End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.742971658706665,
-    "eval_runtime": 62.6586,
+    "eval_loss": 1.7481727600097656,
+    "eval_runtime": 63.6288,
     "eval_samples": 1680,
-    "eval_samples_per_second": 26.812,
-    "eval_steps_per_second": 26.812,
-    "eval_wer": 0.808972503617945,
-    "train_loss": 2.308333969445064,
-    "train_runtime": 4034.046,
+    "eval_samples_per_second": 26.403,
+    "eval_steps_per_second": 26.403,
+    "eval_wer": 0.7987044311212184,
+    "train_loss": 2.305506002491918,
+    "train_runtime": 4265.6036,
     "train_samples": 4620,
-    "train_samples_per_second": 22.905,
-    "train_steps_per_second": 0.719
+    "train_samples_per_second": 21.662,
+    "train_steps_per_second": 0.68
 }

emissions.csv

@@ -1,2 +1,3 @@
 timestamp,experiment_id,project_name,duration,emissions,energy_consumed,country_name,country_iso_code,region,on_cloud,cloud_provider,cloud_region
 2021-10-28T14:07:16,c1359372-7a2f-4c35-8894-485368db7654,codecarbon,4031.0849454402924,0.009986100533709014,0.04744533240091166,United States,USA,new york,N,,
+2021-10-28T15:24:13,dfd85941-742c-4041-8873-eb0fc7f517aa,codecarbon,4262.816283941269,0.010563398417267292,0.05018815377421022,United States,USA,new york,N,,

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.742971658706665,
-    "eval_runtime": 62.6586,
+    "eval_loss": 1.7481727600097656,
+    "eval_runtime": 63.6288,
     "eval_samples": 1680,
-    "eval_samples_per_second": 26.812,
-    "eval_steps_per_second": 26.812,
-    "eval_wer": 0.808972503617945
+    "eval_samples_per_second": 26.403,
+    "eval_steps_per_second": 26.403,
+    "eval_wer": 0.7987044311212184
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27eaa3571c7de058069859c3ca0a47cf9c3332f5e005ace596cd29c0ac5cedde
+oid sha256:d055e70a0bb80565ab20248a8c7638e55789cc3b48c9dbd0dcd6ade7cfc11286
 size 164046635

runs/Oct28_14-11-54_brutasse/events.out.tfevents.1635430384.brutasse.22330.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e81d12c8e29e2fab6bc5a1eed454a0235c0254addbbe58a68bca6e90fda5456
-size 57216
+oid sha256:6005f973af9755164116075861924729cec76e698b42253d172b8b95c933d37f
+size 59458

runs/Oct28_14-11-54_brutasse/events.out.tfevents.1635434718.brutasse.22330.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c505fce74b247973919ab606b9f69d61c41d39206fb6d8fb8944f307ee8415bf
+size 358

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "train_loss": 2.308333969445064,
-    "train_runtime": 4034.046,
+    "train_loss": 2.305506002491918,
+    "train_runtime": 4265.6036,
     "train_samples": 4620,
-    "train_samples_per_second": 22.905,
-    "train_steps_per_second": 0.719
+    "train_samples_per_second": 21.662,
+    "train_steps_per_second": 0.68
 }

trainer_state.json

@@ -34,25 +34,25 @@
     {
       "epoch": 0.34,
       "learning_rate": 4.800000000000001e-06,
-      "loss": 7.1211,
+      "loss": 7.121,
       "step": 50
     },
     {
       "epoch": 0.41,
       "learning_rate": 5.8e-06,
-      "loss": 6.1102,
+      "loss": 6.1101,
       "step": 60
     },
     {
       "epoch": 0.48,
       "learning_rate": 6.800000000000001e-06,
-      "loss": 5.2533,
+      "loss": 5.2532,
       "step": 70
     },
     {
       "epoch": 0.55,
       "learning_rate": 7.7e-06,
-      "loss": 5.3192,
+      "loss": 5.3191,
       "step": 80
     },
     {
@@ -69,10 +69,10 @@
     },
     {
       "epoch": 0.69,
-      "eval_loss": 4.080197811126709,
-      "eval_runtime": 61.9535,
-      "eval_samples_per_second": 27.117,
-      "eval_steps_per_second": 27.117,
+      "eval_loss": 4.080183982849121,
+      "eval_runtime": 61.3907,
+      "eval_samples_per_second": 27.366,
+      "eval_steps_per_second": 27.366,
       "eval_wer": 1.0,
       "step": 100
     },
@@ -85,7 +85,7 @@
     {
       "epoch": 0.83,
       "learning_rate": 1.1700000000000001e-05,
-      "loss": 3.7734,
+      "loss": 3.7733,
       "step": 120
     },
     {
@@ -133,15 +133,15 @@
     {
       "epoch": 1.38,
       "learning_rate": 1.97e-05,
-      "loss": 2.9806,
+      "loss": 2.9805,
       "step": 200
     },
     {
       "epoch": 1.38,
-      "eval_loss": 2.979235887527466,
-      "eval_runtime": 61.729,
-      "eval_samples_per_second": 27.216,
-      "eval_steps_per_second": 27.216,
+      "eval_loss": 2.9792349338531494,
+      "eval_runtime": 61.348,
+      "eval_samples_per_second": 27.385,
+      "eval_steps_per_second": 27.385,
       "eval_wer": 1.0,
       "step": 200
     },
@@ -207,10 +207,10 @@
     },
     {
       "epoch": 2.07,
-      "eval_loss": 2.9407591819763184,
-      "eval_runtime": 62.0192,
-      "eval_samples_per_second": 27.088,
-      "eval_steps_per_second": 27.088,
+      "eval_loss": 2.9407615661621094,
+      "eval_runtime": 61.4623,
+      "eval_samples_per_second": 27.334,
+      "eval_steps_per_second": 27.334,
       "eval_wer": 1.0,
       "step": 300
     },
@@ -271,15 +271,15 @@
     {
       "epoch": 2.76,
       "learning_rate": 3.97e-05,
-      "loss": 2.9656,
+      "loss": 2.9655,
       "step": 400
     },
     {
       "epoch": 2.76,
-      "eval_loss": 2.914335012435913,
-      "eval_runtime": 61.8283,
-      "eval_samples_per_second": 27.172,
-      "eval_steps_per_second": 27.172,
+      "eval_loss": 2.9143309593200684,
+      "eval_runtime": 61.4594,
+      "eval_samples_per_second": 27.335,
+      "eval_steps_per_second": 27.335,
       "eval_wer": 1.0,
       "step": 400
     },
@@ -322,13 +322,13 @@
     {
       "epoch": 3.24,
       "learning_rate": 4.6700000000000003e-05,
-      "loss": 2.8962,
+      "loss": 2.8961,
       "step": 470
     },
     {
       "epoch": 3.31,
       "learning_rate": 4.77e-05,
-      "loss": 2.9122,
+      "loss": 2.9121,
       "step": 480
     },
     {
@@ -345,10 +345,10 @@
     },
     {
       "epoch": 3.45,
-      "eval_loss": 2.8774032592773438,
-      "eval_runtime": 61.89,
-      "eval_samples_per_second": 27.145,
-      "eval_steps_per_second": 27.145,
+      "eval_loss": 2.8774638175964355,
+      "eval_runtime": 62.6239,
+      "eval_samples_per_second": 26.827,
+      "eval_steps_per_second": 26.827,
       "eval_wer": 1.0,
       "step": 500
     },
@@ -361,7 +361,7 @@
     {
       "epoch": 3.59,
       "learning_rate": 5.17e-05,
-      "loss": 2.8849,
+      "loss": 2.885,
       "step": 520
     },
     {
@@ -373,1024 +373,1024 @@
     {
       "epoch": 3.72,
       "learning_rate": 5.3700000000000004e-05,
-      "loss": 2.8881,
+      "loss": 2.888,
       "step": 540
     },
     {
       "epoch": 3.79,
       "learning_rate": 5.470000000000001e-05,
