End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 99.98,
-    "eval_loss": 0.43458130955696106,
-    "eval_runtime": 2.5109,
+    "eval_loss": 0.42036938667297363,
+    "eval_runtime": 2.4874,
     "eval_samples": 17,
-    "eval_samples_per_second": 6.77,
-    "eval_steps_per_second": 1.991,
-    "eval_wer": 0.22418879056047197,
-    "train_loss": 0.679512545520609,
-    "train_runtime": 4168.4023,
+    "eval_samples_per_second": 6.835,
+    "eval_steps_per_second": 2.01,
+    "eval_wer": 0.18436578171091444,
+    "train_loss": 0.2150352662259882,
+    "train_runtime": 4203.5944,
     "train_samples": 179,
-    "train_samples_per_second": 4.294,
-    "train_steps_per_second": 0.264
+    "train_samples_per_second": 4.258,
+    "train_steps_per_second": 0.262
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 99.98,
-    "eval_loss": 0.43458130955696106,
-    "eval_runtime": 2.5109,
+    "eval_loss": 0.42036938667297363,
+    "eval_runtime": 2.4874,
     "eval_samples": 17,
-    "eval_samples_per_second": 6.77,
-    "eval_steps_per_second": 1.991,
-    "eval_wer": 0.22418879056047197
+    "eval_samples_per_second": 6.835,
+    "eval_steps_per_second": 2.01,
+    "eval_wer": 0.18436578171091444
 }

runs/Nov25_06-38-15_4b2aca464fe8/events.out.tfevents.1669362984.4b2aca464fe8.10347.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed0cc73049cc1851efdfe8936e8b5634f576b073c7d9335da8de9c67b26e2962
+size 358

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 99.98,
-    "train_loss": 0.679512545520609,
-    "train_runtime": 4168.4023,
+    "train_loss": 0.2150352662259882,
+    "train_runtime": 4203.5944,
     "train_samples": 179,
-    "train_samples_per_second": 4.294,
-    "train_steps_per_second": 0.264
+    "train_samples_per_second": 4.258,
+    "train_steps_per_second": 0.262
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.43458130955696106,
-  "best_model_checkpoint": "ai-light-dance_drums_ft_pretrain_wav2vec2-base-new_onset-idmt-mdb-2/checkpoint-638",
+  "best_metric": 0.42036938667297363,
+  "best_model_checkpoint": "ai-light-dance_drums_ft_pretrain_wav2vec2-base-new_onset-idmt-mdb-2/checkpoint-374",
   "epoch": 99.97777777777777,
   "global_step": 1100,
   "is_hyper_param_search": false,
@@ -9,1572 +9,1572 @@
   "log_history": [
     {
       "epoch": 0.89,
-      "learning_rate": 3.9999999999999996e-05,
-      "loss": 22.7802,
+      "learning_rate": 9.999999999999999e-05,
+      "loss": 0.2599,
       "step": 10
     },
     {
       "epoch": 0.98,
-      "eval_loss": 60.154850006103516,
-      "eval_runtime": 2.4092,
-      "eval_samples_per_second": 7.056,
-      "eval_steps_per_second": 2.075,
-      "eval_wer": 0.9882005899705014,
+      "eval_loss": 0.42808040976524353,
+      "eval_runtime": 2.3932,
+      "eval_samples_per_second": 7.103,
+      "eval_steps_per_second": 2.089,
+      "eval_wer": 0.21976401179941002,
       "step": 11
     },
     {
       "epoch": 1.8,
-      "learning_rate": 0.00014,
-      "loss": 13.7635,
+      "learning_rate": 0.00019999999999999998,
+      "loss": 0.2491,
       "step": 20
     },
     {
       "epoch": 1.98,
-      "eval_loss": 18.182188034057617,
-      "eval_runtime": 2.5023,
-      "eval_samples_per_second": 6.794,
-      "eval_steps_per_second": 1.998,
-      "eval_wer": 0.9985250737463127,
+      "eval_loss": 0.48912692070007324,
+      "eval_runtime": 2.4857,
+      "eval_samples_per_second": 6.839,
+      "eval_steps_per_second": 2.011,
+      "eval_wer": 0.19469026548672566,
       "step": 22
     },
     {
       "epoch": 2.71,
-      "learning_rate": 0.00023999999999999998,
-      "loss": 3.4364,
+      "learning_rate": 0.0003,
+      "loss": 0.2619,
       "step": 30
     },
     {
       "epoch": 2.98,
-      "eval_loss": 1.2338517904281616,
-      "eval_runtime": 2.5133,
-      "eval_samples_per_second": 6.764,
-      "eval_steps_per_second": 1.989,
-      "eval_wer": 0.7315634218289085,
+      "eval_loss": 0.5496053099632263,
+      "eval_runtime": 2.4967,
+      "eval_samples_per_second": 6.809,
+      "eval_steps_per_second": 2.003,
+      "eval_wer": 0.2182890855457227,
       "step": 33
     },
     {
       "epoch": 3.62,
-      "learning_rate": 0.00029887850467289715,
-      "loss": 1.0479,
+      "learning_rate": 0.00029747663551401866,
+      "loss": 0.3354,
       "step": 40
     },
     {
       "epoch": 3.98,
-      "eval_loss": 0.8432507514953613,
-      "eval_runtime": 2.5729,
-      "eval_samples_per_second": 6.607,
-      "eval_steps_per_second": 1.943,
-      "eval_wer": 0.4085545722713864,
+      "eval_loss": 0.5201741456985474,
+      "eval_runtime": 2.5002,
+      "eval_samples_per_second": 6.799,
+      "eval_steps_per_second": 2.0,
+      "eval_wer": 0.20943952802359883,
       "step": 44
     },
     {
       "epoch": 4.53,
-      "learning_rate": 0.00029607476635514017,
-      "loss": 0.739,
+      "learning_rate": 0.0002946728971962617,
+      "loss": 0.277,
       "step": 50
     },
     {
       "epoch": 4.98,
-      "eval_loss": 0.7656970620155334,
-      "eval_runtime": 2.5779,
-      "eval_samples_per_second": 6.595,
-      "eval_steps_per_second": 1.94,
-      "eval_wer": 0.30973451327433627,
+      "eval_loss": 0.4573761224746704,
+      "eval_runtime": 2.5614,
+      "eval_samples_per_second": 6.637,
+      "eval_steps_per_second": 1.952,
+      "eval_wer": 0.2079646017699115,
       "step": 55
     },
     {
       "epoch": 5.44,
-      "learning_rate": 0.00029327102803738313,
-      "loss": 0.6492,
+      "learning_rate": 0.00029186915887850464,
+      "loss": 0.3065,
       "step": 60
     },
     {
       "epoch": 5.98,
-      "eval_loss": 0.8034406900405884,
-      "eval_runtime": 2.5383,
-      "eval_samples_per_second": 6.697,
-      "eval_steps_per_second": 1.97,
-      "eval_wer": 0.2994100294985251,
+      "eval_loss": 0.47488853335380554,
+      "eval_runtime": 2.5674,
+      "eval_samples_per_second": 6.621,
+      "eval_steps_per_second": 1.947,
+      "eval_wer": 0.2079646017699115,
       "step": 66
     },
     {
       "epoch": 6.36,
-      "learning_rate": 0.00029046728971962615,
-      "loss": 0.6044,
+      "learning_rate": 0.00028906542056074766,
+      "loss": 0.2669,
       "step": 70
     },
     {
       "epoch": 6.98,
-      "eval_loss": 0.6401463150978088,
-      "eval_runtime": 2.5648,
-      "eval_samples_per_second": 6.628,
-      "eval_steps_per_second": 1.949,
-      "eval_wer": 0.3333333333333333,
+      "eval_loss": 0.5901708602905273,
+      "eval_runtime": 2.5163,
+      "eval_samples_per_second": 6.756,
+      "eval_steps_per_second": 1.987,
+      "eval_wer": 0.2182890855457227,
       "step": 77
     },
     {
       "epoch": 7.27,
-      "learning_rate": 0.0002876635514018691,
-      "loss": 0.5662,
+      "learning_rate": 0.0002862616822429906,
+      "loss": 0.2829,
       "step": 80
     },
     {
       "epoch": 7.98,
-      "eval_loss": 0.7298213243484497,
-      "eval_runtime": 2.5349,
-      "eval_samples_per_second": 6.706,
-      "eval_steps_per_second": 1.972,
-      "eval_wer": 0.2610619469026549,
+      "eval_loss": 0.8560299873352051,
+      "eval_runtime": 2.4999,
+      "eval_samples_per_second": 6.8,
+      "eval_steps_per_second": 2.0,
+      "eval_wer": 0.20501474926253688,
       "step": 88
     },
     {
       "epoch": 8.18,
-      "learning_rate": 0.00028485981308411213,
-      "loss": 0.5321,
+      "learning_rate": 0.00028345794392523364,
+      "loss": 0.2509,
       "step": 90
     },
     {
       "epoch": 8.98,
-      "eval_loss": 0.8125590085983276,
-      "eval_runtime": 2.5294,
-      "eval_samples_per_second": 6.721,
-      "eval_steps_per_second": 1.977,
-      "eval_wer": 0.29793510324483774,
+      "eval_loss": 0.6189576983451843,
+      "eval_runtime": 2.597,
+      "eval_samples_per_second": 6.546,
+      "eval_steps_per_second": 1.925,
+      "eval_wer": 0.20353982300884957,
       "step": 99
     },
     {
       "epoch": 9.09,
-      "learning_rate": 0.0002820560747663551,
-      "loss": 0.5423,
+      "learning_rate": 0.0002806542056074766,
+      "loss": 0.2754,
       "step": 100
     },
     {
       "epoch": 9.98,
-      "learning_rate": 0.0002792523364485981,
-      "loss": 0.5037,
+      "learning_rate": 0.0002778504672897196,
+      "loss": 0.2728,
       "step": 110
     },
     {
       "epoch": 9.98,
-      "eval_loss": 0.7135093212127686,
-      "eval_runtime": 2.5473,
-      "eval_samples_per_second": 6.674,
-      "eval_steps_per_second": 1.963,
