End of training

README.md

@@ -2,6 +2,8 @@
 license: apache-2.0
 base_model: facebook/wav2vec2-large-xlsr-53
 tags:
+- automatic-speech-recognition
+- DewiBrynJones/banc-trawsgrifiadau-bangor-normalized
 - generated_from_trainer
 metrics:
 - wer
@@ -15,7 +17,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # wav2vec2-xlsr-53-ft-btb-cy
 
-This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on an unknown dataset.
+This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-NORMALIZED - DEFAULT dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.3847
 - Wer: 0.2967

all_results.json

@@ -1,15 +1,15 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 0.9668737053871155,
-    "eval_runtime": 156.2848,
+    "epoch": 3.536067892503536,
+    "eval_loss": 0.3846580684185028,
+    "eval_runtime": 154.4496,
     "eval_samples": 5656,
-    "eval_samples_per_second": 36.19,
-    "eval_steps_per_second": 4.524,
-    "eval_wer": 0.6124682055958152,
-    "total_flos": 4.036432448909298e+19,
-    "train_loss": 1.1022147617942597,
-    "train_runtime": 28048.236,
+    "eval_samples_per_second": 36.62,
+    "eval_steps_per_second": 4.578,
+    "eval_wer": 0.29666778646958136,
+    "total_flos": 9.55129266706546e+18,
+    "train_loss": 1.1546670959472656,
+    "train_runtime": 6557.446,
     "train_samples": 22621,
-    "train_samples_per_second": 12.098,
-    "train_steps_per_second": 0.378
+    "train_samples_per_second": 12.2,
+    "train_steps_per_second": 0.381
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 0.9668737053871155,
-    "eval_runtime": 156.2848,
+    "epoch": 3.536067892503536,
+    "eval_loss": 0.3846580684185028,
+    "eval_runtime": 154.4496,
     "eval_samples": 5656,
-    "eval_samples_per_second": 36.19,
-    "eval_steps_per_second": 4.524,
-    "eval_wer": 0.6124682055958152
+    "eval_samples_per_second": 36.62,
+    "eval_steps_per_second": 4.578,
+    "eval_wer": 0.29666778646958136
 }

runs/May10_18-59-42_09e070d6a7b1/events.out.tfevents.1715370937.09e070d6a7b1.452.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eee0599976dbd99cbdb80f8b30b81cf8b4d0feaf90c1cb5c5e34bf537275fa6a
+size 406

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 15.0,
-    "total_flos": 4.036432448909298e+19,
-    "train_loss": 1.1022147617942597,
-    "train_runtime": 28048.236,
+    "epoch": 3.536067892503536,
+    "total_flos": 9.55129266706546e+18,
+    "train_loss": 1.1546670959472656,
+    "train_runtime": 6557.446,
     "train_samples": 22621,
-    "train_samples_per_second": 12.098,
-    "train_steps_per_second": 0.378
+    "train_samples_per_second": 12.2,
+    "train_steps_per_second": 0.381
 }

trainer_state.json

@@ -1,1130 +1,289 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 15.0,
+  "epoch": 3.536067892503536,
   "eval_steps": 100,
-  "global_step": 10605,
+  "global_step": 2500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.14144271570014144,
-      "eval_loss": 3.7426819801330566,
-      "eval_runtime": 154.1912,
-      "eval_samples_per_second": 36.682,
-      "eval_steps_per_second": 4.585,
+      "eval_loss": 3.7463672161102295,
+      "eval_runtime": 153.5435,
+      "eval_samples_per_second": 36.836,
+      "eval_steps_per_second": 4.605,
       "eval_wer": 1.0,
       "step": 100
     },
     {
       "epoch": 0.2828854314002829,
-      "eval_loss": 2.9179046154022217,
-      "eval_runtime": 149.8513,
-      "eval_samples_per_second": 37.744,
-      "eval_steps_per_second": 4.718,
+      "eval_loss": 2.9399216175079346,
+      "eval_runtime": 150.7703,
+      "eval_samples_per_second": 37.514,
+      "eval_steps_per_second": 4.689,
       "eval_wer": 1.0,
       "step": 200
     },
     {
       "epoch": 0.4243281471004243,
-      "eval_loss": 2.8035507202148438,
-      "eval_runtime": 151.9434,
-      "eval_samples_per_second": 37.224,
-      "eval_steps_per_second": 4.653,
-      "eval_wer": 1.0,
+      "eval_loss": 2.5961458683013916,
+      "eval_runtime": 151.4236,
+      "eval_samples_per_second": 37.352,
+      "eval_steps_per_second": 4.669,
+      "eval_wer": 0.9991041576682503,
       "step": 300
     },
     {
       "epoch": 0.5657708628005658,
-      "eval_loss": 1.2195814847946167,
-      "eval_runtime": 152.8075,
-      "eval_samples_per_second": 37.014,
-      "eval_steps_per_second": 4.627,
-      "eval_wer": 0.8933627681528051,
+      "eval_loss": 1.1618728637695312,
+      "eval_runtime": 152.5862,
+      "eval_samples_per_second": 37.068,
+      "eval_steps_per_second": 4.633,
+      "eval_wer": 0.7905328662155461,
       "step": 400
     },
     {
       "epoch": 0.7072135785007072,
-      "grad_norm": 2.533352851867676,
+      "grad_norm": 1.7047498226165771,
       "learning_rate": 0.00029699999999999996,
-      "loss": 3.574,
+      "loss": 3.5448,
       "step": 500
     },
     {
       "epoch": 0.7072135785007072,
-      "eval_loss": 0.9860211610794067,
-      "eval_runtime": 151.7482,
-      "eval_samples_per_second": 37.272,