-      "loss": 2.9057,
+      "loss": 2.9058,
       "step": 550
     },
     {
       "epoch": 3.86,
       "learning_rate": 5.5700000000000005e-05,
-      "loss": 2.8631,
+      "loss": 2.8629,
       "step": 560
     },
     {
       "epoch": 3.93,
       "learning_rate": 5.6699999999999996e-05,
-      "loss": 2.844,
+      "loss": 2.8436,
       "step": 570
     },
     {
       "epoch": 4.0,
       "learning_rate": 5.77e-05,
-      "loss": 2.8526,
+      "loss": 2.8511,
       "step": 580
     },
     {
       "epoch": 4.07,
       "learning_rate": 5.87e-05,
-      "loss": 2.8439,
+      "loss": 2.8431,
       "step": 590
     },
     {
       "epoch": 4.14,
       "learning_rate": 5.97e-05,
-      "loss": 2.7714,
+      "loss": 2.7719,
       "step": 600
     },
     {
       "epoch": 4.14,
-      "eval_loss": 2.777448892593384,
-      "eval_runtime": 61.9349,
-      "eval_samples_per_second": 27.125,
-      "eval_steps_per_second": 27.125,
+      "eval_loss": 2.7815191745758057,
+      "eval_runtime": 62.3898,
+      "eval_samples_per_second": 26.927,
+      "eval_steps_per_second": 26.927,
       "eval_wer": 0.9999310867617669,
       "step": 600
     },
     {
       "epoch": 4.21,
       "learning_rate": 6.07e-05,
-      "loss": 2.772,
+      "loss": 2.7765,
       "step": 610
     },
     {
       "epoch": 4.28,
       "learning_rate": 6.170000000000001e-05,
-      "loss": 2.8175,
+      "loss": 2.8196,
       "step": 620
     },
     {
       "epoch": 4.34,
       "learning_rate": 6.27e-05,
-      "loss": 2.7637,
+      "loss": 2.7596,
       "step": 630
     },
     {
       "epoch": 4.41,
       "learning_rate": 6.37e-05,
-      "loss": 2.7476,
+      "loss": 2.7299,
       "step": 640
     },
     {
       "epoch": 4.48,
       "learning_rate": 6.47e-05,
-      "loss": 2.7451,
+      "loss": 2.7264,
       "step": 650
     },
     {
       "epoch": 4.55,
       "learning_rate": 6.570000000000001e-05,
-      "loss": 2.7698,
+      "loss": 2.7499,
       "step": 660
     },
     {
       "epoch": 4.62,
       "learning_rate": 6.670000000000001e-05,
-      "loss": 2.6918,
+      "loss": 2.6799,
       "step": 670
     },
     {
       "epoch": 4.69,
       "learning_rate": 6.77e-05,
-      "loss": 2.7045,
+      "loss": 2.682,
       "step": 680
     },
     {
       "epoch": 4.76,
       "learning_rate": 6.87e-05,
-      "loss": 2.775,
+      "loss": 2.7446,
       "step": 690
     },
     {
       "epoch": 4.83,
       "learning_rate": 6.97e-05,
-      "loss": 2.6883,
+      "loss": 2.6531,
       "step": 700
     },
     {
       "epoch": 4.83,
-      "eval_loss": 2.655168294906616,
-      "eval_runtime": 61.9601,
-      "eval_samples_per_second": 27.114,
-      "eval_steps_per_second": 27.114,
-      "eval_wer": 0.9853903934945903,
+      "eval_loss": 2.6374878883361816,
+      "eval_runtime": 63.0358,
+      "eval_samples_per_second": 26.652,
+      "eval_steps_per_second": 26.652,
+      "eval_wer": 1.006477844393908,
       "step": 700
     },
     {
       "epoch": 4.9,
       "learning_rate": 7.07e-05,
-      "loss": 2.6235,
+      "loss": 2.6098,
       "step": 710
     },
     {
       "epoch": 4.97,
       "learning_rate": 7.17e-05,
-      "loss": 2.6454,
+      "loss": 2.6314,
       "step": 720
     },
     {
       "epoch": 5.03,
       "learning_rate": 7.27e-05,
-      "loss": 2.7134,
+      "loss": 2.7144,
       "step": 730
     },
     {
       "epoch": 5.1,
       "learning_rate": 7.37e-05,
-      "loss": 2.6114,
+      "loss": 2.6205,
       "step": 740
     },
     {
       "epoch": 5.17,
       "learning_rate": 7.47e-05,
-      "loss": 2.5768,
+      "loss": 2.5682,
       "step": 750
     },
     {
       "epoch": 5.24,
       "learning_rate": 7.570000000000001e-05,
-      "loss": 2.6549,
+      "loss": 2.6472,
       "step": 760
     },
     {
       "epoch": 5.31,
       "learning_rate": 7.670000000000001e-05,
-      "loss": 2.6445,
+      "loss": 2.6284,
       "step": 770
     },
     {
       "epoch": 5.38,
       "learning_rate": 7.77e-05,
-      "loss": 2.5427,
+      "loss": 2.547,
       "step": 780
     },
     {
       "epoch": 5.45,
       "learning_rate": 7.87e-05,
-      "loss": 2.5433,
+      "loss": 2.5544,
       "step": 790
     },
     {
       "epoch": 5.52,
       "learning_rate": 7.970000000000001e-05,
-      "loss": 2.6477,
+      "loss": 2.6425,
       "step": 800
     },
     {
       "epoch": 5.52,
-      "eval_loss": 2.5424375534057617,
-      "eval_runtime": 62.1641,
-      "eval_samples_per_second": 27.025,
-      "eval_steps_per_second": 27.025,
-      "eval_wer": 1.0113706843084556,
+      "eval_loss": 2.5602073669433594,
+      "eval_runtime": 63.2769,
+      "eval_samples_per_second": 26.55,
+      "eval_steps_per_second": 26.55,
+      "eval_wer": 1.0210185376610847,
       "step": 800
     },
     {
       "epoch": 5.59,
       "learning_rate": 8.070000000000001e-05,
-      "loss": 2.5609,
+      "loss": 2.5776,
       "step": 810
     },
     {
       "epoch": 5.66,
       "learning_rate": 8.17e-05,
-      "loss": 2.4538,
+      "loss": 2.4746,
       "step": 820
     },
     {
       "epoch": 5.72,
       "learning_rate": 8.27e-05,
-      "loss": 2.521,
+      "loss": 2.5199,
       "step": 830
     },
     {
       "epoch": 5.79,
       "learning_rate": 8.37e-05,
-      "loss": 2.6379,
+      "loss": 2.6432,
       "step": 840
     },
     {
       "epoch": 5.86,
       "learning_rate": 8.47e-05,
-      "loss": 2.4359,
+      "loss": 2.4396,
       "step": 850
     },
     {
       "epoch": 5.93,
       "learning_rate": 8.57e-05,
-      "loss": 2.4555,
+      "loss": 2.4539,
       "step": 860
     },
     {
       "epoch": 6.0,
       "learning_rate": 8.67e-05,
-      "loss": 2.542,
+      "loss": 2.5582,
       "step": 870
     },
     {
       "epoch": 6.07,
       "learning_rate": 8.77e-05,
-      "loss": 2.5067,
+      "loss": 2.5219,
       "step": 880
     },
     {
       "epoch": 6.14,
       "learning_rate": 8.87e-05,
-      "loss": 2.3783,
+      "loss": 2.4092,
       "step": 890
     },
     {
       "epoch": 6.21,
       "learning_rate": 8.970000000000001e-05,
-      "loss": 2.3607,
+      "loss": 2.3963,
       "step": 900
     },
     {
       "epoch": 6.21,
-      "eval_loss": 2.4291484355926514,
-      "eval_runtime": 62.2969,
-      "eval_samples_per_second": 26.968,
-      "eval_steps_per_second": 26.968,
-      "eval_wer": 1.101233546964372,
+      "eval_loss": 2.466503858566284,
+      "eval_runtime": 63.3462,
+      "eval_samples_per_second": 26.521,
