End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.3688374757766724,
-    "eval_runtime": 49.4666,
+    "eval_loss": 1.3600618839263916,
+    "eval_runtime": 50.525,
     "eval_samples": 1680,
-    "eval_samples_per_second": 33.962,
-    "eval_steps_per_second": 33.962,
-    "eval_wer": 0.6817586658397078,
-    "train_loss": 0.938711966728342,
-    "train_runtime": 4007.4157,
+    "eval_samples_per_second": 33.251,
+    "eval_steps_per_second": 33.251,
+    "eval_wer": 0.6776238715457239,
+    "train_loss": 0.9487250096427984,
+    "train_runtime": 4020.0745,
     "train_samples": 4620,
-    "train_samples_per_second": 23.057,
-    "train_steps_per_second": 0.724
+    "train_samples_per_second": 22.985,
+    "train_steps_per_second": 0.721
 }

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
 2021-10-27T10:35:59,1041b9e5-8f0f-43d4-9424-cc7eff53732e,codecarbon,4004.539319753647,0.009920645469192451,0.04713434641766139,United States,USA,new york,N,,
+2021-10-28T00:30:31,912953b8-a980-4cab-8157-37666b2dbc9d,codecarbon,4017.2396454811096,0.00995231552467445,0.04728481519217719,United States,USA,new york,N,,

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.3688374757766724,
-    "eval_runtime": 49.4666,
+    "eval_loss": 1.3600618839263916,
+    "eval_runtime": 50.525,
     "eval_samples": 1680,
-    "eval_samples_per_second": 33.962,
-    "eval_steps_per_second": 33.962,
-    "eval_wer": 0.6817586658397078
+    "eval_samples_per_second": 33.251,
+    "eval_steps_per_second": 33.251,
+    "eval_wer": 0.6776238715457239
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9a64bd40598afe47c4bfb209c83bbb9f3ee40d64fd5fc5a0107161c71a11af0d
+oid sha256:c7441e7588c191aa050f6e17e77c7bb12b02d18099677a48d44f1a501f35d21a
 size 94084922

runs/Oct27_23-22-06_brutasse/events.out.tfevents.1635377008.brutasse.8534.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3bf58c7a1823434d50f8eabd0e7d5cdef53fd54ffd603c5a3cc418ed9e5c775
-size 56769
+oid sha256:43b9385ab7e8da796d15482d78b20a5b4bee103a2a0226a4a9d1968fc8b9f866
+size 59011

runs/Oct27_23-22-06_brutasse/events.out.tfevents.1635381083.brutasse.8534.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed901ec653638b6ed5f1d206f6d8f48ba63ac816478bd6ede38c4ba6212555cd
+size 358

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "train_loss": 0.938711966728342,
-    "train_runtime": 4007.4157,
+    "train_loss": 0.9487250096427984,
+    "train_runtime": 4020.0745,
     "train_samples": 4620,
-    "train_samples_per_second": 23.057,
-    "train_steps_per_second": 0.724
+    "train_samples_per_second": 22.985,
+    "train_steps_per_second": 0.721
 }

trainer_state.json

@@ -10,2017 +10,2017 @@
     {
       "epoch": 0.07,
       "learning_rate": 8.000000000000001e-07,
-      "loss": 9.653,
+      "loss": 9.5936,
       "step": 10
     },
     {
       "epoch": 0.14,
       "learning_rate": 1.8e-06,
-      "loss": 8.4639,
+      "loss": 8.4469,
       "step": 20
     },
     {
       "epoch": 0.21,
       "learning_rate": 2.8000000000000003e-06,
-      "loss": 7.9446,
+      "loss": 7.9584,
       "step": 30
     },
     {
       "epoch": 0.28,
       "learning_rate": 3.7e-06,
-      "loss": 8.6256,
+      "loss": 8.7061,
       "step": 40
     },
     {
       "epoch": 0.34,
       "learning_rate": 4.7e-06,
-      "loss": 8.0756,
+      "loss": 8.267,
       "step": 50
     },
     {
       "epoch": 0.41,
       "learning_rate": 5.7000000000000005e-06,
-      "loss": 7.4432,
+      "loss": 7.7959,
       "step": 60
     },
     {
       "epoch": 0.48,
       "learning_rate": 6.700000000000001e-06,
-      "loss": 6.5524,
+      "loss": 7.1516,
       "step": 70
     },
     {
       "epoch": 0.55,
       "learning_rate": 7.7e-06,
-      "loss": 6.5869,
+      "loss": 7.8574,
       "step": 80
     },
     {
       "epoch": 0.62,
       "learning_rate": 8.7e-06,
-      "loss": 5.2277,
+      "loss": 6.5821,
       "step": 90
     },
     {
       "epoch": 0.69,
       "learning_rate": 9.7e-06,
-      "loss": 4.2247,
+      "loss": 5.4447,
       "step": 100
     },
     {
       "epoch": 0.69,
-      "eval_loss": 3.8606975078582764,
-      "eval_runtime": 48.7411,
-      "eval_samples_per_second": 34.468,
-      "eval_steps_per_second": 34.468,
+      "eval_loss": 4.9545793533325195,
+      "eval_runtime": 50.1326,
+      "eval_samples_per_second": 33.511,
+      "eval_steps_per_second": 33.511,
       "eval_wer": 1.0,
       "step": 100
     },
     {
       "epoch": 0.76,
       "learning_rate": 1.0700000000000001e-05,
-      "loss": 3.6454,
+      "loss": 4.5955,
       "step": 110
     },
     {
       "epoch": 0.83,
       "learning_rate": 1.1700000000000001e-05,
-      "loss": 3.3306,
+      "loss": 3.8067,
       "step": 120
     },
     {
       "epoch": 0.9,
       "learning_rate": 1.27e-05,
-      "loss": 3.1763,
+      "loss": 3.3454,
       "step": 130
     },
     {
       "epoch": 0.97,
       "learning_rate": 1.3700000000000001e-05,
-      "loss": 3.0708,
+      "loss": 3.1785,
       "step": 140
     },
     {
       "epoch": 1.03,
       "learning_rate": 1.47e-05,
-      "loss": 3.0836,
+      "loss": 3.1751,
       "step": 150
     },
     {
       "epoch": 1.1,
       "learning_rate": 1.5700000000000002e-05,
-      "loss": 2.999,
+      "loss": 3.0387,
       "step": 160
     },
     {
       "epoch": 1.17,
       "learning_rate": 1.6700000000000003e-05,
-      "loss": 2.9717,
+      "loss": 2.9955,
       "step": 170
     },
     {
       "epoch": 1.24,
       "learning_rate": 1.77e-05,
-      "loss": 2.9846,
+      "loss": 2.9967,
       "step": 180
     },
     {
       "epoch": 1.31,
       "learning_rate": 1.87e-05,
-      "loss": 2.9912,
+      "loss": 2.992,
       "step": 190
     },
     {
       "epoch": 1.38,
       "learning_rate": 1.97e-05,
-      "loss": 2.9444,
+      "loss": 2.9499,
       "step": 200
     },
     {
       "epoch": 1.38,
-      "eval_loss": 2.950913667678833,
-      "eval_runtime": 49.1765,
-      "eval_samples_per_second": 34.163,
-      "eval_steps_per_second": 34.163,
+      "eval_loss": 2.951915979385376,
+      "eval_runtime": 50.1317,
+      "eval_samples_per_second": 33.512,
+      "eval_steps_per_second": 33.512,
       "eval_wer": 1.0,
       "step": 200
     },
     {
       "epoch": 1.45,
       "learning_rate": 2.07e-05,
-      "loss": 2.9386,
+      "loss": 2.9419,
       "step": 210
     },
     {
       "epoch": 1.52,
       "learning_rate": 2.1700000000000002e-05,
-      "loss": 2.9601,
+      "loss": 2.9594,
       "step": 220
     },
     {
       "epoch": 1.59,
       "learning_rate": 2.2700000000000003e-05,
-      "loss": 2.9395,
+      "loss": 2.942,
       "step": 230
     },
     {
       "epoch": 1.66,
       "learning_rate": 2.37e-05,
-      "loss": 2.9159,
+      "loss": 2.9188,
       "step": 240
     },
     {
       "epoch": 1.72,
       "learning_rate": 2.47e-05,
-      "loss": 2.9124,
+      "loss": 2.9152,
       "step": 250
     },
     {
       "epoch": 1.79,
       "learning_rate": 2.57e-05,
-      "loss": 2.9412,
+      "loss": 2.9417,
       "step": 260
     },
     {
       "epoch": 1.86,
       "learning_rate": 2.6700000000000002e-05,
-      "loss": 2.8826,
+      "loss": 2.8942,
       "step": 270
     },
     {
       "epoch": 1.93,
