End of training

README.md

@@ -3,11 +3,24 @@ license: apache-2.0
 base_model: openai/whisper-large-v3
 tags:
 - generated_from_trainer
+datasets:
+- DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
 metrics:
 - wer
 model-index:
 - name: whisper-large-v3-ft-btb-ccv-cy
-  results: []
+  results:
+  - task:
+      name: Automatic Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv default
+      type: DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
+      args: default
+    metrics:
+    - name: Wer
+      type: wer
+      value: 0.29472481531011024
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -15,7 +28,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # whisper-large-v3-ft-btb-ccv-cy
 
-This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on an unknown dataset.
+This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv default dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.4479
 - Wer: 0.2947

all_results.json

@@ -1,15 +1,15 @@
 {
     "epoch": 3.861003861003861,
-    "eval_loss": 0.4226454198360443,
-    "eval_runtime": 2917.5486,
+    "eval_loss": 0.447899729013443,
+    "eval_runtime": 2897.5042,
     "eval_samples": 7022,
-    "eval_samples_per_second": 2.407,
-    "eval_steps_per_second": 0.15,
-    "eval_wer": 0.2671921259024568,
+    "eval_samples_per_second": 2.423,
+    "eval_steps_per_second": 0.152,
+    "eval_wer": 0.29472481531011024,
     "total_flos": 5.435487665750016e+20,
-    "train_loss": 0.2543565913915634,
-    "train_runtime": 60445.9813,
+    "train_loss": 0.2695653033494949,
+    "train_runtime": 64824.1008,
     "train_samples": 41435,
-    "train_samples_per_second": 2.647,
-    "train_steps_per_second": 0.083
+    "train_samples_per_second": 2.468,
+    "train_steps_per_second": 0.077
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.861003861003861,
-    "eval_loss": 0.4226454198360443,
-    "eval_runtime": 2917.5486,
+    "eval_loss": 0.447899729013443,
+    "eval_runtime": 2897.5042,
     "eval_samples": 7022,
-    "eval_samples_per_second": 2.407,
-    "eval_steps_per_second": 0.15,
-    "eval_wer": 0.2671921259024568
+    "eval_samples_per_second": 2.423,
+    "eval_steps_per_second": 0.152,
+    "eval_wer": 0.29472481531011024
 }

runs/Aug21_09-58-49_1030eff7f038/events.out.tfevents.1724299633.1030eff7f038.141.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c94fce2d390f3934b0a7b6039f1f6d97844a67da465809b27a8b8d2a5562b713
+size 406

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.861003861003861,
     "total_flos": 5.435487665750016e+20,
-    "train_loss": 0.2543565913915634,
-    "train_runtime": 60445.9813,
+    "train_loss": 0.2695653033494949,
+    "train_runtime": 64824.1008,
     "train_samples": 41435,
-    "train_samples_per_second": 2.647,
-    "train_steps_per_second": 0.083
+    "train_samples_per_second": 2.468,
+    "train_steps_per_second": 0.077
 }

trainer_state.json

@@ -10,1457 +10,1457 @@
   "log_history": [
     {
       "epoch": 0.019305019305019305,
-      "grad_norm": 7.541379928588867,
+      "grad_norm": 7.365023136138916,
       "learning_rate": 5.000000000000001e-07,
-      "loss": 1.5011,
+      "loss": 1.4936,
       "step": 25
     },
     {
       "epoch": 0.03861003861003861,
-      "grad_norm": 5.699951648712158,
+      "grad_norm": 5.869426727294922,
       "learning_rate": 1.0000000000000002e-06,
-      "loss": 1.1952,
+      "loss": 1.2163,
       "step": 50
     },
     {
       "epoch": 0.05791505791505792,
-      "grad_norm": 4.949509143829346,
+      "grad_norm": 5.041717052459717,
       "learning_rate": 1.5e-06,
-      "loss": 0.8322,
+      "loss": 0.866,
       "step": 75
     },
     {
       "epoch": 0.07722007722007722,
-      "grad_norm": 4.566027641296387,
+      "grad_norm": 4.53203010559082,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 0.7302,
+      "loss": 0.7562,
       "step": 100
     },
     {
       "epoch": 0.09652509652509653,
-      "grad_norm": 5.739590167999268,
+      "grad_norm": 5.75096321105957,
       "learning_rate": 2.5e-06,
-      "loss": 0.658,
+      "loss": 0.6816,
       "step": 125
     },
     {
       "epoch": 0.11583011583011583,
-      "grad_norm": 5.6211347579956055,
+      "grad_norm": 5.387002944946289,
       "learning_rate": 3e-06,
-      "loss": 0.6784,
+      "loss": 0.7012,
       "step": 150
     },
     {
       "epoch": 0.13513513513513514,
-      "grad_norm": 5.0315046310424805,
+      "grad_norm": 5.036501884460449,
       "learning_rate": 3.5e-06,
-      "loss": 0.6025,
+      "loss": 0.6297,
       "step": 175
     },
     {
       "epoch": 0.15444015444015444,
-      "grad_norm": 4.241130828857422,
+      "grad_norm": 4.358458042144775,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.5837,
+      "loss": 0.6081,
       "step": 200
     },
     {
       "epoch": 0.17374517374517376,
-      "grad_norm": 4.572744369506836,
+      "grad_norm": 4.636687278747559,
       "learning_rate": 4.5e-06,
-      "loss": 0.5771,
+      "loss": 0.6034,
       "step": 225
     },
     {
       "epoch": 0.19305019305019305,
-      "grad_norm": 4.686110496520996,
+      "grad_norm": 4.5222554206848145,
       "learning_rate": 5e-06,
-      "loss": 0.5647,
+      "loss": 0.5915,
       "step": 250
     },
     {
       "epoch": 0.21235521235521235,
-      "grad_norm": 4.762725353240967,
+      "grad_norm": 4.840209007263184,
       "learning_rate": 5.500000000000001e-06,
-      "loss": 0.5817,
+      "loss": 0.6055,
       "step": 275
     },
     {
       "epoch": 0.23166023166023167,
-      "grad_norm": 4.6283278465271,
+      "grad_norm": 4.3733673095703125,
       "learning_rate": 6e-06,
-      "loss": 0.5409,
+      "loss": 0.565,
       "step": 300
     },
     {
       "epoch": 0.25096525096525096,
-      "grad_norm": 3.748809576034546,
+      "grad_norm": 3.8710694313049316,
       "learning_rate": 6.5000000000000004e-06,
-      "loss": 0.4916,
+      "loss": 0.5128,
       "step": 325
     },
     {
       "epoch": 0.2702702702702703,
-      "grad_norm": 4.640991687774658,
+      "grad_norm": 4.533177375793457,
       "learning_rate": 7e-06,
-      "loss": 0.5268,
+      "loss": 0.5532,
       "step": 350
     },
     {
       "epoch": 0.28957528957528955,
-      "grad_norm": 4.591019630432129,
+      "grad_norm": 4.626265048980713,
       "learning_rate": 7.500000000000001e-06,
-      "loss": 0.5422,
+      "loss": 0.5708,
       "step": 375
     },
     {
       "epoch": 0.3088803088803089,
-      "grad_norm": 5.206230640411377,
+      "grad_norm": 5.235887050628662,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.5214,
+      "loss": 0.5413,
       "step": 400
     },
     {
       "epoch": 0.3281853281853282,
