523, "step": 7276 }, { "epoch": 0.91, "learning_rate": 1.0525094970320392e-05, "loss": 1.2295, "step": 7277 }, { "epoch": 0.91, "learning_rate": 1.0496205366701734e-05, "loss": 1.1991, "step": 7278 }, { "epoch": 0.91, "learning_rate": 1.0467354615937675e-05, "loss": 1.2235, "step": 7279 }, { "epoch": 0.91, "learning_rate": 1.0438542722708445e-05, "loss": 1.2388, "step": 7280 }, { "epoch": 0.91, "learning_rate": 1.0409769691687998e-05, "loss": 1.1614, "step": 7281 }, { "epoch": 0.91, "learning_rate": 1.0381035527543936e-05, "loss": 1.2202, "step": 7282 }, { "epoch": 0.91, "learning_rate": 1.0352340234937558e-05, "loss": 1.2171, "step": 7283 }, { "epoch": 0.91, "learning_rate": 1.0323683818523944e-05, "loss": 1.1816, "step": 7284 }, { "epoch": 0.91, "learning_rate": 1.0295066282951737e-05, "loss": 1.1832, "step": 7285 }, { "epoch": 0.91, "learning_rate": 1.0266487632863392e-05, "loss": 1.228, "step": 7286 }, { "epoch": 0.91, "learning_rate": 1.0237947872894977e-05, "loss": 1.2172, "step": 7287 }, { "epoch": 0.91, "learning_rate": 1.020944700767626e-05, "loss": 1.2041, "step": 7288 }, { "epoch": 0.91, "learning_rate": 1.0180985041830765e-05, "loss": 1.2274, "step": 7289 }, { "epoch": 0.91, "learning_rate": 1.0152561979975632e-05, "loss": 1.2012, "step": 7290 }, { "epoch": 0.91, "learning_rate": 1.0124177826721676e-05, "loss": 1.2134, "step": 7291 }, { "epoch": 0.91, "learning_rate": 1.0095832586673515e-05, "loss": 1.198, "step": 7292 }, { "epoch": 0.91, "learning_rate": 1.0067526264429306e-05, "loss": 1.1894, "step": 7293 }, { "epoch": 0.91, "learning_rate": 1.0039258864581041e-05, "loss": 1.2207, "step": 7294 }, { "epoch": 0.91, "learning_rate": 1.0011030391714276e-05, "loss": 1.2195, "step": 7295 }, { "epoch": 0.91, "learning_rate": 9.982840850408292e-06, "loss": 1.2804, "step": 7296 }, { "epoch": 0.91, "learning_rate": 9.954690245236042e-06, "loss": 1.1587, "step": 7297 }, { "epoch": 0.91, "learning_rate": 9.926578580764234e-06, "loss": 1.2195, "step": 7298 }, { "epoch": 0.91, "learning_rate": 9.898505861553191e-06, "loss": 1.1558, "step": 7299 }, { "epoch": 0.91, "learning_rate": 9.87047209215694e-06, "loss": 1.1635, "step": 7300 }, { "epoch": 0.91, "eval_acceptance_rate": 0.6998124718666077, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2176937519257405, "eval_expected_tokens": 2.7721171899071755, "eval_loss": 1.349610686302185, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44302929747679387, "eval_runtime": 50.2075, "eval_samples_per_second": 9.7, "eval_steps_per_second": 0.319, "step": 7300 }, { "epoch": 0.91, "learning_rate": 9.842477277123175e-06, "loss": 1.2121, "step": 7301 }, { "epoch": 0.91, "learning_rate": 9.814521420993233e-06, "loss": 1.2359, "step": 7302 }, { "epoch": 0.91, "learning_rate": 9.786604528302267e-06, "loss": 1.2209, "step": 7303 }, { "epoch": 0.91, "learning_rate": 9.758726603578932e-06, "loss": 1.2141, "step": 7304 }, { "epoch": 0.91, "learning_rate": 9.73088765134572e-06, "loss": 1.19, "step": 7305 }, { "epoch": 0.91, "learning_rate": 9.703087676118688e-06, "loss": 1.2034, "step": 7306 }, { "epoch": 0.91, "learning_rate": 9.675326682407588e-06, "loss": 1.209, "step": 7307 }, { "epoch": 0.91, "learning_rate": 9.647604674715954e-06, "loss": 1.2062, "step": 7308 }, { "epoch": 0.91, "learning_rate": 9.619921657540859e-06, "loss": 1.2718, "step": 7309 }, { "epoch": 0.91, "learning_rate": 9.592277635373098e-06, "loss": 1.2203, "step": 7310 }, { "epoch": 0.91, "learning_rate": 9.564672612697168e-06, "loss": 1.1581, "step": 7311 }, { "epoch": 0.91, "learning_rate": 9.537106593991207e-06, "loss": 1.2142, "step": 7312 }, { "epoch": 0.91, "learning_rate": 9.509579583727057e-06, "loss": 1.2794, "step": 7313 }, { "epoch": 0.91, "learning_rate": 9.4820915863702e-06, "loss": 1.2096, "step": 7314 }, { "epoch": 0.91, "learning_rate": 9.454642606379793e-06, "loss": 1.2117, "step": 7315 }, { "epoch": 0.91, "learning_rate": 9.42723264820866e-06, "loss": 1.1836, "step": 7316 }, { "epoch": 0.91, "learning_rate": 9.399861716303332e-06, "loss": 1.2003, "step": 7317 }, { "epoch": 0.91, "learning_rate": 9.37252981510403e-06, "loss": 1.2662, "step": 7318 }, { "epoch": 0.91, "learning_rate": 9.345236949044517e-06, "loss": 1.1952, "step": 7319 }, { "epoch": 0.92, "learning_rate": 9.317983122552332e-06, "loss": 1.1638, "step": 7320 }, { "epoch": 0.92, "learning_rate": 9.290768340048688e-06, "loss": 1.1409, "step": 7321 }, { "epoch": 0.92, "learning_rate": 9.263592605948335e-06, "loss": 1.1666, "step": 7322 }, { "epoch": 0.92, "learning_rate": 9.23645592465988e-06, "loss": 1.1872, "step": 7323 }, { "epoch": 0.92, "learning_rate": 9.209358300585474e-06, "loss": 1.2377, "step": 7324 }, { "epoch": 0.92, "learning_rate": 9.18229973812093e-06, "loss": 1.1933, "step": 7325 }, { "epoch": 0.92, "learning_rate": 9.155280241655795e-06, "loss": 1.1479, "step": 7326 }, { "epoch": 0.92, "learning_rate": 9.128299815573143e-06, "loss": 1.1926, "step": 7327 }, { "epoch": 0.92, "learning_rate": 9.10135846424992e-06, "loss": 1.2365, "step": 7328 }, { "epoch": 0.92, "learning_rate": 9.074456192056519e-06, "loss": 1.1674, "step": 7329 }, { "epoch": 0.92, "learning_rate": 9.047593003357141e-06, "loss": 1.1562, "step": 7330 }, { "epoch": 0.92, "learning_rate": 9.02076890250958e-06, "loss": 1.2702, "step": 7331 }, { "epoch": 0.92, "learning_rate": 8.993983893865271e-06, "loss": 1.216, "step": 7332 }, { "epoch": 0.92, "learning_rate": 8.967237981769405e-06, "loss": 1.2462, "step": 7333 }, { "epoch": 0.92, "learning_rate": 8.940531170560735e-06, "loss": 1.2555, "step": 7334 }, { "epoch": 0.92, "learning_rate": 8.913863464571654e-06, "loss": 1.2272, "step": 7335 }, { "epoch": 0.92, "learning_rate": 8.887234868128342e-06, "loss": 1.1895, "step": 7336 }, { "epoch": 0.92, "learning_rate": 8.860645385550481e-06, "loss": 1.2164, "step": 7337 }, { "epoch": 0.92, "learning_rate": 8.83409502115151e-06, "loss": 1.2359, "step": 7338 }, { "epoch": 0.92, "learning_rate": 8.807583779238514e-06, "loss": 1.1926, "step": 7339 }, { "epoch": 0.92, "learning_rate": 8.781111664112162e-06, "loss": 1.1972, "step": 7340 }, { "epoch": 0.92, "learning_rate": 8.754678680066825e-06, "loss": 1.2052, "step": 7341 }, { "epoch": 0.92, "learning_rate": 8.728284831390516e-06, "loss": 1.1639, "step": 7342 }, { "epoch": 0.92, "learning_rate": 8.701930122364976e-06, "loss": 1.2526, "step": 7343 }, { "epoch": 0.92, "learning_rate": 8.675614557265455e-06, "loss": 1.2595, "step": 7344 }, { "epoch": 0.92, "learning_rate": 8.649338140360951e-06, "loss": 1.2038, "step": 7345 }, { "epoch": 0.92, "learning_rate": 8.623100875914059e-06, "loss": 1.2266, "step": 7346 }, { "epoch": 0.92, "learning_rate": 8.596902768181064e-06, "loss": 1.2266, "step": 7347 }, { "epoch": 0.92, "learning_rate": 8.570743821411903e-06, "loss": 1.2214, "step": 7348 }, { "epoch": 0.92, "learning_rate": 8.544624039850124e-06, "loss": 1.1744, "step": 7349 }, { "epoch": 0.92, "learning_rate": 8.51854342773295e-06, "loss": 1.2132, "step": 7350 }, { "epoch": 0.92, "learning_rate": 8.49250198929119e-06, "loss": 1.1585, "step": 7351 }, { "epoch": 0.92, "learning_rate": 8.46649972874941e-06, "loss": 1.2171, "step": 7352 }, { "epoch": 0.92, "learning_rate": 8.440536650325737e-06, "loss": 1.2061, "step": 7353 }, { "epoch": 0.92, "learning_rate": 8.41461275823191e-06, "loss": 1.2234, "step": 7354 }, { "epoch": 0.92, "learning_rate": 8.388728056673428e-06, "loss": 1.1467, "step": 7355 }, { "epoch": 0.92, "learning_rate": 8.362882549849349e-06, "loss": 1.2801, "step": 7356 }, { "epoch": 0.92, "learning_rate": 8.337076241952374e-06, "loss": 1.2091, "step": 7357 }, { "epoch": 0.92, "learning_rate": 8.311309137168877e-06, "loss": 1.2293, "step": 7358 }, { "epoch": 0.92, "learning_rate": 8.285581239678874e-06, "loss": 1.2178, "step": 7359 }, { "epoch": 0.92, "learning_rate": 8.259892553655946e-06, "loss": 1.1837, "step": 7360 }, { "epoch": 0.92, "learning_rate": 8.234243083267395e-06, "loss": 1.271, "step": 7361 }, { "epoch": 0.92, "learning_rate": 8.208632832674174e-06, "loss": 1.125, "step": 7362 }, { "epoch": 0.92, "learning_rate": 8.183061806030818e-06, "loss": 1.1847, "step": 7363 }, { "epoch": 0.92, "learning_rate": 8.157530007485508e-06, "loss": 1.1852, "step": 7364 }, { "epoch": 0.92, "learning_rate": 8.132037441180045e-06, "loss": 1.2017, "step": 7365 }, { "epoch": 0.92, "learning_rate": 8.106584111249926e-06, "loss": 1.2335, "step": 7366 }, { "epoch": 0.92, "learning_rate": 8.081170021824235e-06, "loss": 1.1798, "step": 7367 }, { "epoch": 