-      "eval_wer": 0.2994100294985251,
+      "eval_loss": 0.6561536192893982,
+      "eval_runtime": 2.5862,
+      "eval_samples_per_second": 6.573,
+      "eval_steps_per_second": 1.933,
+      "eval_wer": 0.21091445427728614,
       "step": 110
     },
     {
       "epoch": 10.89,
-      "learning_rate": 0.0002764485981308411,
-      "loss": 0.4823,
+      "learning_rate": 0.0002750467289719626,
+      "loss": 0.2615,
       "step": 120
     },
     {
       "epoch": 10.98,
-      "eval_loss": 0.5975809097290039,
-      "eval_runtime": 2.4821,
-      "eval_samples_per_second": 6.849,
+      "eval_loss": 0.6290748119354248,
+      "eval_runtime": 2.4827,
+      "eval_samples_per_second": 6.847,
       "eval_steps_per_second": 2.014,
-      "eval_wer": 0.26548672566371684,
+      "eval_wer": 0.20648967551622419,
       "step": 121
     },
     {
       "epoch": 11.8,
-      "learning_rate": 0.0002736448598130841,
-      "loss": 0.4622,
+      "learning_rate": 0.0002722429906542056,
+      "loss": 0.2586,
       "step": 130
     },
     {
       "epoch": 11.98,
-      "eval_loss": 0.6875410079956055,
-      "eval_runtime": 2.4654,
-      "eval_samples_per_second": 6.895,
-      "eval_steps_per_second": 2.028,
-      "eval_wer": 0.2448377581120944,
+      "eval_loss": 0.6167445778846741,
+      "eval_runtime": 2.4578,
+      "eval_samples_per_second": 6.917,
+      "eval_steps_per_second": 2.034,
+      "eval_wer": 0.18436578171091444,
       "step": 132
     },
     {
       "epoch": 12.71,
-      "learning_rate": 0.00027084112149532705,
-      "loss": 0.4761,
+      "learning_rate": 0.00026943925233644856,
+      "loss": 0.2441,
       "step": 140
     },
     {
       "epoch": 12.98,
-      "eval_loss": 0.6402226090431213,
-      "eval_runtime": 2.4808,
-      "eval_samples_per_second": 6.853,
-      "eval_steps_per_second": 2.015,
-      "eval_wer": 0.24631268436578171,
+      "eval_loss": 0.6735682487487793,
+      "eval_runtime": 2.4754,
+      "eval_samples_per_second": 6.867,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.19616519174041297,
       "step": 143
     },
     {
       "epoch": 13.62,
-      "learning_rate": 0.00026803738317757007,
-      "loss": 0.4296,
+      "learning_rate": 0.0002666355140186916,
+      "loss": 0.233,
       "step": 150
     },
     {
       "epoch": 13.98,
-      "eval_loss": 0.82173752784729,
-      "eval_runtime": 2.4283,
-      "eval_samples_per_second": 7.001,
-      "eval_steps_per_second": 2.059,
-      "eval_wer": 0.2448377581120944,
+      "eval_loss": 0.5727431774139404,
+      "eval_runtime": 2.4735,
+      "eval_samples_per_second": 6.873,
+      "eval_steps_per_second": 2.021,
+      "eval_wer": 0.20501474926253688,
       "step": 154
     },
     {
       "epoch": 14.53,
-      "learning_rate": 0.0002652336448598131,
-      "loss": 0.4655,
+      "learning_rate": 0.00026383177570093454,
+      "loss": 0.2567,
       "step": 160
     },
     {
       "epoch": 14.98,
-      "eval_loss": 0.7825008630752563,
-      "eval_runtime": 2.4304,
-      "eval_samples_per_second": 6.995,
-      "eval_steps_per_second": 2.057,
-      "eval_wer": 0.2551622418879056,
+      "eval_loss": 0.6164781451225281,
+      "eval_runtime": 2.4905,
+      "eval_samples_per_second": 6.826,
+      "eval_steps_per_second": 2.008,
+      "eval_wer": 0.1873156342182891,
       "step": 165
     },
     {
       "epoch": 15.44,
-      "learning_rate": 0.00026242990654205605,
-      "loss": 0.4122,
+      "learning_rate": 0.00026102803738317756,
+      "loss": 0.2264,
       "step": 170
     },
     {
       "epoch": 15.98,
-      "eval_loss": 0.712116539478302,
-      "eval_runtime": 2.4336,
-      "eval_samples_per_second": 6.986,
-      "eval_steps_per_second": 2.055,
-      "eval_wer": 0.2448377581120944,
+      "eval_loss": 0.7506299018859863,
+      "eval_runtime": 2.5014,
+      "eval_samples_per_second": 6.796,
+      "eval_steps_per_second": 1.999,
+      "eval_wer": 0.2079646017699115,
       "step": 176
     },
     {
       "epoch": 16.36,
-      "learning_rate": 0.00025962616822429907,
-      "loss": 0.4234,
+      "learning_rate": 0.0002582242990654205,
+      "loss": 0.2346,
       "step": 180
     },
     {
       "epoch": 16.98,
-      "eval_loss": 0.8301025032997131,
-      "eval_runtime": 2.4483,
-      "eval_samples_per_second": 6.944,
-      "eval_steps_per_second": 2.042,
-      "eval_wer": 0.2669616519174041,
+      "eval_loss": 0.701734185218811,
+      "eval_runtime": 2.4973,
+      "eval_samples_per_second": 6.807,
+      "eval_steps_per_second": 2.002,
+      "eval_wer": 0.1887905604719764,
       "step": 187
     },
     {
       "epoch": 17.27,
-      "learning_rate": 0.00025682242990654204,
-      "loss": 0.441,
+      "learning_rate": 0.00025542056074766354,
+      "loss": 0.2343,
       "step": 190
     },
     {
       "epoch": 17.98,
-      "eval_loss": 0.7342628240585327,
-      "eval_runtime": 2.4341,
-      "eval_samples_per_second": 6.984,
-      "eval_steps_per_second": 2.054,
-      "eval_wer": 0.2640117994100295,
+      "eval_loss": 0.5930327773094177,
+      "eval_runtime": 2.5093,
+      "eval_samples_per_second": 6.775,
+      "eval_steps_per_second": 1.993,
+      "eval_wer": 0.20943952802359883,
       "step": 198
     },
     {
       "epoch": 18.18,
-      "learning_rate": 0.00025401869158878505,
-      "loss": 0.4781,
+      "learning_rate": 0.0002526168224299065,
+      "loss": 0.2638,
       "step": 200
     },
     {
       "epoch": 18.98,
-      "eval_loss": 0.7387515902519226,
-      "eval_runtime": 2.4267,
-      "eval_samples_per_second": 7.005,
-      "eval_steps_per_second": 2.06,
-      "eval_wer": 0.21386430678466076,
+      "eval_loss": 0.5729739665985107,
+      "eval_runtime": 2.4782,
+      "eval_samples_per_second": 6.86,
+      "eval_steps_per_second": 2.018,
+      "eval_wer": 0.20058997050147492,
       "step": 209
     },
     {
       "epoch": 19.09,
-      "learning_rate": 0.000251214953271028,
-      "loss": 0.3873,
+      "learning_rate": 0.0002498130841121495,
+      "loss": 0.2332,
       "step": 210
     },
     {
       "epoch": 19.98,
-      "learning_rate": 0.00024841121495327103,
-      "loss": 0.4006,
+      "learning_rate": 0.0002470093457943925,
+      "loss": 0.2543,
       "step": 220
     },
     {
       "epoch": 19.98,
-      "eval_loss": 0.6700401306152344,
-      "eval_runtime": 2.427,
-      "eval_samples_per_second": 7.005,
-      "eval_steps_per_second": 2.06,
-      "eval_wer": 0.252212389380531,
+      "eval_loss": 0.4991203248500824,
+      "eval_runtime": 2.4588,
+      "eval_samples_per_second": 6.914,
+      "eval_steps_per_second": 2.034,
+      "eval_wer": 0.21976401179941002,
       "step": 220
     },
     {
       "epoch": 20.89,
-      "learning_rate": 0.000245607476635514,
-      "loss": 0.42,
+      "learning_rate": 0.0002442056074766355,
+      "loss": 0.2476,
       "step": 230
     },
     {
       "epoch": 20.98,
-      "eval_loss": 0.5540403127670288,
-      "eval_runtime": 2.4476,
-      "eval_samples_per_second": 6.946,
-      "eval_steps_per_second": 2.043,
-      "eval_wer": 0.24926253687315633,
+      "eval_loss": 0.6363905668258667,
+      "eval_runtime": 2.4779,
+      "eval_samples_per_second": 6.861,
+      "eval_steps_per_second": 2.018,
+      "eval_wer": 0.20648967551622419,
       "step": 231
     },
     {
       "epoch": 21.8,
-      "learning_rate": 0.000242803738317757,
-      "loss": 0.4289,
+      "learning_rate": 0.00024140186915887847,
+      "loss": 0.2777,
       "step": 240
     },
     {
       "epoch": 21.98,
-      "eval_loss": 0.9949514865875244,
-      "eval_runtime": 2.4445,
-      "eval_samples_per_second": 6.954,
-      "eval_steps_per_second": 2.045,
-      "eval_wer": 0.24926253687315633,
+      "eval_loss": 0.6247313618659973,
+      "eval_runtime": 2.4757,
+      "eval_samples_per_second": 6.867,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.18436578171091444,
       "step": 242
     },
     {
       "epoch": 22.71,
-      "learning_rate": 0.00023999999999999998,
-      "loss": 0.4014,
+      "learning_rate": 0.00023859813084112146,
+      "loss": 0.2661,
       "step": 250
     },
     {
       "epoch": 22.98,
-      "eval_loss": 0.7282938957214355,
-      "eval_runtime": 2.4265,
-      "eval_samples_per_second": 7.006,
-      "eval_steps_per_second": 2.061,
-      "eval_wer": 0.252212389380531,
+      "eval_loss": 0.5588754415512085,