       "learning_rate": 2.7700000000000002e-05,
-      "loss": 2.8915,
+      "loss": 2.9023,
       "step": 280
     },
     {
       "epoch": 2.0,
       "learning_rate": 2.87e-05,
-      "loss": 2.9096,
+      "loss": 2.9168,
       "step": 290
     },
     {
       "epoch": 2.07,
       "learning_rate": 2.97e-05,
-      "loss": 2.8858,
+      "loss": 2.8989,
       "step": 300
     },
     {
       "epoch": 2.07,
-      "eval_loss": 2.844599723815918,
-      "eval_runtime": 49.0987,
-      "eval_samples_per_second": 34.217,
-      "eval_steps_per_second": 34.217,
+      "eval_loss": 2.862382650375366,
+      "eval_runtime": 50.3009,
+      "eval_samples_per_second": 33.399,
+      "eval_steps_per_second": 33.399,
       "eval_wer": 1.0,
       "step": 300
     },
     {
       "epoch": 2.14,
       "learning_rate": 3.07e-05,
-      "loss": 2.8025,
+      "loss": 2.8309,
       "step": 310
     },
     {
       "epoch": 2.21,
       "learning_rate": 3.1700000000000005e-05,
-      "loss": 2.7678,
+      "loss": 2.8001,
       "step": 320
     },
     {
       "epoch": 2.28,
       "learning_rate": 3.27e-05,
-      "loss": 2.7655,
+      "loss": 2.7949,
       "step": 330
     },
     {
       "epoch": 2.34,
       "learning_rate": 3.3700000000000006e-05,
-      "loss": 2.6538,
+      "loss": 2.6851,
       "step": 340
     },
     {
       "epoch": 2.41,
       "learning_rate": 3.4699999999999996e-05,
-      "loss": 2.5774,
+      "loss": 2.591,
       "step": 350
     },
     {
       "epoch": 2.48,
       "learning_rate": 3.57e-05,
-      "loss": 2.5132,
+      "loss": 2.5043,
       "step": 360
     },
     {
       "epoch": 2.55,
       "learning_rate": 3.6700000000000004e-05,
-      "loss": 2.4977,
+      "loss": 2.4854,
       "step": 370
     },
     {
       "epoch": 2.62,
       "learning_rate": 3.77e-05,
-      "loss": 2.3494,
+      "loss": 2.3164,
       "step": 380
     },
     {
       "epoch": 2.69,
       "learning_rate": 3.8700000000000006e-05,
-      "loss": 2.2912,
+      "loss": 2.245,
       "step": 390
     },
     {
       "epoch": 2.76,
       "learning_rate": 3.97e-05,
-      "loss": 2.2804,
+      "loss": 2.2076,
       "step": 400
     },
     {
       "epoch": 2.76,
-      "eval_loss": 2.1984939575195312,
-      "eval_runtime": 49.2048,
-      "eval_samples_per_second": 34.143,
-      "eval_steps_per_second": 34.143,
-      "eval_wer": 1.0013782647646612,
+      "eval_loss": 2.108926296234131,
+      "eval_runtime": 50.3626,
+      "eval_samples_per_second": 33.358,
+      "eval_steps_per_second": 33.358,
+      "eval_wer": 1.0008269588587968,
       "step": 400
     },
     {
       "epoch": 2.83,
       "learning_rate": 4.07e-05,
-      "loss": 2.1559,
+      "loss": 2.0686,
       "step": 410
     },
     {
       "epoch": 2.9,
       "learning_rate": 4.17e-05,
-      "loss": 2.013,
+      "loss": 1.9182,
       "step": 420
     },
     {
       "epoch": 2.97,
       "learning_rate": 4.27e-05,
-      "loss": 1.9527,
+      "loss": 1.8595,
       "step": 430
     },
     {
       "epoch": 3.03,
       "learning_rate": 4.3700000000000005e-05,
-      "loss": 1.9805,
+      "loss": 1.881,
       "step": 440
     },
     {
       "epoch": 3.1,
       "learning_rate": 4.47e-05,
-      "loss": 1.7592,
+      "loss": 1.6557,
       "step": 450
     },
     {
       "epoch": 3.17,
       "learning_rate": 4.5700000000000006e-05,
-      "loss": 1.6803,
+      "loss": 1.5894,
       "step": 460
     },
     {
       "epoch": 3.24,
       "learning_rate": 4.6700000000000003e-05,
-      "loss": 1.674,
+      "loss": 1.5914,
       "step": 470
     },
     {
       "epoch": 3.31,
       "learning_rate": 4.77e-05,
-      "loss": 1.6908,
+      "loss": 1.6008,
       "step": 480
     },
     {
       "epoch": 3.38,
       "learning_rate": 4.87e-05,
-      "loss": 1.4964,
+      "loss": 1.4103,
       "step": 490
     },
     {
       "epoch": 3.45,
       "learning_rate": 4.97e-05,
-      "loss": 1.505,
+      "loss": 1.4186,
       "step": 500
     },
     {
       "epoch": 3.45,
-      "eval_loss": 1.4971957206726074,
-      "eval_runtime": 49.3444,
-      "eval_samples_per_second": 34.046,
-      "eval_steps_per_second": 34.046,
-      "eval_wer": 0.9609261939218524,
+      "eval_loss": 1.4111816883087158,
+      "eval_runtime": 50.441,
+      "eval_samples_per_second": 33.306,
+      "eval_steps_per_second": 33.306,
+      "eval_wer": 0.9164771552615257,
       "step": 500
     },
     {
       "epoch": 3.52,
       "learning_rate": 5.0700000000000006e-05,
-      "loss": 1.5943,
+      "loss": 1.5086,
       "step": 510
     },
     {
       "epoch": 3.59,
       "learning_rate": 5.17e-05,
-      "loss": 1.3954,
+      "loss": 1.3112,
       "step": 520
     },
     {
       "epoch": 3.66,
       "learning_rate": 5.270000000000001e-05,
-      "loss": 1.3033,
+      "loss": 1.2336,
       "step": 530
     },
     {
       "epoch": 3.72,
       "learning_rate": 5.3700000000000004e-05,
-      "loss": 1.4023,
+      "loss": 1.3263,
       "step": 540
     },
     {
       "epoch": 3.79,
       "learning_rate": 5.470000000000001e-05,
-      "loss": 1.4554,
+      "loss": 1.3748,
       "step": 550
     },
     {
       "epoch": 3.86,
       "learning_rate": 5.5700000000000005e-05,
-      "loss": 1.2369,
+      "loss": 1.1614,
       "step": 560
     },
     {
       "epoch": 3.93,
       "learning_rate": 5.6699999999999996e-05,
-      "loss": 1.2277,
+      "loss": 1.1547,
       "step": 570
     },
     {
       "epoch": 4.0,
       "learning_rate": 5.77e-05,
-      "loss": 1.3168,
+      "loss": 1.2548,
       "step": 580
     },
     {
       "epoch": 4.07,
       "learning_rate": 5.87e-05,
-      "loss": 1.2854,
+      "loss": 1.2191,
       "step": 590
     },
     {
       "epoch": 4.14,
       "learning_rate": 5.97e-05,
-      "loss": 1.06,
+      "loss": 0.9951,
       "step": 600
     },
     {
       "epoch": 4.14,
-      "eval_loss": 1.2013553380966187,
-      "eval_runtime": 49.3577,
-      "eval_samples_per_second": 34.037,
-      "eval_steps_per_second": 34.037,
-      "eval_wer": 0.805802494659224,
+      "eval_loss": 1.1378357410430908,
+      "eval_runtime": 50.3227,
+      "eval_samples_per_second": 33.385,
+      "eval_steps_per_second": 33.385,
+      "eval_wer": 0.7701054372544965,
       "step": 600
     },
     {
       "epoch": 4.21,
       "learning_rate": 6.07e-05,
-      "loss": 1.1026,
+      "loss": 1.0463,
       "step": 610
     },
     {
       "epoch": 4.28,
       "learning_rate": 6.170000000000001e-05,
-      "loss": 1.2475,
+      "loss": 1.1918,
       "step": 620
     },
     {
       "epoch": 4.34,
       "learning_rate": 6.27e-05,
-      "loss": 1.0605,
+      "loss": 0.9968,
       "step": 630
     },
     {
       "epoch": 4.41,
       "learning_rate": 6.37e-05,
-      "loss": 1.0361,
+      "loss": 0.9801,
       "step": 640
     },
     {
       "epoch": 4.48,
       "learning_rate": 6.47e-05,
-      "loss": 1.1079,
+      "loss": 1.0596,
       "step": 650
     },
     {
       "epoch": 4.55,
       "learning_rate": 6.570000000000001e-05,
-      "loss": 1.2017,
+      "loss": 1.1484,
       "step": 660
     },
     {
       "epoch": 4.62,
       "learning_rate": 6.670000000000001e-05,
-      "loss": 0.9803,
+      "loss": 0.9331,
       "step": 670