-      "grad_norm": 4.378481388092041,
+      "grad_norm": 4.232483386993408,
       "learning_rate": 8.5e-06,
-      "loss": 0.5085,
+      "loss": 0.5326,
       "step": 425
     },
     {
       "epoch": 0.3474903474903475,
-      "grad_norm": 4.319910526275635,
+      "grad_norm": 4.374448776245117,
       "learning_rate": 9e-06,
-      "loss": 0.4955,
+      "loss": 0.5205,
       "step": 450
     },
     {
       "epoch": 0.3667953667953668,
-      "grad_norm": 4.216291427612305,
+      "grad_norm": 4.335531711578369,
       "learning_rate": 9.5e-06,
-      "loss": 0.5112,
+      "loss": 0.5347,
       "step": 475
     },
     {
       "epoch": 0.3861003861003861,
-      "grad_norm": 4.130216121673584,
+      "grad_norm": 4.169550895690918,
       "learning_rate": 1e-05,
-      "loss": 0.4917,
+      "loss": 0.5148,
       "step": 500
     },
     {
       "epoch": 0.40540540540540543,
-      "grad_norm": 3.5388808250427246,
+      "grad_norm": 3.5508618354797363,
       "learning_rate": 9.944444444444445e-06,
-      "loss": 0.4995,
+      "loss": 0.5227,
       "step": 525
     },
     {
       "epoch": 0.4247104247104247,
-      "grad_norm": 3.57562255859375,
+      "grad_norm": 3.7011241912841797,
       "learning_rate": 9.88888888888889e-06,
-      "loss": 0.4747,
+      "loss": 0.4989,
       "step": 550
     },
     {
       "epoch": 0.444015444015444,
-      "grad_norm": 3.1933300495147705,
+      "grad_norm": 3.37355899810791,
       "learning_rate": 9.833333333333333e-06,
-      "loss": 0.4789,
+      "loss": 0.5067,
       "step": 575
     },
     {
       "epoch": 0.46332046332046334,
-      "grad_norm": 3.9434165954589844,
+      "grad_norm": 4.071038246154785,
       "learning_rate": 9.777777777777779e-06,
-      "loss": 0.4642,
+      "loss": 0.4884,
       "step": 600
     },
     {
       "epoch": 0.4826254826254826,
-      "grad_norm": 3.4227051734924316,
+      "grad_norm": 3.6045796871185303,
       "learning_rate": 9.722222222222223e-06,
-      "loss": 0.4493,
+      "loss": 0.4752,
       "step": 625
     },
     {
       "epoch": 0.5019305019305019,
-      "grad_norm": 3.3387601375579834,
+      "grad_norm": 3.414005994796753,
       "learning_rate": 9.666666666666667e-06,
-      "loss": 0.4561,
+      "loss": 0.4784,
       "step": 650
     },
     {
       "epoch": 0.5212355212355212,
-      "grad_norm": 3.66536021232605,
+      "grad_norm": 3.203928232192993,
       "learning_rate": 9.611111111111112e-06,
-      "loss": 0.4703,
+      "loss": 0.4915,
       "step": 675
     },
     {
       "epoch": 0.5405405405405406,
-      "grad_norm": 3.9533724784851074,
+      "grad_norm": 4.148367404937744,
       "learning_rate": 9.555555555555556e-06,
-      "loss": 0.4602,
+      "loss": 0.4861,
       "step": 700
     },
     {
       "epoch": 0.5598455598455598,
-      "grad_norm": 3.853438377380371,
+      "grad_norm": 3.9569835662841797,
       "learning_rate": 9.5e-06,
-      "loss": 0.4524,
+      "loss": 0.4744,
       "step": 725
     },
     {
       "epoch": 0.5791505791505791,
-      "grad_norm": 3.8207361698150635,
+      "grad_norm": 3.682847261428833,
       "learning_rate": 9.444444444444445e-06,
-      "loss": 0.4422,
+      "loss": 0.4662,
       "step": 750
     },
     {
       "epoch": 0.5984555984555985,
-      "grad_norm": 3.5014588832855225,
+      "grad_norm": 3.498784303665161,
       "learning_rate": 9.38888888888889e-06,
-      "loss": 0.4261,
+      "loss": 0.4479,
       "step": 775
     },
     {
       "epoch": 0.6177606177606177,
-      "grad_norm": 3.4164435863494873,
+      "grad_norm": 3.3415608406066895,
       "learning_rate": 9.333333333333334e-06,
-      "loss": 0.4244,
+      "loss": 0.4422,
       "step": 800
     },
     {
       "epoch": 0.637065637065637,
-      "grad_norm": 3.5216405391693115,
+      "grad_norm": 3.7531261444091797,
       "learning_rate": 9.277777777777778e-06,
-      "loss": 0.4406,
+      "loss": 0.4639,
       "step": 825
     },
     {
       "epoch": 0.6563706563706564,
-      "grad_norm": 3.0401480197906494,
+      "grad_norm": 3.1370999813079834,
       "learning_rate": 9.222222222222224e-06,
-      "loss": 0.4047,
+      "loss": 0.4286,
       "step": 850
     },
     {
       "epoch": 0.6756756756756757,
-      "grad_norm": 3.4229395389556885,
+      "grad_norm": 3.474857807159424,
       "learning_rate": 9.166666666666666e-06,
-      "loss": 0.4205,
+      "loss": 0.4461,
       "step": 875
     },
     {
       "epoch": 0.694980694980695,
-      "grad_norm": 3.6540348529815674,
+      "grad_norm": 3.5445330142974854,
       "learning_rate": 9.111111111111112e-06,
-      "loss": 0.4143,
+      "loss": 0.4364,
       "step": 900
     },
     {
       "epoch": 0.7142857142857143,
-      "grad_norm": 3.199246883392334,
+      "grad_norm": 3.309083938598633,
       "learning_rate": 9.055555555555556e-06,
-      "loss": 0.4062,
+      "loss": 0.4274,
       "step": 925
     },
     {
       "epoch": 0.7335907335907336,
-      "grad_norm": 3.220662832260132,
+      "grad_norm": 3.169285297393799,
       "learning_rate": 9e-06,
-      "loss": 0.405,
+      "loss": 0.4244,
       "step": 950
     },
     {
       "epoch": 0.752895752895753,
-      "grad_norm": 3.5012640953063965,
+      "grad_norm": 3.5987651348114014,
       "learning_rate": 8.944444444444446e-06,
-      "loss": 0.3942,
+      "loss": 0.4189,
       "step": 975
     },
     {
       "epoch": 0.7722007722007722,
-      "grad_norm": 3.518545150756836,
+      "grad_norm": 3.622044563293457,
       "learning_rate": 8.888888888888888e-06,
-      "loss": 0.3936,
+      "loss": 0.4118,
       "step": 1000
     },
     {
       "epoch": 0.7722007722007722,
-      "eval_loss": 0.45976704359054565,
-      "eval_runtime": 2976.7472,
-      "eval_samples_per_second": 2.359,
-      "eval_steps_per_second": 0.147,
-      "eval_wer": 0.34102477535968423,
+      "eval_loss": 0.48980528116226196,
+      "eval_runtime": 2996.853,
+      "eval_samples_per_second": 2.343,
+      "eval_steps_per_second": 0.146,
+      "eval_wer": 0.3626803079679568,
       "step": 1000
     },
     {
       "epoch": 0.7915057915057915,
-      "grad_norm": 3.2509634494781494,
+      "grad_norm": 3.3155651092529297,
       "learning_rate": 8.833333333333334e-06,
-      "loss": 0.4169,
+      "loss": 0.4394,
       "step": 1025
     },
     {
       "epoch": 0.8108108108108109,
-      "grad_norm": 2.5278899669647217,
+      "grad_norm": 2.6345057487487793,
       "learning_rate": 8.777777777777778e-06,
-      "loss": 0.3895,
+      "loss": 0.4143,
       "step": 1050
     },
     {
       "epoch": 0.8301158301158301,
-      "grad_norm": 3.232598066329956,
+      "grad_norm": 3.3825740814208984,