+      "eval_runtime": 2.5016,
+      "eval_samples_per_second": 6.796,
+      "eval_steps_per_second": 1.999,
+      "eval_wer": 0.20058997050147492,
       "step": 253
     },
     {
       "epoch": 23.62,
-      "learning_rate": 0.00023719626168224297,
-      "loss": 0.3397,
+      "learning_rate": 0.00023579439252336445,
+      "loss": 0.2094,
       "step": 260
     },
     {
       "epoch": 23.98,
-      "eval_loss": 0.8326579928398132,
-      "eval_runtime": 2.4497,
-      "eval_samples_per_second": 6.94,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.26548672566371684,
+      "eval_loss": 0.531648576259613,
+      "eval_runtime": 2.4851,
+      "eval_samples_per_second": 6.841,
+      "eval_steps_per_second": 2.012,
+      "eval_wer": 0.2079646017699115,
       "step": 264
     },
     {
       "epoch": 24.53,
-      "learning_rate": 0.00023439252336448596,
-      "loss": 0.3879,
+      "learning_rate": 0.00023299065420560744,
+      "loss": 0.2496,
       "step": 270
     },
     {
       "epoch": 24.98,
-      "eval_loss": 0.9388102889060974,
-      "eval_runtime": 2.5119,
-      "eval_samples_per_second": 6.768,
-      "eval_steps_per_second": 1.991,
-      "eval_wer": 0.29056047197640117,
+      "eval_loss": 0.8821189999580383,
+      "eval_runtime": 2.49,
+      "eval_samples_per_second": 6.827,
+      "eval_steps_per_second": 2.008,
+      "eval_wer": 0.18436578171091444,
       "step": 275
     },
     {
       "epoch": 25.44,
-      "learning_rate": 0.00023158878504672895,
-      "loss": 0.3445,
+      "learning_rate": 0.00023018691588785043,
+      "loss": 0.2302,
       "step": 280
     },
     {
       "epoch": 25.98,
-      "eval_loss": 0.7623295783996582,
-      "eval_runtime": 2.4514,
-      "eval_samples_per_second": 6.935,
-      "eval_steps_per_second": 2.04,
-      "eval_wer": 0.252212389380531,
+      "eval_loss": 0.5408168435096741,
+      "eval_runtime": 2.4887,
+      "eval_samples_per_second": 6.831,
+      "eval_steps_per_second": 2.009,
+      "eval_wer": 0.18141592920353983,
       "step": 286
     },
     {
       "epoch": 26.36,
-      "learning_rate": 0.00022878504672897194,
-      "loss": 0.3933,
+      "learning_rate": 0.00022738317757009342,
+      "loss": 0.2651,
       "step": 290
     },
     {
       "epoch": 26.98,
-      "eval_loss": 0.9125231504440308,
-      "eval_runtime": 2.4489,
-      "eval_samples_per_second": 6.942,
-      "eval_steps_per_second": 2.042,
-      "eval_wer": 0.24188790560471976,
+      "eval_loss": 0.6479179859161377,
+      "eval_runtime": 2.4859,
+      "eval_samples_per_second": 6.838,
+      "eval_steps_per_second": 2.011,
+      "eval_wer": 0.20943952802359883,
       "step": 297
     },
     {
       "epoch": 27.27,
-      "learning_rate": 0.00022598130841121493,
-      "loss": 0.3173,
+      "learning_rate": 0.00022457943925233641,
+      "loss": 0.2119,
       "step": 300
     },
     {
       "epoch": 27.98,
-      "eval_loss": 0.7446789741516113,
-      "eval_runtime": 2.4479,
-      "eval_samples_per_second": 6.945,
-      "eval_steps_per_second": 2.043,
-      "eval_wer": 0.2448377581120944,
+      "eval_loss": 0.5875303745269775,
+      "eval_runtime": 2.5001,
+      "eval_samples_per_second": 6.8,
+      "eval_steps_per_second": 2.0,
+      "eval_wer": 0.18141592920353983,
       "step": 308
     },
     {
       "epoch": 28.18,
-      "learning_rate": 0.00022317757009345792,
-      "loss": 0.3734,
+      "learning_rate": 0.0002217757009345794,
+      "loss": 0.2468,
       "step": 310
     },
     {
       "epoch": 28.98,
-      "eval_loss": 0.6601439118385315,
-      "eval_runtime": 2.4288,
-      "eval_samples_per_second": 6.999,
-      "eval_steps_per_second": 2.059,
-      "eval_wer": 0.2935103244837758,
+      "eval_loss": 0.7614116072654724,
+      "eval_runtime": 2.4857,
+      "eval_samples_per_second": 6.839,
+      "eval_steps_per_second": 2.012,
+      "eval_wer": 0.1976401179941003,
       "step": 319
     },
     {
       "epoch": 29.09,
-      "learning_rate": 0.0002203738317757009,
-      "loss": 0.3681,
+      "learning_rate": 0.0002189719626168224,
+      "loss": 0.267,
       "step": 320
     },
     {
       "epoch": 29.98,
-      "learning_rate": 0.0002175700934579439,
-      "loss": 0.3347,
+      "learning_rate": 0.00021616822429906539,
+      "loss": 0.2239,
       "step": 330
     },
     {
       "epoch": 29.98,
-      "eval_loss": 0.7021532654762268,
-      "eval_runtime": 2.4714,
-      "eval_samples_per_second": 6.879,
-      "eval_steps_per_second": 2.023,
-      "eval_wer": 0.26991150442477874,
+      "eval_loss": 0.4908014237880707,
+      "eval_runtime": 2.5125,
+      "eval_samples_per_second": 6.766,
+      "eval_steps_per_second": 1.99,
+      "eval_wer": 0.1902654867256637,
       "step": 330
     },
     {
       "epoch": 30.89,
-      "learning_rate": 0.0002147663551401869,
-      "loss": 0.3564,
+      "learning_rate": 0.00021336448598130838,
+      "loss": 0.2514,
       "step": 340
     },
     {
       "epoch": 30.98,
-      "eval_loss": 0.748757004737854,
-      "eval_runtime": 2.5209,
-      "eval_samples_per_second": 6.744,
-      "eval_steps_per_second": 1.983,
-      "eval_wer": 0.2920353982300885,
+      "eval_loss": 0.5196467638015747,
+      "eval_runtime": 2.4757,
+      "eval_samples_per_second": 6.867,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.20353982300884957,
       "step": 341
     },
     {
       "epoch": 31.8,
-      "learning_rate": 0.00021196261682242989,
-      "loss": 0.3371,
+      "learning_rate": 0.00021056074766355137,
+      "loss": 0.2244,
       "step": 350
     },
     {
       "epoch": 31.98,
-      "eval_loss": 0.6413426995277405,
-      "eval_runtime": 2.4848,
-      "eval_samples_per_second": 6.842,
-      "eval_steps_per_second": 2.012,
-      "eval_wer": 0.2581120943952802,
+      "eval_loss": 0.5580195784568787,
+      "eval_runtime": 2.5054,
+      "eval_samples_per_second": 6.785,
+      "eval_steps_per_second": 1.996,
+      "eval_wer": 0.19911504424778761,
       "step": 352
     },
     {
       "epoch": 32.71,
-      "learning_rate": 0.00020915887850467288,
-      "loss": 0.355,
+      "learning_rate": 0.00020775700934579436,
+      "loss": 0.2524,
       "step": 360
     },
     {
       "epoch": 32.98,
-      "eval_loss": 0.5131234526634216,
-      "eval_runtime": 2.4307,
-      "eval_samples_per_second": 6.994,
-      "eval_steps_per_second": 2.057,
-      "eval_wer": 0.2374631268436578,
+      "eval_loss": 0.5342008471488953,
+      "eval_runtime": 2.483,
+      "eval_samples_per_second": 6.847,
+      "eval_steps_per_second": 2.014,
+      "eval_wer": 0.20206489675516223,
       "step": 363
     },
     {
       "epoch": 33.62,
-      "learning_rate": 0.00020635514018691587,
-      "loss": 0.3648,
+      "learning_rate": 0.00020495327102803735,
+      "loss": 0.2516,
       "step": 370
     },
     {
       "epoch": 33.98,
-      "eval_loss": 0.5807757377624512,
-      "eval_runtime": 2.4713,
-      "eval_samples_per_second": 6.879,
-      "eval_steps_per_second": 2.023,
-      "eval_wer": 0.22861356932153393,
+      "eval_loss": 0.42036938667297363,
+      "eval_runtime": 2.4952,
+      "eval_samples_per_second": 6.813,
+      "eval_steps_per_second": 2.004,
+      "eval_wer": 0.18436578171091444,
       "step": 374
     },
     {
       "epoch": 34.53,
-      "learning_rate": 0.00020355140186915886,
-      "loss": 0.3209,
+      "learning_rate": 0.00020214953271028034,
+      "loss": 0.2515,
       "step": 380
     },
     {
       "epoch": 34.98,
-      "eval_loss": 0.5391702651977539,
-      "eval_runtime": 2.4562,
-      "eval_samples_per_second": 6.921,
-      "eval_steps_per_second": 2.036,
-      "eval_wer": 0.22566371681415928,
+      "eval_loss": 0.5135067105293274,
+      "eval_runtime": 2.4689,
+      "eval_samples_per_second": 6.886,
+      "eval_steps_per_second": 2.025,
+      "eval_wer": 0.21238938053097345,
       "step": 385
     },
     {
       "epoch": 35.44,
-      "learning_rate": 0.00020074766355140185,
-      "loss": 0.3522,
+      "learning_rate": 0.00019934579439252333,
+      "loss": 0.2542,
       "step": 390
     },
     {
       "epoch": 35.98,
-      "eval_loss": 0.4410792887210846,
-      "eval_runtime": 2.4622,
-      "eval_samples_per_second": 6.904,
-      "eval_steps_per_second": 2.031,
-      "eval_wer": 0.22271386430678466,
+      "eval_loss": 0.8149600028991699,
+      "eval_runtime": 2.5099,
+      "eval_samples_per_second": 6.773,