-      "eval_steps_per_second": 4.659,
-      "eval_wer": 0.7276159395946313,
+      "eval_loss": 0.946560800075531,
+      "eval_runtime": 153.1427,
+      "eval_samples_per_second": 36.933,
+      "eval_steps_per_second": 4.617,
+      "eval_wer": 0.6897506038937147,
       "step": 500
     },
     {
       "epoch": 0.8486562942008486,
-      "eval_loss": 0.8391533493995667,
-      "eval_runtime": 153.0126,
-      "eval_samples_per_second": 36.964,
-      "eval_steps_per_second": 4.621,
-      "eval_wer": 0.6504295244037049,
+      "eval_loss": 0.7894724607467651,
+      "eval_runtime": 151.7162,
+      "eval_samples_per_second": 37.28,
+      "eval_steps_per_second": 4.66,
+      "eval_wer": 0.6110604533602086,
       "step": 600
     },
     {
       "epoch": 0.9900990099009901,
-      "eval_loss": 0.7803803086280823,
-      "eval_runtime": 152.7414,
-      "eval_samples_per_second": 37.03,
-      "eval_steps_per_second": 4.629,
-      "eval_wer": 0.6029178864519844,
+      "eval_loss": 0.6820164918899536,
+      "eval_runtime": 152.0086,
+      "eval_samples_per_second": 37.208,
+      "eval_steps_per_second": 4.651,
+      "eval_wer": 0.5378893314776599,
       "step": 700
     },
     {
       "epoch": 1.1315417256011315,
-      "eval_loss": 0.6122242212295532,
-      "eval_runtime": 153.2031,
-      "eval_samples_per_second": 36.918,
-      "eval_steps_per_second": 4.615,
-      "eval_wer": 0.4909056006142919,
+      "eval_loss": 0.6039016842842102,
+      "eval_runtime": 152.002,
+      "eval_samples_per_second": 37.21,
+      "eval_steps_per_second": 4.651,
+      "eval_wer": 0.47239685815296506,
       "step": 800
     },
     {
       "epoch": 1.272984441301273,
-      "eval_loss": 0.5901117920875549,
-      "eval_runtime": 152.5961,
-      "eval_samples_per_second": 37.065,
-      "eval_steps_per_second": 4.633,
-      "eval_wer": 0.48828206235702515,
+      "eval_loss": 0.5631398558616638,
+      "eval_runtime": 153.0321,
+      "eval_samples_per_second": 36.96,
+      "eval_steps_per_second": 4.62,
+      "eval_wer": 0.46745372814384667,
       "step": 900
     },
     {
       "epoch": 1.4144271570014144,
-      "grad_norm": 0.9860969185829163,
-      "learning_rate": 0.0002853043047996041,
-      "loss": 0.811,
+      "grad_norm": 0.8790757656097412,
+      "learning_rate": 0.00022574999999999996,
+      "loss": 0.7808,
       "step": 1000
     },
     {
       "epoch": 1.4144271570014144,
-      "eval_loss": 0.5500049591064453,
-      "eval_runtime": 153.3918,
-      "eval_samples_per_second": 36.873,
-      "eval_steps_per_second": 4.609,
-      "eval_wer": 0.45078466189950567,
+      "eval_loss": 0.5279428958892822,
+      "eval_runtime": 152.9768,
+      "eval_samples_per_second": 36.973,
+      "eval_steps_per_second": 4.622,
+      "eval_wer": 0.4291084769080642,
       "step": 1000
     },
     {
       "epoch": 1.5558698727015559,
-      "eval_loss": 0.5231846570968628,
-      "eval_runtime": 151.7862,
-      "eval_samples_per_second": 37.263,
-      "eval_steps_per_second": 4.658,
-      "eval_wer": 0.41418310377373585,
+      "eval_loss": 0.5024306178092957,
+      "eval_runtime": 152.8927,
+      "eval_samples_per_second": 36.993,
+      "eval_steps_per_second": 4.624,
+      "eval_wer": 0.39940170529986724,
       "step": 1100
     },
     {
       "epoch": 1.6973125884016973,
-      "eval_loss": 0.518621027469635,
-      "eval_runtime": 152.9089,
-      "eval_samples_per_second": 36.989,
-      "eval_steps_per_second": 4.624,
-      "eval_wer": 0.4064724608468909,
+      "eval_loss": 0.4894837439060211,
+      "eval_runtime": 153.3103,
+      "eval_samples_per_second": 36.892,
+      "eval_steps_per_second": 4.612,
+      "eval_wer": 0.3894514565436483,
       "step": 1200
     },
     {
       "epoch": 1.8387553041018387,
-      "eval_loss": 0.49534252285957336,
-      "eval_runtime": 153.7928,
-      "eval_samples_per_second": 36.777,
-      "eval_steps_per_second": 4.597,
-      "eval_wer": 0.3929388427636736,
+      "eval_loss": 0.4595918357372284,
+      "eval_runtime": 153.0651,
+      "eval_samples_per_second": 36.952,
+      "eval_steps_per_second": 4.619,
+      "eval_wer": 0.3695829534002016,
       "step": 1300
     },
     {
       "epoch": 1.9801980198019802,
-      "eval_loss": 0.48800522089004517,
-      "eval_runtime": 152.9541,
-      "eval_samples_per_second": 36.978,
-      "eval_steps_per_second": 4.622,
-      "eval_wer": 0.39279486810321385,
+      "eval_loss": 0.44729524850845337,
+      "eval_runtime": 154.1998,
+      "eval_samples_per_second": 36.68,
+      "eval_steps_per_second": 4.585,
+      "eval_wer": 0.3610884484330758,
       "step": 1400
     },
     {
       "epoch": 2.1216407355021216,
-      "grad_norm": 0.8855465650558472,
-      "learning_rate": 0.00027054923305294405,
-      "loss": 0.6459,
+      "grad_norm": 0.8205087184906006,
+      "learning_rate": 0.0001512,
+      "loss": 0.6005,
       "step": 1500
     },
     {
       "epoch": 2.1216407355021216,