+      "eval_steps_per_second": 26.521,
+      "eval_wer": 1.0590586451657364,
       "step": 900
     },
     {
       "epoch": 6.28,
       "learning_rate": 9.070000000000001e-05,
-      "loss": 2.584,
+      "loss": 2.5834,
       "step": 910
     },
     {
       "epoch": 6.34,
       "learning_rate": 9.17e-05,
-      "loss": 2.4121,
+      "loss": 2.4385,
       "step": 920
     },
     {
       "epoch": 6.41,
       "learning_rate": 9.27e-05,
-      "loss": 2.363,
+      "loss": 2.3945,
       "step": 930
     },
     {
       "epoch": 6.48,
       "learning_rate": 9.370000000000001e-05,
-      "loss": 2.4513,
+      "loss": 2.4617,
       "step": 940
     },
     {
       "epoch": 6.55,
       "learning_rate": 9.47e-05,
-      "loss": 2.5266,
+      "loss": 2.5169,
       "step": 950
     },
     {
       "epoch": 6.62,
       "learning_rate": 9.57e-05,
-      "loss": 2.261,
+      "loss": 2.2648,
       "step": 960
     },
     {
       "epoch": 6.69,
       "learning_rate": 9.67e-05,
-      "loss": 2.2193,
+      "loss": 2.2494,
       "step": 970
     },
     {
       "epoch": 6.76,
       "learning_rate": 9.77e-05,
-      "loss": 2.5063,
+      "loss": 2.5194,
       "step": 980
     },
     {
       "epoch": 6.83,
       "learning_rate": 9.87e-05,
-      "loss": 2.3737,
+      "loss": 2.3886,
       "step": 990
     },
     {
       "epoch": 6.9,
       "learning_rate": 9.970000000000001e-05,
-      "loss": 2.1048,
+      "loss": 2.1447,
       "step": 1000
     },
     {
       "epoch": 6.9,
-      "eval_loss": 2.2525901794433594,
-      "eval_runtime": 62.0939,
-      "eval_samples_per_second": 27.056,
-      "eval_steps_per_second": 27.056,
-      "eval_wer": 0.9743642753773,
+      "eval_loss": 2.2791571617126465,
+      "eval_runtime": 63.8981,
+      "eval_samples_per_second": 26.292,
+      "eval_steps_per_second": 26.292,
+      "eval_wer": 0.9847701743504927,
       "step": 1000
     },
     {
       "epoch": 6.97,
       "learning_rate": 9.963157894736843e-05,
-      "loss": 2.2737,
+      "loss": 2.2952,
       "step": 1010
     },
     {
       "epoch": 7.03,
       "learning_rate": 9.910526315789475e-05,
-      "loss": 2.4889,
+      "loss": 2.4979,
       "step": 1020
     },
     {
       "epoch": 7.1,
       "learning_rate": 9.857894736842106e-05,
-      "loss": 2.1847,
+      "loss": 2.1919,
       "step": 1030
     },
     {
       "epoch": 7.17,
       "learning_rate": 9.805263157894737e-05,
-      "loss": 2.135,
+      "loss": 2.1653,
       "step": 1040
     },
     {
       "epoch": 7.24,
       "learning_rate": 9.752631578947369e-05,
-      "loss": 2.3103,
+      "loss": 2.3318,
       "step": 1050
     },
     {
       "epoch": 7.31,
       "learning_rate": 9.7e-05,
-      "loss": 2.3921,
+      "loss": 2.3904,
       "step": 1060
     },
     {
       "epoch": 7.38,
       "learning_rate": 9.647368421052631e-05,
-      "loss": 2.0108,
+      "loss": 2.0835,
       "step": 1070
     },
     {
       "epoch": 7.45,
       "learning_rate": 9.594736842105264e-05,
-      "loss": 2.1143,
+      "loss": 2.146,
       "step": 1080
     },
     {
       "epoch": 7.52,
       "learning_rate": 9.542105263157895e-05,
-      "loss": 2.4364,
+      "loss": 2.4565,
       "step": 1090
     },
     {
       "epoch": 7.59,
       "learning_rate": 9.489473684210527e-05,
-      "loss": 2.2448,
+      "loss": 2.2719,
       "step": 1100
     },
     {
       "epoch": 7.59,
-      "eval_loss": 2.169701099395752,
-      "eval_runtime": 62.1876,
-      "eval_samples_per_second": 27.015,
-      "eval_steps_per_second": 27.015,
-      "eval_wer": 0.9332230721521604,
+      "eval_loss": 2.2237350940704346,
+      "eval_runtime": 63.7312,
+      "eval_samples_per_second": 26.361,
+      "eval_steps_per_second": 26.361,
+      "eval_wer": 0.9464544138929089,
       "step": 1100
     },
     {
       "epoch": 7.66,
       "learning_rate": 9.436842105263158e-05,
-      "loss": 1.9274,
+      "loss": 1.9881,
       "step": 1110
     },
     {
       "epoch": 7.72,
       "learning_rate": 9.384210526315789e-05,
-      "loss": 2.1163,
+      "loss": 2.1616,
       "step": 1120
     },
     {
       "epoch": 7.79,
       "learning_rate": 9.331578947368422e-05,
-      "loss": 2.3358,
+      "loss": 2.3746,
       "step": 1130
     },
     {
       "epoch": 7.86,
       "learning_rate": 9.278947368421053e-05,
-      "loss": 1.9315,
+      "loss": 1.9976,
       "step": 1140
     },
     {
       "epoch": 7.93,
       "learning_rate": 9.226315789473686e-05,
-      "loss": 1.9978,
+      "loss": 2.024,
       "step": 1150
     },
     {
       "epoch": 8.0,
       "learning_rate": 9.173684210526317e-05,
-      "loss": 2.338,
+      "loss": 2.363,
       "step": 1160
     },
     {
       "epoch": 8.07,
       "learning_rate": 9.121052631578948e-05,
-      "loss": 2.2001,
+      "loss": 2.2491,
       "step": 1170
     },
     {
       "epoch": 8.14,
       "learning_rate": 9.06842105263158e-05,
-      "loss": 1.894,
+      "loss": 1.9415,
       "step": 1180
     },
     {
       "epoch": 8.21,
       "learning_rate": 9.015789473684211e-05,
-      "loss": 2.0173,
+      "loss": 2.0498,
       "step": 1190
     },
     {
       "epoch": 8.28,
       "learning_rate": 8.963157894736842e-05,
-      "loss": 2.3525,
+      "loss": 2.3629,
       "step": 1200
     },
     {
       "epoch": 8.28,
-      "eval_loss": 2.042870044708252,
-      "eval_runtime": 62.0448,
-      "eval_samples_per_second": 27.077,
-      "eval_steps_per_second": 27.077,
-      "eval_wer": 0.8720970298394322,
+      "eval_loss": 2.1057612895965576,
+      "eval_runtime": 63.7855,
+      "eval_samples_per_second": 26.338,
+      "eval_steps_per_second": 26.338,
+      "eval_wer": 0.8907036041623596,
       "step": 1200
     },
     {
       "epoch": 8.34,
       "learning_rate": 8.910526315789474e-05,
-      "loss": 1.8758,
+      "loss": 1.94,
       "step": 1210
     },
     {
       "epoch": 8.41,
       "learning_rate": 8.857894736842106e-05,
-      "loss": 1.8701,
+      "loss": 1.8996,
       "step": 1220
     },
     {
       "epoch": 8.48,
       "learning_rate": 8.805263157894737e-05,
-      "loss": 2.1876,
+      "loss": 2.2006,
       "step": 1230
     },
     {
       "epoch": 8.55,
       "learning_rate": 8.752631578947369e-05,
-      "loss": 2.2548,
+      "loss": 2.2553,
       "step": 1240
     },
     {
       "epoch": 8.62,
       "learning_rate": 8.7e-05,
-      "loss": 1.8274,
+      "loss": 1.8448,
       "step": 1250
     },
     {
       "epoch": 8.69,
       "learning_rate": 8.647368421052631e-05,