       "learning_rate": 8.722222222222224e-06,
-      "loss": 0.3782,
+      "loss": 0.399,
       "step": 1075
     },
     {
       "epoch": 0.8494208494208494,
-      "grad_norm": 3.384092092514038,
+      "grad_norm": 3.439831256866455,
       "learning_rate": 8.666666666666668e-06,
-      "loss": 0.406,
+      "loss": 0.4269,
       "step": 1100
     },
     {
       "epoch": 0.8687258687258688,
-      "grad_norm": 3.261749267578125,
+      "grad_norm": 2.9175798892974854,
       "learning_rate": 8.611111111111112e-06,
-      "loss": 0.4397,
+      "loss": 0.46,
       "step": 1125
     },
     {
       "epoch": 0.888030888030888,
-      "grad_norm": 3.817667007446289,
+      "grad_norm": 3.6642494201660156,
       "learning_rate": 8.555555555555556e-06,
-      "loss": 0.3846,
+      "loss": 0.4038,
       "step": 1150
     },
     {
       "epoch": 0.9073359073359073,
-      "grad_norm": 3.3195998668670654,
+      "grad_norm": 3.3490536212921143,
       "learning_rate": 8.5e-06,
-      "loss": 0.3748,
+      "loss": 0.3989,
       "step": 1175
     },
     {
       "epoch": 0.9266409266409267,
-      "grad_norm": 3.510660171508789,
+      "grad_norm": 3.563148021697998,
       "learning_rate": 8.444444444444446e-06,
-      "loss": 0.3977,
+      "loss": 0.4221,
       "step": 1200
     },
     {
       "epoch": 0.9459459459459459,
-      "grad_norm": 2.8854782581329346,
+      "grad_norm": 3.2811503410339355,
       "learning_rate": 8.38888888888889e-06,
-      "loss": 0.3768,
+      "loss": 0.3959,
       "step": 1225
     },
     {
       "epoch": 0.9652509652509652,
-      "grad_norm": 3.2596817016601562,
+      "grad_norm": 4.630711078643799,
       "learning_rate": 8.333333333333334e-06,
-      "loss": 0.3791,
+      "loss": 0.3992,
       "step": 1250
     },
     {
       "epoch": 0.9845559845559846,
-      "grad_norm": 2.9499335289001465,
+      "grad_norm": 2.9398629665374756,
       "learning_rate": 8.277777777777778e-06,
-      "loss": 0.3839,
+      "loss": 0.4075,
       "step": 1275
     },
     {
       "epoch": 1.0038610038610039,
-      "grad_norm": 2.457566738128662,
+      "grad_norm": 2.4827306270599365,
       "learning_rate": 8.222222222222222e-06,
-      "loss": 0.3371,
+      "loss": 0.3593,
       "step": 1300
     },
     {
       "epoch": 1.0231660231660231,
-      "grad_norm": 2.8226237297058105,
+      "grad_norm": 2.922198534011841,
       "learning_rate": 8.166666666666668e-06,
-      "loss": 0.2705,
+      "loss": 0.2864,
       "step": 1325
     },
     {
       "epoch": 1.0424710424710424,
-      "grad_norm": 3.664156198501587,
+      "grad_norm": 2.595409393310547,
       "learning_rate": 8.111111111111112e-06,
-      "loss": 0.2729,
+      "loss": 0.2928,
       "step": 1350
     },
     {
       "epoch": 1.0617760617760619,
-      "grad_norm": 2.497749090194702,
+      "grad_norm": 2.6854002475738525,
       "learning_rate": 8.055555555555557e-06,
-      "loss": 0.2613,
+      "loss": 0.2773,
       "step": 1375
     },
     {
       "epoch": 1.0810810810810811,
-      "grad_norm": 2.437830686569214,
+      "grad_norm": 2.5066304206848145,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.2718,
+      "loss": 0.2864,
       "step": 1400
     },
     {
       "epoch": 1.1003861003861004,
-      "grad_norm": 2.5171914100646973,
+      "grad_norm": 2.477524518966675,
       "learning_rate": 7.944444444444445e-06,
-      "loss": 0.2649,
+      "loss": 0.2815,
       "step": 1425
     },
     {
       "epoch": 1.1196911196911197,
-      "grad_norm": 3.023686647415161,
+      "grad_norm": 2.947479724884033,
       "learning_rate": 7.88888888888889e-06,
-      "loss": 0.2662,
+      "loss": 0.2814,
       "step": 1450
     },
     {
       "epoch": 1.138996138996139,
-      "grad_norm": 2.358494520187378,
+      "grad_norm": 2.497398853302002,
       "learning_rate": 7.833333333333333e-06,
-      "loss": 0.2511,
+      "loss": 0.27,
       "step": 1475
     },
     {
       "epoch": 1.1583011583011582,
-      "grad_norm": 2.428818941116333,
+      "grad_norm": 2.5605263710021973,
       "learning_rate": 7.77777777777778e-06,
-      "loss": 0.2626,
+      "loss": 0.2778,
       "step": 1500
     },
     {
       "epoch": 1.1776061776061777,
-      "grad_norm": 3.066359281539917,
+      "grad_norm": 2.8643362522125244,
       "learning_rate": 7.722222222222223e-06,
-      "loss": 0.2504,
+      "loss": 0.2625,
       "step": 1525
     },
     {
       "epoch": 1.196911196911197,
-      "grad_norm": 2.8853089809417725,
+      "grad_norm": 2.8837692737579346,
       "learning_rate": 7.666666666666667e-06,
-      "loss": 0.2565,
+      "loss": 0.2732,
       "step": 1550
     },
     {
       "epoch": 1.2162162162162162,
-      "grad_norm": 2.476994514465332,
+      "grad_norm": 2.499385118484497,
       "learning_rate": 7.611111111111111e-06,
-      "loss": 0.274,
+      "loss": 0.2898,
       "step": 1575
     },
     {
       "epoch": 1.2355212355212355,
-      "grad_norm": 3.0519564151763916,
+      "grad_norm": 3.035726547241211,
       "learning_rate": 7.555555555555556e-06,
-      "loss": 0.2604,
+      "loss": 0.279,
       "step": 1600
     },
     {
       "epoch": 1.2548262548262548,
-      "grad_norm": 2.985853910446167,
+      "grad_norm": 2.872089385986328,
       "learning_rate": 7.500000000000001e-06,
-      "loss": 0.2832,
+      "loss": 0.3005,
       "step": 1625
     },
     {
       "epoch": 1.2741312741312742,
-      "grad_norm": 2.780881643295288,
+      "grad_norm": 2.8907766342163086,
       "learning_rate": 7.444444444444445e-06,
-      "loss": 0.2743,
+      "loss": 0.2941,
       "step": 1650
     },
     {
       "epoch": 1.2934362934362935,
-      "grad_norm": 2.6131482124328613,
+      "grad_norm": 2.8525350093841553,
       "learning_rate": 7.38888888888889e-06,
-      "loss": 0.2483,
+      "loss": 0.2644,
       "step": 1675
     },
     {
       "epoch": 1.3127413127413128,
-      "grad_norm": 3.0259549617767334,
+      "grad_norm": 2.7391347885131836,
       "learning_rate": 7.333333333333333e-06,
-      "loss": 0.2686,
+      "loss": 0.2811,
       "step": 1700
     },
     {
       "epoch": 1.332046332046332,
-      "grad_norm": 2.420754909515381,
+      "grad_norm": 2.4943618774414062,
       "learning_rate": 7.277777777777778e-06,
-      "loss": 0.265,
+      "loss": 0.2786,
       "step": 1725
     },
     {
       "epoch": 1.3513513513513513,
-      "grad_norm": 3.001450300216675,
+      "grad_norm": 3.009016752243042,
       "learning_rate": 7.222222222222223e-06,
-      "loss": 0.2772,
+      "loss": 0.2928,
       "step": 1750
     },
     {
       "epoch": 1.3706563706563706,
-      "grad_norm": 3.064401626586914,