     },
     {
       "epoch": 4.69,
       "learning_rate": 6.77e-05,
-      "loss": 1.0294,
+      "loss": 0.9861,
       "step": 680
     },
     {
       "epoch": 4.76,
       "learning_rate": 6.87e-05,
-      "loss": 1.1393,
+      "loss": 1.0923,
       "step": 690
     },
     {
       "epoch": 4.83,
       "learning_rate": 6.97e-05,
-      "loss": 1.0166,
+      "loss": 0.9754,
       "step": 700
     },
     {
       "epoch": 4.83,
-      "eval_loss": 1.0605226755142212,
-      "eval_runtime": 49.5346,
-      "eval_samples_per_second": 33.916,
-      "eval_steps_per_second": 33.916,
-      "eval_wer": 0.7535662600785611,
+      "eval_loss": 1.0151933431625366,
+      "eval_runtime": 50.6525,
+      "eval_samples_per_second": 33.167,
+      "eval_steps_per_second": 33.167,
+      "eval_wer": 0.7273792295499966,
       "step": 700
     },
     {
       "epoch": 4.9,
       "learning_rate": 7.07e-05,
-      "loss": 0.8619,
+      "loss": 0.8267,
       "step": 710
     },
     {
       "epoch": 4.97,
       "learning_rate": 7.17e-05,
-      "loss": 0.9549,
+      "loss": 0.9163,
       "step": 720
     },
     {
       "epoch": 5.03,
       "learning_rate": 7.27e-05,
-      "loss": 1.1162,
+      "loss": 1.0682,
       "step": 730
     },
     {
       "epoch": 5.1,
       "learning_rate": 7.37e-05,
-      "loss": 0.8634,
+      "loss": 0.8203,
       "step": 740
     },
     {
       "epoch": 5.17,
       "learning_rate": 7.47e-05,
-      "loss": 0.8374,
+      "loss": 0.8,
       "step": 750
     },
     {
       "epoch": 5.24,
       "learning_rate": 7.570000000000001e-05,
-      "loss": 0.9316,
+      "loss": 0.8987,
       "step": 760
     },
     {
       "epoch": 5.31,
       "learning_rate": 7.670000000000001e-05,
-      "loss": 0.9827,
+      "loss": 0.9465,
       "step": 770
     },
     {
       "epoch": 5.38,
       "learning_rate": 7.77e-05,
-      "loss": 0.7866,
+      "loss": 0.7583,
       "step": 780
     },
     {
       "epoch": 5.45,
       "learning_rate": 7.87e-05,
-      "loss": 0.816,
+      "loss": 0.7862,
       "step": 790
     },
     {
       "epoch": 5.52,
       "learning_rate": 7.970000000000001e-05,
-      "loss": 0.966,
+      "loss": 0.9364,
       "step": 800
     },
     {
       "epoch": 5.52,
-      "eval_loss": 0.9963059425354004,
-      "eval_runtime": 49.4475,
-      "eval_samples_per_second": 33.975,
-      "eval_steps_per_second": 33.975,
-      "eval_wer": 0.7100820067534973,
+      "eval_loss": 0.9619123339653015,
+      "eval_runtime": 50.5375,
+      "eval_samples_per_second": 33.243,
+      "eval_steps_per_second": 33.243,
+      "eval_wer": 0.7011232857831989,
       "step": 800
     },
     {
       "epoch": 5.59,
       "learning_rate": 8.070000000000001e-05,
-      "loss": 0.8691,
+      "loss": 0.8322,
       "step": 810
     },
     {
       "epoch": 5.66,
       "learning_rate": 8.17e-05,
-      "loss": 0.748,
+      "loss": 0.724,
       "step": 820
     },
     {
       "epoch": 5.72,
       "learning_rate": 8.27e-05,
-      "loss": 0.8441,
+      "loss": 0.8198,
       "step": 830
     },
     {
       "epoch": 5.79,
       "learning_rate": 8.37e-05,
-      "loss": 1.0217,
+      "loss": 0.9825,
       "step": 840
     },
     {
       "epoch": 5.86,
       "learning_rate": 8.47e-05,
-      "loss": 0.7328,
+      "loss": 0.7021,
       "step": 850
     },
     {
       "epoch": 5.93,
       "learning_rate": 8.57e-05,
-      "loss": 0.7625,
+      "loss": 0.7367,
       "step": 860
     },
     {
       "epoch": 6.0,
       "learning_rate": 8.67e-05,
-      "loss": 0.883,
+      "loss": 0.8597,
       "step": 870
     },
     {
       "epoch": 6.07,
       "learning_rate": 8.77e-05,
-      "loss": 0.8376,
+      "loss": 0.791,
       "step": 880
     },
     {
       "epoch": 6.14,
       "learning_rate": 8.87e-05,
-      "loss": 0.6518,
+      "loss": 0.6308,
       "step": 890
     },
     {
       "epoch": 6.21,
       "learning_rate": 8.970000000000001e-05,
-      "loss": 0.6857,
+      "loss": 0.6557,
       "step": 900
     },
     {
       "epoch": 6.21,
-      "eval_loss": 0.9442995190620422,
-      "eval_runtime": 49.4918,
-      "eval_samples_per_second": 33.945,
-      "eval_steps_per_second": 33.945,
-      "eval_wer": 0.6898215147129764,
+      "eval_loss": 0.9144111275672913,
+      "eval_runtime": 50.5004,
+      "eval_samples_per_second": 33.267,
+      "eval_steps_per_second": 33.267,
+      "eval_wer": 0.6867893322307215,
       "step": 900
     },
     {
       "epoch": 6.28,
       "learning_rate": 9.070000000000001e-05,
-      "loss": 0.8764,
+      "loss": 0.8438,
       "step": 910
     },
     {
       "epoch": 6.34,
       "learning_rate": 9.17e-05,
-      "loss": 0.7431,
+      "loss": 0.7106,
       "step": 920
     },
     {
       "epoch": 6.41,
       "learning_rate": 9.27e-05,
-      "loss": 0.6688,
+      "loss": 0.6506,
       "step": 930
     },
     {
       "epoch": 6.48,
       "learning_rate": 9.370000000000001e-05,
-      "loss": 0.7162,
+      "loss": 0.7062,
       "step": 940
     },
     {
       "epoch": 6.55,
       "learning_rate": 9.47e-05,
-      "loss": 0.8465,
+      "loss": 0.8188,
       "step": 950
     },
     {
       "epoch": 6.62,
       "learning_rate": 9.57e-05,
-      "loss": 0.6191,
+      "loss": 0.6014,
       "step": 960
     },
     {
       "epoch": 6.69,
       "learning_rate": 9.67e-05,
-      "loss": 0.6149,
+      "loss": 0.5934,
       "step": 970
     },
     {
       "epoch": 6.76,
       "learning_rate": 9.77e-05,
-      "loss": 0.8114,
+      "loss": 0.7971,
       "step": 980
     },
     {
       "epoch": 6.83,
       "learning_rate": 9.87e-05,
-      "loss": 0.7832,
+      "loss": 0.7547,
       "step": 990
     },
     {
       "epoch": 6.9,
       "learning_rate": 9.970000000000001e-05,
-      "loss": 0.5859,
+      "loss": 0.5681,
       "step": 1000
     },
     {
       "epoch": 6.9,
-      "eval_loss": 0.9042919278144836,
-      "eval_runtime": 49.5927,
-      "eval_samples_per_second": 33.876,
-      "eval_steps_per_second": 33.876,
-      "eval_wer": 0.6795534422162497,
+      "eval_loss": 0.889886736869812,
+      "eval_runtime": 50.3615,
+      "eval_samples_per_second": 33.359,
+      "eval_steps_per_second": 33.359,
+      "eval_wer": 0.6682516711460271,
       "step": 1000
     },
     {
       "epoch": 6.97,
       "learning_rate": 9.963157894736843e-05,
-      "loss": 0.6366,
+      "loss": 0.6196,
       "step": 1010
     },
     {
       "epoch": 7.03,
       "learning_rate": 9.910526315789475e-05,
-      "loss": 0.8343,
+      "loss": 0.819,
       "step": 1020
     },
     {
       "epoch": 7.1,
       "learning_rate": 9.857894736842106e-05,
-      "loss": 0.5977,
+      "loss": 0.5811,
       "step": 1030
     },
     {
       "epoch": 7.17,
       "learning_rate": 9.805263157894737e-05,
-      "loss": 0.5143,
+      "loss": 0.5121,
       "step": 1040
     },
     {
       "epoch": 7.24,
       "learning_rate": 9.752631578947369e-05,
-      "loss": 0.5725,
+      "loss": 0.5588,
       "step": 1050
     },
     {
       "epoch": 7.31,
       "learning_rate": 9.7e-05,
-      "loss": 0.7447,
+      "loss": 0.7227,
       "step": 1060
     },
     {
       "epoch": 7.38,
       "learning_rate": 9.647368421052631e-05,
-      "loss": 0.4958,