0.92, "learning_rate": 8.055795177025732e-06, "loss": 1.1809, "step": 7368 }, { "epoch": 0.92, "learning_rate": 8.03045958097076e-06, "loss": 1.238, "step": 7369 }, { "epoch": 0.92, "learning_rate": 8.005163237769314e-06, "loss": 1.2189, "step": 7370 }, { "epoch": 0.92, "learning_rate": 7.979906151524997e-06, "loss": 1.2178, "step": 7371 }, { "epoch": 0.92, "learning_rate": 7.954688326335141e-06, "loss": 1.1671, "step": 7372 }, { "epoch": 0.92, "learning_rate": 7.929509766290587e-06, "loss": 1.2089, "step": 7373 }, { "epoch": 0.92, "learning_rate": 7.904370475475875e-06, "loss": 1.2234, "step": 7374 }, { "epoch": 0.92, "learning_rate": 7.879270457969184e-06, "loss": 1.1855, "step": 7375 }, { "epoch": 0.92, "learning_rate": 7.854209717842232e-06, "loss": 1.2712, "step": 7376 }, { "epoch": 0.92, "learning_rate": 7.829188259160486e-06, "loss": 1.2075, "step": 7377 }, { "epoch": 0.92, "learning_rate": 7.804206085982952e-06, "loss": 1.2228, "step": 7378 }, { "epoch": 0.92, "learning_rate": 7.779263202362303e-06, "loss": 1.1487, "step": 7379 }, { "epoch": 0.92, "learning_rate": 7.754359612344857e-06, "loss": 1.17, "step": 7380 }, { "epoch": 0.92, "learning_rate": 7.729495319970498e-06, "loss": 1.2006, "step": 7381 }, { "epoch": 0.92, "learning_rate": 7.704670329272773e-06, "loss": 1.2813, "step": 7382 }, { "epoch": 0.92, "learning_rate": 7.679884644278878e-06, "loss": 1.2063, "step": 7383 }, { "epoch": 0.92, "learning_rate": 7.655138269009598e-06, "loss": 1.1922, "step": 7384 }, { "epoch": 0.92, "learning_rate": 7.630431207479278e-06, "loss": 1.1972, "step": 7385 }, { "epoch": 0.92, "learning_rate": 7.605763463696042e-06, "loss": 1.2251, "step": 7386 }, { "epoch": 0.92, "learning_rate": 7.581135041661497e-06, "loss": 1.2203, "step": 7387 }, { "epoch": 0.92, "learning_rate": 7.556545945370974e-06, "loss": 1.2433, "step": 7388 }, { "epoch": 0.92, "learning_rate": 7.53199617881331e-06, "loss": 1.1581, "step": 7389 }, { "epoch": 0.92, "learning_rate": 7.507485745971044e-06, "loss": 1.1966, "step": 7390 }, { "epoch": 0.92, "learning_rate": 7.483014650820352e-06, "loss": 1.1854, "step": 7391 }, { "epoch": 0.92, "learning_rate": 7.458582897330923e-06, "loss": 1.2733, "step": 7392 }, { "epoch": 0.92, "learning_rate": 7.434190489466197e-06, "loss": 1.2733, "step": 7393 }, { "epoch": 0.92, "learning_rate": 7.409837431183148e-06, "loss": 1.188, "step": 7394 }, { "epoch": 0.92, "learning_rate": 7.3855237264323635e-06, "loss": 1.2251, "step": 7395 }, { "epoch": 0.92, "learning_rate": 7.3612493791580805e-06, "loss": 1.1936, "step": 7396 }, { "epoch": 0.92, "learning_rate": 7.337014393298119e-06, "loss": 1.237, "step": 7397 }, { "epoch": 0.92, "learning_rate": 7.312818772784002e-06, "loss": 1.1932, "step": 7398 }, { "epoch": 0.92, "learning_rate": 7.288662521540729e-06, "loss": 1.2349, "step": 7399 }, { "epoch": 0.93, "learning_rate": 7.2645456434869975e-06, "loss": 1.1575, "step": 7400 }, { "epoch": 0.93, "eval_acceptance_rate": 0.6998339891433716, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.217783950025744, "eval_expected_tokens": 2.77222993753218, "eval_loss": 1.3496590852737427, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44305748438304504, "eval_runtime": 50.24, "eval_samples_per_second": 9.693, "eval_steps_per_second": 0.318, "step": 7400 }, { "epoch": 0.93, "learning_rate": 7.240468142535123e-06, "loss": 1.2191, "step": 7401 }, { "epoch": 0.93, "learning_rate": 7.216430022591009e-06, "loss": 1.1947, "step": 7402 }, { "epoch": 0.93, "learning_rate": 7.192431287554118e-06, "loss": 1.1856, "step": 7403 }, { "epoch": 0.93, "learning_rate": 7.168471941317673e-06, "loss": 1.2205, "step": 7404 }, { "epoch": 0.93, "learning_rate": 7.144551987768338e-06, "loss": 1.1682, "step": 7405 }, { "epoch": 0.93, "learning_rate": 7.120671430786541e-06, "loss": 1.2418, "step": 7406 }, { "epoch": 0.93, "learning_rate": 7.09683027424618e-06, "loss": 1.2083, "step": 7407 }, { "epoch": 0.93, "learning_rate": 7.07302852201483e-06, "loss": 1.2212, "step": 7408 }, { "epoch": 0.93, "learning_rate": 7.049266177953623e-06, "loss": 1.2037, "step": 7409 }, { "epoch": 0.93, "learning_rate": 7.0255432459174205e-06, "loss": 1.1958, "step": 7410 }, { "epoch": 0.93, "learning_rate": 7.001859729754589e-06, "loss": 1.157, "step": 7411 }, { "epoch": 0.93, "learning_rate": 6.978215633307139e-06, "loss": 1.1744, "step": 7412 }, { "epoch": 0.93, "learning_rate": 6.954610960410612e-06, "loss": 1.1971, "step": 7413 }, { "epoch": 0.93, "learning_rate": 6.931045714894279e-06, "loss": 1.1768, "step": 7414 }, { "epoch": 0.93, "learning_rate": 6.907519900580861e-06, "loss": 1.2381, "step": 7415 }, { "epoch": 0.93, "learning_rate": 6.884033521286831e-06, "loss": 1.1659, "step": 7416 }, { "epoch": 0.93, "learning_rate": 6.860586580822198e-06, "loss": 1.2268, "step": 7417 }, { "epoch": 0.93, "learning_rate": 6.837179082990613e-06, "loss": 1.2306, "step": 7418 }, { "epoch": 0.93, "learning_rate": 6.813811031589234e-06, "loss": 1.197, "step": 7419 }, { "epoch": 0.93, "learning_rate": 6.7904824304088895e-06, "loss": 1.2332, "step": 7420 }, { "epoch": 0.93, "learning_rate": 6.767193283234025e-06, "loss": 1.2064, "step": 7421 }, { "epoch": 0.93, "learning_rate": 6.7439435938426165e-06, "loss": 1.1832, "step": 7422 }, { "epoch": 0.93, "learning_rate": 6.720733366006343e-06, "loss": 1.1795, "step": 7423 }, { "epoch": 0.93, "learning_rate": 6.697562603490387e-06, "loss": 1.2829, "step": 7424 }, { "epoch": 0.93, "learning_rate": 6.674431310053519e-06, "loss": 1.1839, "step": 7425 }, { "epoch": 0.93, "learning_rate": 6.651339489448266e-06, "loss": 1.1846, "step": 7426 }, { "epoch": 0.93, "learning_rate": 6.62828714542052e-06, "loss": 1.1982, "step": 7427 }, { "epoch": 0.93, "learning_rate": 6.605274281709927e-06, "loss": 1.2044, "step": 7428 }, { "epoch": 0.93, "learning_rate": 6.582300902049726e-06, "loss": 1.2087, "step": 7429 }, { "epoch": 0.93, "learning_rate": 6.559367010166628e-06, "loss": 1.1586, "step": 7430 }, { "epoch": 0.93, "learning_rate": 6.5364726097811025e-06, "loss": 1.2903, "step": 7431 }, { "epoch": 0.93, "learning_rate": 6.513617704607094e-06, "loss": 1.2044, "step": 7432 }, { "epoch": 0.93, "learning_rate": 6.490802298352194e-06, "loss": 1.2129, "step": 7433 }, { "epoch": 0.93, "learning_rate": 6.4680263947175246e-06, "loss": 1.2583, "step": 7434 }, { "epoch": 0.93, "learning_rate": 6.445289997397879e-06, "loss": 1.1978, "step": 7435 }, { "epoch": 0.93, "learning_rate": 6.422593110081643e-06, "loss": 1.2083, "step": 7436 }, { "epoch": 0.93, "learning_rate": 6.3999357364507295e-06, "loss": 1.1811, "step": 7437 }, { "epoch": 0.93, "learning_rate": 6.3773178801806446e-06, "loss": 1.1714, "step": 7438 }, { "epoch": 0.93, "learning_rate": 6.354739544940508e-06, "loss": 1.2113, "step": 7439 }, { "epoch": 0.93, "learning_rate": 6.332200734393057e-06, "loss": 1.1844, "step": 7440 }, { "epoch": 0.93, "learning_rate": 6.309701452194561e-06, "loss": 1.2355, "step": 7441 }, { "epoch": 0.93, "learning_rate": 6.2872417019949335e-06, "loss": 1.2005, "step": 7442 }, { "epoch": 0.93, "learning_rate": 6.2648214874376495e-06, "loss": 1.247, "step": 7443 }, { "epoch": 0.93, "learning_rate": 6.2424408121597445e-06, "loss": 1.1641, "step": 7444 }, { "epoch": 0.93, "learning_rate": 6.220099679791896e-06, "loss": 1.1723, "step": 7445 }, { "epoch": 0.93, "learning_rate": 6.19779809395829e-06, "loss": 1.1946, "step": 7446 }, { "epoch": 0.93, "learning_rate": 6.175536058276754e-06, "loss": 1.184, "step": 7447 }, { "epoch": 0.93, "learning_rate": 6.153313576358704e-06, "loss": 1.1785, "step": 7448 }, { "epoch": 0.93, "learning_rate": 6.131130651809092e-06, "loss": 1.1943, "step": 7449 }, { "epoch": 0.93, "learning_rate": 6.108987288226536e-06, "loss": 1.2698, "step": 7450 }, { "epoch": 0.93, "learning_rate": 6.086883489203138e-06, "loss": 1.1901, "step": 7451 }, { "epoch": 0.93, "learning_rate": 6.064819258324639e-06, "loss": 1.2408, "step": 7452 }, { "epoch": 0.93, "learning_rate": 6.042794599170343e-06, "loss": 1.2007, "step": 7453 }, { "epoch": 0.93, "learning_rate": 6.020809515313141e-06, "loss": 1.1913, "step": 7454 }, { "epoch": 0.93, "learning_rate": 5.998864010319516e-06, "loss": 1.1702, "step": 7455 }, { "epoch": 0.93, "learning_rate": 5.976958087749534e-06, "loss": 1.1981, "step": 7456 }, { "epoch": 0.93, "learning_rate": 5.9550917511567996e-06, "loss": 1.1827, "step": 7457 }, { "epoch": 0.93, "learning_rate": 5.933265004088501e-06, "loss": 1.3073, "step": 7458 }, { "epoch": 0.93, "learning_rate": 