-      "eval_loss": 0.46446114778518677,
-      "eval_runtime": 152.609,
-      "eval_samples_per_second": 37.062,
-      "eval_steps_per_second": 4.633,
-      "eval_wer": 0.3691990209723089,
+      "eval_loss": 0.43324384093284607,
+      "eval_runtime": 150.9169,
+      "eval_samples_per_second": 37.478,
+      "eval_steps_per_second": 4.685,
+      "eval_wer": 0.3474268528738942,
       "step": 1500
     },
     {
       "epoch": 2.263083451202263,
-      "eval_loss": 0.4666256904602051,
-      "eval_runtime": 153.127,
-      "eval_samples_per_second": 36.937,
-      "eval_steps_per_second": 4.617,
-      "eval_wer": 0.3585928876517733,
+      "eval_loss": 0.4268616735935211,
+      "eval_runtime": 152.3911,
+      "eval_samples_per_second": 37.115,
+      "eval_steps_per_second": 4.639,
+      "eval_wer": 0.3418118411159636,
       "step": 1600
     },
     {
       "epoch": 2.4045261669024045,
-      "eval_loss": 0.45017901062965393,
-      "eval_runtime": 154.5149,
-      "eval_samples_per_second": 36.605,
-      "eval_steps_per_second": 4.576,
-      "eval_wer": 0.35934475532306315,
+      "eval_loss": 0.4155045449733734,
+      "eval_runtime": 153.4832,
+      "eval_samples_per_second": 36.851,
+      "eval_steps_per_second": 4.606,
+      "eval_wer": 0.33606885188206875,
       "step": 1700
     },
     {
       "epoch": 2.545968882602546,
-      "eval_loss": 0.45280084013938904,
-      "eval_runtime": 152.9066,
-      "eval_samples_per_second": 36.99,
-      "eval_steps_per_second": 4.624,
-      "eval_wer": 0.3637759754283246,
+      "eval_loss": 0.4121190905570984,
+      "eval_runtime": 153.7529,
+      "eval_samples_per_second": 36.786,
+      "eval_steps_per_second": 4.598,
+      "eval_wer": 0.32143142806865993,
       "step": 1800
     },
     {
       "epoch": 2.6874115983026874,
-      "eval_loss": 0.46647289395332336,
-      "eval_runtime": 153.103,
-      "eval_samples_per_second": 36.942,
-      "eval_steps_per_second": 4.618,
-      "eval_wer": 0.392618899073763,
+      "eval_loss": 0.4145391285419464,
+      "eval_runtime": 159.3319,
+      "eval_samples_per_second": 35.498,
+      "eval_steps_per_second": 4.437,
+      "eval_wer": 0.3366447505239078,
       "step": 1900
     },
     {
       "epoch": 2.828854314002829,
-      "grad_norm": 0.6227492690086365,
-      "learning_rate": 0.00025579416130628403,
-      "loss": 0.5306,
+      "grad_norm": 0.8615767359733582,
+      "learning_rate": 7.664999999999999e-05,
+      "loss": 0.4666,
       "step": 2000
     },
     {
       "epoch": 2.828854314002829,
-      "eval_loss": 0.4328605532646179,
-      "eval_runtime": 153.5492,
-      "eval_samples_per_second": 36.835,
-      "eval_steps_per_second": 4.604,
-      "eval_wer": 0.3505143094815312,
+      "eval_loss": 0.39387884736061096,
+      "eval_runtime": 153.1343,
+      "eval_samples_per_second": 36.935,
+      "eval_steps_per_second": 4.617,
+      "eval_wer": 0.3114171905744589,
       "step": 2000
     },
     {
       "epoch": 2.9702970297029703,
-      "eval_loss": 0.4245360791683197,
-      "eval_runtime": 153.9467,
-      "eval_samples_per_second": 36.74,
-      "eval_steps_per_second": 4.592,
-      "eval_wer": 0.3373806210107021,
+      "eval_loss": 0.38894009590148926,
+      "eval_runtime": 152.8789,
+      "eval_samples_per_second": 36.997,
+      "eval_steps_per_second": 4.625,
+      "eval_wer": 0.30807377901489336,
       "step": 2100
     },
     {
       "epoch": 3.1117397454031117,
-      "eval_loss": 0.4376748204231262,
-      "eval_runtime": 152.9335,
-      "eval_samples_per_second": 36.983,
-      "eval_steps_per_second": 4.623,
-      "eval_wer": 0.3340372094511366,
+      "eval_loss": 0.3909347653388977,
+      "eval_runtime": 154.4919,
+      "eval_samples_per_second": 36.61,
+      "eval_steps_per_second": 4.576,
+      "eval_wer": 0.30644206619634945,
       "step": 2200
     },
     {
       "epoch": 3.253182461103253,
-      "eval_loss": 0.4271674156188965,
-      "eval_runtime": 153.1738,
-      "eval_samples_per_second": 36.925,
-      "eval_steps_per_second": 4.616,
-      "eval_wer": 0.33373326294572153,
+      "eval_loss": 0.3874327838420868,
+      "eval_runtime": 153.3025,
+      "eval_samples_per_second": 36.894,
+      "eval_steps_per_second": 4.612,
+      "eval_wer": 0.3015469277407176,
       "step": 2300
     },
     {
       "epoch": 3.3946251768033946,
-      "eval_loss": 0.43350949883461,
-      "eval_runtime": 153.2683,
-      "eval_samples_per_second": 36.903,
-      "eval_steps_per_second": 4.613,
-      "eval_wer": 0.332597462846539,
+      "eval_loss": 0.386868953704834,
+      "eval_runtime": 153.0443,
+      "eval_samples_per_second": 36.957,
+      "eval_steps_per_second": 4.62,
+      "eval_wer": 0.29833149365711636,
       "step": 2400
     },
     {
       "epoch": 3.536067892503536,
-      "grad_norm": 0.6211841106414795,
-      "learning_rate": 0.0002410094012864918,
-      "loss": 0.4628,
+      "grad_norm": 0.6678842902183533,