+      "grad_norm": 3.0978188514709473,
       "learning_rate": 7.166666666666667e-06,
-      "loss": 0.2844,
+      "loss": 0.298,
       "step": 1775
     },
     {
       "epoch": 1.3899613899613898,
-      "grad_norm": 2.651357412338257,
+      "grad_norm": 2.746687412261963,
       "learning_rate": 7.111111111111112e-06,
-      "loss": 0.2681,
+      "loss": 0.2848,
       "step": 1800
     },
     {
       "epoch": 1.4092664092664093,
-      "grad_norm": 2.363473415374756,
+      "grad_norm": 2.6545677185058594,
       "learning_rate": 7.055555555555557e-06,
-      "loss": 0.2567,
+      "loss": 0.2755,
       "step": 1825
     },
     {
       "epoch": 1.4285714285714286,
-      "grad_norm": 3.044689178466797,
+      "grad_norm": 3.196070671081543,
       "learning_rate": 7e-06,
-      "loss": 0.2566,
+      "loss": 0.2733,
       "step": 1850
     },
     {
       "epoch": 1.4478764478764479,
-      "grad_norm": 2.3913726806640625,
+      "grad_norm": 2.5609796047210693,
       "learning_rate": 6.944444444444445e-06,
-      "loss": 0.2506,
+      "loss": 0.2659,
       "step": 1875
     },
     {
       "epoch": 1.4671814671814671,
-      "grad_norm": 2.2928853034973145,
+      "grad_norm": 2.339729070663452,
       "learning_rate": 6.88888888888889e-06,
-      "loss": 0.2727,
+      "loss": 0.2924,
       "step": 1900
     },
     {
       "epoch": 1.4864864864864864,
-      "grad_norm": 2.741959571838379,
+      "grad_norm": 2.7920401096343994,
       "learning_rate": 6.833333333333334e-06,
-      "loss": 0.2467,
+      "loss": 0.2628,
       "step": 1925
     },
     {
       "epoch": 1.505791505791506,
-      "grad_norm": 2.0824573040008545,
+      "grad_norm": 2.2347285747528076,
       "learning_rate": 6.777777777777779e-06,
-      "loss": 0.2635,
+      "loss": 0.2816,
       "step": 1950
     },
     {
       "epoch": 1.525096525096525,
-      "grad_norm": 2.4026057720184326,
+      "grad_norm": 2.6325347423553467,
       "learning_rate": 6.7222222222222235e-06,
-      "loss": 0.2494,
+      "loss": 0.2664,
       "step": 1975
     },
     {
       "epoch": 1.5444015444015444,
-      "grad_norm": 3.137629270553589,
+      "grad_norm": 3.093266248703003,
       "learning_rate": 6.666666666666667e-06,
-      "loss": 0.2493,
+      "loss": 0.2629,
       "step": 2000
     },
     {
       "epoch": 1.5444015444015444,
-      "eval_loss": 0.40415582060813904,
-      "eval_runtime": 2884.198,
-      "eval_samples_per_second": 2.435,
-      "eval_steps_per_second": 0.152,
-      "eval_wer": 0.29395418895756503,
+      "eval_loss": 0.4272039532661438,
+      "eval_runtime": 2939.2445,
+      "eval_samples_per_second": 2.389,
+      "eval_steps_per_second": 0.149,
+      "eval_wer": 0.3144726762830916,
       "step": 2000
     },
     {
       "epoch": 1.5637065637065637,
-      "grad_norm": 2.6668949127197266,
+      "grad_norm": 2.7231810092926025,
       "learning_rate": 6.6111111111111115e-06,
-      "loss": 0.2522,
+      "loss": 0.2671,
       "step": 2025
     },
     {
       "epoch": 1.583011583011583,
-      "grad_norm": 2.5405499935150146,
+      "grad_norm": 2.4896557331085205,
       "learning_rate": 6.555555555555556e-06,
-      "loss": 0.2747,
+      "loss": 0.291,
       "step": 2050
     },
     {
       "epoch": 1.6023166023166024,
-      "grad_norm": 2.4073848724365234,
+      "grad_norm": 2.549482583999634,
       "learning_rate": 6.5000000000000004e-06,
-      "loss": 0.2563,
+      "loss": 0.2697,
       "step": 2075
     },
     {
       "epoch": 1.6216216216216215,
-      "grad_norm": 3.3368024826049805,
+      "grad_norm": 3.055518627166748,
       "learning_rate": 6.444444444444445e-06,
-      "loss": 0.2477,
+      "loss": 0.2612,
       "step": 2100
     },
     {
       "epoch": 1.640926640926641,
-      "grad_norm": 2.741755723953247,
+      "grad_norm": 2.8023324012756348,
       "learning_rate": 6.3888888888888885e-06,
-      "loss": 0.2444,
+      "loss": 0.2612,
       "step": 2125
     },
     {
       "epoch": 1.6602316602316602,
-      "grad_norm": 2.5123753547668457,
+      "grad_norm": 2.5764496326446533,
       "learning_rate": 6.333333333333333e-06,
-      "loss": 0.253,
+      "loss": 0.2718,
       "step": 2150
     },
     {
       "epoch": 1.6795366795366795,
-      "grad_norm": 2.8450229167938232,
+      "grad_norm": 3.0034117698669434,
       "learning_rate": 6.277777777777778e-06,
-      "loss": 0.2695,
+      "loss": 0.2836,
       "step": 2175
     },
     {
       "epoch": 1.698841698841699,
-      "grad_norm": 2.2329864501953125,
+      "grad_norm": 3.285400629043579,
       "learning_rate": 6.222222222222223e-06,
-      "loss": 0.2531,
+      "loss": 0.267,
       "step": 2200
     },
     {
       "epoch": 1.718146718146718,
-      "grad_norm": 2.8518948554992676,
+      "grad_norm": 2.8904147148132324,
       "learning_rate": 6.166666666666667e-06,
-      "loss": 0.2671,
+      "loss": 0.2844,
       "step": 2225
     },
     {
       "epoch": 1.7374517374517375,
-      "grad_norm": 2.6565253734588623,
+      "grad_norm": 2.7520534992218018,
       "learning_rate": 6.111111111111112e-06,
-      "loss": 0.2477,
+      "loss": 0.2641,
       "step": 2250
     },
     {
       "epoch": 1.7567567567567568,
-      "grad_norm": 2.2332699298858643,
+      "grad_norm": 2.262890338897705,
       "learning_rate": 6.055555555555555e-06,
-      "loss": 0.2461,
+      "loss": 0.2636,
       "step": 2275
     },
     {
       "epoch": 1.776061776061776,
-      "grad_norm": 2.5508103370666504,
+      "grad_norm": 2.410085439682007,
       "learning_rate": 6e-06,
-      "loss": 0.2574,
+      "loss": 0.2753,
       "step": 2300
     },
     {
       "epoch": 1.7953667953667953,
-      "grad_norm": 2.47121000289917,
+      "grad_norm": 2.5720226764678955,
       "learning_rate": 5.944444444444445e-06,
-      "loss": 0.2519,
+      "loss": 0.2702,
       "step": 2325
     },
     {
       "epoch": 1.8146718146718146,
-      "grad_norm": 2.798379898071289,
+      "grad_norm": 2.999687433242798,
       "learning_rate": 5.88888888888889e-06,
-      "loss": 0.2774,
+      "loss": 0.2998,
       "step": 2350
     },
     {
       "epoch": 1.833976833976834,
-      "grad_norm": 2.6081783771514893,
+      "grad_norm": 2.5406405925750732,
       "learning_rate": 5.833333333333334e-06,
-      "loss": 0.2271,
+      "loss": 0.2444,
       "step": 2375
     },
     {
       "epoch": 1.8532818532818531,
-      "grad_norm": 2.355163335800171,
+      "grad_norm": 2.314110517501831,
       "learning_rate": 5.777777777777778e-06,
-      "loss": 0.248,
+      "loss": 0.2664,
       "step": 2400
     },
     {
       "epoch": 1.8725868725868726,