5.911477850085445e-06, "loss": 1.199, "step": 7459 }, { "epoch": 0.93, "learning_rate": 5.8897302926819715e-06, "loss": 1.2084, "step": 7460 }, { "epoch": 0.93, "learning_rate": 5.868022335406065e-06, "loss": 1.2037, "step": 7461 }, { "epoch": 0.93, "learning_rate": 5.84635398177924e-06, "loss": 1.2677, "step": 7462 }, { "epoch": 0.93, "learning_rate": 5.824725235316547e-06, "loss": 1.2368, "step": 7463 }, { "epoch": 0.93, "learning_rate": 5.803136099526623e-06, "loss": 1.2557, "step": 7464 }, { "epoch": 0.93, "learning_rate": 5.781586577911807e-06, "loss": 1.196, "step": 7465 }, { "epoch": 0.93, "learning_rate": 5.760076673967829e-06, "loss": 1.1492, "step": 7466 }, { "epoch": 0.93, "learning_rate": 5.738606391184093e-06, "loss": 1.2012, "step": 7467 }, { "epoch": 0.93, "learning_rate": 5.717175733043589e-06, "loss": 1.2052, "step": 7468 }, { "epoch": 0.93, "learning_rate": 5.695784703022788e-06, "loss": 1.2419, "step": 7469 }, { "epoch": 0.93, "learning_rate": 5.674433304591858e-06, "loss": 1.1869, "step": 7470 }, { "epoch": 0.93, "learning_rate": 5.653121541214445e-06, "loss": 1.1982, "step": 7471 }, { "epoch": 0.93, "learning_rate": 5.631849416347756e-06, "loss": 1.2406, "step": 7472 }, { "epoch": 0.93, "learning_rate": 5.610616933442669e-06, "loss": 1.2282, "step": 7473 }, { "epoch": 0.93, "learning_rate": 5.589424095943513e-06, "loss": 1.1797, "step": 7474 }, { "epoch": 0.93, "learning_rate": 5.568270907288286e-06, "loss": 1.1948, "step": 7475 }, { "epoch": 0.93, "learning_rate": 5.5471573709084676e-06, "loss": 1.2447, "step": 7476 }, { "epoch": 0.93, "learning_rate": 5.52608349022915e-06, "loss": 1.2336, "step": 7477 }, { "epoch": 0.93, "learning_rate": 5.505049268668988e-06, "loss": 1.2212, "step": 7478 }, { "epoch": 0.93, "learning_rate": 5.484054709640196e-06, "loss": 1.2369, "step": 7479 }, { "epoch": 0.94, "learning_rate": 5.463099816548578e-06, "loss": 1.201, "step": 7480 }, { "epoch": 0.94, "learning_rate": 5.442184592793497e-06, "loss": 1.2113, "step": 7481 }, { "epoch": 0.94, "learning_rate": 5.4213090417678525e-06, "loss": 1.1564, "step": 7482 }, { "epoch": 0.94, "learning_rate": 5.4004731668581e-06, "loss": 1.2598, "step": 7483 }, { "epoch": 0.94, "learning_rate": 5.3796769714442865e-06, "loss": 1.1972, "step": 7484 }, { "epoch": 0.94, "learning_rate": 5.358920458900046e-06, "loss": 1.2081, "step": 7485 }, { "epoch": 0.94, "learning_rate": 5.3382036325925474e-06, "loss": 1.172, "step": 7486 }, { "epoch": 0.94, "learning_rate": 5.317526495882519e-06, "loss": 1.2029, "step": 7487 }, { "epoch": 0.94, "learning_rate": 5.2968890521242205e-06, "loss": 1.2054, "step": 7488 }, { "epoch": 0.94, "learning_rate": 5.276291304665531e-06, "loss": 1.2379, "step": 7489 }, { "epoch": 0.94, "learning_rate": 5.255733256847889e-06, "loss": 1.2114, "step": 7490 }, { "epoch": 0.94, "learning_rate": 5.235214912006209e-06, "loss": 1.228, "step": 7491 }, { "epoch": 0.94, "learning_rate": 5.214736273469078e-06, "loss": 1.2536, "step": 7492 }, { "epoch": 0.94, "learning_rate": 5.194297344558535e-06, "loss": 1.2554, "step": 7493 }, { "epoch": 0.94, "learning_rate": 5.173898128590316e-06, "loss": 1.217, "step": 7494 }, { "epoch": 0.94, "learning_rate": 5.15353862887355e-06, "loss": 1.2496, "step": 7495 }, { "epoch": 0.94, "learning_rate": 5.133218848711013e-06, "loss": 1.2233, "step": 7496 }, { "epoch": 0.94, "learning_rate": 5.112938791399041e-06, "loss": 1.1853, "step": 7497 }, { "epoch": 0.94, "learning_rate": 5.0926984602275015e-06, "loss": 1.2069, "step": 7498 }, { "epoch": 0.94, "learning_rate": 5.072497858479852e-06, "loss": 1.2888, "step": 7499 }, { "epoch": 0.94, "learning_rate": 5.052336989433082e-06, "loss": 1.241, "step": 7500 }, { "epoch": 0.94, "eval_acceptance_rate": 0.6998406052589417, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.217811684966682, "eval_expected_tokens": 2.7722646062083522, "eval_loss": 1.3496109247207642, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44306615155208806, "eval_runtime": 50.2836, "eval_samples_per_second": 9.685, "eval_steps_per_second": 0.318, "step": 7500 }, { "epoch": 0.94, "learning_rate": 5.032215856357714e-06, "loss": 1.2715, "step": 7501 }, { "epoch": 0.94, "learning_rate": 5.012134462517831e-06, "loss": 1.252, "step": 7502 }, { "epoch": 0.94, "learning_rate": 4.992092811171078e-06, "loss": 1.2316, "step": 7503 }, { "epoch": 0.94, "learning_rate": 4.97209090556866e-06, "loss": 1.1801, "step": 7504 }, { "epoch": 0.94, "learning_rate": 4.952128748955398e-06, "loss": 1.2141, "step": 7505 }, { "epoch": 0.94, "learning_rate": 4.932206344569562e-06, "loss": 1.2282, "step": 7506 }, { "epoch": 0.94, "learning_rate": 4.912323695642956e-06, "loss": 1.2405, "step": 7507 }, { "epoch": 0.94, "learning_rate": 4.8924808054010564e-06, "loss": 1.2722, "step": 7508 }, { "epoch": 0.94, "learning_rate": 4.8726776770627865e-06, "loss": 1.2621, "step": 7509 }, { "epoch": 0.94, "learning_rate": 4.852914313840662e-06, "loss": 1.2457, "step": 7510 }, { "epoch": 0.94, "learning_rate": 4.833190718940756e-06, "loss": 1.2816, "step": 7511 }, { "epoch": 0.94, "learning_rate": 4.8135068955626474e-06, "loss": 1.1478, "step": 7512 }, { "epoch": 0.94, "learning_rate": 4.793862846899505e-06, "loss": 1.2206, "step": 7513 }, { "epoch": 0.94, "learning_rate": 4.774258576138058e-06, "loss": 1.2213, "step": 7514 }, { "epoch": 0.94, "learning_rate": 4.754694086458538e-06, "loss": 1.2347, "step": 7515 }, { "epoch": 0.94, "learning_rate": 4.7351693810347154e-06, "loss": 1.2462, "step": 7516 }, { "epoch": 0.94, "learning_rate": 4.715684463033942e-06, "loss": 1.196, "step": 7517 }, { "epoch": 0.94, "learning_rate": 4.6962393356171364e-06, "loss": 1.2055, "step": 7518 }, { "epoch": 0.94, "learning_rate": 4.676834001938718e-06, "loss": 1.3169, "step": 7519 }, { "epoch": 0.94, "learning_rate": 4.657468465146642e-06, "loss": 1.2452, "step": 7520 }, { "epoch": 0.94, "learning_rate": 4.638142728382477e-06, "loss": 1.2668, "step": 7521 }, { "epoch": 0.94, "learning_rate": 4.618856794781217e-06, "loss": 1.1992, "step": 7522 }, { "epoch": 0.94, "learning_rate": 4.599610667471527e-06, "loss": 1.2121, "step": 7523 }, { "epoch": 0.94, "learning_rate": 4.580404349575545e-06, "loss": 1.2467, "step": 7524 }, { "epoch": 0.94, "learning_rate": 4.561237844208977e-06, "loss": 1.1978, "step": 7525 }, { "epoch": 0.94, "learning_rate": 4.542111154481027e-06, "loss": 1.2103, "step": 7526 }, { "epoch": 0.94, "learning_rate": 4.523024283494493e-06, "loss": 1.2055, "step": 7527 }, { "epoch": 0.94, "learning_rate": 4.503977234345647e-06, "loss": 1.1805, "step": 7528 }, { "epoch": 0.94, "learning_rate": 4.484970010124406e-06, "loss": 1.1852, "step": 7529 }, { "epoch": 0.94, "learning_rate": 4.4660026139141384e-06, "loss": 1.2025, "step": 7530 }, { "epoch": 0.94, "learning_rate": 4.447075048791799e-06, "loss": 1.2217, "step": 7531 }, { "epoch": 0.94, "learning_rate": 4.4281873178278475e-06, "loss": 1.2538, "step": 7532 }, { "epoch": 0.94, "learning_rate": 4.409339424086251e-06, "loss": 1.2304, "step": 7533 }, { "epoch": 0.94, "learning_rate": 4.390531370624645e-06, "loss": 1.154, "step": 7534 }, { "epoch": 0.94, "learning_rate": 4.371763160494008e-06, "loss": 1.2558, "step": 7535 }, { "epoch": 0.94, "learning_rate": 4.353034796739069e-06, "loss": 1.1481, "step": 7536 }, { "epoch": 0.94, "learning_rate": 4.334346282397928e-06, "loss": 1.192, "step": 7537 }, { "epoch": 0.94, "learning_rate": 4.315697620502296e-06, "loss": 1.2242, "step": 7538 }, { "epoch": 0.94, "learning_rate": 4.297088814077394e-06, "loss": 1.2246, "step": 7539 }, { "epoch": 0.94, "learning_rate": 4.2785198661419725e-06, "loss": 1.2328, "step": 7540 }, { "epoch": 0.94, "learning_rate": 4.259990779708344e-06, "loss": 1.2421, "step": 7541 }, { "epoch": 0.94, "learning_rate": 4.241501557782329e-06, "loss": 1.2901, "step": 7542 }, { "epoch": 0.94, "learning_rate": 4.223052203363303e-06, "loss": 1.1811, "step": 7543 }, { "epoch": 0.94, "learning_rate": 4.204642719444179e-06, "loss": 1.2439, "step": 7544 }, { "epoch": 0.94, "learning_rate": 4.186273109011374e-06, "loss": 1.2239, "step": 7545 }, { "epoch": 0.94, "learning_rate": 4.167943375044836e-06, "loss": 1.211, "step": 7546 }, { "epoch": 0.94, "learning_rate": 4.149653520518021e-06, "loss": 1.279, "step": 7547 }, { "epoch": 0.94, "learning_rate": 4.131403548397999e-06, "loss": 1.1655, "step": 7548 }, { "epoch": 0.94, "learning_rate": 4.113193461645348e-06, "loss": 1.2026, "step": 7549 }, { "epoch": 0.94, "learning_rate": 4.095023263214121e-06, "loss": 