-      "loss": 1.9417,
+      "loss": 1.9506,
       "step": 1260
     },
     {
       "epoch": 8.76,
       "learning_rate": 8.594736842105263e-05,
-      "loss": 2.3158,
+      "loss": 2.3542,
       "step": 1270
     },
     {
       "epoch": 8.83,
       "learning_rate": 8.542105263157894e-05,
-      "loss": 2.0282,
+      "loss": 2.0478,
       "step": 1280
     },
     {
       "epoch": 8.9,
       "learning_rate": 8.489473684210527e-05,
-      "loss": 1.7798,
+      "loss": 1.7948,
       "step": 1290
     },
     {
       "epoch": 8.97,
       "learning_rate": 8.436842105263158e-05,
-      "loss": 2.099,
+      "loss": 2.0913,
       "step": 1300
     },
     {
       "epoch": 8.97,
-      "eval_loss": 2.0076003074645996,
-      "eval_runtime": 62.3855,
-      "eval_samples_per_second": 26.929,
-      "eval_steps_per_second": 26.929,
-      "eval_wer": 0.9182688994555854,
+      "eval_loss": 2.0112950801849365,
+      "eval_runtime": 63.7547,
+      "eval_samples_per_second": 26.351,
+      "eval_steps_per_second": 26.351,
+      "eval_wer": 0.9069671283853629,
       "step": 1300
     },
     {
       "epoch": 9.03,
       "learning_rate": 8.38421052631579e-05,
-      "loss": 2.2983,
+      "loss": 2.2877,
       "step": 1310
     },
     {
       "epoch": 9.1,
       "learning_rate": 8.331578947368422e-05,
-      "loss": 1.7612,
+      "loss": 1.7645,
       "step": 1320
     },
     {
       "epoch": 9.17,
       "learning_rate": 8.278947368421053e-05,
-      "loss": 1.7982,
+      "loss": 1.7994,
       "step": 1330
     },
     {
       "epoch": 9.24,
       "learning_rate": 8.226315789473684e-05,
-      "loss": 2.1745,
+      "loss": 2.166,
       "step": 1340
     },
     {
       "epoch": 9.31,
       "learning_rate": 8.173684210526317e-05,
-      "loss": 2.1586,
+      "loss": 2.1645,
       "step": 1350
     },
     {
       "epoch": 9.38,
       "learning_rate": 8.121052631578948e-05,
-      "loss": 1.8252,
+      "loss": 1.8409,
       "step": 1360
     },
     {
       "epoch": 9.45,
       "learning_rate": 8.06842105263158e-05,
-      "loss": 1.9301,
+      "loss": 1.9361,
       "step": 1370
     },
     {
       "epoch": 9.52,
       "learning_rate": 8.015789473684211e-05,
-      "loss": 2.2251,
+      "loss": 2.2377,
       "step": 1380
     },
     {
       "epoch": 9.59,
       "learning_rate": 7.963157894736842e-05,
-      "loss": 1.8685,
+      "loss": 1.8667,
       "step": 1390
     },
     {
       "epoch": 9.66,
       "learning_rate": 7.910526315789474e-05,
-      "loss": 1.8236,
+      "loss": 1.8334,
       "step": 1400
     },
     {
       "epoch": 9.66,
-      "eval_loss": 1.9694862365722656,
-      "eval_runtime": 62.2724,
-      "eval_samples_per_second": 26.978,
-      "eval_steps_per_second": 26.978,
-      "eval_wer": 0.8262697264144442,
+      "eval_loss": 1.946588397026062,
+      "eval_runtime": 63.4101,
+      "eval_samples_per_second": 26.494,
+      "eval_steps_per_second": 26.494,
+      "eval_wer": 0.8177244848735442,
       "step": 1400
     },
     {
       "epoch": 9.72,
       "learning_rate": 7.857894736842105e-05,
-      "loss": 2.1524,
+      "loss": 2.1079,
       "step": 1410
     },
     {
       "epoch": 9.79,
       "learning_rate": 7.805263157894738e-05,
-      "loss": 2.2737,
+      "loss": 2.2388,
       "step": 1420
     },
     {
       "epoch": 9.86,
       "learning_rate": 7.752631578947369e-05,
-      "loss": 1.7647,
+      "loss": 1.7457,
       "step": 1430
     },
     {
       "epoch": 9.93,
       "learning_rate": 7.7e-05,
-      "loss": 1.857,
+      "loss": 1.8413,
       "step": 1440
     },
     {
       "epoch": 10.0,
       "learning_rate": 7.647368421052631e-05,
-      "loss": 2.1688,
+      "loss": 2.1608,
       "step": 1450
     },
     {
       "epoch": 10.07,
       "learning_rate": 7.594736842105263e-05,
-      "loss": 2.0013,
+      "loss": 2.023,
       "step": 1460
     },
     {
       "epoch": 10.14,
       "learning_rate": 7.542105263157895e-05,
-      "loss": 1.6881,
+      "loss": 1.6968,
       "step": 1470
     },
     {
       "epoch": 10.21,
       "learning_rate": 7.489473684210527e-05,
-      "loss": 1.8441,
+      "loss": 1.8525,
       "step": 1480
     },
     {
       "epoch": 10.28,
       "learning_rate": 7.43684210526316e-05,
-      "loss": 2.2788,
+      "loss": 2.2802,
       "step": 1490
     },
     {
       "epoch": 10.34,
       "learning_rate": 7.38421052631579e-05,
-      "loss": 1.6602,
+      "loss": 1.6608,
       "step": 1500
     },
     {
       "epoch": 10.34,
-      "eval_loss": 1.932761311531067,
-      "eval_runtime": 62.3067,
-      "eval_samples_per_second": 26.963,
-      "eval_steps_per_second": 26.963,
-      "eval_wer": 0.8536971952312039,
+      "eval_loss": 1.921738862991333,
+      "eval_runtime": 63.4326,
+      "eval_samples_per_second": 26.485,
+      "eval_steps_per_second": 26.485,
+      "eval_wer": 0.869753979739508,
       "step": 1500
     },
     {
       "epoch": 10.41,
       "learning_rate": 7.331578947368422e-05,
-      "loss": 1.7309,
+      "loss": 1.7294,
       "step": 1510
     },
     {
       "epoch": 10.48,
       "learning_rate": 7.278947368421053e-05,
-      "loss": 2.0171,
+      "loss": 2.0241,
       "step": 1520
     },
     {
       "epoch": 10.55,
       "learning_rate": 7.226315789473685e-05,
-      "loss": 2.1863,
+      "loss": 2.1875,
       "step": 1530
     },
     {
       "epoch": 10.62,
       "learning_rate": 7.173684210526316e-05,
-      "loss": 1.7332,
+      "loss": 1.7529,
       "step": 1540
     },
     {
       "epoch": 10.69,
       "learning_rate": 7.121052631578947e-05,
-      "loss": 1.9525,
+      "loss": 1.9517,
       "step": 1550
     },
     {
       "epoch": 10.76,
       "learning_rate": 7.06842105263158e-05,
-      "loss": 2.2204,
+      "loss": 2.2139,
       "step": 1560
     },
     {
       "epoch": 10.83,
       "learning_rate": 7.015789473684211e-05,
-      "loss": 1.8776,
+      "loss": 1.8942,
       "step": 1570
     },
     {
       "epoch": 10.9,
       "learning_rate": 6.963157894736842e-05,
-      "loss": 1.7232,
+      "loss": 1.7269,
       "step": 1580
     },
     {
       "epoch": 10.97,
       "learning_rate": 6.910526315789474e-05,
-      "loss": 1.8677,
+      "loss": 1.8851,
       "step": 1590
     },
     {
       "epoch": 11.03,
       "learning_rate": 6.857894736842105e-05,
-      "loss": 2.2146,
+      "loss": 2.2194,
       "step": 1600
     },
     {
       "epoch": 11.03,
-      "eval_loss": 1.879468321800232,