-      "grad_norm": 2.8276679515838623,
+      "grad_norm": 2.745534896850586,
       "learning_rate": 5.722222222222222e-06,
-      "loss": 0.2575,
+      "loss": 0.2676,
       "step": 2425
     },
     {
       "epoch": 1.8918918918918919,
-      "grad_norm": 2.9946837425231934,
+      "grad_norm": 3.1628456115722656,
       "learning_rate": 5.666666666666667e-06,
-      "loss": 0.234,
+      "loss": 0.2475,
       "step": 2450
     },
     {
       "epoch": 1.9111969111969112,
-      "grad_norm": 2.660792350769043,
+      "grad_norm": 2.6415674686431885,
       "learning_rate": 5.611111111111112e-06,
-      "loss": 0.2373,
+      "loss": 0.2565,
       "step": 2475
     },
     {
       "epoch": 1.9305019305019306,
-      "grad_norm": 2.4244725704193115,
+      "grad_norm": 2.4771909713745117,
       "learning_rate": 5.555555555555557e-06,
-      "loss": 0.2443,
+      "loss": 0.2621,
       "step": 2500
     },
     {
       "epoch": 1.9498069498069497,
-      "grad_norm": 2.6997570991516113,
+      "grad_norm": 2.7251639366149902,
       "learning_rate": 5.500000000000001e-06,
-      "loss": 0.2553,
+      "loss": 0.2703,
       "step": 2525
     },
     {
       "epoch": 1.9691119691119692,
-      "grad_norm": 2.6614644527435303,
+      "grad_norm": 2.73382830619812,
       "learning_rate": 5.444444444444445e-06,
-      "loss": 0.2526,
+      "loss": 0.2728,
       "step": 2550
     },
     {
       "epoch": 1.9884169884169884,
-      "grad_norm": 2.8852739334106445,
+      "grad_norm": 3.0017306804656982,
       "learning_rate": 5.388888888888889e-06,
-      "loss": 0.2363,
+      "loss": 0.2521,
       "step": 2575
     },
     {
       "epoch": 2.0077220077220077,
-      "grad_norm": 1.8691250085830688,
+      "grad_norm": 1.9712167978286743,
       "learning_rate": 5.333333333333334e-06,
-      "loss": 0.2197,
+      "loss": 0.2352,
       "step": 2600
     },
     {
       "epoch": 2.027027027027027,
-      "grad_norm": 1.8070666790008545,
+      "grad_norm": 1.8378448486328125,
       "learning_rate": 5.2777777777777785e-06,
-      "loss": 0.1556,
+      "loss": 0.1696,
       "step": 2625
     },
     {
       "epoch": 2.0463320463320462,
-      "grad_norm": 2.2826385498046875,
+      "grad_norm": 2.419578790664673,
       "learning_rate": 5.2222222222222226e-06,
-      "loss": 0.1476,
+      "loss": 0.1592,
       "step": 2650
     },
     {
       "epoch": 2.0656370656370657,
-      "grad_norm": 2.1625537872314453,
+      "grad_norm": 2.145256757736206,
       "learning_rate": 5.1666666666666675e-06,
-      "loss": 0.1628,
+      "loss": 0.1719,
       "step": 2675
     },
     {
       "epoch": 2.0849420849420848,
-      "grad_norm": 2.158252000808716,
+      "grad_norm": 2.147104024887085,
       "learning_rate": 5.1111111111111115e-06,
-      "loss": 0.1353,
+      "loss": 0.1487,
       "step": 2700
     },
     {
       "epoch": 2.1042471042471043,
-      "grad_norm": 2.2190427780151367,
+      "grad_norm": 2.280949592590332,
       "learning_rate": 5.0555555555555555e-06,
-      "loss": 0.1697,
+      "loss": 0.1785,
       "step": 2725
     },
     {
       "epoch": 2.1235521235521237,
-      "grad_norm": 1.9213645458221436,
+      "grad_norm": 2.359393835067749,
       "learning_rate": 5e-06,
-      "loss": 0.1535,
+      "loss": 0.1656,
       "step": 2750
     },
     {
       "epoch": 2.142857142857143,
-      "grad_norm": 2.329991340637207,
+      "grad_norm": 2.4849932193756104,
       "learning_rate": 4.944444444444445e-06,
-      "loss": 0.1582,
+      "loss": 0.1697,
       "step": 2775
     },
     {
       "epoch": 2.1621621621621623,
-      "grad_norm": 2.3484299182891846,
+      "grad_norm": 2.478525400161743,
       "learning_rate": 4.888888888888889e-06,
-      "loss": 0.1464,
+      "loss": 0.1586,
       "step": 2800
     },
     {
       "epoch": 2.1814671814671813,
-      "grad_norm": 2.418144941329956,
+      "grad_norm": 2.258152723312378,
       "learning_rate": 4.833333333333333e-06,
-      "loss": 0.1557,
+      "loss": 0.1677,
       "step": 2825
     },
     {
       "epoch": 2.200772200772201,
-      "grad_norm": 2.182532787322998,
+      "grad_norm": 2.2826361656188965,
       "learning_rate": 4.777777777777778e-06,
-      "loss": 0.1463,
+      "loss": 0.1603,
       "step": 2850
     },
     {
       "epoch": 2.2200772200772203,
-      "grad_norm": 1.8456135988235474,
+      "grad_norm": 2.3771138191223145,
       "learning_rate": 4.722222222222222e-06,
-      "loss": 0.1497,
+      "loss": 0.159,
       "step": 2875
     },
     {
       "epoch": 2.2393822393822393,
-      "grad_norm": 2.3882248401641846,
+      "grad_norm": 2.353187322616577,
       "learning_rate": 4.666666666666667e-06,
-      "loss": 0.1487,
+      "loss": 0.1617,
       "step": 2900
     },
     {
       "epoch": 2.258687258687259,
-      "grad_norm": 2.5126445293426514,
+      "grad_norm": 2.2030959129333496,
       "learning_rate": 4.611111111111112e-06,
-      "loss": 0.1577,
+      "loss": 0.17,
       "step": 2925
     },
     {
       "epoch": 2.277992277992278,
-      "grad_norm": 2.2082295417785645,
+      "grad_norm": 2.1587460041046143,
       "learning_rate": 4.555555555555556e-06,
-      "loss": 0.1491,
+      "loss": 0.1618,
       "step": 2950
     },
     {
       "epoch": 2.2972972972972974,
-      "grad_norm": 2.6382830142974854,
+      "grad_norm": 2.918156623840332,
       "learning_rate": 4.5e-06,
-      "loss": 0.1428,
+      "loss": 0.1545,
       "step": 2975
     },
     {
       "epoch": 2.3166023166023164,
-      "grad_norm": 2.447270631790161,
+      "grad_norm": 2.4041364192962646,
       "learning_rate": 4.444444444444444e-06,
-      "loss": 0.1548,
+      "loss": 0.1677,
       "step": 3000
     },
     {
       "epoch": 2.3166023166023164,
-      "eval_loss": 0.39975878596305847,
-      "eval_runtime": 2877.301,
-      "eval_samples_per_second": 2.44,
-      "eval_steps_per_second": 0.153,
-      "eval_wer": 0.28095621461590403,
+      "eval_loss": 0.42425960302352905,
+      "eval_runtime": 2930.947,
+      "eval_samples_per_second": 2.396,
+      "eval_steps_per_second": 0.15,
+      "eval_wer": 0.3038651796263357,
       "step": 3000
     },
     {
       "epoch": 2.335907335907336,
-      "grad_norm": 1.664686679840088,
+      "grad_norm": 1.8374619483947754,
       "learning_rate": 4.388888888888889e-06,
-      "loss": 0.1407,
+      "loss": 0.1505,
       "step": 3025
     },
     {
       "epoch": 2.3552123552123554,
-      "grad_norm": 2.261843681335449,
+      "grad_norm": 2.2818222045898438,
       "learning_rate": 4.333333333333334e-06,
-      "loss": 0.1511,
+      "loss": 0.1635,
       "step": 3050
     },
     {
       "epoch": 2.3745173745173744,