+      "loss": 0.4808,
       "step": 1070
     },
     {
       "epoch": 7.45,
       "learning_rate": 9.594736842105264e-05,
-      "loss": 0.5129,
+      "loss": 0.5031,
       "step": 1080
     },
     {
       "epoch": 7.52,
       "learning_rate": 9.542105263157895e-05,
-      "loss": 0.6971,
+      "loss": 0.6767,
       "step": 1090
     },
     {
       "epoch": 7.59,
       "learning_rate": 9.489473684210527e-05,
-      "loss": 0.6812,
+      "loss": 0.66,
       "step": 1100
     },
     {
       "epoch": 7.59,
-      "eval_loss": 0.9094838500022888,
-      "eval_runtime": 49.4365,
-      "eval_samples_per_second": 33.983,
-      "eval_steps_per_second": 33.983,
-      "eval_wer": 0.6715595065812142,
+      "eval_loss": 0.8991804122924805,
+      "eval_runtime": 50.3017,
+      "eval_samples_per_second": 33.399,
+      "eval_steps_per_second": 33.399,
+      "eval_wer": 0.6653573151402384,
       "step": 1100
     },
     {
       "epoch": 7.66,
       "learning_rate": 9.436842105263158e-05,
-      "loss": 0.4968,
+      "loss": 0.487,
       "step": 1110
     },
     {
       "epoch": 7.72,
       "learning_rate": 9.384210526315789e-05,
-      "loss": 0.5241,
+      "loss": 0.5132,
       "step": 1120
     },
     {
       "epoch": 7.79,
       "learning_rate": 9.331578947368422e-05,
-      "loss": 0.7348,
+      "loss": 0.704,
       "step": 1130
     },
     {
       "epoch": 7.86,
       "learning_rate": 9.278947368421053e-05,
-      "loss": 0.5148,
+      "loss": 0.5007,
       "step": 1140
     },
     {
       "epoch": 7.93,
       "learning_rate": 9.226315789473686e-05,
-      "loss": 0.518,
+      "loss": 0.513,
       "step": 1150
     },
     {
       "epoch": 8.0,
       "learning_rate": 9.173684210526317e-05,
-      "loss": 0.6009,
+      "loss": 0.5883,
       "step": 1160
     },
     {
       "epoch": 8.07,
       "learning_rate": 9.121052631578948e-05,
-      "loss": 0.6677,
+      "loss": 0.6605,
       "step": 1170
     },
     {
       "epoch": 8.14,
       "learning_rate": 9.06842105263158e-05,
-      "loss": 0.4171,
+      "loss": 0.4155,
       "step": 1180
     },
     {
       "epoch": 8.21,
       "learning_rate": 9.015789473684211e-05,
-      "loss": 0.4129,
+      "loss": 0.414,
       "step": 1190
     },
     {
       "epoch": 8.28,
       "learning_rate": 8.963157894736842e-05,
-      "loss": 0.6088,
+      "loss": 0.6144,
       "step": 1200
     },
     {
       "epoch": 8.28,
-      "eval_loss": 0.9421910047531128,
-      "eval_runtime": 50.1077,
-      "eval_samples_per_second": 33.528,
-      "eval_steps_per_second": 33.528,
-      "eval_wer": 0.6677003652401626,
+      "eval_loss": 0.9299214482307434,
+      "eval_runtime": 50.3304,
+      "eval_samples_per_second": 33.379,
+      "eval_steps_per_second": 33.379,
+      "eval_wer": 0.6898215147129764,
       "step": 1200
     },
     {
       "epoch": 8.34,
       "learning_rate": 8.910526315789474e-05,
-      "loss": 0.4936,
+      "loss": 0.4794,
       "step": 1210
     },
     {
       "epoch": 8.41,
       "learning_rate": 8.857894736842106e-05,
-      "loss": 0.4166,
+      "loss": 0.4113,
       "step": 1220
     },
     {
       "epoch": 8.48,
       "learning_rate": 8.805263157894737e-05,
-      "loss": 0.4619,
+      "loss": 0.4479,
       "step": 1230
     },
     {
       "epoch": 8.55,
       "learning_rate": 8.752631578947369e-05,
-      "loss": 0.6438,
+      "loss": 0.6263,
       "step": 1240
     },
     {
       "epoch": 8.62,
       "learning_rate": 8.7e-05,
-      "loss": 0.4375,
+      "loss": 0.4205,
       "step": 1250
     },
     {
       "epoch": 8.69,
       "learning_rate": 8.647368421052631e-05,
-      "loss": 0.3933,
+      "loss": 0.4036,
       "step": 1260
     },
     {
       "epoch": 8.76,
       "learning_rate": 8.594736842105263e-05,
-      "loss": 0.5451,
+      "loss": 0.5458,
       "step": 1270
     },
     {
       "epoch": 8.83,
       "learning_rate": 8.542105263157894e-05,
-      "loss": 0.5696,
+      "loss": 0.5463,
       "step": 1280
     },
     {
       "epoch": 8.9,
       "learning_rate": 8.489473684210527e-05,
-      "loss": 0.3988,
+      "loss": 0.3955,
       "step": 1290
     },
     {
       "epoch": 8.97,
       "learning_rate": 8.436842105263158e-05,
-      "loss": 0.4162,
+      "loss": 0.4099,
       "step": 1300
     },
     {
       "epoch": 8.97,
-      "eval_loss": 0.9547951817512512,
-      "eval_runtime": 49.7551,
-      "eval_samples_per_second": 33.765,
-      "eval_steps_per_second": 33.765,
-      "eval_wer": 0.6657018813314037,
+      "eval_loss": 0.9509925246238708,
+      "eval_runtime": 50.5682,
+      "eval_samples_per_second": 33.222,
+      "eval_steps_per_second": 33.222,
+      "eval_wer": 0.6673557990489973,
       "step": 1300
     },
     {
       "epoch": 9.03,
       "learning_rate": 8.38421052631579e-05,
-      "loss": 0.581,
+      "loss": 0.5702,
       "step": 1310
     },
     {
       "epoch": 9.1,
       "learning_rate": 8.331578947368422e-05,
-      "loss": 0.3721,
+      "loss": 0.3673,
       "step": 1320
     },
     {
       "epoch": 9.17,
       "learning_rate": 8.278947368421053e-05,
-      "loss": 0.3251,
+      "loss": 0.314,
       "step": 1330
     },
     {
       "epoch": 9.24,
       "learning_rate": 8.226315789473684e-05,
-      "loss": 0.3486,
+      "loss": 0.341,
       "step": 1340
     },
     {
       "epoch": 9.31,
       "learning_rate": 8.173684210526317e-05,
-      "loss": 0.5894,
+      "loss": 0.5948,
       "step": 1350
     },
     {
       "epoch": 9.38,
       "learning_rate": 8.121052631578948e-05,
-      "loss": 0.3616,
+      "loss": 0.3596,
       "step": 1360
     },
     {
       "epoch": 9.45,
       "learning_rate": 8.06842105263158e-05,
-      "loss": 0.3189,
+      "loss": 0.3284,
       "step": 1370
     },
     {
       "epoch": 9.52,
       "learning_rate": 8.015789473684211e-05,
-      "loss": 0.4903,
+      "loss": 0.4988,
       "step": 1380
     },
     {
       "epoch": 9.59,
       "learning_rate": 7.963157894736842e-05,
-      "loss": 0.4564,
+      "loss": 0.4438,
       "step": 1390
     },
     {
       "epoch": 9.66,
       "learning_rate": 7.910526315789474e-05,
-      "loss": 0.3411,
+      "loss": 0.3384,
       "step": 1400
     },
     {
       "epoch": 9.66,
-      "eval_loss": 0.9900805950164795,
-      "eval_runtime": 49.6853,
-      "eval_samples_per_second": 33.813,
-      "eval_steps_per_second": 33.813,
-      "eval_wer": 0.6688718902901247,
+      "eval_loss": 0.9598137736320496,
+      "eval_runtime": 50.3681,
+      "eval_samples_per_second": 33.354,
+      "eval_steps_per_second": 33.354,
+      "eval_wer": 0.6611536076080216,
       "step": 1400
     },
     {
       "epoch": 9.72,
       "learning_rate": 7.857894736842105e-05,
-      "loss": 0.3415,
+      "loss": 0.3422,
       "step": 1410
     },
     {
       "epoch": 9.79,
       "learning_rate": 7.805263157894738e-05,
-      "loss": 0.5735,
+      "loss": 0.5592,
       "step": 1420
     },
     {
       "epoch": 9.86,
       "learning_rate": 7.752631578947369e-05,
-      "loss": 0.3583,
+      "loss": 0.3536,
       "step": 1430
     },
     {
       "epoch": 9.93,
       "learning_rate": 7.7e-05,
-      "loss": 0.2923,
+      "loss": 0.3111,
       "step": 1440