1.2383, "step": 7550 }, { "epoch": 0.94, "learning_rate": 4.076892956051903e-06, "loss": 1.2486, "step": 7551 }, { "epoch": 0.94, "learning_rate": 4.058802543099844e-06, "loss": 1.2056, "step": 7552 }, { "epoch": 0.94, "learning_rate": 4.040752027292649e-06, "loss": 1.1775, "step": 7553 }, { "epoch": 0.94, "learning_rate": 4.0227414115584785e-06, "loss": 1.2001, "step": 7554 }, { "epoch": 0.94, "learning_rate": 4.004770698819077e-06, "loss": 1.2406, "step": 7555 }, { "epoch": 0.94, "learning_rate": 3.986839891989669e-06, "loss": 1.2535, "step": 7556 }, { "epoch": 0.94, "learning_rate": 3.968948993979038e-06, "loss": 1.1827, "step": 7557 }, { "epoch": 0.94, "learning_rate": 3.95109800768953e-06, "loss": 1.2028, "step": 7558 }, { "epoch": 0.94, "learning_rate": 3.933286936016911e-06, "loss": 1.1615, "step": 7559 }, { "epoch": 0.94, "learning_rate": 3.915515781850565e-06, "loss": 1.1653, "step": 7560 }, { "epoch": 0.95, "learning_rate": 3.89778454807338e-06, "loss": 1.228, "step": 7561 }, { "epoch": 0.95, "learning_rate": 3.880093237561721e-06, "loss": 1.2159, "step": 7562 }, { "epoch": 0.95, "learning_rate": 3.862441853185544e-06, "loss": 1.264, "step": 7563 }, { "epoch": 0.95, "learning_rate": 3.8448303978082796e-06, "loss": 1.1725, "step": 7564 }, { "epoch": 0.95, "learning_rate": 3.827258874286921e-06, "loss": 1.2279, "step": 7565 }, { "epoch": 0.95, "learning_rate": 3.8097272854719366e-06, "loss": 1.2134, "step": 7566 }, { "epoch": 0.95, "learning_rate": 3.7922356342073306e-06, "loss": 1.233, "step": 7567 }, { "epoch": 0.95, "learning_rate": 3.7747839233306936e-06, "loss": 1.2147, "step": 7568 }, { "epoch": 0.95, "learning_rate": 3.757372155673039e-06, "loss": 1.2167, "step": 7569 }, { "epoch": 0.95, "learning_rate": 3.7400003340589682e-06, "loss": 1.2547, "step": 7570 }, { "epoch": 0.95, "learning_rate": 3.722668461306533e-06, "loss": 1.2554, "step": 7571 }, { "epoch": 0.95, "learning_rate": 3.705376540227373e-06, "loss": 1.193, "step": 7572 }, { "epoch": 0.95, "learning_rate": 3.6881245736266333e-06, "loss": 1.2313, "step": 7573 }, { "epoch": 0.95, "learning_rate": 3.670912564302992e-06, "loss": 1.1738, "step": 7574 }, { "epoch": 0.95, "learning_rate": 3.653740515048576e-06, "loss": 1.2202, "step": 7575 }, { "epoch": 0.95, "learning_rate": 3.6366084286490743e-06, "loss": 1.1867, "step": 7576 }, { "epoch": 0.95, "learning_rate": 3.619516307883708e-06, "loss": 1.1698, "step": 7577 }, { "epoch": 0.95, "learning_rate": 3.6024641555252315e-06, "loss": 1.1961, "step": 7578 }, { "epoch": 0.95, "learning_rate": 3.5854519743398206e-06, "loss": 1.1729, "step": 7579 }, { "epoch": 0.95, "learning_rate": 3.568479767087296e-06, "loss": 1.2095, "step": 7580 }, { "epoch": 0.95, "learning_rate": 3.5515475365208718e-06, "loss": 1.1765, "step": 7581 }, { "epoch": 0.95, "learning_rate": 3.5346552853873504e-06, "loss": 1.2472, "step": 7582 }, { "epoch": 0.95, "learning_rate": 3.517803016427068e-06, "loss": 1.2476, "step": 7583 }, { "epoch": 0.95, "learning_rate": 3.5009907323737824e-06, "loss": 1.2423, "step": 7584 }, { "epoch": 0.95, "learning_rate": 3.484218435954839e-06, "loss": 1.2217, "step": 7585 }, { "epoch": 0.95, "learning_rate": 3.46748612989109e-06, "loss": 1.1991, "step": 7586 }, { "epoch": 0.95, "learning_rate": 3.450793816896891e-06, "loss": 1.2107, "step": 7587 }, { "epoch": 0.95, "learning_rate": 3.4341414996801325e-06, "loss": 1.1715, "step": 7588 }, { "epoch": 0.95, "learning_rate": 3.417529180942125e-06, "loss": 1.2753, "step": 7589 }, { "epoch": 0.95, "learning_rate": 3.4009568633777966e-06, "loss": 1.1963, "step": 7590 }, { "epoch": 0.95, "learning_rate": 3.384424549675552e-06, "loss": 1.1576, "step": 7591 }, { "epoch": 0.95, "learning_rate": 3.3679322425172464e-06, "loss": 1.2212, "step": 7592 }, { "epoch": 0.95, "learning_rate": 3.3514799445784063e-06, "loss": 1.2236, "step": 7593 }, { "epoch": 0.95, "learning_rate": 3.3350676585278682e-06, "loss": 1.1628, "step": 7594 }, { "epoch": 0.95, "learning_rate": 3.3186953870281143e-06, "loss": 1.2515, "step": 7595 }, { "epoch": 0.95, "learning_rate": 3.302363132735048e-06, "loss": 1.1657, "step": 7596 }, { "epoch": 0.95, "learning_rate": 3.286070898298188e-06, "loss": 1.1216, "step": 7597 }, { "epoch": 0.95, "learning_rate": 3.2698186863604483e-06, "loss": 1.2212, "step": 7598 }, { "epoch": 0.95, "learning_rate": 3.2536064995583315e-06, "loss": 1.2004, "step": 7599 }, { "epoch": 0.95, "learning_rate": 3.2374343405217886e-06, "loss": 1.2702, "step": 7600 }, { "epoch": 0.95, "eval_acceptance_rate": 0.6998440623283386, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2178261772661996, "eval_expected_tokens": 2.7722827215827497, "eval_loss": 1.3495378494262695, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44307068039568737, "eval_runtime": 50.4417, "eval_samples_per_second": 9.655, "eval_steps_per_second": 0.317, "step": 7600 }, { "epoch": 0.95, "learning_rate": 3.2213022118743053e-06, "loss": 1.2206, "step": 7601 }, { "epoch": 0.95, "learning_rate": 3.205210116232926e-06, "loss": 1.2232, "step": 7602 }, { "epoch": 0.95, "learning_rate": 3.1891580562080912e-06, "loss": 1.1887, "step": 7603 }, { "epoch": 0.95, "learning_rate": 3.173146034403801e-06, "loss": 1.2549, "step": 7604 }, { "epoch": 0.95, "learning_rate": 3.1571740534175888e-06, "loss": 1.2025, "step": 7605 }, { "epoch": 0.95, "learning_rate": 3.1412421158404368e-06, "loss": 1.2127, "step": 7606 }, { "epoch": 0.95, "learning_rate": 3.1253502242569174e-06, "loss": 1.2223, "step": 7607 }, { "epoch": 0.95, "learning_rate": 3.1094983812449962e-06, "loss": 1.1828, "step": 7608 }, { "epoch": 0.95, "learning_rate": 3.0936865893761987e-06, "loss": 1.1603, "step": 7609 }, { "epoch": 0.95, "learning_rate": 3.077914851215585e-06, "loss": 1.2447, "step": 7610 }, { "epoch": 0.95, "learning_rate": 3.062183169321636e-06, "loss": 1.1756, "step": 7611 }, { "epoch": 0.95, "learning_rate": 3.0464915462464214e-06, "loss": 1.2398, "step": 7612 }, { "epoch": 0.95, "learning_rate": 3.030839984535461e-06, "loss": 1.219, "step": 7613 }, { "epoch": 0.95, "learning_rate": 3.015228486727778e-06, "loss": 1.2078, "step": 7614 }, { "epoch": 0.95, "learning_rate": 2.9996570553559034e-06, "loss": 1.2526, "step": 7615 }, { "epoch": 0.95, "learning_rate": 2.9841256929458715e-06, "loss": 1.2395, "step": 7616 }, { "epoch": 0.95, "learning_rate": 2.968634402017223e-06, "loss": 1.1921, "step": 7617 }, { "epoch": 0.95, "learning_rate": 2.9531831850830027e-06, "loss": 1.214, "step": 7618 }, { "epoch": 0.95, "learning_rate": 2.937772044649706e-06, "loss": 1.1794, "step": 7619 }, { "epoch": 0.95, "learning_rate": 2.922400983217416e-06, "loss": 1.1852, "step": 7620 }, { "epoch": 0.95, "learning_rate": 2.9070700032796105e-06, "loss": 1.2024, "step": 7621 }, { "epoch": 0.95, "learning_rate": 2.891779107323356e-06, "loss": 1.1411, "step": 7622 }, { "epoch": 0.95, "learning_rate": 2.876528297829112e-06, "loss": 1.2802, "step": 7623 }, { "epoch": 0.95, "learning_rate": 2.8613175772709834e-06, "loss": 1.2334, "step": 7624 }, { "epoch": 0.95, "learning_rate": 2.846146948116468e-06, "loss": 1.2707, "step": 7625 }, { "epoch": 0.95, "learning_rate": 2.8310164128265147e-06, "loss": 1.2277, "step": 7626 }, { "epoch": 0.95, "learning_rate": 2.815925973855715e-06, "loss": 1.2646, "step": 7627 }, { "epoch": 0.95, "learning_rate": 2.800875633652028e-06, "loss": 1.2435, "step": 7628 }, { "epoch": 0.95, "learning_rate": 2.7858653946569722e-06, "loss": 1.2333, "step": 7629 }, { "epoch": 0.95, "learning_rate": 2.770895259305517e-06, "loss": 1.1996, "step": 7630 }, { "epoch": 0.95, "learning_rate": 2.7559652300261918e-06, "loss": 1.2461, "step": 7631 }, { "epoch": 0.95, "learning_rate": 2.741075309240976e-06, "loss": 1.1534, "step": 7632 }, { "epoch": 0.95, "learning_rate": 2.726225499365326e-06, "loss": 1.1674, "step": 7633 }, { "epoch": 0.95, "learning_rate": 2.711415802808176e-06, "loss": 1.1954, "step": 7634 }, { "epoch": 0.95, "learning_rate": 2.6966462219720766e-06, "loss": 1.1722, "step": 7635 }, { "epoch": 0.95, "learning_rate": 2.681916759252917e-06, "loss": 1.1745, "step": 7636 }, { "epoch": 0.95, "learning_rate": 2.6672274170401747e-06, "loss": 1.1864, "step": 7637 }, { "epoch": 0.95, "learning_rate": 2.652578197716776e-06, "loss": 1.2262, "step": 7638 }, { "epoch": 0.95, "learning_rate": 2.6379691036591545e-06, "loss": 1.2485, "step": 7639 }, { "epoch": 0.95, "learning_rate": 2.6234001372372197e-06, "loss": 1.2172, "step": 7640 }, { "epoch": 0.96, "learning_rate": 2.608871300814386e-06, "loss": 