+      "eval_steps_per_second": 1.992,
+      "eval_wer": 0.19616519174041297,
       "step": 396
     },
     {
       "epoch": 36.36,
-      "learning_rate": 0.00019794392523364484,
-      "loss": 0.3252,
+      "learning_rate": 0.00019654205607476632,
+      "loss": 0.2269,
       "step": 400
     },
     {
       "epoch": 36.98,
-      "eval_loss": 0.46930572390556335,
-      "eval_runtime": 2.419,
-      "eval_samples_per_second": 7.028,
-      "eval_steps_per_second": 2.067,
-      "eval_wer": 0.21091445427728614,
+      "eval_loss": 0.8832700252532959,
+      "eval_runtime": 2.5448,
+      "eval_samples_per_second": 6.68,
+      "eval_steps_per_second": 1.965,
+      "eval_wer": 0.20943952802359883,
       "step": 407
     },
     {
       "epoch": 37.27,
-      "learning_rate": 0.00019514018691588783,
-      "loss": 0.3216,
+      "learning_rate": 0.0001937383177570093,
+      "loss": 0.212,
       "step": 410
     },
     {
       "epoch": 37.98,
-      "eval_loss": 0.4621430039405823,
-      "eval_runtime": 2.4503,
-      "eval_samples_per_second": 6.938,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.20648967551622419,
+      "eval_loss": 1.3234660625457764,
+      "eval_runtime": 2.4758,
+      "eval_samples_per_second": 6.867,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.2182890855457227,
       "step": 418
     },
     {
       "epoch": 38.18,
-      "learning_rate": 0.00019233644859813082,
-      "loss": 0.3119,
+      "learning_rate": 0.0001909345794392523,
+      "loss": 0.2119,
       "step": 420
     },
     {
       "epoch": 38.98,
-      "eval_loss": 0.5093832612037659,
-      "eval_runtime": 2.4699,
-      "eval_samples_per_second": 6.883,
-      "eval_steps_per_second": 2.024,
-      "eval_wer": 0.2168141592920354,
+      "eval_loss": 0.6919201016426086,
+      "eval_runtime": 2.478,
+      "eval_samples_per_second": 6.86,
+      "eval_steps_per_second": 2.018,
+      "eval_wer": 0.20206489675516223,
       "step": 429
     },
     {
       "epoch": 39.09,
-      "learning_rate": 0.0001895327102803738,
-      "loss": 0.3266,
+      "learning_rate": 0.0001881308411214953,
+      "loss": 0.2196,
       "step": 430
     },
     {
       "epoch": 39.98,
-      "learning_rate": 0.0001867289719626168,
-      "loss": 0.3247,
+      "learning_rate": 0.00018532710280373828,
+      "loss": 0.2228,
       "step": 440
     },
     {
       "epoch": 39.98,
-      "eval_loss": 0.4897085130214691,
-      "eval_runtime": 2.4522,
-      "eval_samples_per_second": 6.933,
-      "eval_steps_per_second": 2.039,
-      "eval_wer": 0.23156342182890854,
+      "eval_loss": 0.6712406873703003,
+      "eval_runtime": 2.4899,
+      "eval_samples_per_second": 6.827,
+      "eval_steps_per_second": 2.008,
+      "eval_wer": 0.20206489675516223,
       "step": 440
     },
     {
       "epoch": 40.89,
-      "learning_rate": 0.0001839252336448598,
-      "loss": 0.3246,
+      "learning_rate": 0.00018252336448598127,
+      "loss": 0.2127,
       "step": 450
     },
     {
       "epoch": 40.98,
-      "eval_loss": 0.6471104621887207,
-      "eval_runtime": 2.42,
-      "eval_samples_per_second": 7.025,
-      "eval_steps_per_second": 2.066,
-      "eval_wer": 0.22123893805309736,
+      "eval_loss": 0.755742073059082,
+      "eval_runtime": 2.4974,
+      "eval_samples_per_second": 6.807,
+      "eval_steps_per_second": 2.002,
+      "eval_wer": 0.1976401179941003,
       "step": 451
     },
     {
       "epoch": 41.8,
-      "learning_rate": 0.00018112149532710278,
-      "loss": 0.2997,
+      "learning_rate": 0.00017971962616822426,
+      "loss": 0.2064,
       "step": 460
     },
     {
       "epoch": 41.98,
-      "eval_loss": 0.5569115877151489,
-      "eval_runtime": 2.4575,
-      "eval_samples_per_second": 6.918,
-      "eval_steps_per_second": 2.035,
-      "eval_wer": 0.2153392330383481,
+      "eval_loss": 0.5917617082595825,
+      "eval_runtime": 2.4802,
+      "eval_samples_per_second": 6.854,
+      "eval_steps_per_second": 2.016,
+      "eval_wer": 0.19469026548672566,
       "step": 462
     },
     {
       "epoch": 42.71,
-      "learning_rate": 0.00017831775700934577,
-      "loss": 0.2969,
+      "learning_rate": 0.00017691588785046725,
+      "loss": 0.2147,
       "step": 470
     },
     {
       "epoch": 42.98,
-      "eval_loss": 0.47663480043411255,
-      "eval_runtime": 2.444,
-      "eval_samples_per_second": 6.956,
-      "eval_steps_per_second": 2.046,
-      "eval_wer": 0.20943952802359883,
+      "eval_loss": 0.804908275604248,
+      "eval_runtime": 2.4373,
+      "eval_samples_per_second": 6.975,
+      "eval_steps_per_second": 2.051,
+      "eval_wer": 0.19616519174041297,
       "step": 473
     },
     {
       "epoch": 43.62,
-      "learning_rate": 0.00017551401869158876,
-      "loss": 0.3202,
+      "learning_rate": 0.00017411214953271025,
+      "loss": 0.193,
       "step": 480
     },
     {
       "epoch": 43.98,
-      "eval_loss": 0.4978266954421997,
-      "eval_runtime": 2.4301,
-      "eval_samples_per_second": 6.996,
-      "eval_steps_per_second": 2.058,
-      "eval_wer": 0.23156342182890854,
+      "eval_loss": 0.7116745710372925,
+      "eval_runtime": 2.4658,
+      "eval_samples_per_second": 6.894,
+      "eval_steps_per_second": 2.028,
+      "eval_wer": 0.1976401179941003,
       "step": 484
     },
     {
       "epoch": 44.53,
-      "learning_rate": 0.00017271028037383175,
-      "loss": 0.3093,
+      "learning_rate": 0.00017130841121495324,
+      "loss": 0.2063,
       "step": 490
     },
     {
       "epoch": 44.98,
-      "eval_loss": 0.4776196777820587,
-      "eval_runtime": 2.4293,
-      "eval_samples_per_second": 6.998,
-      "eval_steps_per_second": 2.058,
-      "eval_wer": 0.2182890855457227,
+      "eval_loss": 0.554362416267395,
+      "eval_runtime": 2.4717,
+      "eval_samples_per_second": 6.878,
+      "eval_steps_per_second": 2.023,
+      "eval_wer": 0.19616519174041297,
       "step": 495
     },
     {
       "epoch": 45.44,
-      "learning_rate": 0.00016990654205607475,
-      "loss": 0.298,
+      "learning_rate": 0.00016850467289719623,
+      "loss": 0.1989,
       "step": 500
     },
     {
       "epoch": 45.98,
-      "eval_loss": 0.5007936954498291,
-      "eval_runtime": 2.452,
-      "eval_samples_per_second": 6.933,
-      "eval_steps_per_second": 2.039,
-      "eval_wer": 0.21976401179941002,
+      "eval_loss": 0.5782185196876526,
+      "eval_runtime": 2.4456,
+      "eval_samples_per_second": 6.951,
+      "eval_steps_per_second": 2.044,
+      "eval_wer": 0.1887905604719764,
       "step": 506
     },
     {
       "epoch": 46.36,
-      "learning_rate": 0.00016710280373831774,
-      "loss": 0.3151,
+      "learning_rate": 0.00016570093457943922,
+      "loss": 0.2193,
       "step": 510
     },
     {
       "epoch": 46.98,
-      "eval_loss": 0.48113784193992615,
-      "eval_runtime": 2.4479,
-      "eval_samples_per_second": 6.945,
-      "eval_steps_per_second": 2.043,
-      "eval_wer": 0.2079646017699115,
+      "eval_loss": 0.5216302275657654,
+      "eval_runtime": 2.4591,
+      "eval_samples_per_second": 6.913,
+      "eval_steps_per_second": 2.033,
+      "eval_wer": 0.19469026548672566,
       "step": 517
     },
     {
       "epoch": 47.27,
-      "learning_rate": 0.00016429906542056073,
-      "loss": 0.2824,
+      "learning_rate": 0.0001628971962616822,
+      "loss": 0.2012,
       "step": 520
     },
     {
       "epoch": 47.98,
-      "eval_loss": 0.5011029839515686,
-      "eval_runtime": 2.4404,
-      "eval_samples_per_second": 6.966,
-      "eval_steps_per_second": 2.049,
-      "eval_wer": 0.20648967551622419,
+      "eval_loss": 0.5268691778182983,
+      "eval_runtime": 2.4596,
+      "eval_samples_per_second": 6.912,
+      "eval_steps_per_second": 2.033,
+      "eval_wer": 0.19174041297935104,
       "step": 528
     },
     {
       "epoch": 48.18,
-      "learning_rate": 0.00016149532710280372,
-      "loss": 0.3089,
+      "learning_rate": 0.0001600934579439252,
+      "loss": 0.2187,
       "step": 530
     },
     {
       "epoch": 48.98,
-      "eval_loss": 0.5131256580352783,
-      "eval_runtime": 2.4575,
-      "eval_samples_per_second": 6.917,
-      "eval_steps_per_second": 2.035,
-      "eval_wer": 0.21386430678466076,
+      "eval_loss": 0.4635768532752991,
+      "eval_runtime": 2.4781,
+      "eval_samples_per_second": 6.86,
+      "eval_steps_per_second": 2.018,