     },
     {
       "epoch": 10.0,
       "learning_rate": 7.647368421052631e-05,
-      "loss": 0.3396,
+      "loss": 0.3469,
       "step": 1450
     },
     {
       "epoch": 10.07,
       "learning_rate": 7.594736842105263e-05,
-      "loss": 0.493,
+      "loss": 0.4812,
       "step": 1460
     },
     {
       "epoch": 10.14,
       "learning_rate": 7.542105263157895e-05,
-      "loss": 0.2637,
+      "loss": 0.2573,
       "step": 1470
     },
     {
       "epoch": 10.21,
       "learning_rate": 7.489473684210527e-05,
-      "loss": 0.2342,
+      "loss": 0.2372,
       "step": 1480
     },
     {
       "epoch": 10.28,
       "learning_rate": 7.43684210526316e-05,
-      "loss": 0.4601,
+      "loss": 0.4399,
       "step": 1490
     },
     {
       "epoch": 10.34,
       "learning_rate": 7.38421052631579e-05,
-      "loss": 0.3323,
+      "loss": 0.3163,
       "step": 1500
     },
     {
       "epoch": 10.34,
-      "eval_loss": 0.9995749592781067,
-      "eval_runtime": 49.7001,
-      "eval_samples_per_second": 33.803,
-      "eval_steps_per_second": 33.803,
-      "eval_wer": 0.6637723106608779,
+      "eval_loss": 0.9953927993774414,
+      "eval_runtime": 50.4333,
+      "eval_samples_per_second": 33.311,
+      "eval_steps_per_second": 33.311,
+      "eval_wer": 0.6612225208462545,
       "step": 1500
     },
     {
       "epoch": 10.41,
       "learning_rate": 7.331578947368422e-05,
-      "loss": 0.2467,
+      "loss": 0.2451,
       "step": 1510
     },
     {
       "epoch": 10.48,
       "learning_rate": 7.278947368421053e-05,
-      "loss": 0.2583,
+      "loss": 0.2571,
       "step": 1520
     },
     {
       "epoch": 10.55,
       "learning_rate": 7.226315789473685e-05,
-      "loss": 0.5144,
+      "loss": 0.5127,
       "step": 1530
     },
     {
       "epoch": 10.62,
       "learning_rate": 7.173684210526316e-05,
-      "loss": 0.2957,
+      "loss": 0.2847,
       "step": 1540
     },
     {
       "epoch": 10.69,
       "learning_rate": 7.121052631578947e-05,
-      "loss": 0.2613,
+      "loss": 0.2619,
       "step": 1550
     },
     {
       "epoch": 10.76,
       "learning_rate": 7.06842105263158e-05,
-      "loss": 0.3751,
+      "loss": 0.3631,
       "step": 1560
     },
     {
       "epoch": 10.83,
       "learning_rate": 7.015789473684211e-05,
-      "loss": 0.4468,
+      "loss": 0.4317,
       "step": 1570
     },
     {
       "epoch": 10.9,
       "learning_rate": 6.963157894736842e-05,
-      "loss": 0.2497,
+      "loss": 0.2507,
       "step": 1580
     },
     {
       "epoch": 10.97,
       "learning_rate": 6.910526315789474e-05,
-      "loss": 0.2478,
+      "loss": 0.2485,
       "step": 1590
     },
     {
       "epoch": 11.03,
       "learning_rate": 6.857894736842105e-05,
-      "loss": 0.431,
+      "loss": 0.4204,
       "step": 1600
     },
     {
       "epoch": 11.03,
-      "eval_loss": 1.0520888566970825,
-      "eval_runtime": 49.5966,
-      "eval_samples_per_second": 33.873,
-      "eval_steps_per_second": 33.873,
-      "eval_wer": 0.6708014609606505,
+      "eval_loss": 1.0164344310760498,
+      "eval_runtime": 50.3517,
+      "eval_samples_per_second": 33.365,
+      "eval_steps_per_second": 33.365,
+      "eval_wer": 0.66067121494039,
       "step": 1600
     },
     {
       "epoch": 11.1,
       "learning_rate": 6.805263157894736e-05,
-      "loss": 0.2673,
+      "loss": 0.265,
       "step": 1610
     },
     {
       "epoch": 11.17,
       "learning_rate": 6.752631578947368e-05,
-      "loss": 0.1873,
+      "loss": 0.1877,
       "step": 1620
     },
     {
       "epoch": 11.24,
       "learning_rate": 6.7e-05,
-      "loss": 0.1937,
+      "loss": 0.1922,
       "step": 1630
     },
     {
       "epoch": 11.31,
       "learning_rate": 6.647368421052632e-05,
-      "loss": 0.48,
+      "loss": 0.4621,
       "step": 1640
     },
     {
       "epoch": 11.38,
       "learning_rate": 6.594736842105264e-05,
-      "loss": 0.2313,
+      "loss": 0.2252,
       "step": 1650
     },
     {
       "epoch": 11.45,
       "learning_rate": 6.542105263157895e-05,
-      "loss": 0.1876,
+      "loss": 0.1885,
       "step": 1660
     },
     {
       "epoch": 11.52,
       "learning_rate": 6.489473684210527e-05,
-      "loss": 0.3463,
+      "loss": 0.3207,
       "step": 1670
     },
     {
       "epoch": 11.59,
       "learning_rate": 6.436842105263158e-05,
-      "loss": 0.3318,
+      "loss": 0.3149,
       "step": 1680
     },
     {
       "epoch": 11.66,
       "learning_rate": 6.384210526315791e-05,
-      "loss": 0.1915,
+      "loss": 0.1912,
       "step": 1690
     },
     {
       "epoch": 11.72,
       "learning_rate": 6.331578947368422e-05,
-      "loss": 0.2029,
+      "loss": 0.1932,
       "step": 1700
     },
     {
       "epoch": 11.72,
-      "eval_loss": 1.0945649147033691,
-      "eval_runtime": 49.6143,
-      "eval_samples_per_second": 33.861,
-      "eval_steps_per_second": 33.861,
-      "eval_wer": 0.6792777892633175,
+      "eval_loss": 1.0637414455413818,
+      "eval_runtime": 50.3761,
+      "eval_samples_per_second": 33.349,
+      "eval_steps_per_second": 33.349,
+      "eval_wer": 0.6657707945696368,
       "step": 1700
     },
     {
       "epoch": 11.79,
       "learning_rate": 6.278947368421053e-05,
-      "loss": 0.4609,
+      "loss": 0.4378,
       "step": 1710
     },
     {
       "epoch": 11.86,
       "learning_rate": 6.226315789473685e-05,
-      "loss": 0.2301,
+      "loss": 0.2184,
       "step": 1720
     },
     {
       "epoch": 11.93,
       "learning_rate": 6.173684210526316e-05,
-      "loss": 0.2117,
+      "loss": 0.2063,
       "step": 1730
     },
     {
       "epoch": 12.0,
       "learning_rate": 6.121052631578947e-05,
-      "loss": 0.2205,
+      "loss": 0.2094,
       "step": 1740
     },
     {
       "epoch": 12.07,
       "learning_rate": 6.0684210526315785e-05,
-      "loss": 0.396,
+      "loss": 0.3799,
       "step": 1750
     },
     {
       "epoch": 12.14,
       "learning_rate": 6.015789473684211e-05,
-      "loss": 0.1706,
+      "loss": 0.1585,
       "step": 1760
     },
     {
       "epoch": 12.21,
       "learning_rate": 5.9631578947368425e-05,
-      "loss": 0.1486,
+      "loss": 0.1441,
       "step": 1770
     },
     {
       "epoch": 12.28,
       "learning_rate": 5.9105263157894744e-05,
-      "loss": 0.3378,
+      "loss": 0.3064,
       "step": 1780
     },
     {
       "epoch": 12.34,
       "learning_rate": 5.857894736842106e-05,
-      "loss": 0.2552,
+      "loss": 0.2439,
       "step": 1790
     },
     {
       "epoch": 12.41,
       "learning_rate": 5.805263157894737e-05,
-      "loss": 0.1424,
+      "loss": 0.1449,
       "step": 1800
     },
     {
       "epoch": 12.41,
-      "eval_loss": 1.1287890672683716,
-      "eval_runtime": 49.6023,
-      "eval_samples_per_second": 33.869,
-      "eval_steps_per_second": 33.869,
-      "eval_wer": 0.6712149403900489,
+      "eval_loss": 1.1190370321273804,
+      "eval_runtime": 50.5074,
+      "eval_samples_per_second": 33.262,
+      "eval_steps_per_second": 33.262,
+      "eval_wer": 0.6652194886637723,
       "step": 1800
     },
     {
       "epoch": 12.48,
       "learning_rate": 5.752631578947368e-05,
-      "loss": 0.1614,
+      "loss": 0.1513,