-      "eval_runtime": 62.2267,
-      "eval_samples_per_second": 26.998,
-      "eval_steps_per_second": 26.998,
-      "eval_wer": 0.8443249948315071,
+      "eval_loss": 1.9090718030929565,
+      "eval_runtime": 63.6141,
+      "eval_samples_per_second": 26.409,
+      "eval_steps_per_second": 26.409,
+      "eval_wer": 0.8727172489835298,
       "step": 1600
     },
     {
       "epoch": 11.1,
       "learning_rate": 6.805263157894736e-05,
-      "loss": 1.7176,
+      "loss": 1.7286,
       "step": 1610
     },
     {
       "epoch": 11.17,
       "learning_rate": 6.752631578947368e-05,
-      "loss": 1.7268,
+      "loss": 1.7089,
       "step": 1620
     },
     {
       "epoch": 11.24,
       "learning_rate": 6.7e-05,
-      "loss": 2.0447,
+      "loss": 2.0165,
       "step": 1630
     },
     {
       "epoch": 11.31,
       "learning_rate": 6.647368421052632e-05,
-      "loss": 2.0887,
+      "loss": 2.08,
       "step": 1640
     },
     {
       "epoch": 11.38,
       "learning_rate": 6.594736842105264e-05,
-      "loss": 1.7193,
+      "loss": 1.7116,
       "step": 1650
     },
     {
       "epoch": 11.45,
       "learning_rate": 6.542105263157895e-05,
-      "loss": 1.8665,
+      "loss": 1.8577,
       "step": 1660
     },
     {
       "epoch": 11.52,
       "learning_rate": 6.489473684210527e-05,
-      "loss": 2.2249,
+      "loss": 2.2062,
       "step": 1670
     },
     {
       "epoch": 11.59,
       "learning_rate": 6.436842105263158e-05,
-      "loss": 1.7308,
+      "loss": 1.7554,
       "step": 1680
     },
     {
       "epoch": 11.66,
       "learning_rate": 6.384210526315791e-05,
-      "loss": 1.6378,
+      "loss": 1.6673,
       "step": 1690
     },
     {
       "epoch": 11.72,
       "learning_rate": 6.331578947368422e-05,
-      "loss": 1.9278,
+      "loss": 1.9002,
       "step": 1700
     },
     {
       "epoch": 11.72,
-      "eval_loss": 1.874898076057434,
-      "eval_runtime": 62.4749,
-      "eval_samples_per_second": 26.891,
-      "eval_steps_per_second": 26.891,
-      "eval_wer": 0.8074564123768175,
+      "eval_loss": 1.8745884895324707,
+      "eval_runtime": 63.6985,
+      "eval_samples_per_second": 26.374,
+      "eval_steps_per_second": 26.374,
+      "eval_wer": 0.8332299634759838,
       "step": 1700
     },
     {
       "epoch": 11.79,
       "learning_rate": 6.278947368421053e-05,
-      "loss": 2.167,
+      "loss": 2.1405,
       "step": 1710
     },
     {
       "epoch": 11.86,
       "learning_rate": 6.226315789473685e-05,
-      "loss": 1.6498,
+      "loss": 1.6339,
       "step": 1720
     },
     {
       "epoch": 11.93,
       "learning_rate": 6.173684210526316e-05,
-      "loss": 1.7843,
+      "loss": 1.7832,
       "step": 1730
     },
     {
       "epoch": 12.0,
       "learning_rate": 6.121052631578947e-05,
-      "loss": 2.1545,
+      "loss": 2.1309,
       "step": 1740
     },
     {
       "epoch": 12.07,
       "learning_rate": 6.0684210526315785e-05,
-      "loss": 1.9012,
+      "loss": 1.895,
       "step": 1750
     },
     {
       "epoch": 12.14,
       "learning_rate": 6.015789473684211e-05,
-      "loss": 1.6388,
+      "loss": 1.6297,
       "step": 1760
     },
     {
       "epoch": 12.21,
       "learning_rate": 5.9631578947368425e-05,
-      "loss": 1.8509,
+      "loss": 1.8337,
       "step": 1770
     },
     {
       "epoch": 12.28,
       "learning_rate": 5.9105263157894744e-05,
-      "loss": 2.1512,
+      "loss": 2.1476,
       "step": 1780
     },
     {
       "epoch": 12.34,
       "learning_rate": 5.857894736842106e-05,
-      "loss": 1.6984,
+      "loss": 1.6987,
       "step": 1790
     },
     {
       "epoch": 12.41,
       "learning_rate": 5.805263157894737e-05,
-      "loss": 1.6324,
+      "loss": 1.6268,
       "step": 1800
     },
     {
       "epoch": 12.41,
-      "eval_loss": 1.850366234779358,
-      "eval_runtime": 62.1197,
-      "eval_samples_per_second": 27.045,
-      "eval_steps_per_second": 27.045,
-      "eval_wer": 0.8091103300944111,
+      "eval_loss": 1.878185510635376,
+      "eval_runtime": 63.513,
+      "eval_samples_per_second": 26.451,
+      "eval_steps_per_second": 26.451,
+      "eval_wer": 0.795052029494866,
       "step": 1800
     },
     {
       "epoch": 12.48,
       "learning_rate": 5.752631578947368e-05,
-      "loss": 2.0102,
+      "loss": 1.9851,
       "step": 1810
     },
     {
       "epoch": 12.55,
       "learning_rate": 5.6999999999999996e-05,
-      "loss": 2.0616,
+      "loss": 2.0705,
       "step": 1820
     },
     {
       "epoch": 12.62,
       "learning_rate": 5.647368421052632e-05,
-      "loss": 1.6063,
+      "loss": 1.609,
       "step": 1830
     },
     {
       "epoch": 12.69,
       "learning_rate": 5.5947368421052636e-05,
-      "loss": 1.718,
+      "loss": 1.7078,
       "step": 1840
     },
     {
       "epoch": 12.76,
       "learning_rate": 5.542105263157895e-05,
-      "loss": 2.0991,
+      "loss": 2.0835,
       "step": 1850
     },
     {
       "epoch": 12.83,
       "learning_rate": 5.489473684210527e-05,
-      "loss": 1.841,
+      "loss": 1.8212,
       "step": 1860
     },
     {
       "epoch": 12.9,
       "learning_rate": 5.436842105263158e-05,
-      "loss": 1.6345,
+      "loss": 1.6314,
       "step": 1870
     },
     {
       "epoch": 12.97,
       "learning_rate": 5.3842105263157895e-05,
-      "loss": 1.8546,
+      "loss": 1.8577,
       "step": 1880
     },
     {
       "epoch": 13.03,
       "learning_rate": 5.331578947368421e-05,
-      "loss": 2.1289,
+      "loss": 2.1196,
       "step": 1890
     },
     {
       "epoch": 13.1,
       "learning_rate": 5.2789473684210534e-05,
-      "loss": 1.6517,
+      "loss": 1.6455,
       "step": 1900
     },
     {
       "epoch": 13.1,
-      "eval_loss": 1.8227797746658325,
-      "eval_runtime": 62.3551,
-      "eval_samples_per_second": 26.942,
-      "eval_steps_per_second": 26.942,
-      "eval_wer": 0.8233064571704224,
+      "eval_loss": 1.8229954242706299,
+      "eval_runtime": 64.0654,
+      "eval_samples_per_second": 26.223,
+      "eval_steps_per_second": 26.223,
+      "eval_wer": 0.8224794983116257,
       "step": 1900
     },
     {
       "epoch": 13.17,
       "learning_rate": 5.226315789473685e-05,
-      "loss": 1.7168,
+      "loss": 1.7046,
       "step": 1910
     },
     {
       "epoch": 13.24,
       "learning_rate": 5.173684210526316e-05,
-      "loss": 1.924,
+      "loss": 1.9167,
       "step": 1920
     },
     {
       "epoch": 13.31,
       "learning_rate": 5.121052631578947e-05,
-      "loss": 1.9177,