-      "grad_norm": 2.157227039337158,
+      "grad_norm": 2.3476946353912354,
       "learning_rate": 4.277777777777778e-06,
-      "loss": 0.1433,
+      "loss": 0.1572,
       "step": 3075
     },
     {
       "epoch": 2.393822393822394,
-      "grad_norm": 2.434943199157715,
+      "grad_norm": 2.727679491043091,
       "learning_rate": 4.222222222222223e-06,
-      "loss": 0.1476,
+      "loss": 0.1567,
       "step": 3100
     },
     {
       "epoch": 2.413127413127413,
-      "grad_norm": 2.106074333190918,
+      "grad_norm": 2.148946523666382,
       "learning_rate": 4.166666666666667e-06,
-      "loss": 0.1496,
+      "loss": 0.1629,
       "step": 3125
     },
     {
       "epoch": 2.4324324324324325,
-      "grad_norm": 2.263333320617676,
+      "grad_norm": 2.3598289489746094,
       "learning_rate": 4.111111111111111e-06,
-      "loss": 0.1409,
+      "loss": 0.1537,
       "step": 3150
     },
     {
       "epoch": 2.4517374517374515,
-      "grad_norm": 2.9890952110290527,
+      "grad_norm": 2.627530813217163,
       "learning_rate": 4.055555555555556e-06,
-      "loss": 0.1497,
+      "loss": 0.1617,
       "step": 3175
     },
     {
       "epoch": 2.471042471042471,
-      "grad_norm": 2.174651622772217,
+      "grad_norm": 2.1012682914733887,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.1418,
+      "loss": 0.1541,
       "step": 3200
     },
     {
       "epoch": 2.4903474903474905,
-      "grad_norm": 2.380537509918213,
+      "grad_norm": 2.719054937362671,
       "learning_rate": 3.944444444444445e-06,
-      "loss": 0.1369,
+      "loss": 0.1479,
       "step": 3225
     },
     {
       "epoch": 2.5096525096525095,
-      "grad_norm": 2.1118268966674805,
+      "grad_norm": 1.9170578718185425,
       "learning_rate": 3.88888888888889e-06,
-      "loss": 0.1465,
+      "loss": 0.1558,
       "step": 3250
     },
     {
       "epoch": 2.528957528957529,
-      "grad_norm": 2.221266746520996,
+      "grad_norm": 2.209770917892456,
       "learning_rate": 3.833333333333334e-06,
-      "loss": 0.1392,
+      "loss": 0.1494,
       "step": 3275
     },
     {
       "epoch": 2.5482625482625485,
-      "grad_norm": 2.216095209121704,
+      "grad_norm": 2.3125557899475098,
       "learning_rate": 3.777777777777778e-06,
-      "loss": 0.143,
+      "loss": 0.1549,
       "step": 3300
     },
     {
       "epoch": 2.5675675675675675,
-      "grad_norm": 2.23201322555542,
+      "grad_norm": 2.2932839393615723,
       "learning_rate": 3.7222222222222225e-06,
-      "loss": 0.1399,
+      "loss": 0.1529,
       "step": 3325
     },
     {
       "epoch": 2.586872586872587,
-      "grad_norm": 2.174283742904663,
+      "grad_norm": 2.347710371017456,
       "learning_rate": 3.6666666666666666e-06,
-      "loss": 0.1531,
+      "loss": 0.1637,
       "step": 3350
     },
     {
       "epoch": 2.606177606177606,
-      "grad_norm": 2.455362558364868,
+      "grad_norm": 2.479405641555786,
       "learning_rate": 3.6111111111111115e-06,
-      "loss": 0.1466,
+      "loss": 0.162,
       "step": 3375
     },
     {
       "epoch": 2.6254826254826256,
-      "grad_norm": 2.6259799003601074,
+      "grad_norm": 2.7078042030334473,
       "learning_rate": 3.555555555555556e-06,
-      "loss": 0.1458,
+      "loss": 0.1606,
       "step": 3400
     },
     {
       "epoch": 2.6447876447876446,
-      "grad_norm": 2.1615617275238037,
+      "grad_norm": 2.097592830657959,
       "learning_rate": 3.5e-06,
-      "loss": 0.145,
+      "loss": 0.1554,
       "step": 3425
     },
     {
       "epoch": 2.664092664092664,
-      "grad_norm": 2.36684513092041,
+      "grad_norm": 2.405099630355835,
       "learning_rate": 3.444444444444445e-06,
-      "loss": 0.1462,
+      "loss": 0.1611,
       "step": 3450
     },
     {
       "epoch": 2.6833976833976836,
-      "grad_norm": 2.027125835418701,
+      "grad_norm": 2.120976686477661,
       "learning_rate": 3.3888888888888893e-06,
-      "loss": 0.1445,
+      "loss": 0.1537,
       "step": 3475
     },
     {
       "epoch": 2.7027027027027026,
-      "grad_norm": 2.1827645301818848,
+      "grad_norm": 2.165222406387329,
       "learning_rate": 3.3333333333333333e-06,
-      "loss": 0.1466,
+      "loss": 0.156,
       "step": 3500
     },
     {
       "epoch": 2.722007722007722,
-      "grad_norm": 2.1014013290405273,
+      "grad_norm": 2.3541574478149414,
       "learning_rate": 3.277777777777778e-06,
-      "loss": 0.1453,
+      "loss": 0.1571,
       "step": 3525
     },
     {
       "epoch": 2.741312741312741,
-      "grad_norm": 1.92685866355896,
+      "grad_norm": 2.5069708824157715,
       "learning_rate": 3.2222222222222227e-06,
-      "loss": 0.1341,
+      "loss": 0.1467,
       "step": 3550
     },
     {
       "epoch": 2.7606177606177607,
-      "grad_norm": 2.028932571411133,
+      "grad_norm": 1.9154902696609497,
       "learning_rate": 3.1666666666666667e-06,
-      "loss": 0.1399,
+      "loss": 0.1556,
       "step": 3575
     },
     {
       "epoch": 2.7799227799227797,
-      "grad_norm": 1.9122258424758911,
+      "grad_norm": 1.964211106300354,
       "learning_rate": 3.1111111111111116e-06,
-      "loss": 0.1433,
+      "loss": 0.1534,
       "step": 3600
     },
     {
       "epoch": 2.799227799227799,
-      "grad_norm": 2.2335093021392822,
+      "grad_norm": 2.3793370723724365,
       "learning_rate": 3.055555555555556e-06,
-      "loss": 0.1438,
+      "loss": 0.1578,
       "step": 3625
     },
     {
       "epoch": 2.8185328185328187,
-      "grad_norm": 2.2860329151153564,
+      "grad_norm": 2.7694272994995117,
       "learning_rate": 3e-06,
-      "loss": 0.1392,
+      "loss": 0.1518,
       "step": 3650
     },
     {
       "epoch": 2.8378378378378377,
-      "grad_norm": 1.873155951499939,
+      "grad_norm": 2.1063687801361084,
       "learning_rate": 2.944444444444445e-06,
-      "loss": 0.1396,
+      "loss": 0.1534,
       "step": 3675
     },
     {
       "epoch": 2.857142857142857,
-      "grad_norm": 2.6895735263824463,
+      "grad_norm": 2.708767890930176,
       "learning_rate": 2.888888888888889e-06,
-      "loss": 0.1419,
+      "loss": 0.153,
       "step": 3700
     },
     {
       "epoch": 2.8764478764478767,
-      "grad_norm": 1.6812546253204346,
+      "grad_norm": 1.7426552772521973,
       "learning_rate": 2.8333333333333335e-06,
-      "loss": 0.138,
+      "loss": 0.1501,
       "step": 3725
     },
     {
       "epoch": 2.8957528957528957,
-      "grad_norm": 2.893087387084961,
+      "grad_norm": 2.721165180206299,
       "learning_rate": 2.7777777777777783e-06,
-      "loss": 0.1479,
+      "loss": 0.1613,
       "step": 3750
     },
     {