       "step": 1810
     },
     {
       "epoch": 12.55,
       "learning_rate": 5.6999999999999996e-05,
-      "loss": 0.3994,
+      "loss": 0.3656,
       "step": 1820
     },
     {
       "epoch": 12.62,
       "learning_rate": 5.647368421052632e-05,
-      "loss": 0.1802,
+      "loss": 0.171,
       "step": 1830
     },
     {
       "epoch": 12.69,
       "learning_rate": 5.5947368421052636e-05,
-      "loss": 0.1493,
+      "loss": 0.1463,
       "step": 1840
     },
     {
       "epoch": 12.76,
       "learning_rate": 5.542105263157895e-05,
-      "loss": 0.2477,
+      "loss": 0.2383,
       "step": 1850
     },
     {
       "epoch": 12.83,
       "learning_rate": 5.489473684210527e-05,
-      "loss": 0.3343,
+      "loss": 0.3031,
       "step": 1860
     },
     {
       "epoch": 12.9,
       "learning_rate": 5.436842105263158e-05,
-      "loss": 0.1604,
+      "loss": 0.1513,
       "step": 1870
     },
     {
       "epoch": 12.97,
       "learning_rate": 5.3842105263157895e-05,
-      "loss": 0.1511,
+      "loss": 0.1484,
       "step": 1880
     },
     {
       "epoch": 13.03,
       "learning_rate": 5.331578947368421e-05,
-      "loss": 0.3277,
+      "loss": 0.3051,
       "step": 1890
     },
     {
       "epoch": 13.1,
       "learning_rate": 5.2789473684210534e-05,
-      "loss": 0.1922,
+      "loss": 0.1803,
       "step": 1900
     },
     {
       "epoch": 13.1,
-      "eval_loss": 1.1456407308578491,
-      "eval_runtime": 49.7045,
-      "eval_samples_per_second": 33.8,
-      "eval_steps_per_second": 33.8,
-      "eval_wer": 0.6739714699193715,
+      "eval_loss": 1.1260483264923096,
+      "eval_runtime": 50.4412,
+      "eval_samples_per_second": 33.306,
+      "eval_steps_per_second": 33.306,
+      "eval_wer": 0.6689408035283578,
       "step": 1900
     },
     {
       "epoch": 13.17,
       "learning_rate": 5.226315789473685e-05,
-      "loss": 0.1204,
+      "loss": 0.1158,
       "step": 1910
     },
     {
       "epoch": 13.24,
       "learning_rate": 5.173684210526316e-05,
-      "loss": 0.1144,
+      "loss": 0.1038,
       "step": 1920
     },
     {
       "epoch": 13.31,
       "learning_rate": 5.121052631578947e-05,
-      "loss": 0.373,
+      "loss": 0.3329,
       "step": 1930
     },
     {
       "epoch": 13.38,
       "learning_rate": 5.068421052631579e-05,
-      "loss": 0.1471,
+      "loss": 0.1411,
       "step": 1940
     },
     {
       "epoch": 13.45,
       "learning_rate": 5.0157894736842106e-05,
-      "loss": 0.1108,
+      "loss": 0.1082,
       "step": 1950
     },
     {
       "epoch": 13.52,
       "learning_rate": 4.9631578947368426e-05,
-      "loss": 0.2611,
+      "loss": 0.2466,
       "step": 1960
     },
     {
       "epoch": 13.59,
       "learning_rate": 4.910526315789474e-05,
-      "loss": 0.2552,
+      "loss": 0.2415,
       "step": 1970
     },
     {
       "epoch": 13.66,
       "learning_rate": 4.857894736842106e-05,
-      "loss": 0.1166,
+      "loss": 0.1175,
       "step": 1980
     },
     {
       "epoch": 13.72,
       "learning_rate": 4.805263157894737e-05,
-      "loss": 0.124,
+      "loss": 0.121,
       "step": 1990
     },
     {
       "epoch": 13.79,
       "learning_rate": 4.7526315789473684e-05,
-      "loss": 0.326,
+      "loss": 0.328,
       "step": 2000
     },
     {
       "epoch": 13.79,
-      "eval_loss": 1.2077220678329468,
-      "eval_runtime": 49.5752,
-      "eval_samples_per_second": 33.888,
-      "eval_steps_per_second": 33.888,
-      "eval_wer": 0.6914754324305699,
+      "eval_loss": 1.218603253364563,
+      "eval_runtime": 50.4012,
+      "eval_samples_per_second": 33.333,
+      "eval_steps_per_second": 33.333,
+      "eval_wer": 0.6751429949693336,
       "step": 2000
     },
     {
       "epoch": 13.86,
       "learning_rate": 4.7e-05,
-      "loss": 0.1467,
+      "loss": 0.1445,
       "step": 2010
     },
     {
       "epoch": 13.93,
       "learning_rate": 4.647368421052632e-05,
-      "loss": 0.1127,
+      "loss": 0.1145,
       "step": 2020
     },
     {
       "epoch": 14.0,
       "learning_rate": 4.594736842105264e-05,
-      "loss": 0.1298,
+      "loss": 0.122,
       "step": 2030
     },
     {
       "epoch": 14.07,
       "learning_rate": 4.542105263157895e-05,
-      "loss": 0.3033,
+      "loss": 0.2931,
       "step": 2040
     },
     {
       "epoch": 14.14,
       "learning_rate": 4.489473684210527e-05,
-      "loss": 0.0986,
+      "loss": 0.0965,
       "step": 2050
     },
     {
       "epoch": 14.21,
       "learning_rate": 4.436842105263158e-05,
-      "loss": 0.0835,
+      "loss": 0.084,
       "step": 2060
     },
     {
       "epoch": 14.28,
       "learning_rate": 4.3842105263157895e-05,
-      "loss": 0.2537,
+      "loss": 0.2362,
       "step": 2070
     },
     {
       "epoch": 14.34,
       "learning_rate": 4.3315789473684215e-05,
-      "loss": 0.1715,
+      "loss": 0.16,
       "step": 2080
     },
     {
       "epoch": 14.41,
       "learning_rate": 4.278947368421053e-05,
-      "loss": 0.1022,
+      "loss": 0.1034,
       "step": 2090
     },
     {
       "epoch": 14.48,
       "learning_rate": 4.226315789473684e-05,
-      "loss": 0.0892,
+      "loss": 0.0838,
       "step": 2100
     },
     {
       "epoch": 14.48,
-      "eval_loss": 1.2525088787078857,
-      "eval_runtime": 49.6787,
-      "eval_samples_per_second": 33.817,
-      "eval_steps_per_second": 33.817,
-      "eval_wer": 0.6795534422162497,
+      "eval_loss": 1.2591233253479004,
+      "eval_runtime": 50.4613,
+      "eval_samples_per_second": 33.293,
+      "eval_steps_per_second": 33.293,
+      "eval_wer": 0.6908552132864724,
       "step": 2100
     },
     {
       "epoch": 14.55,
       "learning_rate": 4.1736842105263154e-05,
-      "loss": 0.2858,
+      "loss": 0.2733,
       "step": 2110
     },
     {
       "epoch": 14.62,
       "learning_rate": 4.1210526315789474e-05,
-      "loss": 0.1159,
+      "loss": 0.1126,
       "step": 2120
     },
     {
       "epoch": 14.69,
       "learning_rate": 4.0684210526315794e-05,
-      "loss": 0.0943,
+      "loss": 0.0986,
       "step": 2130
     },
     {
       "epoch": 14.76,
       "learning_rate": 4.015789473684211e-05,
-      "loss": 0.1748,
+      "loss": 0.1728,
       "step": 2140
     },
     {
       "epoch": 14.83,
       "learning_rate": 3.9631578947368426e-05,
-      "loss": 0.2513,
+      "loss": 0.2317,
       "step": 2150
     },
     {
       "epoch": 14.9,
       "learning_rate": 3.910526315789474e-05,
-      "loss": 0.0915,
+      "loss": 0.0932,
       "step": 2160
     },
     {
       "epoch": 14.97,
       "learning_rate": 3.857894736842105e-05,
-      "loss": 0.0812,
+      "loss": 0.0831,
       "step": 2170
     },
     {
       "epoch": 15.03,
       "learning_rate": 3.8052631578947365e-05,
-      "loss": 0.2623,
+      "loss": 0.2388,
       "step": 2180
     },
     {
       "epoch": 15.1,
       "learning_rate": 3.7526315789473685e-05,
-      "loss": 0.1312,
+      "loss": 0.129,
       "step": 2190
     },
     {
       "epoch": 15.17,
       "learning_rate": 3.7e-05,
-      "loss": 0.0769,
+      "loss": 0.0766,
       "step": 2200
     },
     {
       "epoch": 15.17,
-      "eval_loss": 1.2313077449798584,
-      "eval_runtime": 49.7254,