+      "loss": 1.909,
       "step": 1930
     },
     {
       "epoch": 13.38,
       "learning_rate": 5.068421052631579e-05,
-      "loss": 1.64,
+      "loss": 1.6338,
       "step": 1940
     },
     {
       "epoch": 13.45,
       "learning_rate": 5.0157894736842106e-05,
-      "loss": 1.6953,
+      "loss": 1.7069,
       "step": 1950
     },
     {
       "epoch": 13.52,
       "learning_rate": 4.9631578947368426e-05,
-      "loss": 2.1591,
+      "loss": 2.1442,
       "step": 1960
     },
     {
       "epoch": 13.59,
       "learning_rate": 4.910526315789474e-05,
-      "loss": 1.7455,
+      "loss": 1.7517,
       "step": 1970
     },
     {
       "epoch": 13.66,
       "learning_rate": 4.857894736842106e-05,
-      "loss": 1.5742,
+      "loss": 1.5841,
       "step": 1980
     },
     {
       "epoch": 13.72,
       "learning_rate": 4.805263157894737e-05,
-      "loss": 1.9511,
+      "loss": 1.9493,
       "step": 1990
     },
     {
       "epoch": 13.79,
       "learning_rate": 4.7526315789473684e-05,
-      "loss": 2.0463,
+      "loss": 2.0308,
       "step": 2000
     },
     {
       "epoch": 13.79,
-      "eval_loss": 1.819372296333313,
-      "eval_runtime": 62.2592,
-      "eval_samples_per_second": 26.984,
-      "eval_steps_per_second": 26.984,
-      "eval_wer": 0.8563848115222934,
+      "eval_loss": 1.8066883087158203,
+      "eval_runtime": 63.421,
+      "eval_samples_per_second": 26.49,
+      "eval_steps_per_second": 26.49,
+      "eval_wer": 0.8560402453311281,
       "step": 2000
     },
     {
       "epoch": 13.86,
       "learning_rate": 4.7e-05,
-      "loss": 1.5744,
+      "loss": 1.5627,
       "step": 2010
     },
     {
@@ -1402,620 +1402,620 @@
     {
       "epoch": 14.0,
       "learning_rate": 4.594736842105264e-05,
-      "loss": 1.9124,
+      "loss": 1.9153,
       "step": 2030
     },
     {
       "epoch": 14.07,
       "learning_rate": 4.542105263157895e-05,
-      "loss": 1.8303,
+      "loss": 1.8294,
       "step": 2040
     },
     {
       "epoch": 14.14,
       "learning_rate": 4.489473684210527e-05,
-      "loss": 1.5034,
+      "loss": 1.513,
       "step": 2050
     },
     {
       "epoch": 14.21,
       "learning_rate": 4.436842105263158e-05,
-      "loss": 1.7662,
+      "loss": 1.7689,
       "step": 2060
     },
     {
       "epoch": 14.28,
       "learning_rate": 4.3842105263157895e-05,
-      "loss": 2.1269,
+      "loss": 2.1114,
       "step": 2070
     },
     {
       "epoch": 14.34,
       "learning_rate": 4.3315789473684215e-05,
-      "loss": 1.6707,
+      "loss": 1.6554,
       "step": 2080
     },
     {
       "epoch": 14.41,
       "learning_rate": 4.278947368421053e-05,
-      "loss": 1.6393,
+      "loss": 1.639,
       "step": 2090
     },
     {
       "epoch": 14.48,
       "learning_rate": 4.226315789473684e-05,
-      "loss": 1.8736,
+      "loss": 1.855,
       "step": 2100
     },
     {
       "epoch": 14.48,
-      "eval_loss": 1.7949970960617065,
-      "eval_runtime": 62.2709,
-      "eval_samples_per_second": 26.979,
-      "eval_steps_per_second": 26.979,
-      "eval_wer": 0.8224794983116257,
+      "eval_loss": 1.8128886222839355,
+      "eval_runtime": 63.7285,
+      "eval_samples_per_second": 26.362,
+      "eval_steps_per_second": 26.362,
+      "eval_wer": 0.8176555716353111,
       "step": 2100
     },
     {
       "epoch": 14.55,
       "learning_rate": 4.1736842105263154e-05,
-      "loss": 1.9396,
+      "loss": 1.9404,
       "step": 2110
     },
     {
       "epoch": 14.62,
       "learning_rate": 4.1210526315789474e-05,
-      "loss": 1.555,
+      "loss": 1.5515,
       "step": 2120
     },
     {
       "epoch": 14.69,
       "learning_rate": 4.0684210526315794e-05,
-      "loss": 1.7577,
+      "loss": 1.7568,
       "step": 2130
     },
     {
       "epoch": 14.76,
       "learning_rate": 4.015789473684211e-05,
-      "loss": 2.0354,
+      "loss": 2.0269,
       "step": 2140
     },
     {
       "epoch": 14.83,
       "learning_rate": 3.9631578947368426e-05,
-      "loss": 1.7073,
+      "loss": 1.7064,
       "step": 2150
     },
     {
       "epoch": 14.9,
       "learning_rate": 3.910526315789474e-05,
-      "loss": 1.4961,
+      "loss": 1.4851,
       "step": 2160
     },
     {
       "epoch": 14.97,
       "learning_rate": 3.857894736842105e-05,
-      "loss": 1.7068,
+      "loss": 1.69,
       "step": 2170
     },
     {
       "epoch": 15.03,
       "learning_rate": 3.8052631578947365e-05,
-      "loss": 2.0397,
+      "loss": 2.0147,
       "step": 2180
     },
     {
       "epoch": 15.1,
       "learning_rate": 3.7526315789473685e-05,
-      "loss": 1.5849,
+      "loss": 1.5575,
       "step": 2190
     },
     {
       "epoch": 15.17,
       "learning_rate": 3.7e-05,
-      "loss": 1.6233,
+      "loss": 1.5901,
       "step": 2200
     },
     {
       "epoch": 15.17,
-      "eval_loss": 1.7896100282669067,
-      "eval_runtime": 62.05,
-      "eval_samples_per_second": 27.075,
-      "eval_steps_per_second": 27.075,
-      "eval_wer": 0.8429467300668458,
+      "eval_loss": 1.7891418933868408,
+      "eval_runtime": 63.2021,
+      "eval_samples_per_second": 26.581,
+      "eval_steps_per_second": 26.581,
+      "eval_wer": 0.8367445386258701,
       "step": 2200
     },
     {
       "epoch": 15.24,
       "learning_rate": 3.647368421052632e-05,
-      "loss": 1.8936,
+      "loss": 1.8573,
       "step": 2210
     },
     {
       "epoch": 15.31,
       "learning_rate": 3.594736842105264e-05,
-      "loss": 1.8176,
+      "loss": 1.7881,
       "step": 2220
     },
     {
       "epoch": 15.38,
       "learning_rate": 3.542105263157895e-05,
-      "loss": 1.5963,
+      "loss": 1.5679,
       "step": 2230
     },
     {
       "epoch": 15.45,
       "learning_rate": 3.4894736842105264e-05,
-      "loss": 1.6365,
+      "loss": 1.6162,
       "step": 2240
     },
     {
       "epoch": 15.52,
       "learning_rate": 3.436842105263158e-05,
-      "loss": 2.0577,
+      "loss": 2.0334,
       "step": 2250
     },
     {
       "epoch": 15.59,
       "learning_rate": 3.3842105263157896e-05,
-      "loss": 1.5942,
+      "loss": 1.5751,
       "step": 2260
     },
     {
       "epoch": 15.66,
       "learning_rate": 3.331578947368421e-05,
-      "loss": 1.5867,
+      "loss": 1.5643,
       "step": 2270
     },
     {
       "epoch": 15.72,
       "learning_rate": 3.278947368421052e-05,
-      "loss": 1.8733,
+      "loss": 1.8435,
       "step": 2280
     },
     {
       "epoch": 15.79,
       "learning_rate": 3.226315789473684e-05,