       "epoch": 2.915057915057915,
-      "grad_norm": 2.508882999420166,
+      "grad_norm": 2.525545358657837,
       "learning_rate": 2.7222222222222224e-06,
-      "loss": 0.1427,
+      "loss": 0.1534,
       "step": 3775
     },
     {
       "epoch": 2.9343629343629343,
-      "grad_norm": 2.536713123321533,
+      "grad_norm": 2.6334362030029297,
       "learning_rate": 2.666666666666667e-06,
-      "loss": 0.1476,
+      "loss": 0.158,
       "step": 3800
     },
     {
       "epoch": 2.9536679536679538,
-      "grad_norm": 2.256779432296753,
+      "grad_norm": 2.1491916179656982,
       "learning_rate": 2.6111111111111113e-06,
-      "loss": 0.1357,
+      "loss": 0.15,
       "step": 3825
     },
     {
       "epoch": 2.972972972972973,
-      "grad_norm": 2.0813608169555664,
+      "grad_norm": 2.0909786224365234,
       "learning_rate": 2.5555555555555557e-06,
-      "loss": 0.1362,
+      "loss": 0.147,
       "step": 3850
     },
     {
       "epoch": 2.9922779922779923,
-      "grad_norm": 2.6842668056488037,
+      "grad_norm": 2.8027398586273193,
       "learning_rate": 2.5e-06,
-      "loss": 0.1527,
+      "loss": 0.1642,
       "step": 3875
     },
     {
       "epoch": 3.011583011583012,
-      "grad_norm": 1.5669656991958618,
+      "grad_norm": 1.8371341228485107,
       "learning_rate": 2.4444444444444447e-06,
-      "loss": 0.1045,
+      "loss": 0.1176,
       "step": 3900
     },
     {
       "epoch": 3.030888030888031,
-      "grad_norm": 1.9614019393920898,
+      "grad_norm": 2.1273913383483887,
       "learning_rate": 2.388888888888889e-06,
-      "loss": 0.0861,
+      "loss": 0.0937,
       "step": 3925
     },
     {
       "epoch": 3.0501930501930503,
-      "grad_norm": 1.6240154504776,
+      "grad_norm": 1.8209782838821411,
       "learning_rate": 2.3333333333333336e-06,
-      "loss": 0.0811,
+      "loss": 0.089,
       "step": 3950
     },
     {
       "epoch": 3.0694980694980694,
-      "grad_norm": 1.675820231437683,
+      "grad_norm": 2.118212938308716,
       "learning_rate": 2.277777777777778e-06,
-      "loss": 0.0808,
+      "loss": 0.0888,
       "step": 3975
     },
     {
       "epoch": 3.088803088803089,
-      "grad_norm": 1.5205894708633423,
+      "grad_norm": 1.675663709640503,
       "learning_rate": 2.222222222222222e-06,
-      "loss": 0.0755,
+      "loss": 0.0841,
       "step": 4000
     },
     {
       "epoch": 3.088803088803089,
-      "eval_loss": 0.42321887612342834,
-      "eval_runtime": 2878.9611,
-      "eval_samples_per_second": 2.439,
-      "eval_steps_per_second": 0.152,
-      "eval_wer": 0.27565833896016206,
+      "eval_loss": 0.4492855668067932,
+      "eval_runtime": 2954.4178,
+      "eval_samples_per_second": 2.377,
+      "eval_steps_per_second": 0.149,
+      "eval_wer": 0.2968411212526454,
       "step": 4000
     },
     {
       "epoch": 3.108108108108108,
-      "grad_norm": 1.8952137231826782,
+      "grad_norm": 2.0060861110687256,
       "learning_rate": 2.166666666666667e-06,
-      "loss": 0.0774,
+      "loss": 0.084,
       "step": 4025
     },
     {
       "epoch": 3.1274131274131274,
-      "grad_norm": 1.7920564413070679,
+      "grad_norm": 1.8335351943969727,
       "learning_rate": 2.1111111111111114e-06,
-      "loss": 0.0796,
+      "loss": 0.0871,
       "step": 4050
     },
     {
       "epoch": 3.146718146718147,
-      "grad_norm": 1.4591903686523438,
+      "grad_norm": 2.165919303894043,
       "learning_rate": 2.0555555555555555e-06,
-      "loss": 0.0784,
+      "loss": 0.0892,
       "step": 4075
     },
     {
       "epoch": 3.166023166023166,
-      "grad_norm": 1.7589879035949707,
+      "grad_norm": 1.5656211376190186,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 0.08,
+      "loss": 0.0869,
       "step": 4100
     },
     {
       "epoch": 3.1853281853281854,
-      "grad_norm": 2.2422068119049072,
+      "grad_norm": 1.9687210321426392,
       "learning_rate": 1.944444444444445e-06,
-      "loss": 0.0784,
+      "loss": 0.0878,
       "step": 4125
     },
     {
       "epoch": 3.2046332046332044,
-      "grad_norm": 2.1973836421966553,
+      "grad_norm": 2.017395496368408,
       "learning_rate": 1.888888888888889e-06,
-      "loss": 0.0747,
+      "loss": 0.0823,
       "step": 4150
     },
     {
       "epoch": 3.223938223938224,
-      "grad_norm": 1.9121280908584595,
+      "grad_norm": 2.3069615364074707,
       "learning_rate": 1.8333333333333333e-06,
-      "loss": 0.08,
+      "loss": 0.0894,
       "step": 4175
     },
     {
       "epoch": 3.2432432432432434,
-      "grad_norm": 1.950451135635376,
+      "grad_norm": 1.8950186967849731,
       "learning_rate": 1.777777777777778e-06,
-      "loss": 0.0787,
+      "loss": 0.0889,
       "step": 4200
     },
     {
       "epoch": 3.2625482625482625,
-      "grad_norm": 1.5729962587356567,
+      "grad_norm": 1.9198052883148193,
       "learning_rate": 1.7222222222222224e-06,
-      "loss": 0.0708,
+      "loss": 0.081,
       "step": 4225
     },
     {
       "epoch": 3.281853281853282,
-      "grad_norm": 1.7916944026947021,
+      "grad_norm": 2.25732421875,
       "learning_rate": 1.6666666666666667e-06,
-      "loss": 0.074,
+      "loss": 0.0872,
       "step": 4250
     },
     {
       "epoch": 3.301158301158301,
-      "grad_norm": 2.091259717941284,
+      "grad_norm": 1.6695570945739746,
       "learning_rate": 1.6111111111111113e-06,
-      "loss": 0.0734,
+      "loss": 0.0827,
       "step": 4275
     },
     {
       "epoch": 3.3204633204633205,
-      "grad_norm": 2.2973198890686035,
+      "grad_norm": 2.1860039234161377,
       "learning_rate": 1.5555555555555558e-06,
-      "loss": 0.0763,
+      "loss": 0.0864,
       "step": 4300
     },
     {
       "epoch": 3.33976833976834,
-      "grad_norm": 2.0894815921783447,
+      "grad_norm": 2.0618574619293213,
       "learning_rate": 1.5e-06,
-      "loss": 0.0748,
+      "loss": 0.0843,
       "step": 4325
     },
     {
       "epoch": 3.359073359073359,
-      "grad_norm": 1.7151942253112793,
+      "grad_norm": 1.952563762664795,
       "learning_rate": 1.4444444444444445e-06,
-      "loss": 0.0719,
+      "loss": 0.0822,
       "step": 4350
     },
     {
       "epoch": 3.3783783783783785,
-      "grad_norm": 2.2781429290771484,
+      "grad_norm": 2.2584211826324463,
       "learning_rate": 1.3888888888888892e-06,
-      "loss": 0.0795,
+      "loss": 0.087,
       "step": 4375
     },
     {
       "epoch": 3.3976833976833976,
-      "grad_norm": 1.7880635261535645,
+      "grad_norm": 1.7842656373977661,
       "learning_rate": 1.3333333333333334e-06,
-      "loss": 0.0765,
+      "loss": 0.0856,
       "step": 4400
     },