+      "learning_rate": 1.9499999999999995e-06,
+      "loss": 0.3805,
       "step": 2500
     },
     {
       "epoch": 3.536067892503536,
-      "eval_loss": 0.42679545283317566,
-      "eval_runtime": 154.4777,
-      "eval_samples_per_second": 36.614,
-      "eval_steps_per_second": 4.577,
-      "eval_wer": 0.3274783638079698,
-      "step": 2500
-    },
-    {
-      "epoch": 3.6775106082036775,
-      "eval_loss": 0.4502430558204651,
-      "eval_runtime": 154.6864,
-      "eval_samples_per_second": 36.564,
-      "eval_steps_per_second": 4.571,
-      "eval_wer": 0.34091599878421397,
-      "step": 2600
-    },
-    {
-      "epoch": 3.818953323903819,
-      "eval_loss": 0.6344878673553467,
-      "eval_runtime": 153.1131,
-      "eval_samples_per_second": 36.94,
-      "eval_steps_per_second": 4.618,
-      "eval_wer": 0.43904272847978754,
-      "step": 2700
-    },
-    {
-      "epoch": 3.9603960396039604,
-      "eval_loss": 1.0202795267105103,
-      "eval_runtime": 153.5324,
-      "eval_samples_per_second": 36.839,
-      "eval_steps_per_second": 4.605,
-      "eval_wer": 0.6403193038025308,
-      "step": 2800
-    },
-    {
-      "epoch": 4.101838755304102,
-      "eval_loss": 1.2207801342010498,
-      "eval_runtime": 154.3697,
-      "eval_samples_per_second": 36.639,
-      "eval_steps_per_second": 4.58,
-      "eval_wer": 0.7921805762185855,
-      "step": 2900
-    },
-    {
-      "epoch": 4.243281471004243,
-      "grad_norm": 2.4671809673309326,
-      "learning_rate": 0.00022619495299356753,
-      "loss": 0.8685,
-      "step": 3000
-    },
-    {
-      "epoch": 4.243281471004243,
-      "eval_loss": 1.101838231086731,
-      "eval_runtime": 155.2474,
-      "eval_samples_per_second": 36.432,
-      "eval_steps_per_second": 4.554,
-      "eval_wer": 0.7387019884500328,
-      "step": 3000
-    },
-    {
-      "epoch": 4.384724186704385,
-      "eval_loss": 1.24972665309906,
-      "eval_runtime": 154.4564,
-      "eval_samples_per_second": 36.619,
-      "eval_steps_per_second": 4.577,
-      "eval_wer": 0.8061941098366687,
-      "step": 3100
-    },
-    {
-      "epoch": 4.526166902404526,
-      "eval_loss": 1.6164859533309937,
-      "eval_runtime": 153.9159,
-      "eval_samples_per_second": 36.747,
-      "eval_steps_per_second": 4.593,
-      "eval_wer": 0.9616227543952265,
-      "step": 3200
-    },
-    {
-      "epoch": 4.667609618104668,
-      "eval_loss": 1.4655081033706665,
-      "eval_runtime": 153.645,
-      "eval_samples_per_second": 36.812,
-      "eval_steps_per_second": 4.602,
-      "eval_wer": 0.9216617875253955,
-      "step": 3300
-    },
-    {
-      "epoch": 4.809052333804809,
-      "eval_loss": 1.0288450717926025,
-      "eval_runtime": 152.6496,
-      "eval_samples_per_second": 37.052,
-      "eval_steps_per_second": 4.632,
-      "eval_wer": 0.7464606229303643,
-      "step": 3400
-    },
-    {
-      "epoch": 4.9504950495049505,
-      "grad_norm": 2.009023666381836,
-      "learning_rate": 0.00021143988124690746,
-      "loss": 1.3918,
-      "step": 3500
-    },
-    {
-      "epoch": 4.9504950495049505,
-      "eval_loss": 0.9067263603210449,
-      "eval_runtime": 156.1428,
-      "eval_samples_per_second": 36.223,
-      "eval_steps_per_second": 4.528,
-      "eval_wer": 0.5948393082817424,
-      "step": 3500
-    },
-    {
-      "epoch": 5.091937765205092,
-      "eval_loss": 0.9486163258552551,
-      "eval_runtime": 154.3538,
-      "eval_samples_per_second": 36.643,
-      "eval_steps_per_second": 4.58,
-      "eval_wer": 0.6352801906864392,
-      "step": 3600
-    },
-    {
-      "epoch": 5.233380480905233,
-      "eval_loss": 0.8674383163452148,
-      "eval_runtime": 155.8543,
-      "eval_samples_per_second": 36.29,
-      "eval_steps_per_second": 4.536,
-      "eval_wer": 0.5428324614867783,
-      "step": 3700
-    },
-    {
-      "epoch": 5.374823196605375,
-      "eval_loss": 0.9402504563331604,
-      "eval_runtime": 154.1915,
-      "eval_samples_per_second": 36.682,
-      "eval_steps_per_second": 4.585,
-      "eval_wer": 0.5792900449520885,
-      "step": 3800
-    },
-    {
-      "epoch": 5.516265912305516,
-      "eval_loss": 0.948098361492157,
-      "eval_runtime": 159.4305,
-      "eval_samples_per_second": 35.476,
-      "eval_steps_per_second": 4.435,
-      "eval_wer": 0.5763625601894067,
-      "step": 3900
-    },
-    {
-      "epoch": 5.657708628005658,
-      "grad_norm": 1.2370288372039795,
-      "learning_rate": 0.0001966848095002474,
-      "loss": 1.0402,
-      "step": 4000
-    },
-    {
-      "epoch": 5.657708628005658,
-      "eval_loss": 1.0175639390945435,
-      "eval_runtime": 154.2179,
-      "eval_samples_per_second": 36.675,
-      "eval_steps_per_second": 4.584,
-      "eval_wer": 0.8256946777367183,
-      "step": 4000
-    },
-    {
-      "epoch": 5.799151343705799,
-      "eval_loss": 0.9857003092765808,
-      "eval_runtime": 154.4099,
-      "eval_samples_per_second": 36.63,