1.2056, "step": 7641 }, { "epoch": 0.96, "learning_rate": 2.594382596747602e-06, "loss": 1.2226, "step": 7642 }, { "epoch": 0.96, "learning_rate": 2.5799340273871817e-06, "loss": 1.2069, "step": 7643 }, { "epoch": 0.96, "learning_rate": 2.5655255950770563e-06, "loss": 1.1792, "step": 7644 }, { "epoch": 0.96, "learning_rate": 2.5511573021545777e-06, "loss": 1.2242, "step": 7645 }, { "epoch": 0.96, "learning_rate": 2.5368291509506036e-06, "loss": 1.196, "step": 7646 }, { "epoch": 0.96, "learning_rate": 2.5225411437894697e-06, "loss": 1.2773, "step": 7647 }, { "epoch": 0.96, "learning_rate": 2.5082932829890437e-06, "loss": 1.199, "step": 7648 }, { "epoch": 0.96, "learning_rate": 2.494085570860616e-06, "loss": 1.1648, "step": 7649 }, { "epoch": 0.96, "learning_rate": 2.479918009708981e-06, "loss": 1.1972, "step": 7650 }, { "epoch": 0.96, "learning_rate": 2.465790601832468e-06, "loss": 1.1582, "step": 7651 }, { "epoch": 0.96, "learning_rate": 2.4517033495228547e-06, "loss": 1.2361, "step": 7652 }, { "epoch": 0.96, "learning_rate": 2.437656255065396e-06, "loss": 1.256, "step": 7653 }, { "epoch": 0.96, "learning_rate": 2.4236493207388242e-06, "loss": 1.2277, "step": 7654 }, { "epoch": 0.96, "learning_rate": 2.4096825488154062e-06, "loss": 1.2223, "step": 7655 }, { "epoch": 0.96, "learning_rate": 2.3957559415608567e-06, "loss": 1.2677, "step": 7656 }, { "epoch": 0.96, "learning_rate": 2.3818695012343683e-06, "loss": 1.237, "step": 7657 }, { "epoch": 0.96, "learning_rate": 2.3680232300886672e-06, "loss": 1.152, "step": 7658 }, { "epoch": 0.96, "learning_rate": 2.3542171303698734e-06, "loss": 1.225, "step": 7659 }, { "epoch": 0.96, "learning_rate": 2.3404512043176673e-06, "loss": 1.2226, "step": 7660 }, { "epoch": 0.96, "learning_rate": 2.3267254541652073e-06, "loss": 1.1682, "step": 7661 }, { "epoch": 0.96, "learning_rate": 2.3130398821391008e-06, "loss": 1.2373, "step": 7662 }, { "epoch": 0.96, "learning_rate": 2.299394490459461e-06, "loss": 1.2604, "step": 7663 }, { "epoch": 0.96, "learning_rate": 2.2857892813398785e-06, "loss": 1.2094, "step": 7664 }, { "epoch": 0.96, "learning_rate": 2.272224256987365e-06, "loss": 1.1748, "step": 7665 }, { "epoch": 0.96, "learning_rate": 2.25869941960255e-06, "loss": 1.191, "step": 7666 }, { "epoch": 0.96, "learning_rate": 2.245214771379428e-06, "loss": 1.1917, "step": 7667 }, { "epoch": 0.96, "learning_rate": 2.231770314505499e-06, "loss": 1.1495, "step": 7668 }, { "epoch": 0.96, "learning_rate": 2.2183660511617964e-06, "loss": 1.205, "step": 7669 }, { "epoch": 0.96, "learning_rate": 2.20500198352272e-06, "loss": 1.2645, "step": 7670 }, { "epoch": 0.96, "learning_rate": 2.1916781137562846e-06, "loss": 1.2638, "step": 7671 }, { "epoch": 0.96, "learning_rate": 2.1783944440238736e-06, "loss": 1.1912, "step": 7672 }, { "epoch": 0.96, "learning_rate": 2.165150976480429e-06, "loss": 1.2253, "step": 7673 }, { "epoch": 0.96, "learning_rate": 2.151947713274316e-06, "loss": 1.1956, "step": 7674 }, { "epoch": 0.96, "learning_rate": 2.1387846565474044e-06, "loss": 1.2201, "step": 7675 }, { "epoch": 0.96, "learning_rate": 2.1256618084350422e-06, "loss": 1.1758, "step": 7676 }, { "epoch": 0.96, "learning_rate": 2.1125791710660256e-06, "loss": 1.243, "step": 7677 }, { "epoch": 0.96, "learning_rate": 2.0995367465626582e-06, "loss": 1.1845, "step": 7678 }, { "epoch": 0.96, "learning_rate": 2.086534537040746e-06, "loss": 1.3013, "step": 7679 }, { "epoch": 0.96, "learning_rate": 2.073572544609492e-06, "loss": 1.2756, "step": 7680 }, { "epoch": 0.96, "learning_rate": 2.0606507713716584e-06, "loss": 1.2075, "step": 7681 }, { "epoch": 0.96, "learning_rate": 2.047769219423401e-06, "loss": 1.1787, "step": 7682 }, { "epoch": 0.96, "learning_rate": 2.0349278908544377e-06, "loss": 1.2353, "step": 7683 }, { "epoch": 0.96, "learning_rate": 2.02212678774788e-06, "loss": 1.1691, "step": 7684 }, { "epoch": 0.96, "learning_rate": 2.009365912180372e-06, "loss": 1.2294, "step": 7685 }, { "epoch": 0.96, "learning_rate": 1.996645266222036e-06, "loss": 1.2244, "step": 7686 }, { "epoch": 0.96, "learning_rate": 1.983964851936415e-06, "loss": 1.2254, "step": 7687 }, { "epoch": 0.96, "learning_rate": 1.9713246713805587e-06, "loss": 1.2116, "step": 7688 }, { "epoch": 0.96, "learning_rate": 1.958724726604966e-06, "loss": 1.2089, "step": 7689 }, { "epoch": 0.96, "learning_rate": 1.946165019653667e-06, "loss": 1.1927, "step": 7690 }, { "epoch": 0.96, "learning_rate": 1.9336455525640893e-06, "loss": 1.2222, "step": 7691 }, { "epoch": 0.96, "learning_rate": 1.9211663273672187e-06, "loss": 1.1454, "step": 7692 }, { "epoch": 0.96, "learning_rate": 1.908727346087408e-06, "loss": 1.2187, "step": 7693 }, { "epoch": 0.96, "learning_rate": 1.8963286107425703e-06, "loss": 1.196, "step": 7694 }, { "epoch": 0.96, "learning_rate": 1.8839701233440699e-06, "loss": 1.2061, "step": 7695 }, { "epoch": 0.96, "learning_rate": 1.8716518858966914e-06, "loss": 1.2543, "step": 7696 }, { "epoch": 0.96, "learning_rate": 1.8593739003987532e-06, "loss": 1.2512, "step": 7697 }, { "epoch": 0.96, "learning_rate": 1.847136168842023e-06, "loss": 1.2439, "step": 7698 }, { "epoch": 0.96, "learning_rate": 1.8349386932116908e-06, "loss": 1.2882, "step": 7699 }, { "epoch": 0.96, "learning_rate": 1.822781475486507e-06, "loss": 1.2552, "step": 7700 }, { "epoch": 0.96, "eval_acceptance_rate": 0.6998319029808044, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2177752048617916, "eval_expected_tokens": 2.7722190060772394, "eval_loss": 1.349587082862854, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4430547515193099, "eval_runtime": 50.2961, "eval_samples_per_second": 9.683, "eval_steps_per_second": 0.318, "step": 7700 }, { "epoch": 0.96, "learning_rate": 1.810664517638616e-06, "loss": 1.2353, "step": 7701 }, { "epoch": 0.96, "learning_rate": 1.7985878216336958e-06, "loss": 1.1587, "step": 7702 }, { "epoch": 0.96, "learning_rate": 1.7865513894307905e-06, "loss": 1.2336, "step": 7703 }, { "epoch": 0.96, "learning_rate": 1.7745552229825323e-06, "loss": 1.2319, "step": 7704 }, { "epoch": 0.96, "learning_rate": 1.762599324234948e-06, "loss": 1.2976, "step": 7705 }, { "epoch": 0.96, "learning_rate": 1.7506836951275696e-06, "loss": 1.2032, "step": 7706 }, { "epoch": 0.96, "learning_rate": 1.7388083375933506e-06, "loss": 1.2069, "step": 7707 }, { "epoch": 0.96, "learning_rate": 1.7269732535587224e-06, "loss": 1.2665, "step": 7708 }, { "epoch": 0.96, "learning_rate": 1.7151784449436492e-06, "loss": 1.2573, "step": 7709 }, { "epoch": 0.96, "learning_rate": 1.7034239136614892e-06, "loss": 1.2075, "step": 7710 }, { "epoch": 0.96, "learning_rate": 1.6917096616190786e-06, "loss": 1.2065, "step": 7711 }, { "epoch": 0.96, "learning_rate": 1.6800356907167579e-06, "loss": 1.234, "step": 7712 }, { "epoch": 0.96, "learning_rate": 1.6684020028482904e-06, "loss": 1.2252, "step": 7713 }, { "epoch": 0.96, "learning_rate": 1.6568085999008887e-06, "loss": 1.2107, "step": 7714 }, { "epoch": 0.96, "learning_rate": 1.6452554837553257e-06, "loss": 1.2472, "step": 7715 }, { "epoch": 0.96, "learning_rate": 1.6337426562857138e-06, "loss": 1.2037, "step": 7716 }, { "epoch": 0.96, "learning_rate": 1.6222701193597256e-06, "loss": 1.1816, "step": 7717 }, { "epoch": 0.96, "learning_rate": 1.6108378748384555e-06, "loss": 1.2324, "step": 7718 }, { "epoch": 0.96, "learning_rate": 1.599445924576448e-06, "loss": 1.2563, "step": 7719 }, { "epoch": 0.96, "learning_rate": 1.5880942704217528e-06, "loss": 1.2472, "step": 7720 }, { "epoch": 0.97, "learning_rate": 1.5767829142158684e-06, "loss": 1.2899, "step": 7721 }, { "epoch": 0.97, "learning_rate": 1.5655118577937167e-06, "loss": 1.2177, "step": 7722 }, { "epoch": 0.97, "learning_rate": 1.5542811029837512e-06, "loss": 1.1546, "step": 7723 }, { "epoch": 0.97, "learning_rate": 1.5430906516078202e-06, "loss": 1.2448, "step": 7724 }, { "epoch": 0.97, "learning_rate": 1.531940505481305e-06, "loss": 1.253, "step": 7725 }, { "epoch": 0.97, "learning_rate": 1.5208306664129813e-06, "loss": 1.1807, "step": 7726 }, { "epoch": 0.97, "learning_rate": 1.5097611362051012e-06, "loss": 1.2171, "step": 7727 }, { "epoch": 0.97, "learning_rate": 1.4987319166533674e-06, "loss": 1.2258, "step": 7728 }, { "epoch": 0.97, "learning_rate": 1.4877430095470156e-06, "loss": 1.2107, "step": 7729 }, { "epoch": 0.97, "learning_rate": 1.4767944166687031e-06, "loss": 1.2614, "step": 7730 }, { "epoch": 0.97, "learning_rate": 1.4658861397944811e-06, "loss": 1.2906, "step": 7731 }, { "epoch": 0.97, "learning_rate": 