+      "eval_wer": 0.18436578171091444,
       "step": 539
     },
     {
       "epoch": 49.09,
-      "learning_rate": 0.0001586915887850467,
-      "loss": 0.2795,
+      "learning_rate": 0.0001572897196261682,
+      "loss": 0.1857,
       "step": 540
     },
     {
       "epoch": 49.98,
-      "learning_rate": 0.0001558878504672897,
-      "loss": 0.3064,
+      "learning_rate": 0.00015448598130841118,
+      "loss": 0.2128,
       "step": 550
     },
     {
       "epoch": 49.98,
-      "eval_loss": 0.47494271397590637,
-      "eval_runtime": 2.4456,
-      "eval_samples_per_second": 6.951,
-      "eval_steps_per_second": 2.045,
-      "eval_wer": 0.22271386430678466,
+      "eval_loss": 0.49676355719566345,
+      "eval_runtime": 2.4581,
+      "eval_samples_per_second": 6.916,
+      "eval_steps_per_second": 2.034,
+      "eval_wer": 0.1887905604719764,
       "step": 550
     },
     {
       "epoch": 50.89,
-      "learning_rate": 0.0001530841121495327,
-      "loss": 0.2734,
+      "learning_rate": 0.00015168224299065417,
+      "loss": 0.2041,
       "step": 560
     },
     {
       "epoch": 50.98,
-      "eval_loss": 0.5396826863288879,
-      "eval_runtime": 2.4349,
-      "eval_samples_per_second": 6.982,
-      "eval_steps_per_second": 2.053,
-      "eval_wer": 0.2079646017699115,
+      "eval_loss": 0.47835999727249146,
+      "eval_runtime": 2.4475,
+      "eval_samples_per_second": 6.946,
+      "eval_steps_per_second": 2.043,
+      "eval_wer": 0.1887905604719764,
       "step": 561
     },
     {
       "epoch": 51.8,
-      "learning_rate": 0.00015028037383177568,
-      "loss": 0.2911,
+      "learning_rate": 0.0001488785046728972,
+      "loss": 0.1993,
       "step": 570
     },
     {
       "epoch": 51.98,
-      "eval_loss": 0.4975402355194092,
-      "eval_runtime": 2.4591,
-      "eval_samples_per_second": 6.913,
-      "eval_steps_per_second": 2.033,
-      "eval_wer": 0.20353982300884957,
+      "eval_loss": 0.5591913461685181,
+      "eval_runtime": 2.4775,
+      "eval_samples_per_second": 6.862,
+      "eval_steps_per_second": 2.018,
+      "eval_wer": 0.17551622418879056,
       "step": 572
     },
     {
       "epoch": 52.71,
-      "learning_rate": 0.00014747663551401867,
-      "loss": 0.2889,
+      "learning_rate": 0.00014607476635514018,
+      "loss": 0.1981,
       "step": 580
     },
     {
       "epoch": 52.98,
-      "eval_loss": 0.46327924728393555,
-      "eval_runtime": 2.4538,
-      "eval_samples_per_second": 6.928,
-      "eval_steps_per_second": 2.038,
-      "eval_wer": 0.2168141592920354,
+      "eval_loss": 0.4870879352092743,
+      "eval_runtime": 2.4711,
+      "eval_samples_per_second": 6.88,
+      "eval_steps_per_second": 2.023,
+      "eval_wer": 0.17846607669616518,
       "step": 583
     },
     {
       "epoch": 53.62,
-      "learning_rate": 0.00014467289719626166,
-      "loss": 0.2523,
+      "learning_rate": 0.00014327102803738317,
+      "loss": 0.1808,
       "step": 590
     },
     {
       "epoch": 53.98,
-      "eval_loss": 0.4589496850967407,
-      "eval_runtime": 2.4386,
-      "eval_samples_per_second": 6.971,
-      "eval_steps_per_second": 2.05,
-      "eval_wer": 0.22418879056047197,
+      "eval_loss": 0.47705358266830444,
+      "eval_runtime": 2.4668,
+      "eval_samples_per_second": 6.891,
+      "eval_steps_per_second": 2.027,
+      "eval_wer": 0.17404129793510326,
       "step": 594
     },
     {
       "epoch": 54.53,
-      "learning_rate": 0.00014186915887850465,
-      "loss": 0.272,
+      "learning_rate": 0.00014046728971962616,
+      "loss": 0.2317,
       "step": 600
     },
     {
       "epoch": 54.98,
-      "eval_loss": 0.4855732023715973,
-      "eval_runtime": 2.4564,
-      "eval_samples_per_second": 6.921,
-      "eval_steps_per_second": 2.035,
-      "eval_wer": 0.21238938053097345,
+      "eval_loss": 0.5285036563873291,
+      "eval_runtime": 2.4608,
+      "eval_samples_per_second": 6.908,
+      "eval_steps_per_second": 2.032,
+      "eval_wer": 0.18141592920353983,
       "step": 605
     },
     {
       "epoch": 55.44,
-      "learning_rate": 0.00013906542056074764,
-      "loss": 0.2733,
+      "learning_rate": 0.00013766355140186915,
+      "loss": 0.1906,
       "step": 610
     },
     {
       "epoch": 55.98,
-      "eval_loss": 0.4474070966243744,
-      "eval_runtime": 2.4335,
-      "eval_samples_per_second": 6.986,
-      "eval_steps_per_second": 2.055,
-      "eval_wer": 0.22418879056047197,
+      "eval_loss": 0.5485416650772095,
+      "eval_runtime": 2.4937,
+      "eval_samples_per_second": 6.817,
+      "eval_steps_per_second": 2.005,
+      "eval_wer": 0.18436578171091444,
       "step": 616
     },
     {
       "epoch": 56.36,
-      "learning_rate": 0.00013626168224299063,
-      "loss": 0.2856,
+      "learning_rate": 0.00013485981308411214,
+      "loss": 0.1924,
       "step": 620
     },
     {
       "epoch": 56.98,
-      "eval_loss": 0.4533959627151489,
-      "eval_runtime": 2.4476,
-      "eval_samples_per_second": 6.946,
-      "eval_steps_per_second": 2.043,
-      "eval_wer": 0.22713864306784662,
+      "eval_loss": 0.5615388751029968,
+      "eval_runtime": 2.4859,
+      "eval_samples_per_second": 6.839,
+      "eval_steps_per_second": 2.011,
+      "eval_wer": 0.18141592920353983,
       "step": 627
     },
     {
       "epoch": 57.27,
-      "learning_rate": 0.00013345794392523362,
-      "loss": 0.2402,
+      "learning_rate": 0.00013205607476635513,
+      "loss": 0.1761,
       "step": 630
     },
     {
       "epoch": 57.98,
-      "eval_loss": 0.43458130955696106,
-      "eval_runtime": 2.4576,
-      "eval_samples_per_second": 6.917,
-      "eval_steps_per_second": 2.034,
-      "eval_wer": 0.22418879056047197,
+      "eval_loss": 0.46036362648010254,
+      "eval_runtime": 2.4852,
+      "eval_samples_per_second": 6.84,
+      "eval_steps_per_second": 2.012,
+      "eval_wer": 0.17994100294985252,
       "step": 638
     },
     {
       "epoch": 58.18,
-      "learning_rate": 0.00013065420560747661,
-      "loss": 0.2567,
+      "learning_rate": 0.00012925233644859812,
+      "loss": 0.2047,
       "step": 640
     },
     {
       "epoch": 58.98,
-      "eval_loss": 0.501366376876831,
-      "eval_runtime": 2.4401,
-      "eval_samples_per_second": 6.967,
-      "eval_steps_per_second": 2.049,
-      "eval_wer": 0.22861356932153393,
+      "eval_loss": 0.422275573015213,
+      "eval_runtime": 2.4837,
+      "eval_samples_per_second": 6.845,
+      "eval_steps_per_second": 2.013,
+      "eval_wer": 0.18289085545722714,
       "step": 649
     },
     {
       "epoch": 59.09,
-      "learning_rate": 0.0001278504672897196,
-      "loss": 0.294,
+      "learning_rate": 0.00012644859813084111,
+      "loss": 0.2295,
       "step": 650
     },
     {
       "epoch": 59.98,
-      "learning_rate": 0.0001250467289719626,
-      "loss": 0.28,
+      "learning_rate": 0.0001236448598130841,
+      "loss": 0.1992,
       "step": 660
     },
     {
       "epoch": 59.98,
-      "eval_loss": 0.44283339381217957,
-      "eval_runtime": 2.4528,
-      "eval_samples_per_second": 6.931,
-      "eval_steps_per_second": 2.038,
-      "eval_wer": 0.2182890855457227,
+      "eval_loss": 0.47056129574775696,
+      "eval_runtime": 2.4575,
+      "eval_samples_per_second": 6.918,
+      "eval_steps_per_second": 2.035,
+      "eval_wer": 0.1873156342182891,
       "step": 660
     },
     {
       "epoch": 60.89,
-      "learning_rate": 0.00012224299065420559,
-      "loss": 0.2541,
+      "learning_rate": 0.0001208411214953271,
+      "loss": 0.1949,
       "step": 670
     },
     {
       "epoch": 60.98,
-      "eval_loss": 0.4875503480434418,
-      "eval_runtime": 2.4588,
-      "eval_samples_per_second": 6.914,
-      "eval_steps_per_second": 2.033,
-      "eval_wer": 0.22271386430678466,
+      "eval_loss": 0.4633275866508484,
+      "eval_runtime": 2.458,
+      "eval_samples_per_second": 6.916,
+      "eval_steps_per_second": 2.034,
+      "eval_wer": 0.18436578171091444,
       "step": 671
     },
     {
       "epoch": 61.8,
-      "learning_rate": 0.00011943925233644859,
-      "loss": 0.2544,
+      "learning_rate": 0.00011803738317757009,
+      "loss": 0.2034,
       "step": 680
     },
     {
       "epoch": 61.98,
-      "eval_loss": 0.4705180823802948,
-      "eval_runtime": 2.4501,
-      "eval_samples_per_second": 6.939,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.20501474926253688,