-      "eval_steps_per_second": 4.579,
-      "eval_wer": 0.6342883652477164,
-      "step": 4100
-    },
-    {
-      "epoch": 5.9405940594059405,
-      "eval_loss": 1.3289028406143188,
-      "eval_runtime": 155.7394,
-      "eval_samples_per_second": 36.317,
-      "eval_steps_per_second": 4.54,
-      "eval_wer": 0.9014093519540561,
-      "step": 4200
-    },
-    {
-      "epoch": 6.082036775106082,
-      "eval_loss": 2.0890820026397705,
-      "eval_runtime": 154.2749,
-      "eval_samples_per_second": 36.662,
-      "eval_steps_per_second": 4.583,
-      "eval_wer": 0.7125305946153477,
-      "step": 4300
-    },
-    {
-      "epoch": 6.223479490806223,
-      "eval_loss": 1.256324291229248,
-      "eval_runtime": 154.5375,
-      "eval_samples_per_second": 36.6,
-      "eval_steps_per_second": 4.575,
-      "eval_wer": 0.7696085488953944,
-      "step": 4400
-    },
-    {
-      "epoch": 6.364922206506365,
-      "grad_norm": 0.9906980395317078,
-      "learning_rate": 0.0001819000494804552,
-      "loss": 1.2886,
-      "step": 4500
-    },
-    {
-      "epoch": 6.364922206506365,
-      "eval_loss": 1.1441457271575928,
-      "eval_runtime": 155.0449,
-      "eval_samples_per_second": 36.48,
-      "eval_steps_per_second": 4.56,
-      "eval_wer": 0.692726080209883,
-      "step": 4500
-    },
-    {
-      "epoch": 6.506364922206506,
-      "eval_loss": 1.0626095533370972,
-      "eval_runtime": 155.5445,
-      "eval_samples_per_second": 36.363,
-      "eval_steps_per_second": 4.545,
-      "eval_wer": 0.6573083137367823,
-      "step": 4600
-    },
-    {
-      "epoch": 6.647807637906648,
-      "eval_loss": 0.9997339248657227,
-      "eval_runtime": 155.8362,
-      "eval_samples_per_second": 36.295,
-      "eval_steps_per_second": 4.537,
-      "eval_wer": 0.6422869574954808,
-      "step": 4700
-    },
-    {
-      "epoch": 6.789250353606789,
-      "eval_loss": 0.9813728928565979,
-      "eval_runtime": 155.2999,
-      "eval_samples_per_second": 36.42,
-      "eval_steps_per_second": 4.552,
-      "eval_wer": 0.6380317064196701,
-      "step": 4800
-    },
-    {
-      "epoch": 6.930693069306931,
-      "eval_loss": 1.0955251455307007,
-      "eval_runtime": 153.746,
-      "eval_samples_per_second": 36.788,
-      "eval_steps_per_second": 4.598,
-      "eval_wer": 0.7651133400521508,
-      "step": 4900
-    },
-    {
-      "epoch": 7.072135785007072,
-      "grad_norm": 1.6002442836761475,
-      "learning_rate": 0.00016708560118753091,
-      "loss": 1.0984,
-      "step": 5000
-    },
-    {
-      "epoch": 7.072135785007072,
-      "eval_loss": 0.9212619066238403,
-      "eval_runtime": 155.7654,
-      "eval_samples_per_second": 36.311,
-      "eval_steps_per_second": 4.539,
-      "eval_wer": 0.5882964598230711,
-      "step": 5000
-    },
-    {
-      "epoch": 7.2135785007072135,
-      "eval_loss": 0.8884870409965515,
-      "eval_runtime": 153.6804,
-      "eval_samples_per_second": 36.804,
-      "eval_steps_per_second": 4.6,
-      "eval_wer": 0.5932715842011806,
-      "step": 5100
-    },
-    {
-      "epoch": 7.355021216407355,
-      "eval_loss": 0.900116503238678,
-      "eval_runtime": 155.054,
-      "eval_samples_per_second": 36.478,
-      "eval_steps_per_second": 4.56,
-      "eval_wer": 0.5898641839036329,
-      "step": 5200
-    },
-    {
-      "epoch": 7.496463932107496,
-      "eval_loss": 0.8783684372901917,
-      "eval_runtime": 156.2294,
-      "eval_samples_per_second": 36.203,
-      "eval_steps_per_second": 4.525,
-      "eval_wer": 0.5858968821487418,
-      "step": 5300
-    },
-    {
-      "epoch": 7.637906647807638,
-      "eval_loss": 0.9072028398513794,
-      "eval_runtime": 154.0671,
-      "eval_samples_per_second": 36.711,
-      "eval_steps_per_second": 4.589,
-      "eval_wer": 0.5897522036121643,
-      "step": 5400
-    },
-    {
-      "epoch": 7.779349363507779,
-      "grad_norm": 1.3490877151489258,
-      "learning_rate": 0.00015230084116773872,
-      "loss": 0.9659,
-      "step": 5500
-    },
-    {
-      "epoch": 7.779349363507779,
-      "eval_loss": 0.8811922669410706,
-      "eval_runtime": 153.8934,
-      "eval_samples_per_second": 36.753,
-      "eval_steps_per_second": 4.594,
-      "eval_wer": 0.5841051974852426,
-      "step": 5500
-    },
-    {
-      "epoch": 7.920792079207921,
-      "eval_loss": 0.891165018081665,
-      "eval_runtime": 153.7325,
-      "eval_samples_per_second": 36.791,
-      "eval_steps_per_second": 4.599,
-      "eval_wer": 0.5855129497208491,
-      "step": 5600
-    },
-    {
-      "epoch": 8.062234794908063,
-      "eval_loss": 0.8815582990646362,
-      "eval_runtime": 154.6799,
-      "eval_samples_per_second": 36.566,
-      "eval_steps_per_second": 4.571,
-      "eval_wer": 0.5807137943721905,
-      "step": 5700
-    },
-    {
-      "epoch": 8.203677510608204,
-      "eval_loss": 0.891440749168396,
-      "eval_runtime": 155.4972,
-      "eval_samples_per_second": 36.374,