1.4550181806939622e-06, "loss": 1.2526, "step": 7732 }, { "epoch": 0.97, "learning_rate": 1.4441905411301248e-06, "loss": 1.1884, "step": 7733 }, { "epoch": 0.97, "learning_rate": 1.4334032228595084e-06, "loss": 1.2793, "step": 7734 }, { "epoch": 0.97, "learning_rate": 1.422656227631991e-06, "loss": 1.2855, "step": 7735 }, { "epoch": 0.97, "learning_rate": 1.411949557191039e-06, "loss": 1.2724, "step": 7736 }, { "epoch": 0.97, "learning_rate": 1.4012832132734854e-06, "loss": 1.2568, "step": 7737 }, { "epoch": 0.97, "learning_rate": 1.3906571976096405e-06, "loss": 1.2352, "step": 7738 }, { "epoch": 0.97, "learning_rate": 1.380071511923292e-06, "loss": 1.2231, "step": 7739 }, { "epoch": 0.97, "learning_rate": 1.3695261579316775e-06, "loss": 1.1568, "step": 7740 }, { "epoch": 0.97, "learning_rate": 1.3590211373454564e-06, "loss": 1.1573, "step": 7741 }, { "epoch": 0.97, "learning_rate": 1.348556451868821e-06, "loss": 1.2356, "step": 7742 }, { "epoch": 0.97, "learning_rate": 1.3381321031993577e-06, "loss": 1.2631, "step": 7743 }, { "epoch": 0.97, "learning_rate": 1.3277480930281305e-06, "loss": 1.2068, "step": 7744 }, { "epoch": 0.97, "learning_rate": 1.3174044230396531e-06, "loss": 1.1419, "step": 7745 }, { "epoch": 0.97, "learning_rate": 1.3071010949119166e-06, "loss": 1.2189, "step": 7746 }, { "epoch": 0.97, "learning_rate": 1.2968381103163062e-06, "loss": 1.2552, "step": 7747 }, { "epoch": 0.97, "learning_rate": 1.286615470917768e-06, "loss": 1.1781, "step": 7748 }, { "epoch": 0.97, "learning_rate": 1.2764331783745864e-06, "loss": 1.2539, "step": 7749 }, { "epoch": 0.97, "learning_rate": 1.2662912343386067e-06, "loss": 1.1907, "step": 7750 }, { "epoch": 0.97, "learning_rate": 1.256189640455041e-06, "loss": 1.2139, "step": 7751 }, { "epoch": 0.97, "learning_rate": 1.246128398362606e-06, "loss": 1.2203, "step": 7752 }, { "epoch": 0.97, "learning_rate": 1.2361075096934404e-06, "loss": 1.2888, "step": 7753 }, { "epoch": 0.97, "learning_rate": 1.2261269760731608e-06, "loss": 1.1905, "step": 7754 }, { "epoch": 0.97, "learning_rate": 1.2161867991208885e-06, "loss": 1.2014, "step": 7755 }, { "epoch": 0.97, "learning_rate": 1.2062869804490562e-06, "loss": 1.246, "step": 7756 }, { "epoch": 0.97, "learning_rate": 1.1964275216637122e-06, "loss": 1.1944, "step": 7757 }, { "epoch": 0.97, "learning_rate": 1.1866084243642162e-06, "loss": 1.2075, "step": 7758 }, { "epoch": 0.97, "learning_rate": 1.1768296901434883e-06, "loss": 1.2372, "step": 7759 }, { "epoch": 0.97, "learning_rate": 1.167091320587843e-06, "loss": 1.2301, "step": 7760 }, { "epoch": 0.97, "learning_rate": 1.1573933172770722e-06, "loss": 1.2315, "step": 7761 }, { "epoch": 0.97, "learning_rate": 1.147735681784362e-06, "loss": 1.1724, "step": 7762 }, { "epoch": 0.97, "learning_rate": 1.138118415676459e-06, "loss": 1.2196, "step": 7763 }, { "epoch": 0.97, "learning_rate": 1.1285415205134486e-06, "loss": 1.2145, "step": 7764 }, { "epoch": 0.97, "learning_rate": 1.119004997848977e-06, "loss": 1.2056, "step": 7765 }, { "epoch": 0.97, "learning_rate": 1.109508849230001e-06, "loss": 1.2376, "step": 7766 }, { "epoch": 0.97, "learning_rate": 1.1000530761970661e-06, "loss": 1.2197, "step": 7767 }, { "epoch": 0.97, "learning_rate": 1.0906376802840846e-06, "loss": 1.247, "step": 7768 }, { "epoch": 0.97, "learning_rate": 1.0812626630184453e-06, "loss": 1.2649, "step": 7769 }, { "epoch": 0.97, "learning_rate": 1.0719280259210151e-06, "loss": 1.167, "step": 7770 }, { "epoch": 0.97, "learning_rate": 1.0626337705060273e-06, "loss": 1.2512, "step": 7771 }, { "epoch": 0.97, "learning_rate": 1.053379898281248e-06, "loss": 1.262, "step": 7772 }, { "epoch": 0.97, "learning_rate": 1.0441664107478655e-06, "loss": 1.2685, "step": 7773 }, { "epoch": 0.97, "learning_rate": 1.034993309400517e-06, "loss": 1.2499, "step": 7774 }, { "epoch": 0.97, "learning_rate": 1.0258605957272626e-06, "loss": 1.2748, "step": 7775 }, { "epoch": 0.97, "learning_rate": 1.0167682712096393e-06, "loss": 1.187, "step": 7776 }, { "epoch": 0.97, "learning_rate": 1.007716337322634e-06, "loss": 1.2188, "step": 7777 }, { "epoch": 0.97, "learning_rate": 9.98704795534655e-07, "loss": 1.2161, "step": 7778 }, { "epoch": 0.97, "learning_rate": 9.897336473076167e-07, "loss": 1.1695, "step": 7779 }, { "epoch": 0.97, "learning_rate": 9.808028940967717e-07, "loss": 1.235, "step": 7780 }, { "epoch": 0.97, "learning_rate": 9.719125373509607e-07, "loss": 1.2061, "step": 7781 }, { "epoch": 0.97, "learning_rate": 9.630625785123358e-07, "loss": 1.214, "step": 7782 }, { "epoch": 0.97, "learning_rate": 9.542530190165822e-07, "loss": 1.2243, "step": 7783 }, { "epoch": 0.97, "learning_rate": 9.45483860292834e-07, "loss": 1.2352, "step": 7784 }, { "epoch": 0.97, "learning_rate": 9.367551037635924e-07, "loss": 1.2614, "step": 7785 }, { "epoch": 0.97, "learning_rate": 9.280667508448915e-07, "loss": 1.1946, "step": 7786 }, { "epoch": 0.97, "learning_rate": 9.194188029461592e-07, "loss": 1.2182, "step": 7787 }, { "epoch": 0.97, "learning_rate": 9.108112614703013e-07, "loss": 1.2398, "step": 7788 }, { "epoch": 0.97, "learning_rate": 9.02244127813645e-07, "loss": 1.2156, "step": 7789 }, { "epoch": 0.97, "learning_rate": 8.937174033659956e-07, "loss": 1.1895, "step": 7790 }, { "epoch": 0.97, "learning_rate": 8.852310895105243e-07, "loss": 1.2405, "step": 7791 }, { "epoch": 0.97, "learning_rate": 8.767851876239075e-07, "loss": 1.1861, "step": 7792 }, { "epoch": 0.97, "learning_rate": 8.683796990762993e-07, "loss": 1.1904, "step": 7793 }, { "epoch": 0.97, "learning_rate": 8.60014625231248e-07, "loss": 1.1572, "step": 7794 }, { "epoch": 0.97, "learning_rate": 8.516899674457513e-07, "loss": 1.1338, "step": 7795 }, { "epoch": 0.97, "learning_rate": 8.434057270702288e-07, "loss": 1.1671, "step": 7796 }, { "epoch": 0.97, "learning_rate": 8.35161905448606e-07, "loss": 1.134, "step": 7797 }, { "epoch": 0.97, "learning_rate": 8.269585039181738e-07, "loss": 1.2507, "step": 7798 }, { "epoch": 0.97, "learning_rate": 8.187955238097288e-07, "loss": 1.2661, "step": 7799 }, { "epoch": 0.97, "learning_rate": 8.106729664475177e-07, "loss": 1.2004, "step": 7800 }, { "epoch": 0.97, "eval_acceptance_rate": 0.6998359560966492, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2177921955045523, "eval_expected_tokens": 2.7722402443806904, "eval_loss": 1.3495434522628784, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44306006109517254, "eval_runtime": 50.2597, "eval_samples_per_second": 9.69, "eval_steps_per_second": 0.318, "step": 7800 }, { "epoch": 0.98, "learning_rate": 8.025908331491527e-07, "loss": 1.2235, "step": 7801 }, { "epoch": 0.98, "learning_rate": 7.945491252257519e-07, "loss": 1.1228, "step": 7802 }, { "epoch": 0.98, "learning_rate": 7.865478439818552e-07, "loss": 1.1954, "step": 7803 }, { "epoch": 0.98, "learning_rate": 7.785869907154797e-07, "loss": 1.2011, "step": 7804 }, { "epoch": 0.98, "learning_rate": 7.70666566718009e-07, "loss": 1.1948, "step": 7805 }, { "epoch": 0.98, "learning_rate": 7.627865732743322e-07, "loss": 1.296, "step": 7806 }, { "epoch": 0.98, "learning_rate": 7.549470116627876e-07, "loss": 1.1576, "step": 7807 }, { "epoch": 0.98, "learning_rate": 7.471478831550805e-07, "loss": 1.2674, "step": 7808 }, { "epoch": 0.98, "learning_rate": 7.393891890164206e-07, "loss": 1.1618, "step": 7809 }, { "epoch": 0.98, "learning_rate": 7.316709305054404e-07, "loss": 1.2776, "step": 7810 }, { "epoch": 0.98, "learning_rate": 7.239931088742213e-07, "loss": 1.1817, "step": 7811 }, { "epoch": 0.98, "learning_rate": 7.163557253682673e-07, "loss": 1.1776, "step": 7812 }, { "epoch": 0.98, "learning_rate": 7.087587812265595e-07, "loss": 1.1706, "step": 7813 }, { "epoch": 0.98, "learning_rate": 7.01202277681473e-07, "loss": 1.235, "step": 7814 }, { "epoch": 0.98, "learning_rate": 6.936862159588053e-07, "loss": 1.1833, "step": 7815 }, { "epoch": 0.98, "learning_rate": 6.862105972778865e-07, "loss": 1.2086, "step": 7816 }, { "epoch": 0.98, "learning_rate": 6.787754228513854e-07, "loss": 1.1907, "step": 7817 }, { "epoch": 0.98, "learning_rate": 6.713806938854761e-07, "loss": 1.1968, "step": 7818 }, { "epoch": 0.98, "learning_rate": 6.640264115797546e-07, "loss": 1.2029, "step": 7819 }, { "epoch": 0.98, "learning_rate": 6.567125771272386e-07, "loss": 1.2159, "step": 7820 }, { "epoch": 0.98, "learning_rate": 6.494391917143683e-07, "loss": 1.1865, "step": 7821 }, { "epoch": 0.98, "learning_rate": 6.422062565210884e-07, "loss": 1.2328, "step": 7822 }, { "epoch": 0.98, "learning_rate": 