+      "eval_loss": 0.48540210723876953,
+      "eval_runtime": 2.4576,
+      "eval_samples_per_second": 6.917,
+      "eval_steps_per_second": 2.035,
+      "eval_wer": 0.18141592920353983,
       "step": 682
     },
     {
       "epoch": 62.71,
-      "learning_rate": 0.00011663551401869158,
-      "loss": 0.2786,
+      "learning_rate": 0.00011523364485981308,
+      "loss": 0.2147,
       "step": 690
     },
     {
       "epoch": 62.98,
-      "eval_loss": 0.44488102197647095,
-      "eval_runtime": 2.4756,
-      "eval_samples_per_second": 6.867,
-      "eval_steps_per_second": 2.02,
-      "eval_wer": 0.20206489675516223,
+      "eval_loss": 0.4488876163959503,
+      "eval_runtime": 2.4838,
+      "eval_samples_per_second": 6.844,
+      "eval_steps_per_second": 2.013,
+      "eval_wer": 0.18436578171091444,
       "step": 693
     },
     {
       "epoch": 63.62,
-      "learning_rate": 0.00011383177570093457,
-      "loss": 0.2524,
+      "learning_rate": 0.00011242990654205607,
+      "loss": 0.2135,
       "step": 700
     },
     {
       "epoch": 63.98,
-      "eval_loss": 0.558536946773529,
-      "eval_runtime": 2.4336,
-      "eval_samples_per_second": 6.986,
-      "eval_steps_per_second": 2.055,
-      "eval_wer": 0.20943952802359883,
+      "eval_loss": 0.48737475275993347,
+      "eval_runtime": 2.4621,
+      "eval_samples_per_second": 6.905,
+      "eval_steps_per_second": 2.031,
+      "eval_wer": 0.17256637168141592,
       "step": 704
     },
     {
       "epoch": 64.53,
-      "learning_rate": 0.00011102803738317756,
-      "loss": 0.2524,
+      "learning_rate": 0.00010962616822429906,
+      "loss": 0.2021,
       "step": 710
     },
     {
       "epoch": 64.98,
-      "eval_loss": 0.5178933143615723,
-      "eval_runtime": 2.4703,
-      "eval_samples_per_second": 6.882,
-      "eval_steps_per_second": 2.024,
-      "eval_wer": 0.21091445427728614,
+      "eval_loss": 0.4635281562805176,
+      "eval_runtime": 2.4487,
+      "eval_samples_per_second": 6.942,
+      "eval_steps_per_second": 2.042,
+      "eval_wer": 0.18141592920353983,
       "step": 715
     },
     {
       "epoch": 65.44,
-      "learning_rate": 0.00010822429906542055,
-      "loss": 0.2852,
+      "learning_rate": 0.00010682242990654205,
+      "loss": 0.1822,
       "step": 720
     },
     {
       "epoch": 65.98,
-      "eval_loss": 0.5062990188598633,
-      "eval_runtime": 2.427,
-      "eval_samples_per_second": 7.004,
-      "eval_steps_per_second": 2.06,
-      "eval_wer": 0.21976401179941002,
+      "eval_loss": 0.48127830028533936,
+      "eval_runtime": 2.4625,
+      "eval_samples_per_second": 6.904,
+      "eval_steps_per_second": 2.03,
+      "eval_wer": 0.17846607669616518,
       "step": 726
     },
     {
       "epoch": 66.36,
-      "learning_rate": 0.00010542056074766354,
-      "loss": 0.2393,
+      "learning_rate": 0.00010401869158878504,
+      "loss": 0.1882,
       "step": 730
     },
     {
       "epoch": 66.98,
-      "eval_loss": 0.4768075942993164,
-      "eval_runtime": 2.4351,
-      "eval_samples_per_second": 6.981,
-      "eval_steps_per_second": 2.053,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.5075557827949524,
+      "eval_runtime": 2.4867,
+      "eval_samples_per_second": 6.836,
+      "eval_steps_per_second": 2.011,
+      "eval_wer": 0.17994100294985252,
       "step": 737
     },
     {
       "epoch": 67.27,
-      "learning_rate": 0.00010261682242990653,
-      "loss": 0.2522,
+      "learning_rate": 0.00010121495327102803,
+      "loss": 0.2014,
       "step": 740
     },
     {
       "epoch": 67.98,
-      "eval_loss": 0.44732481241226196,
-      "eval_runtime": 2.4715,
-      "eval_samples_per_second": 6.878,
-      "eval_steps_per_second": 2.023,
-      "eval_wer": 0.19321533923303835,
+      "eval_loss": 0.5182777047157288,
+      "eval_runtime": 2.4749,
+      "eval_samples_per_second": 6.869,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.1887905604719764,
       "step": 748
     },
     {
       "epoch": 68.18,
-      "learning_rate": 9.981308411214952e-05,
-      "loss": 0.2768,
+      "learning_rate": 9.841121495327102e-05,
+      "loss": 0.1869,
       "step": 750
     },
     {
       "epoch": 68.98,
-      "eval_loss": 0.4713572859764099,
-      "eval_runtime": 2.4588,
-      "eval_samples_per_second": 6.914,
-      "eval_steps_per_second": 2.034,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.5035397410392761,
+      "eval_runtime": 2.4735,
+      "eval_samples_per_second": 6.873,
+      "eval_steps_per_second": 2.021,
+      "eval_wer": 0.17994100294985252,
       "step": 759
     },
     {
       "epoch": 69.09,
-      "learning_rate": 9.700934579439252e-05,
-      "loss": 0.2484,
+      "learning_rate": 9.560747663551401e-05,
+      "loss": 0.2118,
       "step": 760
     },
     {
       "epoch": 69.98,
-      "learning_rate": 9.42056074766355e-05,
-      "loss": 0.2463,
+      "learning_rate": 9.2803738317757e-05,
+      "loss": 0.1914,
       "step": 770
     },
     {
       "epoch": 69.98,
-      "eval_loss": 0.4947798550128937,
-      "eval_runtime": 2.5096,
-      "eval_samples_per_second": 6.774,
-      "eval_steps_per_second": 1.992,
-      "eval_wer": 0.19469026548672566,
+      "eval_loss": 0.4693587124347687,
+      "eval_runtime": 2.5588,
+      "eval_samples_per_second": 6.644,
+      "eval_steps_per_second": 1.954,
+      "eval_wer": 0.18436578171091444,
       "step": 770
     },
     {
       "epoch": 70.89,
-      "learning_rate": 9.14018691588785e-05,
-      "loss": 0.2379,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 0.1972,
       "step": 780
     },
     {
       "epoch": 70.98,
-      "eval_loss": 0.4977743923664093,
-      "eval_runtime": 2.4614,
-      "eval_samples_per_second": 6.907,
-      "eval_steps_per_second": 2.031,
-      "eval_wer": 0.19321533923303835,
+      "eval_loss": 0.4484705626964569,
+      "eval_runtime": 2.5088,
+      "eval_samples_per_second": 6.776,
+      "eval_steps_per_second": 1.993,
+      "eval_wer": 0.18436578171091444,
       "step": 781
     },
     {
       "epoch": 71.8,
-      "learning_rate": 8.859813084112149e-05,
-      "loss": 0.2343,
+      "learning_rate": 8.719626168224298e-05,
+      "loss": 0.1724,
       "step": 790
     },
     {
       "epoch": 71.98,
-      "eval_loss": 0.45256489515304565,
-      "eval_runtime": 2.4455,
-      "eval_samples_per_second": 6.952,
-      "eval_steps_per_second": 2.045,
-      "eval_wer": 0.1902654867256637,
+      "eval_loss": 0.457947701215744,
+      "eval_runtime": 2.4982,
+      "eval_samples_per_second": 6.805,
+      "eval_steps_per_second": 2.001,
+      "eval_wer": 0.18289085545722714,
       "step": 792
     },
     {
       "epoch": 72.71,
-      "learning_rate": 8.579439252336448e-05,
-      "loss": 0.3377,
+      "learning_rate": 8.439252336448597e-05,
+      "loss": 0.195,
       "step": 800
     },
     {
       "epoch": 72.98,
-      "eval_loss": 0.451813668012619,
-      "eval_runtime": 2.4509,
-      "eval_samples_per_second": 6.936,
-      "eval_steps_per_second": 2.04,
-      "eval_wer": 0.19616519174041297,
+      "eval_loss": 0.5178218483924866,
+      "eval_runtime": 2.4797,
+      "eval_samples_per_second": 6.856,
+      "eval_steps_per_second": 2.016,
+      "eval_wer": 0.18141592920353983,
       "step": 803
     },
     {
       "epoch": 73.62,
-      "learning_rate": 8.299065420560747e-05,
-      "loss": 0.2683,
+      "learning_rate": 8.158878504672896e-05,
+      "loss": 0.2017,
       "step": 810
     },
     {
       "epoch": 73.98,
-      "eval_loss": 0.44569432735443115,
-      "eval_runtime": 2.4492,
-      "eval_samples_per_second": 6.941,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.21091445427728614,
+      "eval_loss": 0.49777284264564514,
+      "eval_runtime": 2.4694,
+      "eval_samples_per_second": 6.884,
+      "eval_steps_per_second": 2.025,
+      "eval_wer": 0.18289085545722714,
       "step": 814
     },
     {
       "epoch": 74.53,
-      "learning_rate": 8.018691588785046e-05,
-      "loss": 0.2371,
+      "learning_rate": 7.878504672897195e-05,
+      "loss": 0.1874,
       "step": 820
     },
     {
       "epoch": 74.98,
-      "eval_loss": 0.4563811719417572,
-      "eval_runtime": 2.4581,
-      "eval_samples_per_second": 6.916,
-      "eval_steps_per_second": 2.034,
-      "eval_wer": 0.20206489675516223,
+      "eval_loss": 0.5034937858581543,
+      "eval_runtime": 2.4472,