-      "eval_steps_per_second": 4.547,
-      "eval_wer": 0.5803138647598023,
-      "step": 5800
-    },
-    {
-      "epoch": 8.345120226308346,
-      "eval_loss": 0.8956438899040222,
-      "eval_runtime": 154.6865,
-      "eval_samples_per_second": 36.564,
-      "eval_steps_per_second": 4.571,
-      "eval_wer": 0.5810337380621011,
-      "step": 5900
-    },
-    {
-      "epoch": 8.486562942008486,
-      "grad_norm": 1.257432460784912,
-      "learning_rate": 0.00013754576942107867,
-      "loss": 0.9679,
-      "step": 6000
-    },
-    {
-      "epoch": 8.486562942008486,
-      "eval_loss": 0.9162164330482483,
-      "eval_runtime": 155.1949,
-      "eval_samples_per_second": 36.444,
-      "eval_steps_per_second": 4.556,
-      "eval_wer": 0.5780262673769416,
-      "step": 6000
-    },
-    {
-      "epoch": 8.628005657708629,
-      "eval_loss": 0.9409377574920654,
-      "eval_runtime": 154.6732,
-      "eval_samples_per_second": 36.567,
-      "eval_steps_per_second": 4.571,
-      "eval_wer": 0.5810177408776055,
-      "step": 6100
-    },
-    {
-      "epoch": 8.76944837340877,
-      "eval_loss": 0.9370973706245422,
-      "eval_runtime": 155.5326,
-      "eval_samples_per_second": 36.365,
-      "eval_steps_per_second": 4.546,
-      "eval_wer": 0.5780742589304283,
-      "step": 6200
-    },
-    {
-      "epoch": 8.910891089108912,
-      "eval_loss": 0.941677987575531,
-      "eval_runtime": 155.4822,
-      "eval_samples_per_second": 36.377,
-      "eval_steps_per_second": 4.547,
-      "eval_wer": 0.5790020956311689,
-      "step": 6300
-    },
-    {
-      "epoch": 9.052333804809052,
-      "eval_loss": 0.9663541913032532,
-      "eval_runtime": 155.844,
-      "eval_samples_per_second": 36.293,
-      "eval_steps_per_second": 4.537,
-      "eval_wer": 0.5783782054358433,
-      "step": 6400
-    },
-    {
-      "epoch": 9.193776520509195,
-      "grad_norm": 1.5867938995361328,
-      "learning_rate": 0.00012273132112815437,
-      "loss": 1.0241,
-      "step": 6500
-    },
-    {
-      "epoch": 9.193776520509195,
-      "eval_loss": 0.9720383286476135,
-      "eval_runtime": 155.604,
-      "eval_samples_per_second": 36.349,
-      "eval_steps_per_second": 4.544,
-      "eval_wer": 0.5775143574730848,
-      "step": 6500
-    },
-    {
-      "epoch": 9.335219236209335,
-      "eval_loss": 0.9840742349624634,
-      "eval_runtime": 155.8175,
-      "eval_samples_per_second": 36.299,
-      "eval_steps_per_second": 4.537,
-      "eval_wer": 0.5783942026203388,
-      "step": 6600
-    },
-    {
-      "epoch": 9.476661951909477,
-      "eval_loss": 0.9574136137962341,
-      "eval_runtime": 155.2609,
-      "eval_samples_per_second": 36.429,
-      "eval_steps_per_second": 4.554,
-      "eval_wer": 0.5886803922509638,
-      "step": 6700
-    },
-    {
-      "epoch": 9.618104667609618,
-      "eval_loss": 1.0725222826004028,
-      "eval_runtime": 154.2708,
-      "eval_samples_per_second": 36.663,
-      "eval_steps_per_second": 4.583,
-      "eval_wer": 0.606837196653389,
-      "step": 6800
-    },
-    {
-      "epoch": 9.75954738330976,
-      "eval_loss": 1.0362112522125244,
-      "eval_runtime": 155.3381,
-      "eval_samples_per_second": 36.411,
-      "eval_steps_per_second": 4.551,
-      "eval_wer": 0.5999584073203116,
-      "step": 6900
-    },
-    {
-      "epoch": 9.900990099009901,
-      "grad_norm": 0.6058325171470642,
-      "learning_rate": 0.00010794656110836219,
-      "loss": 1.0797,
-      "step": 7000
-    },
-    {
-      "epoch": 9.900990099009901,
-      "eval_loss": 1.0116764307022095,
-      "eval_runtime": 155.5442,
-      "eval_samples_per_second": 36.363,
-      "eval_steps_per_second": 4.545,
-      "eval_wer": 0.5914319079841948,
-      "step": 7000
-    },
-    {
-      "epoch": 10.042432814710043,
-      "eval_loss": 0.9563263058662415,
-      "eval_runtime": 155.5047,
-      "eval_samples_per_second": 36.372,
-      "eval_steps_per_second": 4.546,
-      "eval_wer": 0.6058293740301707,
-      "step": 7100
-    },
-    {
-      "epoch": 10.183875530410184,
-      "eval_loss": 0.9663692116737366,
-      "eval_runtime": 155.3578,
-      "eval_samples_per_second": 36.406,
-      "eval_steps_per_second": 4.551,
-      "eval_wer": 0.5978307817824063,
-      "step": 7200
-    },
-    {
-      "epoch": 10.325318246110326,
-      "eval_loss": 1.0209406614303589,
-      "eval_runtime": 155.4685,
-      "eval_samples_per_second": 36.38,
-      "eval_steps_per_second": 4.548,
-      "eval_wer": 0.6022140103341812,
-      "step": 7300
-    },
-    {
-      "epoch": 10.466760961810467,
-      "eval_loss": 0.9848981499671936,
-      "eval_runtime": 156.2008,
-      "eval_samples_per_second": 36.21,
-      "eval_steps_per_second": 4.526,
-      "eval_wer": 0.5974788437235047,
-      "step": 7400
-    },
-    {
-      "epoch": 10.608203677510609,
-      "grad_norm": 0.0,
-      "learning_rate": 9.319148936170212e-05,
-      "loss": 1.0701,
-      "step": 7500