6.350137727207384e-07, "loss": 1.2162, "step": 7823 }, { "epoch": 0.98, "learning_rate": 6.278617414800791e-07, "loss": 1.2715, "step": 7824 }, { "epoch": 0.98, "learning_rate": 6.207501639593493e-07, "loss": 1.2145, "step": 7825 }, { "epoch": 0.98, "learning_rate": 6.136790413122095e-07, "loss": 1.2487, "step": 7826 }, { "epoch": 0.98, "learning_rate": 6.066483746857143e-07, "loss": 1.2187, "step": 7827 }, { "epoch": 0.98, "learning_rate": 5.996581652204514e-07, "loss": 1.212, "step": 7828 }, { "epoch": 0.98, "learning_rate": 5.927084140503469e-07, "loss": 1.233, "step": 7829 }, { "epoch": 0.98, "learning_rate": 5.857991223028047e-07, "loss": 1.2144, "step": 7830 }, { "epoch": 0.98, "learning_rate": 5.789302910986782e-07, "loss": 1.1794, "step": 7831 }, { "epoch": 0.98, "learning_rate": 5.721019215522427e-07, "loss": 1.224, "step": 7832 }, { "epoch": 0.98, "learning_rate": 5.653140147712232e-07, "loss": 1.2856, "step": 7833 }, { "epoch": 0.98, "learning_rate": 5.585665718567667e-07, "loss": 1.1964, "step": 7834 }, { "epoch": 0.98, "learning_rate": 5.518595939034421e-07, "loss": 1.1811, "step": 7835 }, { "epoch": 0.98, "learning_rate": 5.451930819992957e-07, "loss": 1.2431, "step": 7836 }, { "epoch": 0.98, "learning_rate": 5.385670372257401e-07, "loss": 1.162, "step": 7837 }, { "epoch": 0.98, "learning_rate": 5.319814606577211e-07, "loss": 1.2202, "step": 7838 }, { "epoch": 0.98, "learning_rate": 5.254363533635509e-07, "loss": 1.2241, "step": 7839 }, { "epoch": 0.98, "learning_rate": 5.189317164049634e-07, "loss": 1.2104, "step": 7840 }, { "epoch": 0.98, "learning_rate": 5.124675508371701e-07, "loss": 1.2378, "step": 7841 }, { "epoch": 0.98, "learning_rate": 5.060438577088322e-07, "loss": 1.2354, "step": 7842 }, { "epoch": 0.98, "learning_rate": 4.996606380619773e-07, "loss": 1.248, "step": 7843 }, { "epoch": 0.98, "learning_rate": 4.933178929321103e-07, "loss": 1.2218, "step": 7844 }, { "epoch": 0.98, "learning_rate": 4.870156233481859e-07, "loss": 1.2277, "step": 7845 }, { "epoch": 0.98, "learning_rate": 4.807538303325531e-07, "loss": 1.2809, "step": 7846 }, { "epoch": 0.98, "learning_rate": 4.745325149010382e-07, "loss": 1.2037, "step": 7847 }, { "epoch": 0.98, "learning_rate": 4.6835167806286163e-07, "loss": 1.209, "step": 7848 }, { "epoch": 0.98, "learning_rate": 4.622113208206935e-07, "loss": 1.1948, "step": 7849 }, { "epoch": 0.98, "learning_rate": 4.56111444170626e-07, "loss": 1.1912, "step": 7850 }, { "epoch": 0.98, "learning_rate": 4.500520491022286e-07, "loss": 1.2799, "step": 7851 }, { "epoch": 0.98, "learning_rate": 4.4403313659843733e-07, "loss": 1.2067, "step": 7852 }, { "epoch": 0.98, "learning_rate": 4.3805470763566557e-07, "loss": 1.2002, "step": 7853 }, { "epoch": 0.98, "learning_rate": 4.3211676318372086e-07, "loss": 1.2013, "step": 7854 }, { "epoch": 0.98, "learning_rate": 4.262193042059437e-07, "loss": 1.2379, "step": 7855 }, { "epoch": 0.98, "learning_rate": 4.2036233165893e-07, "loss": 1.2493, "step": 7856 }, { "epoch": 0.98, "learning_rate": 4.145458464929197e-07, "loss": 1.1856, "step": 7857 }, { "epoch": 0.98, "learning_rate": 4.0876984965140805e-07, "loss": 1.1621, "step": 7858 }, { "epoch": 0.98, "learning_rate": 4.0303434207139556e-07, "loss": 1.1942, "step": 7859 }, { "epoch": 0.98, "learning_rate": 3.9733932468333236e-07, "loss": 1.2083, "step": 7860 }, { "epoch": 0.98, "learning_rate": 3.91684798411035e-07, "loss": 1.2038, "step": 7861 }, { "epoch": 0.98, "learning_rate": 3.8607076417185307e-07, "loss": 1.2353, "step": 7862 }, { "epoch": 0.98, "learning_rate": 3.804972228764747e-07, "loss": 1.2298, "step": 7863 }, { "epoch": 0.98, "learning_rate": 3.7496417542906557e-07, "loss": 1.172, "step": 7864 }, { "epoch": 0.98, "learning_rate": 3.694716227271855e-07, "loss": 1.241, "step": 7865 }, { "epoch": 0.98, "learning_rate": 3.6401956566187166e-07, "loss": 1.2117, "step": 7866 }, { "epoch": 0.98, "learning_rate": 3.5860800511755554e-07, "loss": 1.224, "step": 7867 }, { "epoch": 0.98, "learning_rate": 3.532369419721182e-07, "loss": 1.2206, "step": 7868 }, { "epoch": 0.98, "learning_rate": 3.479063770968627e-07, "loss": 1.1587, "step": 7869 }, { "epoch": 0.98, "learning_rate": 3.426163113565417e-07, "loss": 1.1786, "step": 7870 }, { "epoch": 0.98, "learning_rate": 3.3736674560930215e-07, "loss": 1.2252, "step": 7871 }, { "epoch": 0.98, "learning_rate": 3.3215768070674055e-07, "loss": 1.1804, "step": 7872 }, { "epoch": 0.98, "learning_rate": 3.2698911749390324e-07, "loss": 1.2295, "step": 7873 }, { "epoch": 0.98, "learning_rate": 3.218610568092306e-07, "loss": 1.2315, "step": 7874 }, { "epoch": 0.98, "learning_rate": 3.167734994846128e-07, "loss": 1.2312, "step": 7875 }, { "epoch": 0.98, "learning_rate": 3.117264463453617e-07, "loss": 1.2209, "step": 7876 }, { "epoch": 0.98, "learning_rate": 3.0671989821021154e-07, "loss": 1.2404, "step": 7877 }, { "epoch": 0.98, "learning_rate": 3.0175385589134576e-07, "loss": 1.2316, "step": 7878 }, { "epoch": 0.98, "learning_rate": 2.9682832019439775e-07, "loss": 1.2177, "step": 7879 }, { "epoch": 0.98, "learning_rate": 2.9194329191833955e-07, "loss": 1.2232, "step": 7880 }, { "epoch": 0.99, "learning_rate": 2.870987718557039e-07, "loss": 1.2538, "step": 7881 }, { "epoch": 0.99, "learning_rate": 2.8229476079230655e-07, "loss": 1.1428, "step": 7882 }, { "epoch": 0.99, "learning_rate": 2.7753125950752413e-07, "loss": 1.221, "step": 7883 }, { "epoch": 0.99, "learning_rate": 2.728082687740441e-07, "loss": 1.186, "step": 7884 }, { "epoch": 0.99, "learning_rate": 2.6812578935808686e-07, "loss": 1.2017, "step": 7885 }, { "epoch": 0.99, "learning_rate": 2.6348382201923926e-07, "loss": 1.273, "step": 7886 }, { "epoch": 0.99, "learning_rate": 2.5888236751053784e-07, "loss": 1.1814, "step": 7887 }, { "epoch": 0.99, "learning_rate": 2.54321426578441e-07, "loss": 1.1654, "step": 7888 }, { "epoch": 0.99, "learning_rate": 2.4980099996282923e-07, "loss": 1.1399, "step": 7889 }, { "epoch": 0.99, "learning_rate": 2.453210883970325e-07, "loss": 1.2432, "step": 7890 }, { "epoch": 0.99, "learning_rate": 2.4088169260774727e-07, "loss": 1.1775, "step": 7891 }, { "epoch": 0.99, "learning_rate": 2.3648281331520306e-07, "loss": 1.1703, "step": 7892 }, { "epoch": 0.99, "learning_rate": 2.3212445123294014e-07, "loss": 1.2165, "step": 7893 }, { "epoch": 0.99, "learning_rate": 2.278066070680318e-07, "loss": 1.2096, "step": 7894 }, { "epoch": 0.99, "learning_rate": 2.2352928152089003e-07, "loss": 1.1695, "step": 7895 }, { "epoch": 0.99, "learning_rate": 2.1929247528540418e-07, "loss": 1.2686, "step": 7896 }, { "epoch": 0.99, "learning_rate": 2.150961890488856e-07, "loss": 1.2375, "step": 7897 }, { "epoch": 0.99, "learning_rate": 2.109404234920953e-07, "loss": 1.2491, "step": 7898 }, { "epoch": 0.99, "learning_rate": 2.068251792891329e-07, "loss": 1.2059, "step": 7899 }, { "epoch": 0.99, "learning_rate": 2.027504571076033e-07, "loss": 1.201, "step": 7900 }, { "epoch": 0.99, "eval_acceptance_rate": 0.6998305916786194, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.217769707923113, "eval_expected_tokens": 2.7722121349038917, "eval_loss": 1.3495228290557861, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.44305303372597293, "eval_runtime": 50.3088, "eval_samples_per_second": 9.68, "eval_steps_per_second": 0.318, "step": 7900 }, { "epoch": 0.99, "learning_rate": 1.9871625760850553e-07, "loss": 1.2019, "step": 7901 }, { "epoch": 0.99, "learning_rate": 1.9472258144631604e-07, "loss": 1.2238, "step": 7902 }, { "epoch": 0.99, "learning_rate": 1.9076942926887775e-07, "loss": 1.246, "step": 7903 }, { "epoch": 0.99, "learning_rate": 1.8685680171745544e-07, "loss": 1.1885, "step": 7904 }, { "epoch": 0.99, "learning_rate": 1.8298469942681917e-07, "loss": 1.2354, "step": 7905 }, { "epoch": 0.99, "learning_rate": 1.7915312302504982e-07, "loss": 1.2488, "step": 7906 }, { "epoch": 0.99, "learning_rate": 1.7536207313376128e-07, "loss": 1.225, "step": 7907 }, { "epoch": 0.99, "learning_rate": 1.716115503679061e-07, "loss": 1.2501, "step": 7908 }, { "epoch": 0.99, "learning_rate": 1.67901555335942e-07, "loss": 1.2045, "step": 7909 }, { "epoch": 0.99, "learning_rate": 1.6423208863969309e-07, "loss": 1.2002, "step": 7910 }, { "epoch": 0.99, "learning_rate": 1.6060315087440547e-07, "loss": 1.2143, "step": 7911 }, { "epoch": 0.99, "learning_rate": 1.5701474262883043e-07, "loss": 1.1422, "step": 7912 }, { "epoch": 0.99, "learning_rate": 1.5346686448503011e-07, "loss": 1.2295, "step": 7913 }, { "epoch": 0.99, "learning_rate": 