+      "eval_samples_per_second": 6.947,
+      "eval_steps_per_second": 2.043,
+      "eval_wer": 0.1873156342182891,
       "step": 825
     },
     {
       "epoch": 75.44,
-      "learning_rate": 7.738317757009345e-05,
-      "loss": 0.2438,
+      "learning_rate": 7.598130841121495e-05,
+      "loss": 0.1925,
       "step": 830
     },
     {
       "epoch": 75.98,
-      "eval_loss": 0.48755523562431335,
-      "eval_runtime": 2.4767,
-      "eval_samples_per_second": 6.864,
-      "eval_steps_per_second": 2.019,
-      "eval_wer": 0.20943952802359883,
+      "eval_loss": 0.5495364665985107,
+      "eval_runtime": 2.4633,
+      "eval_samples_per_second": 6.901,
+      "eval_steps_per_second": 2.03,
+      "eval_wer": 0.18289085545722714,
       "step": 836
     },
     {
       "epoch": 76.36,
-      "learning_rate": 7.457943925233644e-05,
-      "loss": 0.2408,
+      "learning_rate": 7.317757009345794e-05,
+      "loss": 0.1845,
       "step": 840
     },
     {
       "epoch": 76.98,
-      "eval_loss": 0.43862077593803406,
-      "eval_runtime": 2.4583,
-      "eval_samples_per_second": 6.915,
-      "eval_steps_per_second": 2.034,
-      "eval_wer": 0.20206489675516223,
+      "eval_loss": 0.5394346117973328,
+      "eval_runtime": 2.4749,
+      "eval_samples_per_second": 6.869,
+      "eval_steps_per_second": 2.02,
+      "eval_wer": 0.17994100294985252,
       "step": 847
     },
     {
       "epoch": 77.27,
-      "learning_rate": 7.177570093457943e-05,
-      "loss": 0.2323,
+      "learning_rate": 7.037383177570093e-05,
+      "loss": 0.1718,
       "step": 850
     },
     {
       "epoch": 77.98,
-      "eval_loss": 0.4512799382209778,
-      "eval_runtime": 2.4462,
-      "eval_samples_per_second": 6.95,
-      "eval_steps_per_second": 2.044,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.507049024105072,
+      "eval_runtime": 2.4857,
+      "eval_samples_per_second": 6.839,
+      "eval_steps_per_second": 2.012,
+      "eval_wer": 0.1710914454277286,
       "step": 858
     },
     {
       "epoch": 78.18,
-      "learning_rate": 6.897196261682242e-05,
-      "loss": 0.271,
+      "learning_rate": 6.757009345794392e-05,
+      "loss": 0.1824,
       "step": 860
     },
     {
       "epoch": 78.98,
-      "eval_loss": 0.487374484539032,
-      "eval_runtime": 2.4852,
-      "eval_samples_per_second": 6.841,
-      "eval_steps_per_second": 2.012,
-      "eval_wer": 0.20206489675516223,
+      "eval_loss": 0.49118277430534363,
+      "eval_runtime": 2.4521,
+      "eval_samples_per_second": 6.933,
+      "eval_steps_per_second": 2.039,
+      "eval_wer": 0.17699115044247787,
       "step": 869
     },
     {
       "epoch": 79.09,
-      "learning_rate": 6.616822429906541e-05,
-      "loss": 0.2406,
+      "learning_rate": 6.476635514018691e-05,
+      "loss": 0.1904,
       "step": 870
     },
     {
       "epoch": 79.98,
-      "learning_rate": 6.33644859813084e-05,
-      "loss": 0.229,
+      "learning_rate": 6.19626168224299e-05,
+      "loss": 0.1702,
       "step": 880
     },
     {
       "epoch": 79.98,
-      "eval_loss": 0.4882209002971649,
-      "eval_runtime": 2.454,
-      "eval_samples_per_second": 6.927,
-      "eval_steps_per_second": 2.037,
-      "eval_wer": 0.20648967551622419,
+      "eval_loss": 0.46315157413482666,
+      "eval_runtime": 2.4682,
+      "eval_samples_per_second": 6.888,
+      "eval_steps_per_second": 2.026,
+      "eval_wer": 0.17256637168141592,
       "step": 880
     },
     {
       "epoch": 80.89,
-      "learning_rate": 6.05607476635514e-05,
-      "loss": 0.224,
+      "learning_rate": 5.9158878504672896e-05,
+      "loss": 0.1563,
       "step": 890
     },
     {
       "epoch": 80.98,
-      "eval_loss": 0.49814578890800476,
-      "eval_runtime": 2.4571,
-      "eval_samples_per_second": 6.919,
-      "eval_steps_per_second": 2.035,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.44120773673057556,
+      "eval_runtime": 2.4609,
+      "eval_samples_per_second": 6.908,
+      "eval_steps_per_second": 2.032,
+      "eval_wer": 0.17256637168141592,
       "step": 891
     },
     {
       "epoch": 81.8,
-      "learning_rate": 5.775700934579439e-05,
-      "loss": 0.2442,
+      "learning_rate": 5.635514018691589e-05,
+      "loss": 0.1858,
       "step": 900
     },
     {
       "epoch": 81.98,
-      "eval_loss": 0.5448421835899353,
-      "eval_runtime": 2.4539,
-      "eval_samples_per_second": 6.928,
-      "eval_steps_per_second": 2.038,
-      "eval_wer": 0.20206489675516223,
+      "eval_loss": 0.46348273754119873,
+      "eval_runtime": 2.4786,
+      "eval_samples_per_second": 6.859,
+      "eval_steps_per_second": 2.017,
+      "eval_wer": 0.16666666666666666,
       "step": 902
     },
     {
       "epoch": 82.71,
-      "learning_rate": 5.495327102803738e-05,
-      "loss": 0.2075,
+      "learning_rate": 5.355140186915888e-05,
+      "loss": 0.1701,
       "step": 910
     },
     {
       "epoch": 82.98,
-      "eval_loss": 0.4626116156578064,
-      "eval_runtime": 2.4651,
-      "eval_samples_per_second": 6.896,
-      "eval_steps_per_second": 2.028,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.4837586581707001,
+      "eval_runtime": 2.4341,
+      "eval_samples_per_second": 6.984,
+      "eval_steps_per_second": 2.054,
+      "eval_wer": 0.17256637168141592,
       "step": 913
     },
     {
       "epoch": 83.62,
-      "learning_rate": 5.214953271028037e-05,
-      "loss": 0.2314,
+      "learning_rate": 5.074766355140187e-05,
+      "loss": 0.188,
       "step": 920
     },
     {
       "epoch": 83.98,
-      "eval_loss": 0.47063305974006653,
-      "eval_runtime": 2.4549,
-      "eval_samples_per_second": 6.925,
-      "eval_steps_per_second": 2.037,
-      "eval_wer": 0.20648967551622419,
+      "eval_loss": 0.4775279462337494,
+      "eval_runtime": 2.5362,
+      "eval_samples_per_second": 6.703,
+      "eval_steps_per_second": 1.971,
+      "eval_wer": 0.18141592920353983,
       "step": 924
     },
     {
       "epoch": 84.53,
-      "learning_rate": 4.934579439252336e-05,
-      "loss": 0.2208,
+      "learning_rate": 4.794392523364486e-05,
+      "loss": 0.1789,
       "step": 930
     },
     {
       "epoch": 84.98,
-      "eval_loss": 0.5073422789573669,
-      "eval_runtime": 2.4571,
-      "eval_samples_per_second": 6.919,
-      "eval_steps_per_second": 2.035,
-      "eval_wer": 0.20353982300884957,
+      "eval_loss": 0.4800940752029419,
+      "eval_runtime": 2.4737,
+      "eval_samples_per_second": 6.872,
+      "eval_steps_per_second": 2.021,
+      "eval_wer": 0.17404129793510326,
       "step": 935
     },
     {
       "epoch": 85.44,
-      "learning_rate": 4.6542056074766354e-05,
-      "loss": 0.2547,
+      "learning_rate": 4.514018691588785e-05,
+      "loss": 0.2134,
       "step": 940
     },
     {
       "epoch": 85.98,
-      "eval_loss": 0.48184993863105774,
-      "eval_runtime": 2.4595,
-      "eval_samples_per_second": 6.912,
-      "eval_steps_per_second": 2.033,
-      "eval_wer": 0.19616519174041297,
+      "eval_loss": 0.45418649911880493,
+      "eval_runtime": 2.4552,
+      "eval_samples_per_second": 6.924,
+      "eval_steps_per_second": 2.037,
+      "eval_wer": 0.17846607669616518,
       "step": 946
     },
     {
       "epoch": 86.36,
-      "learning_rate": 4.3738317757009345e-05,
-      "loss": 0.2895,
+      "learning_rate": 4.233644859813084e-05,
+      "loss": 0.2141,
       "step": 950
     },
     {
       "epoch": 86.98,
-      "eval_loss": 0.49310269951820374,
-      "eval_runtime": 2.4502,
-      "eval_samples_per_second": 6.938,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.4498932659626007,
+      "eval_runtime": 2.4831,
+      "eval_samples_per_second": 6.846,
+      "eval_steps_per_second": 2.014,
+      "eval_wer": 0.17846607669616518,
       "step": 957
     },
     {
       "epoch": 87.27,
-      "learning_rate": 4.0934579439252335e-05,
-      "loss": 0.1988,
+      "learning_rate": 3.953271028037383e-05,
+      "loss": 0.1599,
       "step": 960
     },
     {
       "epoch": 87.98,
-      "eval_loss": 0.47016218304634094,
-      "eval_runtime": 2.4764,
-      "eval_samples_per_second": 6.865,
-      "eval_steps_per_second": 2.019,
-      "eval_wer": 0.20058997050147492,
+      "eval_loss": 0.4594888687133789,
+      "eval_runtime": 2.4694,
+      "eval_samples_per_second": 6.884,
+      "eval_steps_per_second": 2.025,
+      "eval_wer": 0.17699115044247787,