-      "eval_samples_per_second": 33.786,
-      "eval_steps_per_second": 33.786,
-      "eval_wer": 0.6736269037282062,
+      "eval_loss": 1.252910852432251,
+      "eval_runtime": 50.3936,
+      "eval_samples_per_second": 33.338,
+      "eval_steps_per_second": 33.338,
+      "eval_wer": 0.6779684377368893,
       "step": 2200
     },
     {
       "epoch": 15.24,
       "learning_rate": 3.647368421052632e-05,
-      "loss": 0.0657,
+      "loss": 0.0677,
       "step": 2210
     },
     {
       "epoch": 15.31,
       "learning_rate": 3.594736842105264e-05,
-      "loss": 0.2479,
+      "loss": 0.2474,
       "step": 2220
     },
     {
       "epoch": 15.38,
       "learning_rate": 3.542105263157895e-05,
-      "loss": 0.1003,
+      "loss": 0.0998,
       "step": 2230
     },
     {
       "epoch": 15.45,
       "learning_rate": 3.4894736842105264e-05,
-      "loss": 0.0655,
+      "loss": 0.0651,
       "step": 2240
     },
     {
       "epoch": 15.52,
       "learning_rate": 3.436842105263158e-05,
-      "loss": 0.1793,
+      "loss": 0.1772,
       "step": 2250
     },
     {
       "epoch": 15.59,
       "learning_rate": 3.3842105263157896e-05,
-      "loss": 0.162,
+      "loss": 0.1637,
       "step": 2260
     },
     {
       "epoch": 15.66,
       "learning_rate": 3.331578947368421e-05,
-      "loss": 0.0824,
+      "loss": 0.0886,
       "step": 2270
     },
     {
       "epoch": 15.72,
       "learning_rate": 3.278947368421052e-05,
-      "loss": 0.0805,
+      "loss": 0.0785,
       "step": 2280
     },
     {
       "epoch": 15.79,
       "learning_rate": 3.226315789473684e-05,
-      "loss": 0.2688,
+      "loss": 0.2612,
       "step": 2290
     },
     {
       "epoch": 15.86,
       "learning_rate": 3.173684210526316e-05,
-      "loss": 0.0927,
+      "loss": 0.0956,
       "step": 2300
     },
     {
       "epoch": 15.86,
-      "eval_loss": 1.300140619277954,
-      "eval_runtime": 49.6316,
-      "eval_samples_per_second": 33.849,
-      "eval_steps_per_second": 33.849,
-      "eval_wer": 0.6863758528013232,
+      "eval_loss": 1.253722906112671,
+      "eval_runtime": 50.2864,
+      "eval_samples_per_second": 33.409,
+      "eval_steps_per_second": 33.409,
+      "eval_wer": 0.6668044931431328,
       "step": 2300
     },
     {
       "epoch": 15.93,
       "learning_rate": 3.1210526315789475e-05,
-      "loss": 0.0762,
+      "loss": 0.0747,
       "step": 2310
     },
     {
       "epoch": 16.0,
       "learning_rate": 3.0684210526315795e-05,
-      "loss": 0.0758,
+      "loss": 0.0816,
       "step": 2320
     },
     {
       "epoch": 16.07,
       "learning_rate": 3.0157894736842108e-05,
-      "loss": 0.2356,
+      "loss": 0.2428,
       "step": 2330
     },
     {
       "epoch": 16.14,
       "learning_rate": 2.963157894736842e-05,
-      "loss": 0.0667,
+      "loss": 0.0704,
       "step": 2340
     },
     {
       "epoch": 16.21,
       "learning_rate": 2.910526315789474e-05,
-      "loss": 0.0607,
+      "loss": 0.0637,
       "step": 2350
     },
     {
       "epoch": 16.28,
       "learning_rate": 2.8578947368421057e-05,
-      "loss": 0.1887,
+      "loss": 0.2066,
       "step": 2360
     },
     {
       "epoch": 16.34,
       "learning_rate": 2.805263157894737e-05,
-      "loss": 0.1209,
+      "loss": 0.1234,
       "step": 2370
     },
     {
       "epoch": 16.41,
       "learning_rate": 2.7526315789473683e-05,
-      "loss": 0.0688,
+      "loss": 0.0793,
       "step": 2380
     },
     {
       "epoch": 16.48,
       "learning_rate": 2.7000000000000002e-05,
-      "loss": 0.0572,
+      "loss": 0.0579,
       "step": 2390
     },
     {
       "epoch": 16.55,
       "learning_rate": 2.647368421052632e-05,
-      "loss": 0.232,
+      "loss": 0.2339,
       "step": 2400
     },
     {
       "epoch": 16.55,
-      "eval_loss": 1.349038004875183,
-      "eval_runtime": 49.7195,
-      "eval_samples_per_second": 33.79,
-      "eval_steps_per_second": 33.79,
-      "eval_wer": 0.6962993591068845,
+      "eval_loss": 1.3210315704345703,
+      "eval_runtime": 50.3899,
+      "eval_samples_per_second": 33.34,
+      "eval_steps_per_second": 33.34,
+      "eval_wer": 0.6796912686927159,
       "step": 2400
     },
     {
       "epoch": 16.62,
       "learning_rate": 2.5947368421052632e-05,
-      "loss": 0.0775,
+      "loss": 0.0771,
       "step": 2410
     },
     {
       "epoch": 16.69,
       "learning_rate": 2.542105263157895e-05,
-      "loss": 0.0604,
+      "loss": 0.0636,
       "step": 2420
     },
     {
       "epoch": 16.76,
       "learning_rate": 2.4894736842105264e-05,
-      "loss": 0.1308,
+      "loss": 0.1349,
       "step": 2430
     },
     {
       "epoch": 16.83,
       "learning_rate": 2.436842105263158e-05,
-      "loss": 0.1677,
+      "loss": 0.1704,
       "step": 2440
     },
     {
       "epoch": 16.9,
       "learning_rate": 2.3842105263157897e-05,
-      "loss": 0.0647,
+      "loss": 0.0735,
       "step": 2450
     },
     {
       "epoch": 16.97,
       "learning_rate": 2.331578947368421e-05,
-      "loss": 0.0554,
+      "loss": 0.0561,
       "step": 2460
     },
     {
       "epoch": 17.03,
       "learning_rate": 2.2789473684210527e-05,
-      "loss": 0.1842,
+      "loss": 0.1833,
       "step": 2470
     },
     {
       "epoch": 17.1,
       "learning_rate": 2.2263157894736843e-05,
-      "loss": 0.0917,
+      "loss": 0.0899,
       "step": 2480
     },
     {
       "epoch": 17.17,
       "learning_rate": 2.173684210526316e-05,
-      "loss": 0.0469,
+      "loss": 0.0514,
       "step": 2490
     },
     {
       "epoch": 17.24,
       "learning_rate": 2.1210526315789476e-05,
-      "loss": 0.0485,
+      "loss": 0.0431,
       "step": 2500
     },
     {
       "epoch": 17.24,
-      "eval_loss": 1.3268318176269531,
-      "eval_runtime": 49.6454,
-      "eval_samples_per_second": 33.84,
-      "eval_steps_per_second": 33.84,
-      "eval_wer": 0.6763145200192957,
+      "eval_loss": 1.3241108655929565,
+      "eval_runtime": 50.2912,
+      "eval_samples_per_second": 33.405,
+      "eval_steps_per_second": 33.405,
+      "eval_wer": 0.6781062642133554,
       "step": 2500
     },
     {
       "epoch": 17.31,
       "learning_rate": 2.068421052631579e-05,
-      "loss": 0.2096,
+      "loss": 0.2135,
       "step": 2510
     },
     {
       "epoch": 17.38,
       "learning_rate": 2.0157894736842105e-05,
-      "loss": 0.0561,
+      "loss": 0.0591,
       "step": 2520
     },
     {
       "epoch": 17.45,
       "learning_rate": 1.963157894736842e-05,
-      "loss": 0.0476,
+      "loss": 0.053,
       "step": 2530
     },
     {
       "epoch": 17.52,
       "learning_rate": 1.9105263157894738e-05,
-      "loss": 0.1363,
+      "loss": 0.1288,
       "step": 2540
     },
     {
       "epoch": 17.59,
       "learning_rate": 1.8578947368421054e-05,
-      "loss": 0.1129,
+      "loss": 0.1119,
       "step": 2550
     },
     {
       "epoch": 17.66,
       "learning_rate": 1.8052631578947367e-05,
-      "loss": 0.0508,
+      "loss": 0.0529,
       "step": 2560
     },
     {
       "epoch": 17.72,
       "learning_rate": 1.7526315789473683e-05,
-      "loss": 0.0426,
+      "loss": 0.0454,
       "step": 2570
     },
     {
       "epoch": 17.79,
       "learning_rate": 1.7000000000000003e-05,