-    },
-    {
-      "epoch": 10.608203677510609,
-      "eval_loss": 0.9718888401985168,
-      "eval_runtime": 156.5494,
-      "eval_samples_per_second": 36.129,
-      "eval_steps_per_second": 4.516,
-      "eval_wer": 0.6057013965542064,
-      "step": 7500
-    },
-    {
-      "epoch": 10.74964639321075,
-      "eval_loss": 0.966968834400177,
-      "eval_runtime": 155.045,
-      "eval_samples_per_second": 36.48,
-      "eval_steps_per_second": 4.56,
-      "eval_wer": 0.6122602421973733,
-      "step": 7600
-    },
-    {
-      "epoch": 10.891089108910892,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.2212,
-      "eval_samples_per_second": 36.205,
-      "eval_steps_per_second": 4.526,
-      "eval_wer": 0.6124682055958152,
-      "step": 7700
-    },
-    {
-      "epoch": 11.032531824611032,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.7045,
-      "eval_samples_per_second": 36.093,
-      "eval_steps_per_second": 4.512,
-      "eval_wer": 0.6124682055958152,
-      "step": 7800
-    },
-    {
-      "epoch": 11.173974540311175,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.3156,
-      "eval_samples_per_second": 36.183,
-      "eval_steps_per_second": 4.523,
-      "eval_wer": 0.6124682055958152,
-      "step": 7900
-    },
-    {
-      "epoch": 11.315417256011315,
-      "grad_norm": 0.0,
-      "learning_rate": 7.843641761504205e-05,
-      "loss": 1.0518,
-      "step": 8000
-    },
-    {
-      "epoch": 11.315417256011315,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.4393,
-      "eval_samples_per_second": 36.155,
-      "eval_steps_per_second": 4.519,
-      "eval_wer": 0.6124682055958152,
-      "step": 8000
-    },
-    {
-      "epoch": 11.456859971711458,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.12,
-      "eval_samples_per_second": 36.462,
-      "eval_steps_per_second": 4.558,
-      "eval_wer": 0.6124682055958152,
-      "step": 8100
-    },
-    {
-      "epoch": 11.598302687411598,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.4065,
-      "eval_samples_per_second": 36.395,
-      "eval_steps_per_second": 4.549,
-      "eval_wer": 0.6124682055958152,
-      "step": 8200
-    },
-    {
-      "epoch": 11.73974540311174,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.2056,
-      "eval_samples_per_second": 36.678,
-      "eval_steps_per_second": 4.585,
-      "eval_wer": 0.6124682055958152,
-      "step": 8300
-    },
-    {
-      "epoch": 11.881188118811881,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.9204,
-      "eval_samples_per_second": 36.275,
-      "eval_steps_per_second": 4.534,
-      "eval_wer": 0.6124682055958152,
-      "step": 8400
-    },
-    {
-      "epoch": 12.022630834512023,
-      "grad_norm": 0.0,
-      "learning_rate": 6.362196932211776e-05,
-      "loss": 1.0594,
-      "step": 8500
-    },
-    {
-      "epoch": 12.022630834512023,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.4966,
-      "eval_samples_per_second": 36.609,
-      "eval_steps_per_second": 4.576,
-      "eval_wer": 0.6124682055958152,
-      "step": 8500
-    },
-    {
-      "epoch": 12.164073550212164,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.0012,
-      "eval_samples_per_second": 36.727,
-      "eval_steps_per_second": 4.591,
-      "eval_wer": 0.6124682055958152,
-      "step": 8600
-    },
-    {
-      "epoch": 12.305516265912306,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.297,
-      "eval_samples_per_second": 36.421,
-      "eval_steps_per_second": 4.553,
-      "eval_wer": 0.6124682055958152,
-      "step": 8700
-    },
-    {
-      "epoch": 12.446958981612447,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.7077,
-      "eval_samples_per_second": 36.093,
-      "eval_steps_per_second": 4.512,
-      "eval_wer": 0.6124682055958152,
-      "step": 8800
-    },
-    {
-      "epoch": 12.58840169731259,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.8279,
-      "eval_samples_per_second": 36.296,
-      "eval_steps_per_second": 4.537,
-      "eval_wer": 0.6124682055958152,
-      "step": 8900
-    },
-    {
-      "epoch": 12.72984441301273,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8866897575457694e-05,
-      "loss": 1.0584,
-      "step": 9000
-    },
-    {
-      "epoch": 12.72984441301273,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.3599,
-      "eval_samples_per_second": 36.406,
-      "eval_steps_per_second": 4.551,
-      "eval_wer": 0.6124682055958152,
-      "step": 9000
-    },
-    {
-      "epoch": 12.871287128712872,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.3076,
-      "eval_samples_per_second": 36.418,
-      "eval_steps_per_second": 4.552,
-      "eval_wer": 0.6124682055958152,
-      "step": 9100
-    },
-    {
-      "epoch": 13.012729844413013,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.448,
-      "eval_samples_per_second": 36.385,
-      "eval_steps_per_second": 4.548,
-      "eval_wer": 0.6124682055958152,
-      "step": 9200