1.4995951701859967e-07, "loss": 1.1715, "step": 7914 }, { "epoch": 0.99, "learning_rate": 1.4649270079847288e-07, "loss": 1.235, "step": 7915 }, { "epoch": 0.99, "learning_rate": 1.4306641638706096e-07, "loss": 1.238, "step": 7916 }, { "epoch": 0.99, "learning_rate": 1.3968066434016935e-07, "loss": 1.2007, "step": 7917 }, { "epoch": 0.99, "learning_rate": 1.363354452070531e-07, "loss": 1.207, "step": 7918 }, { "epoch": 0.99, "learning_rate": 1.3303075953038924e-07, "loss": 1.1991, "step": 7919 }, { "epoch": 0.99, "learning_rate": 1.297666078462767e-07, "loss": 1.1904, "step": 7920 }, { "epoch": 0.99, "learning_rate": 1.2654299068418084e-07, "loss": 1.1226, "step": 7921 }, { "epoch": 0.99, "learning_rate": 1.233599085671e-07, "loss": 1.2414, "step": 7922 }, { "epoch": 0.99, "learning_rate": 1.2021736201139888e-07, "loss": 1.2147, "step": 7923 }, { "epoch": 0.99, "learning_rate": 1.1711535152683639e-07, "loss": 1.2612, "step": 7924 }, { "epoch": 0.99, "learning_rate": 1.1405387761664887e-07, "loss": 1.227, "step": 7925 }, { "epoch": 0.99, "learning_rate": 1.1103294077749459e-07, "loss": 1.2367, "step": 7926 }, { "epoch": 0.99, "learning_rate": 1.080525414993705e-07, "loss": 1.2602, "step": 7927 }, { "epoch": 0.99, "learning_rate": 1.0511268026583421e-07, "loss": 1.194, "step": 7928 }, { "epoch": 0.99, "learning_rate": 1.0221335755378203e-07, "loss": 1.2532, "step": 7929 }, { "epoch": 0.99, "learning_rate": 9.935457383350443e-08, "loss": 1.1804, "step": 7930 }, { "epoch": 0.99, "learning_rate": 9.653632956882486e-08, "loss": 1.1685, "step": 7931 }, { "epoch": 0.99, "learning_rate": 9.375862521687761e-08, "loss": 1.2395, "step": 7932 }, { "epoch": 0.99, "learning_rate": 9.102146122827449e-08, "loss": 1.2948, "step": 7933 }, { "epoch": 0.99, "learning_rate": 8.832483804704916e-08, "loss": 1.2224, "step": 7934 }, { "epoch": 0.99, "learning_rate": 8.566875611068504e-08, "loss": 1.2408, "step": 7935 }, { "epoch": 0.99, "learning_rate": 8.305321585000414e-08, "loss": 1.1882, "step": 7936 }, { "epoch": 0.99, "learning_rate": 8.047821768936148e-08, "loss": 1.1889, "step": 7937 }, { "epoch": 0.99, "learning_rate": 7.794376204642295e-08, "loss": 1.22, "step": 7938 }, { "epoch": 0.99, "learning_rate": 7.544984933235966e-08, "loss": 1.2443, "step": 7939 }, { "epoch": 0.99, "learning_rate": 7.299647995176462e-08, "loss": 1.2221, "step": 7940 }, { "epoch": 0.99, "learning_rate": 7.058365430259727e-08, "loss": 1.2358, "step": 7941 }, { "epoch": 0.99, "learning_rate": 6.821137277626677e-08, "loss": 1.2267, "step": 7942 }, { "epoch": 0.99, "learning_rate": 6.587963575763189e-08, "loss": 1.1755, "step": 7943 }, { "epoch": 0.99, "learning_rate": 6.358844362494565e-08, "loss": 1.1679, "step": 7944 }, { "epoch": 0.99, "learning_rate": 6.133779674988293e-08, "loss": 1.2462, "step": 7945 }, { "epoch": 0.99, "learning_rate": 5.91276954975406e-08, "loss": 1.182, "step": 7946 }, { "epoch": 0.99, "learning_rate": 5.695814022649293e-08, "loss": 1.2626, "step": 7947 }, { "epoch": 0.99, "learning_rate": 5.48291312886251e-08, "loss": 1.2326, "step": 7948 }, { "epoch": 0.99, "learning_rate": 5.2740669029355256e-08, "loss": 1.2174, "step": 7949 }, { "epoch": 0.99, "learning_rate": 5.069275378746796e-08, "loss": 1.2174, "step": 7950 }, { "epoch": 0.99, "learning_rate": 4.8685385895169687e-08, "loss": 1.1782, "step": 7951 }, { "epoch": 0.99, "learning_rate": 4.671856567811661e-08, "loss": 1.2096, "step": 7952 }, { "epoch": 0.99, "learning_rate": 4.479229345535907e-08, "loss": 1.2657, "step": 7953 }, { "epoch": 0.99, "learning_rate": 4.2906569539397086e-08, "loss": 1.2019, "step": 7954 }, { "epoch": 0.99, "learning_rate": 4.106139423612487e-08, "loss": 1.2372, "step": 7955 }, { "epoch": 0.99, "learning_rate": 3.925676784485854e-08, "loss": 1.2544, "step": 7956 }, { "epoch": 0.99, "learning_rate": 3.749269065836391e-08, "loss": 1.1728, "step": 7957 }, { "epoch": 0.99, "learning_rate": 3.576916296280097e-08, "loss": 1.2248, "step": 7958 }, { "epoch": 0.99, "learning_rate": 3.4086185037779384e-08, "loss": 1.1482, "step": 7959 }, { "epoch": 0.99, "learning_rate": 3.2443757156330744e-08, "loss": 1.2207, "step": 7960 }, { "epoch": 1.0, "learning_rate": 3.084187958485307e-08, "loss": 1.2674, "step": 7961 }, { "epoch": 1.0, "learning_rate": 2.928055258322182e-08, "loss": 1.2679, "step": 7962 }, { "epoch": 1.0, "learning_rate": 2.775977640473437e-08, "loss": 1.2272, "step": 7963 }, { "epoch": 1.0, "learning_rate": 2.6279551296082283e-08, "loss": 1.2498, "step": 7964 }, { "epoch": 1.0, "learning_rate": 2.4839877497406794e-08, "loss": 1.2134, "step": 7965 }, { "epoch": 1.0, "learning_rate": 2.3440755242243316e-08, "loss": 1.2336, "step": 7966 }, { "epoch": 1.0, "learning_rate": 2.2082184757521438e-08, "loss": 1.186, "step": 7967 }, { "epoch": 1.0, "learning_rate": 2.07641662637037e-08, "loss": 1.2338, "step": 7968 }, { "epoch": 1.0, "learning_rate": 1.948669997456354e-08, "loss": 1.181, "step": 7969 }, { "epoch": 1.0, "learning_rate": 1.8249786097324085e-08, "loss": 1.1817, "step": 7970 }, { "epoch": 1.0, "learning_rate": 1.705342483265815e-08, "loss": 1.3268, "step": 7971 }, { "epoch": 1.0, "learning_rate": 1.589761637466047e-08, "loss": 1.2372, "step": 7972 }, { "epoch": 1.0, "learning_rate": 1.4782360910764459e-08, "loss": 1.1866, "step": 7973 }, { "epoch": 1.0, "learning_rate": 1.3707658621964214e-08, "loss": 1.2008, "step": 7974 }, { "epoch": 1.0, "learning_rate": 1.2673509682564754e-08, "loss": 1.1932, "step": 7975 }, { "epoch": 1.0, "learning_rate": 1.1679914260320778e-08, "loss": 1.175, "step": 7976 }, { "epoch": 1.0, "learning_rate": 1.0726872516436669e-08, "loss": 1.2699, "step": 7977 }, { "epoch": 1.0, "learning_rate": 9.814384605510985e-09, "loss": 1.1923, "step": 7978 }, { "epoch": 1.0, "learning_rate": 8.94245067553645e-09, "loss": 1.2027, "step": 7979 }, { "epoch": 1.0, "learning_rate": 8.111070868010995e-09, "loss": 1.194, "step": 7980 }, { "epoch": 1.0, "learning_rate": 7.320245317798957e-09, "loss": 1.1902, "step": 7981 }, { "epoch": 1.0, "learning_rate": 6.569974153131097e-09, "loss": 1.1503, "step": 7982 }, { "epoch": 1.0, "learning_rate": 5.860257495798882e-09, "loss": 1.1691, "step": 7983 }, { "epoch": 1.0, "learning_rate": 5.191095460876927e-09, "loss": 1.1984, "step": 7984 }, { "epoch": 1.0, "learning_rate": 4.562488156917289e-09, "loss": 1.2005, "step": 7985 }, { "epoch": 1.0, "learning_rate": 3.974435685949463e-09, "loss": 1.1903, "step": 7986 }, { "epoch": 1.0, "learning_rate": 3.4269381433138515e-09, "loss": 1.1507, "step": 7987 }, { "epoch": 1.0, "learning_rate": 2.919995617828297e-09, "loss": 1.2, "step": 7988 }, { "epoch": 1.0, "learning_rate": 2.4536081917603257e-09, "loss": 1.2447, "step": 7989 }, { "epoch": 1.0, "learning_rate": 2.0277759407438814e-09, "loss": 1.192, "step": 7990 }, { "epoch": 1.0, "learning_rate": 1.6424989338903463e-09, "loss": 1.2855, "step": 7991 }, { "epoch": 1.0, "learning_rate": 1.2977772336775219e-09, "loss": 1.1999, "step": 7992 }, { "epoch": 1.0, "learning_rate": 9.936108960051371e-10, "loss": 1.1748, "step": 7993 }, { "epoch": 1.0, "learning_rate": 7.299999702503612e-10, "loss": 1.1783, "step": 7994 }, { "epoch": 1.0, "learning_rate": 5.069444991845362e-10, "loss": 1.1831, "step": 7995 }, { "epoch": 1.0, "learning_rate": 3.244445189454215e-10, "loss": 1.2574, "step": 7996 }, { "epoch": 1.0, "learning_rate": 1.825000591759718e-10, "loss": 1.2587, "step": 7997 }, { "epoch": 1.0, "learning_rate": 8.111114288555932e-11, "loss": 1.1693, "step": 7998 }, { "epoch": 1.0, "learning_rate": 2.0277786533240416e-11, "loss": 1.2539, "step": 7999 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.1838, "step": 8000 }, { "epoch": 1.0, "eval_acceptance_rate": 0.6998428702354431, "eval_avg_generation_length": 219.1566619873047, "eval_avg_input_length": 366.3496398925781, "eval_expected_speedup": 2.2178211799084955, "eval_expected_tokens": 2.7722764748856195, "eval_loss": 1.3495230674743652, "eval_max_generation_length": 427.55841064453125, "eval_max_input_length": 628.2479248046875, "eval_min_generation_length": 63.83913803100586, "eval_min_input_length": 164.192626953125, "eval_pulsar_acceptance_rate": 0.4430691187214048, "eval_runtime": 50.2194, "eval_samples_per_second": 9.697, "eval_steps_per_second": 0.319, "step": 8000 } ], "logging_steps": 1.0, "max_steps": 8000, "num_input_tokens_seen": 0, "num_train_epochs": 9223372036854775807, "save_steps": 1000, "total_flos": 4.880022587100365e+19, "train_batch_size": 4, "trial_name": null, "trial_params": null }