-      "loss": 0.1895,
+      "loss": 0.1928,
       "step": 2580
     },
     {
       "epoch": 17.86,
       "learning_rate": 1.6473684210526316e-05,
-      "loss": 0.0756,
+      "loss": 0.074,
       "step": 2590
     },
     {
       "epoch": 17.93,
       "learning_rate": 1.5947368421052633e-05,
-      "loss": 0.0487,
+      "loss": 0.0508,
       "step": 2600
     },
     {
       "epoch": 17.93,
-      "eval_loss": 1.3375909328460693,
-      "eval_runtime": 49.815,
-      "eval_samples_per_second": 33.725,
-      "eval_steps_per_second": 33.725,
-      "eval_wer": 0.6779684377368893,
+      "eval_loss": 1.3183706998825073,
+      "eval_runtime": 50.4573,
+      "eval_samples_per_second": 33.295,
+      "eval_steps_per_second": 33.295,
+      "eval_wer": 0.6683205843842602,
       "step": 2600
     },
     {
       "epoch": 18.0,
       "learning_rate": 1.5421052631578946e-05,
-      "loss": 0.0497,
+      "loss": 0.053,
       "step": 2610
     },
     {
       "epoch": 18.07,
       "learning_rate": 1.4894736842105264e-05,
-      "loss": 0.1801,
+      "loss": 0.1823,
       "step": 2620
     },
     {
       "epoch": 18.14,
       "learning_rate": 1.4368421052631582e-05,
-      "loss": 0.047,
+      "loss": 0.0447,
       "step": 2630
     },
     {
       "epoch": 18.21,
       "learning_rate": 1.3842105263157895e-05,
-      "loss": 0.044,
+      "loss": 0.0422,
       "step": 2640
     },
     {
       "epoch": 18.28,
       "learning_rate": 1.3315789473684213e-05,
-      "loss": 0.1409,
+      "loss": 0.1376,
       "step": 2650
     },
     {
       "epoch": 18.34,
       "learning_rate": 1.2789473684210526e-05,
-      "loss": 0.0887,
+      "loss": 0.0875,
       "step": 2660
     },
     {
       "epoch": 18.41,
       "learning_rate": 1.2263157894736844e-05,
-      "loss": 0.0456,
+      "loss": 0.0485,
       "step": 2670
     },
     {
       "epoch": 18.48,
       "learning_rate": 1.1736842105263158e-05,
-      "loss": 0.0403,
+      "loss": 0.0435,
       "step": 2680
     },
     {
       "epoch": 18.55,
       "learning_rate": 1.1210526315789475e-05,
-      "loss": 0.161,
+      "loss": 0.1602,
       "step": 2690
     },
     {
       "epoch": 18.62,
       "learning_rate": 1.068421052631579e-05,
-      "loss": 0.0607,
+      "loss": 0.0616,
       "step": 2700
     },
     {
       "epoch": 18.62,
-      "eval_loss": 1.3700547218322754,
-      "eval_runtime": 49.648,
-      "eval_samples_per_second": 33.838,
-      "eval_steps_per_second": 33.838,
-      "eval_wer": 0.6894769485218111,
+      "eval_loss": 1.372841477394104,
+      "eval_runtime": 50.3934,
+      "eval_samples_per_second": 33.338,
+      "eval_steps_per_second": 33.338,
+      "eval_wer": 0.6888567293777135,
       "step": 2700
     },
     {
       "epoch": 18.69,
       "learning_rate": 1.0157894736842106e-05,
-      "loss": 0.0388,
+      "loss": 0.0399,
       "step": 2710
     },
     {
       "epoch": 18.76,
       "learning_rate": 9.631578947368422e-06,
-      "loss": 0.1065,
+      "loss": 0.1019,
       "step": 2720
     },
     {
       "epoch": 18.83,
       "learning_rate": 9.105263157894737e-06,
-      "loss": 0.1294,
+      "loss": 0.1264,
       "step": 2730
     },
     {
       "epoch": 18.9,
       "learning_rate": 8.578947368421053e-06,
-      "loss": 0.0507,
+      "loss": 0.0499,
       "step": 2740
     },
     {
       "epoch": 18.97,
       "learning_rate": 8.052631578947368e-06,
-      "loss": 0.0382,
+      "loss": 0.0399,
       "step": 2750
     },
     {
       "epoch": 19.03,
       "learning_rate": 7.526315789473684e-06,
-      "loss": 0.1584,
+      "loss": 0.1536,
       "step": 2760
     },
     {
       "epoch": 19.1,
       "learning_rate": 7.000000000000001e-06,
-      "loss": 0.0715,
+      "loss": 0.0706,
       "step": 2770
     },
     {
       "epoch": 19.17,
       "learning_rate": 6.473684210526316e-06,
-      "loss": 0.0443,
+      "loss": 0.0439,
       "step": 2780
     },
     {
       "epoch": 19.24,
       "learning_rate": 5.947368421052632e-06,
-      "loss": 0.041,
+      "loss": 0.0418,
       "step": 2790
     },
     {
       "epoch": 19.31,
       "learning_rate": 5.421052631578947e-06,
-      "loss": 0.1618,
+      "loss": 0.1608,
       "step": 2800
     },
     {
       "epoch": 19.31,
-      "eval_loss": 1.3657355308532715,
-      "eval_runtime": 49.611,
-      "eval_samples_per_second": 33.863,
-      "eval_steps_per_second": 33.863,
-      "eval_wer": 0.6795534422162497,
+      "eval_loss": 1.3572485446929932,
+      "eval_runtime": 50.6618,
+      "eval_samples_per_second": 33.161,
+      "eval_steps_per_second": 33.161,
+      "eval_wer": 0.6770725656398594,
       "step": 2800
     },
     {
       "epoch": 19.38,
       "learning_rate": 4.894736842105263e-06,
-      "loss": 0.0475,
+      "loss": 0.0482,
       "step": 2810
     },
     {
       "epoch": 19.45,
       "learning_rate": 4.368421052631579e-06,
-      "loss": 0.0417,
+      "loss": 0.0388,
       "step": 2820
     },
     {
       "epoch": 19.52,
       "learning_rate": 3.842105263157895e-06,
-      "loss": 0.0996,
+      "loss": 0.0956,
       "step": 2830
     },
     {
       "epoch": 19.59,
       "learning_rate": 3.315789473684211e-06,
-      "loss": 0.0923,
+      "loss": 0.087,
       "step": 2840
     },
     {
       "epoch": 19.66,
       "learning_rate": 2.7894736842105266e-06,
-      "loss": 0.0363,
+      "loss": 0.0349,
       "step": 2850
     },
     {
       "epoch": 19.72,
       "learning_rate": 2.2631578947368426e-06,
-      "loss": 0.0363,
+      "loss": 0.0349,
       "step": 2860
     },
     {
       "epoch": 19.79,
       "learning_rate": 1.7368421052631579e-06,
-      "loss": 0.1545,
+      "loss": 0.1533,
       "step": 2870
     },
     {
       "epoch": 19.86,
       "learning_rate": 1.2105263157894738e-06,
-      "loss": 0.0575,
+      "loss": 0.0562,
       "step": 2880
     },
     {
       "epoch": 19.93,
       "learning_rate": 6.842105263157896e-07,
-      "loss": 0.0372,
+      "loss": 0.0375,
       "step": 2890
     },
     {
       "epoch": 20.0,
       "learning_rate": 1.5789473684210527e-07,
-      "loss": 0.0415,
+      "loss": 0.0378,
       "step": 2900
     },
     {
       "epoch": 20.0,
-      "eval_loss": 1.3688374757766724,
-      "eval_runtime": 49.8972,
-      "eval_samples_per_second": 33.669,
-      "eval_steps_per_second": 33.669,
-      "eval_wer": 0.6817586658397078,
+      "eval_loss": 1.3600618839263916,
+      "eval_runtime": 50.6265,
+      "eval_samples_per_second": 33.184,
+      "eval_steps_per_second": 33.184,
+      "eval_wer": 0.6776238715457239,
       "step": 2900
     },
     {
       "epoch": 20.0,
       "step": 2900,
-      "total_flos": 6.619750382049396e+17,
-      "train_loss": 0.938711966728342,
-      "train_runtime": 4007.4157,
-      "train_samples_per_second": 23.057,
-      "train_steps_per_second": 0.724
+      "total_flos": 6.619462136392861e+17,
+      "train_loss": 0.9487250096427984,
+      "train_runtime": 4020.0745,
+      "train_samples_per_second": 22.985,
+      "train_steps_per_second": 0.721
     }
   ],
   "max_steps": 2900,
   "num_train_epochs": 20,
-  "total_flos": 6.619750382049396e+17,
+  "total_flos": 6.619462136392861e+17,
   "trial_name": null,
   "trial_params": null
 }