       "step": 968
     },
     {
       "epoch": 88.18,
-      "learning_rate": 3.8130841121495326e-05,
-      "loss": 0.2383,
+      "learning_rate": 3.672897196261682e-05,
+      "loss": 0.1927,
       "step": 970
     },
     {
       "epoch": 88.98,
-      "eval_loss": 0.4681670665740967,
-      "eval_runtime": 2.4685,
-      "eval_samples_per_second": 6.887,
-      "eval_steps_per_second": 2.026,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.4772101640701294,
+      "eval_runtime": 2.4616,
+      "eval_samples_per_second": 6.906,
+      "eval_steps_per_second": 2.031,
+      "eval_wer": 0.17551622418879056,
       "step": 979
     },
     {
       "epoch": 89.09,
-      "learning_rate": 3.5327102803738316e-05,
-      "loss": 0.2235,
+      "learning_rate": 3.392523364485981e-05,
+      "loss": 0.1715,
       "step": 980
     },
     {
       "epoch": 89.98,
-      "learning_rate": 3.252336448598131e-05,
-      "loss": 0.2332,
+      "learning_rate": 3.11214953271028e-05,
+      "loss": 0.1709,
       "step": 990
     },
     {
       "epoch": 89.98,
-      "eval_loss": 0.4574868381023407,
-      "eval_runtime": 2.4687,
-      "eval_samples_per_second": 6.886,
-      "eval_steps_per_second": 2.025,
-      "eval_wer": 0.20648967551622419,
+      "eval_loss": 0.458842009305954,
+      "eval_runtime": 2.4896,
+      "eval_samples_per_second": 6.828,
+      "eval_steps_per_second": 2.008,
+      "eval_wer": 0.17699115044247787,
       "step": 990
     },
     {
       "epoch": 90.89,
-      "learning_rate": 2.9719626168224294e-05,
-      "loss": 0.1983,
+      "learning_rate": 2.831775700934579e-05,
+      "loss": 0.1588,
       "step": 1000
     },
     {
       "epoch": 90.98,
-      "eval_loss": 0.47059592604637146,
-      "eval_runtime": 2.441,
-      "eval_samples_per_second": 6.964,
-      "eval_steps_per_second": 2.048,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.4606749713420868,
+      "eval_runtime": 2.4588,
+      "eval_samples_per_second": 6.914,
+      "eval_steps_per_second": 2.033,
+      "eval_wer": 0.17846607669616518,
       "step": 1001
     },
     {
       "epoch": 91.8,
-      "learning_rate": 2.6915887850467285e-05,
-      "loss": 0.2182,
+      "learning_rate": 2.5514018691588784e-05,
+      "loss": 0.1702,
       "step": 1010
     },
     {
       "epoch": 91.98,
-      "eval_loss": 0.4756470322608948,
-      "eval_runtime": 2.4523,
-      "eval_samples_per_second": 6.932,
-      "eval_steps_per_second": 2.039,
-      "eval_wer": 0.19911504424778761,
+      "eval_loss": 0.46561577916145325,
+      "eval_runtime": 2.4745,
+      "eval_samples_per_second": 6.87,
+      "eval_steps_per_second": 2.021,
+      "eval_wer": 0.18289085545722714,
       "step": 1012
     },
     {
       "epoch": 92.71,
-      "learning_rate": 2.411214953271028e-05,
-      "loss": 0.2161,
+      "learning_rate": 2.2710280373831774e-05,
+      "loss": 0.1646,
       "step": 1020
     },
     {
       "epoch": 92.98,
-      "eval_loss": 0.4685567617416382,
-      "eval_runtime": 2.5426,
-      "eval_samples_per_second": 6.686,
-      "eval_steps_per_second": 1.967,
-      "eval_wer": 0.19616519174041297,
+      "eval_loss": 0.46306708455085754,
+      "eval_runtime": 2.4484,
+      "eval_samples_per_second": 6.943,
+      "eval_steps_per_second": 2.042,
+      "eval_wer": 0.18289085545722714,
       "step": 1023
     },
     {
       "epoch": 93.62,
-      "learning_rate": 2.130841121495327e-05,
-      "loss": 0.2215,
+      "learning_rate": 1.9906542056074765e-05,
+      "loss": 0.1867,
       "step": 1030
     },
     {
       "epoch": 93.98,
-      "eval_loss": 0.4688819944858551,
-      "eval_runtime": 2.4691,
-      "eval_samples_per_second": 6.885,
-      "eval_steps_per_second": 2.025,
-      "eval_wer": 0.19321533923303835,
+      "eval_loss": 0.4757649600505829,
+      "eval_runtime": 2.4618,
+      "eval_samples_per_second": 6.906,
+      "eval_steps_per_second": 2.031,
+      "eval_wer": 0.18141592920353983,
       "step": 1034
     },
     {
       "epoch": 94.53,
-      "learning_rate": 1.850467289719626e-05,
-      "loss": 0.2223,
+      "learning_rate": 1.7102803738317756e-05,
+      "loss": 0.1799,
       "step": 1040
     },
     {
       "epoch": 94.98,
-      "eval_loss": 0.451424241065979,
-      "eval_runtime": 2.4423,
-      "eval_samples_per_second": 6.961,
-      "eval_steps_per_second": 2.047,
-      "eval_wer": 0.1887905604719764,
+      "eval_loss": 0.48200085759162903,
+      "eval_runtime": 2.4572,
+      "eval_samples_per_second": 6.919,
+      "eval_steps_per_second": 2.035,
+      "eval_wer": 0.17551622418879056,
       "step": 1045
     },
     {
       "epoch": 95.44,
-      "learning_rate": 1.570093457943925e-05,
-      "loss": 0.2068,
+      "learning_rate": 1.4299065420560746e-05,
+      "loss": 0.1611,
       "step": 1050
     },
     {
       "epoch": 95.98,
-      "eval_loss": 0.44822120666503906,
-      "eval_runtime": 2.455,
-      "eval_samples_per_second": 6.925,
-      "eval_steps_per_second": 2.037,
-      "eval_wer": 0.1887905604719764,
+      "eval_loss": 0.48464781045913696,
+      "eval_runtime": 2.4952,
+      "eval_samples_per_second": 6.813,
+      "eval_steps_per_second": 2.004,
+      "eval_wer": 0.17846607669616518,
       "step": 1056
     },
     {
       "epoch": 96.36,
-      "learning_rate": 1.2897196261682242e-05,
-      "loss": 0.2046,
+      "learning_rate": 1.1495327102803737e-05,
+      "loss": 0.1685,
       "step": 1060
     },
     {
       "epoch": 96.98,
-      "eval_loss": 0.4480588138103485,
-      "eval_runtime": 2.4461,
-      "eval_samples_per_second": 6.95,
-      "eval_steps_per_second": 2.044,
-      "eval_wer": 0.18584070796460178,
+      "eval_loss": 0.48164939880371094,
+      "eval_runtime": 2.5142,
+      "eval_samples_per_second": 6.762,
+      "eval_steps_per_second": 1.989,
+      "eval_wer": 0.17699115044247787,
       "step": 1067
     },
     {
       "epoch": 97.27,
-      "learning_rate": 1.0093457943925232e-05,
-      "loss": 0.2411,
+      "learning_rate": 8.691588785046728e-06,
+      "loss": 0.19,
       "step": 1070
     },
     {
       "epoch": 97.98,
-      "eval_loss": 0.4532119333744049,
-      "eval_runtime": 2.4326,
-      "eval_samples_per_second": 6.988,
-      "eval_steps_per_second": 2.055,
-      "eval_wer": 0.1902654867256637,
+      "eval_loss": 0.4781480133533478,
+      "eval_runtime": 2.5523,
+      "eval_samples_per_second": 6.661,
+      "eval_steps_per_second": 1.959,
+      "eval_wer": 0.17699115044247787,
       "step": 1078
     },
     {
       "epoch": 98.18,
-      "learning_rate": 7.289719626168224e-06,
-      "loss": 0.2296,
+      "learning_rate": 5.887850467289719e-06,
+      "loss": 0.1953,
       "step": 1080
     },
     {
       "epoch": 98.98,
-      "eval_loss": 0.4600735306739807,
-      "eval_runtime": 2.4501,
-      "eval_samples_per_second": 6.939,
-      "eval_steps_per_second": 2.041,
-      "eval_wer": 0.19321533923303835,
+      "eval_loss": 0.47668424248695374,
+      "eval_runtime": 2.4703,
+      "eval_samples_per_second": 6.882,
+      "eval_steps_per_second": 2.024,
+      "eval_wer": 0.17699115044247787,
       "step": 1089
     },
     {
       "epoch": 99.09,
-      "learning_rate": 4.4859813084112145e-06,
-      "loss": 0.2113,
+      "learning_rate": 3.08411214953271e-06,
+      "loss": 0.1635,
       "step": 1090
     },
     {
       "epoch": 99.98,
-      "learning_rate": 1.6822429906542053e-06,
-      "loss": 0.2211,
+      "learning_rate": 2.803738317757009e-07,
+      "loss": 0.188,
       "step": 1100
     },
     {
       "epoch": 99.98,
-      "eval_loss": 0.462471067905426,
-      "eval_runtime": 2.4339,
-      "eval_samples_per_second": 6.985,
-      "eval_steps_per_second": 2.054,
-      "eval_wer": 0.19469026548672566,
+      "eval_loss": 0.47740593552589417,
+      "eval_runtime": 2.4786,
+      "eval_samples_per_second": 6.859,
+      "eval_steps_per_second": 2.017,
+      "eval_wer": 0.17699115044247787,
       "step": 1100
     },
     {
       "epoch": 99.98,
       "step": 1100,
       "total_flos": 1.8082449930602819e+18,
-      "train_loss": 0.679512545520609,
-      "train_runtime": 4168.4023,
-      "train_samples_per_second": 4.294,
-      "train_steps_per_second": 0.264
+      "train_loss": 0.2150352662259882,
+      "train_runtime": 4203.5944,
+      "train_samples_per_second": 4.258,
+      "train_steps_per_second": 0.262
     }
   ],
   "max_steps": 1100,