     {
       "epoch": 3.416988416988417,
-      "grad_norm": 1.8607178926467896,
+      "grad_norm": 2.130709409713745,
       "learning_rate": 1.2777777777777779e-06,
-      "loss": 0.0806,
+      "loss": 0.0886,
       "step": 4425
     },
     {
       "epoch": 3.436293436293436,
-      "grad_norm": 1.5629892349243164,
+      "grad_norm": 1.505071759223938,
       "learning_rate": 1.2222222222222223e-06,
-      "loss": 0.0731,
+      "loss": 0.0795,
       "step": 4450
     },
     {
       "epoch": 3.4555984555984556,
-      "grad_norm": 1.976099967956543,
+      "grad_norm": 2.2736964225769043,
       "learning_rate": 1.1666666666666668e-06,
-      "loss": 0.0727,
+      "loss": 0.0802,
       "step": 4475
     },
     {
       "epoch": 3.474903474903475,
-      "grad_norm": 1.7316856384277344,
+      "grad_norm": 1.685120940208435,
       "learning_rate": 1.111111111111111e-06,
-      "loss": 0.078,
+      "loss": 0.085,
       "step": 4500
     },
     {
       "epoch": 3.494208494208494,
-      "grad_norm": 1.796291470527649,
+      "grad_norm": 1.8761180639266968,
       "learning_rate": 1.0555555555555557e-06,
-      "loss": 0.0653,
+      "loss": 0.0732,
       "step": 4525
     },
     {
       "epoch": 3.5135135135135136,
-      "grad_norm": 1.6075197458267212,
+      "grad_norm": 1.7968302965164185,
       "learning_rate": 1.0000000000000002e-06,
-      "loss": 0.0788,
+      "loss": 0.0868,
       "step": 4550
     },
     {
       "epoch": 3.532818532818533,
-      "grad_norm": 1.6467320919036865,
+      "grad_norm": 1.6335569620132446,
       "learning_rate": 9.444444444444445e-07,
-      "loss": 0.0695,
+      "loss": 0.0787,
       "step": 4575
     },
     {
       "epoch": 3.552123552123552,
-      "grad_norm": 1.1778756380081177,
+      "grad_norm": 1.402917742729187,
       "learning_rate": 8.88888888888889e-07,
-      "loss": 0.0683,
+      "loss": 0.0779,
       "step": 4600
     },
     {
       "epoch": 3.571428571428571,
-      "grad_norm": 1.402613878250122,
+      "grad_norm": 1.4878441095352173,
       "learning_rate": 8.333333333333333e-07,
-      "loss": 0.0698,
+      "loss": 0.0795,
       "step": 4625
     },
     {
       "epoch": 3.5907335907335907,
-      "grad_norm": 2.3424105644226074,
+      "grad_norm": 1.846757411956787,
       "learning_rate": 7.777777777777779e-07,
-      "loss": 0.0788,
+      "loss": 0.0868,
       "step": 4650
     },
     {
       "epoch": 3.61003861003861,
-      "grad_norm": 1.551446795463562,
+      "grad_norm": 1.496010422706604,
       "learning_rate": 7.222222222222222e-07,
-      "loss": 0.0765,
+      "loss": 0.0836,
       "step": 4675
     },
     {
       "epoch": 3.629343629343629,
-      "grad_norm": 2.174612045288086,
+      "grad_norm": 2.0897486209869385,
       "learning_rate": 6.666666666666667e-07,
-      "loss": 0.0759,
+      "loss": 0.0839,
       "step": 4700
     },
     {
       "epoch": 3.6486486486486487,
-      "grad_norm": 1.595508337020874,
+      "grad_norm": 1.6161247491836548,
       "learning_rate": 6.111111111111112e-07,
-      "loss": 0.0783,
+      "loss": 0.0858,
       "step": 4725
     },
     {
       "epoch": 3.667953667953668,
-      "grad_norm": 1.6800012588500977,
+      "grad_norm": 1.9887114763259888,
       "learning_rate": 5.555555555555555e-07,
-      "loss": 0.0768,
+      "loss": 0.0864,
       "step": 4750
     },
     {
       "epoch": 3.687258687258687,
-      "grad_norm": 2.050546884536743,
+      "grad_norm": 2.271867513656616,
       "learning_rate": 5.000000000000001e-07,
-      "loss": 0.0736,
+      "loss": 0.0815,
       "step": 4775
     },
     {
       "epoch": 3.7065637065637067,
-      "grad_norm": 1.9666138887405396,
+      "grad_norm": 1.9788146018981934,
       "learning_rate": 4.444444444444445e-07,
-      "loss": 0.0735,
+      "loss": 0.0821,
       "step": 4800
     },
     {
       "epoch": 3.7258687258687258,
-      "grad_norm": 2.0180180072784424,
+      "grad_norm": 1.8956130743026733,
       "learning_rate": 3.8888888888888895e-07,
-      "loss": 0.071,
+      "loss": 0.0789,
       "step": 4825
     },
     {
       "epoch": 3.7451737451737452,
-      "grad_norm": 2.1212894916534424,
+      "grad_norm": 2.4207746982574463,
       "learning_rate": 3.3333333333333335e-07,
-      "loss": 0.0907,
+      "loss": 0.1009,
       "step": 4850
     },
     {
       "epoch": 3.7644787644787643,
-      "grad_norm": 1.6568444967269897,
+      "grad_norm": 1.7667814493179321,
       "learning_rate": 2.7777777777777776e-07,
-      "loss": 0.0746,
+      "loss": 0.0836,
       "step": 4875
     },
     {
       "epoch": 3.7837837837837838,
-      "grad_norm": 2.122925281524658,
+      "grad_norm": 2.118232488632202,
       "learning_rate": 2.2222222222222224e-07,
-      "loss": 0.0734,
+      "loss": 0.0824,
       "step": 4900
     },
     {
       "epoch": 3.8030888030888033,
-      "grad_norm": 1.5614521503448486,
+      "grad_norm": 1.5285857915878296,
       "learning_rate": 1.6666666666666668e-07,
-      "loss": 0.0737,
+      "loss": 0.0836,
       "step": 4925
     },
     {
       "epoch": 3.8223938223938223,
-      "grad_norm": 3.5003342628479004,
+      "grad_norm": 2.041189670562744,
       "learning_rate": 1.1111111111111112e-07,
-      "loss": 0.0743,
+      "loss": 0.0838,
       "step": 4950
     },
     {
       "epoch": 3.841698841698842,
-      "grad_norm": 1.8813285827636719,
+      "grad_norm": 2.062230110168457,
       "learning_rate": 5.555555555555556e-08,
-      "loss": 0.0712,
+      "loss": 0.0804,
       "step": 4975
     },
     {
       "epoch": 3.861003861003861,
-      "grad_norm": 1.836345911026001,
+      "grad_norm": 1.9036461114883423,
       "learning_rate": 0.0,
-      "loss": 0.0767,
+      "loss": 0.0845,
       "step": 5000
     },
     {
       "epoch": 3.861003861003861,
-      "eval_loss": 0.4226454198360443,
-      "eval_runtime": 2912.2253,
-      "eval_samples_per_second": 2.411,
-      "eval_steps_per_second": 0.151,
-      "eval_wer": 0.2671921259024568,
+      "eval_loss": 0.447899729013443,
+      "eval_runtime": 2978.0577,
+      "eval_samples_per_second": 2.358,
+      "eval_steps_per_second": 0.147,
+      "eval_wer": 0.29472481531011024,
       "step": 5000
     },
     {
       "epoch": 3.861003861003861,
       "step": 5000,
       "total_flos": 5.435487665750016e+20,
-      "train_loss": 0.2543565913915634,
-      "train_runtime": 60445.9813,
-      "train_samples_per_second": 2.647,
-      "train_steps_per_second": 0.083
+      "train_loss": 0.2695653033494949,
+      "train_runtime": 64824.1008,
+      "train_samples_per_second": 2.468,
+      "train_steps_per_second": 0.077
     }
   ],
   "logging_steps": 25,