-      "loss": 2.0624,
+      "loss": 2.0355,
       "step": 2290
     },
     {
       "epoch": 15.86,
       "learning_rate": 3.173684210526316e-05,
-      "loss": 1.4982,
+      "loss": 1.4848,
       "step": 2300
     },
     {
       "epoch": 15.86,
-      "eval_loss": 1.7793470621109009,
-      "eval_runtime": 62.0626,
-      "eval_samples_per_second": 27.069,
-      "eval_steps_per_second": 27.069,
-      "eval_wer": 0.8441182551168079,
+      "eval_loss": 1.7820757627487183,
+      "eval_runtime": 63.5201,
+      "eval_samples_per_second": 26.448,
+      "eval_steps_per_second": 26.448,
+      "eval_wer": 0.8201364482117015,
       "step": 2300
     },
     {
       "epoch": 15.93,
       "learning_rate": 3.1210526315789475e-05,
-      "loss": 1.6152,
+      "loss": 1.5905,
       "step": 2310
     },
     {
       "epoch": 16.0,
       "learning_rate": 3.0684210526315795e-05,
-      "loss": 1.9853,
+      "loss": 1.9714,
       "step": 2320
     },
     {
       "epoch": 16.07,
       "learning_rate": 3.0157894736842108e-05,
-      "loss": 1.8096,
+      "loss": 1.7952,
       "step": 2330
     },
     {
       "epoch": 16.14,
       "learning_rate": 2.963157894736842e-05,
-      "loss": 1.4931,
+      "loss": 1.4772,
       "step": 2340
     },
     {
       "epoch": 16.21,
       "learning_rate": 2.910526315789474e-05,
-      "loss": 1.6762,
+      "loss": 1.6484,
       "step": 2350
     },
     {
       "epoch": 16.28,
       "learning_rate": 2.8578947368421057e-05,
-      "loss": 2.041,
+      "loss": 2.0105,
       "step": 2360
     },
     {
       "epoch": 16.34,
       "learning_rate": 2.805263157894737e-05,
-      "loss": 1.5446,
+      "loss": 1.5344,
       "step": 2370
     },
     {
       "epoch": 16.41,
       "learning_rate": 2.7526315789473683e-05,
-      "loss": 1.5859,
+      "loss": 1.5725,
       "step": 2380
     },
     {
       "epoch": 16.48,
       "learning_rate": 2.7000000000000002e-05,
-      "loss": 1.804,
+      "loss": 1.7699,
       "step": 2390
     },
     {
       "epoch": 16.55,
       "learning_rate": 2.647368421052632e-05,
-      "loss": 1.8955,
+      "loss": 1.8754,
       "step": 2400
     },
     {
       "epoch": 16.55,
-      "eval_loss": 1.7640429735183716,
-      "eval_runtime": 62.2013,
-      "eval_samples_per_second": 27.009,
-      "eval_steps_per_second": 27.009,
-      "eval_wer": 0.805802494659224,
+      "eval_loss": 1.7700049877166748,
+      "eval_runtime": 63.9222,
+      "eval_samples_per_second": 26.282,
+      "eval_steps_per_second": 26.282,
+      "eval_wer": 0.8137275170560264,
       "step": 2400
     },
     {
       "epoch": 16.62,
       "learning_rate": 2.5947368421052632e-05,
-      "loss": 1.4801,
+      "loss": 1.4734,
       "step": 2410
     },
     {
       "epoch": 16.69,
       "learning_rate": 2.542105263157895e-05,
-      "loss": 1.6412,
+      "loss": 1.6307,
       "step": 2420
     },
     {
       "epoch": 16.76,
       "learning_rate": 2.4894736842105264e-05,
-      "loss": 1.9778,
+      "loss": 1.9546,
       "step": 2430
     },
     {
       "epoch": 16.83,
       "learning_rate": 2.436842105263158e-05,
-      "loss": 1.6648,
+      "loss": 1.6538,
       "step": 2440
     },
     {
       "epoch": 16.9,
       "learning_rate": 2.3842105263157897e-05,
-      "loss": 1.5288,
+      "loss": 1.5129,
       "step": 2450
     },
     {
       "epoch": 16.97,
       "learning_rate": 2.331578947368421e-05,
-      "loss": 1.6835,
+      "loss": 1.667,
       "step": 2460
     },
     {
       "epoch": 17.03,
       "learning_rate": 2.2789473684210527e-05,
-      "loss": 1.9791,
+      "loss": 1.966,
       "step": 2470
     },
     {
       "epoch": 17.1,
       "learning_rate": 2.2263157894736843e-05,
-      "loss": 1.478,
+      "loss": 1.4716,
       "step": 2480
     },
     {
       "epoch": 17.17,
       "learning_rate": 2.173684210526316e-05,
-      "loss": 1.5269,
+      "loss": 1.5158,
       "step": 2490
     },
     {
       "epoch": 17.24,
       "learning_rate": 2.1210526315789476e-05,
-      "loss": 1.8253,
+      "loss": 1.7975,
       "step": 2500
     },
     {
       "epoch": 17.24,
-      "eval_loss": 1.766750693321228,
-      "eval_runtime": 62.056,
-      "eval_samples_per_second": 27.072,
-      "eval_steps_per_second": 27.072,
-      "eval_wer": 0.8133829508648611,
+      "eval_loss": 1.779523253440857,
+      "eval_runtime": 64.1655,
+      "eval_samples_per_second": 26.182,
+      "eval_steps_per_second": 26.182,
+      "eval_wer": 0.8171042657294466,
       "step": 2500
     },
     {
       "epoch": 17.31,
       "learning_rate": 2.068421052631579e-05,
-      "loss": 1.8498,
+      "loss": 1.849,
       "step": 2510
     },
     {
       "epoch": 17.38,
       "learning_rate": 2.0157894736842105e-05,
-      "loss": 1.4217,
+      "loss": 1.4139,
       "step": 2520
     },
     {
       "epoch": 17.45,
       "learning_rate": 1.963157894736842e-05,
-      "loss": 1.6156,
+      "loss": 1.6001,
       "step": 2530
     },
     {
       "epoch": 17.52,
       "learning_rate": 1.9105263157894738e-05,
-      "loss": 1.9255,
+      "loss": 1.8943,
       "step": 2540
     },
     {
       "epoch": 17.59,
       "learning_rate": 1.8578947368421054e-05,
-      "loss": 1.6055,
+      "loss": 1.5934,
       "step": 2550
     },
     {
       "epoch": 17.66,
       "learning_rate": 1.8052631578947367e-05,
-      "loss": 1.5202,
+      "loss": 1.5103,
       "step": 2560
     },
     {
       "epoch": 17.72,
       "learning_rate": 1.7526315789473683e-05,
-      "loss": 1.7571,
+      "loss": 1.7289,
       "step": 2570
     },
     {
       "epoch": 17.79,
       "learning_rate": 1.7000000000000003e-05,
-      "loss": 1.893,
+      "loss": 1.8807,
       "step": 2580
     },
     {
       "epoch": 17.86,
       "learning_rate": 1.6473684210526316e-05,
-      "loss": 1.4591,
+      "loss": 1.4627,
       "step": 2590
     },
     {
       "epoch": 17.93,
       "learning_rate": 1.5947368421052633e-05,
-      "loss": 1.5332,
+      "loss": 1.5194,
       "step": 2600
     },
     {
       "epoch": 17.93,
-      "eval_loss": 1.7611440420150757,
-      "eval_runtime": 62.4578,
-      "eval_samples_per_second": 26.898,
-      "eval_steps_per_second": 26.898,
-      "eval_wer": 0.801460960650541,
+      "eval_loss": 1.7605171203613281,
+      "eval_runtime": 63.8146,
+      "eval_samples_per_second": 26.326,
+      "eval_steps_per_second": 26.326,
+      "eval_wer": 0.7976707325477225,
       "step": 2600
     },
     {
       "epoch": 18.0,