-    },
-    {
-      "epoch": 13.154172560113155,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.6022,
-      "eval_samples_per_second": 36.349,
-      "eval_steps_per_second": 4.544,
-      "eval_wer": 0.6124682055958152,
-      "step": 9300
-    },
-    {
-      "epoch": 13.295615275813295,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.7567,
-      "eval_samples_per_second": 36.548,
-      "eval_steps_per_second": 4.568,
-      "eval_wer": 0.6124682055958152,
-      "step": 9400
-    },
-    {
-      "epoch": 13.437057991513438,
-      "grad_norm": 0.0,
-      "learning_rate": 3.408213755566551e-05,
-      "loss": 1.0556,
-      "step": 9500
-    },
-    {
-      "epoch": 13.437057991513438,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.3187,
-      "eval_samples_per_second": 36.651,
-      "eval_steps_per_second": 4.581,
-      "eval_wer": 0.6124682055958152,
-      "step": 9500
-    },
-    {
-      "epoch": 13.578500707213578,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.5249,
-      "eval_samples_per_second": 36.367,
-      "eval_steps_per_second": 4.546,
-      "eval_wer": 0.6124682055958152,
-      "step": 9600
-    },
-    {
-      "epoch": 13.71994342291372,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.212,
-      "eval_samples_per_second": 36.44,
-      "eval_steps_per_second": 4.555,
-      "eval_wer": 0.6124682055958152,
-      "step": 9700
-    },
-    {
-      "epoch": 13.861386138613861,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.6321,
+      "eval_loss": 0.3846580684185028,
+      "eval_runtime": 154.6331,
       "eval_samples_per_second": 36.577,
       "eval_steps_per_second": 4.572,
-      "eval_wer": 0.6124682055958152,
-      "step": 9800
-    },
-    {
-      "epoch": 14.002828854314004,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.2988,
-      "eval_samples_per_second": 36.42,
-      "eval_steps_per_second": 4.553,
-      "eval_wer": 0.6124682055958152,
-      "step": 9900
-    },
-    {
-      "epoch": 14.144271570014144,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9297377535873327e-05,
-      "loss": 1.0511,
-      "step": 10000
-    },
-    {
-      "epoch": 14.144271570014144,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.9222,
-      "eval_samples_per_second": 36.274,
-      "eval_steps_per_second": 4.534,
-      "eval_wer": 0.6124682055958152,
-      "step": 10000
-    },
-    {
-      "epoch": 14.285714285714286,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 156.6698,
-      "eval_samples_per_second": 36.101,
-      "eval_steps_per_second": 4.513,
-      "eval_wer": 0.6124682055958152,
-      "step": 10100
-    },
-    {
-      "epoch": 14.427157001414427,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.8474,
-      "eval_samples_per_second": 36.292,
-      "eval_steps_per_second": 4.536,
-      "eval_wer": 0.6124682055958152,
-      "step": 10200
-    },
-    {
-      "epoch": 14.56859971711457,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.9914,
-      "eval_samples_per_second": 36.492,
-      "eval_steps_per_second": 4.562,
-      "eval_wer": 0.6124682055958152,
-      "step": 10300
-    },
-    {
-      "epoch": 14.71004243281471,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 155.7638,
-      "eval_samples_per_second": 36.311,
-      "eval_steps_per_second": 4.539,
-      "eval_wer": 0.6124682055958152,
-      "step": 10400
-    },
-    {
-      "epoch": 14.851485148514852,
-      "grad_norm": 0.0,
-      "learning_rate": 4.512617516081148e-06,
-      "loss": 1.0585,
-      "step": 10500
-    },
-    {
-      "epoch": 14.851485148514852,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.954,
-      "eval_samples_per_second": 36.501,
-      "eval_steps_per_second": 4.563,
-      "eval_wer": 0.6124682055958152,
-      "step": 10500
-    },
-    {
-      "epoch": 14.992927864214993,
-      "eval_loss": 0.9668737053871155,
-      "eval_runtime": 154.7299,
-      "eval_samples_per_second": 36.554,
-      "eval_steps_per_second": 4.569,
-      "eval_wer": 0.6124682055958152,
-      "step": 10600
+      "eval_wer": 0.29666778646958136,
+      "step": 2500
     },
     {
-      "epoch": 15.0,
-      "step": 10605,
-      "total_flos": 4.036432448909298e+19,
-      "train_loss": 1.1022147617942597,
-      "train_runtime": 28048.236,
-      "train_samples_per_second": 12.098,
-      "train_steps_per_second": 0.378
+      "epoch": 3.536067892503536,
+      "step": 2500,
+      "total_flos": 9.55129266706546e+18,
+      "train_loss": 1.1546670959472656,
+      "train_runtime": 6557.446,
+      "train_samples_per_second": 12.2,
+      "train_steps_per_second": 0.381
     }
   ],
   "logging_steps": 500,
-  "max_steps": 10605,
+  "max_steps": 2500,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 15,
+  "num_train_epochs": 4,
   "save_steps": 400,
-  "total_flos": 4.036432448909298e+19,
+  "total_flos": 9.55129266706546e+18,
   "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null