       "learning_rate": 1.5421052631578946e-05,
-      "loss": 1.9193,
+      "loss": 1.8941,
       "step": 2610
     },
     {
       "epoch": 18.07,
       "learning_rate": 1.4894736842105264e-05,
-      "loss": 1.7022,
+      "loss": 1.6843,
       "step": 2620
     },
     {
       "epoch": 18.14,
       "learning_rate": 1.4368421052631582e-05,
-      "loss": 1.4093,
+      "loss": 1.3952,
       "step": 2630
     },
     {
       "epoch": 18.21,
       "learning_rate": 1.3842105263157895e-05,
-      "loss": 1.6191,
+      "loss": 1.5958,
       "step": 2640
     },
     {
       "epoch": 18.28,
       "learning_rate": 1.3315789473684213e-05,
-      "loss": 1.9511,
+      "loss": 1.9271,
       "step": 2650
     },
     {
       "epoch": 18.34,
       "learning_rate": 1.2789473684210526e-05,
-      "loss": 1.487,
+      "loss": 1.4861,
       "step": 2660
     },
     {
       "epoch": 18.41,
       "learning_rate": 1.2263157894736844e-05,
-      "loss": 1.4701,
+      "loss": 1.47,
       "step": 2670
     },
     {
       "epoch": 18.48,
       "learning_rate": 1.1736842105263158e-05,
-      "loss": 1.7986,
+      "loss": 1.7653,
       "step": 2680
     },
     {
       "epoch": 18.55,
       "learning_rate": 1.1210526315789475e-05,
-      "loss": 1.7998,
+      "loss": 1.7886,
       "step": 2690
     },
     {
       "epoch": 18.62,
       "learning_rate": 1.068421052631579e-05,
-      "loss": 1.452,
+      "loss": 1.4374,
       "step": 2700
     },
     {
       "epoch": 18.62,
-      "eval_loss": 1.7497012615203857,
-      "eval_runtime": 62.4753,
-      "eval_samples_per_second": 26.891,
-      "eval_steps_per_second": 26.891,
-      "eval_wer": 0.8071807594238853,
+      "eval_loss": 1.7529252767562866,
+      "eval_runtime": 63.7187,
+      "eval_samples_per_second": 26.366,
+      "eval_steps_per_second": 26.366,
+      "eval_wer": 0.7978085590241886,
       "step": 2700
     },
     {
       "epoch": 18.69,
       "learning_rate": 1.0157894736842106e-05,
-      "loss": 1.5828,
+      "loss": 1.5556,
       "step": 2710
     },
     {
       "epoch": 18.76,
       "learning_rate": 9.631578947368422e-06,
-      "loss": 1.9231,
+      "loss": 1.8987,
       "step": 2720
     },
     {
       "epoch": 18.83,
       "learning_rate": 9.105263157894737e-06,
-      "loss": 1.5695,
+      "loss": 1.5625,
       "step": 2730
     },
     {
       "epoch": 18.9,
       "learning_rate": 8.578947368421053e-06,
-      "loss": 1.5435,
+      "loss": 1.5209,
       "step": 2740
     },
     {
       "epoch": 18.97,
       "learning_rate": 8.052631578947368e-06,
-      "loss": 1.6403,
+      "loss": 1.6116,
       "step": 2750
     },
     {
       "epoch": 19.03,
       "learning_rate": 7.526315789473684e-06,
-      "loss": 1.85,
+      "loss": 1.8313,
       "step": 2760
     },
     {
       "epoch": 19.1,
       "learning_rate": 7.000000000000001e-06,
-      "loss": 1.4523,
+      "loss": 1.4403,
       "step": 2770
     },
     {
       "epoch": 19.17,
       "learning_rate": 6.473684210526316e-06,
-      "loss": 1.488,
+      "loss": 1.4586,
       "step": 2780
     },
     {
       "epoch": 19.24,
       "learning_rate": 5.947368421052632e-06,
-      "loss": 1.8365,
+      "loss": 1.7928,
       "step": 2790
     },
     {
       "epoch": 19.31,
       "learning_rate": 5.421052631578947e-06,
-      "loss": 1.7609,
+      "loss": 1.7498,
       "step": 2800
     },
     {
       "epoch": 19.31,
-      "eval_loss": 1.7463784217834473,
-      "eval_runtime": 62.7099,
-      "eval_samples_per_second": 26.79,
-      "eval_steps_per_second": 26.79,
-      "eval_wer": 0.8062159740886224,
+      "eval_loss": 1.7522045373916626,
+      "eval_runtime": 63.0263,
+      "eval_samples_per_second": 26.656,
+      "eval_steps_per_second": 26.656,
+      "eval_wer": 0.8022879195093378,
       "step": 2800
     },
     {
       "epoch": 19.38,
       "learning_rate": 4.894736842105263e-06,
-      "loss": 1.4567,
+      "loss": 1.4375,
       "step": 2810
     },
     {
       "epoch": 19.45,
       "learning_rate": 4.368421052631579e-06,
-      "loss": 1.5592,
+      "loss": 1.5455,
       "step": 2820
     },
     {
       "epoch": 19.52,
       "learning_rate": 3.842105263157895e-06,
-      "loss": 1.961,
+      "loss": 1.9311,
       "step": 2830
     },
     {
       "epoch": 19.59,
-      "learning_rate": 3.368421052631579e-06,
-      "loss": 1.4964,
+      "learning_rate": 3.315789473684211e-06,
+      "loss": 1.4828,
       "step": 2840
     },
     {
       "epoch": 19.66,
-      "learning_rate": 2.842105263157895e-06,
-      "loss": 1.4505,
+      "learning_rate": 2.7894736842105266e-06,
+      "loss": 1.4371,
       "step": 2850
     },
     {
       "epoch": 19.72,
-      "learning_rate": 2.315789473684211e-06,
-      "loss": 1.7141,
+      "learning_rate": 2.2631578947368426e-06,
+      "loss": 1.687,
       "step": 2860
     },
     {
       "epoch": 19.79,
-      "learning_rate": 1.7894736842105262e-06,
-      "loss": 1.8701,
+      "learning_rate": 1.7368421052631579e-06,
+      "loss": 1.8411,
       "step": 2870
     },
     {
       "epoch": 19.86,
-      "learning_rate": 1.2631578947368422e-06,
-      "loss": 1.4216,
+      "learning_rate": 1.2105263157894738e-06,
+      "loss": 1.4025,
       "step": 2880
     },
     {
       "epoch": 19.93,
-      "learning_rate": 7.368421052631579e-07,
-      "loss": 1.5857,
+      "learning_rate": 6.842105263157896e-07,
+      "loss": 1.5635,
       "step": 2890
     },
     {
       "epoch": 20.0,
-      "learning_rate": 2.105263157894737e-07,
-      "loss": 1.7743,
+      "learning_rate": 1.5789473684210527e-07,
+      "loss": 1.7452,
       "step": 2900
     },
     {
       "epoch": 20.0,
-      "eval_loss": 1.742971658706665,
-      "eval_runtime": 62.4332,
-      "eval_samples_per_second": 26.909,
-      "eval_steps_per_second": 26.909,
-      "eval_wer": 0.808972503617945,
+      "eval_loss": 1.7481727600097656,
+      "eval_runtime": 63.0435,
+      "eval_samples_per_second": 26.648,
+      "eval_steps_per_second": 26.648,
+      "eval_wer": 0.7987044311212184,
       "step": 2900
     },
     {
       "epoch": 20.0,
       "step": 2900,
       "total_flos": 1.1463619274186412e+18,
-      "train_loss": 2.308333969445064,
-      "train_runtime": 4034.046,
-      "train_samples_per_second": 22.905,
-      "train_steps_per_second": 0.719
+      "train_loss": 2.305506002491918,
+      "train_runtime": 4265.6036,
+      "train_samples_per_second": 21.662,
+      "train_steps_per_second": 0.68
     }
   ],
   "max_steps": 2900,