End of training

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 100.0,
-    "eval_loss": 0.5438345074653625,
-    "eval_runtime": 140.268,
-    "eval_samples": 5076,
-    "eval_samples_per_second": 36.188,
-    "eval_steps_per_second": 2.267,
-    "eval_wer": 0.177349387392344,
-    "train_loss": 0.8928292760036721,
-    "train_runtime": 80759.6589,
-    "train_samples": 20306,
-    "train_samples_per_second": 25.144,
-    "train_steps_per_second": 0.393
+    "epoch": 60.0,
+    "eval_loss": 0.4502493739128113,
+    "eval_runtime": 418.5824,
+    "eval_samples": 20246,
+    "eval_samples_per_second": 48.368,
+    "eval_steps_per_second": 3.024,
+    "eval_wer": 0.23182961772311134,
+    "train_loss": 1.0617500136590194,
+    "train_runtime": 191647.3893,
+    "train_samples": 90777,
+    "train_samples_per_second": 28.42,
+    "train_steps_per_second": 0.444
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 100.0,
-    "train_loss": 0.8928292760036721,
-    "train_runtime": 80759.6589,
-    "train_samples": 20306,
-    "train_samples_per_second": 25.144,
-    "train_steps_per_second": 0.393
+    "epoch": 60.0,
+    "train_loss": 1.0617500136590194,
+    "train_runtime": 191647.3893,
+    "train_samples": 90777,
+    "train_samples_per_second": 28.42,
+    "train_steps_per_second": 0.444
 }

trainer_state.json

@@ -1,394 +1,1042 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 99.99842519685039,
-  "global_step": 31700,
+  "epoch": 59.999647514980616,
+  "global_step": 85080,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 3.15,
-      "eval_loss": 0.5203462243080139,
-      "eval_runtime": 141.5768,
-      "eval_samples_per_second": 35.853,
-      "eval_steps_per_second": 2.246,
-      "eval_wer": 0.4333326279704594,
+      "epoch": 0.7,
+      "eval_loss": 0.5371278524398804,
+      "eval_runtime": 423.1427,
+      "eval_samples_per_second": 47.847,
+      "eval_steps_per_second": 2.992,
+      "eval_wer": 0.3810800894541897,
       "step": 1000
     },
     {
-      "epoch": 6.31,
-      "learning_rate": 0.0009674592833876221,
-      "loss": 1.4284,
+      "epoch": 1.41,
+      "learning_rate": 0.0009881184586108469,
+      "loss": 1.3606,
       "step": 2000
     },
     {
-      "epoch": 6.31,
-      "eval_loss": 0.48156219720840454,
-      "eval_runtime": 142.5636,
-      "eval_samples_per_second": 35.605,
-      "eval_steps_per_second": 2.231,
-      "eval_wer": 0.3950949065746873,
+      "epoch": 1.41,
+      "eval_loss": 0.5247300267219543,
+      "eval_runtime": 419.7941,
+      "eval_samples_per_second": 48.228,
+      "eval_steps_per_second": 3.016,
+      "eval_wer": 0.3902264309176043,
       "step": 2000
     },
     {
-      "epoch": 9.46,
-      "eval_loss": 0.4314565062522888,
-      "eval_runtime": 139.8322,
-      "eval_samples_per_second": 36.301,
-      "eval_steps_per_second": 2.274,
-      "eval_wer": 0.3545506485811626,
+      "epoch": 2.12,
+      "eval_loss": 0.512639582157135,
+      "eval_runtime": 421.2981,
+      "eval_samples_per_second": 48.056,
+      "eval_steps_per_second": 3.005,
+      "eval_wer": 0.38586728632329304,
       "step": 3000
     },
     {
-      "epoch": 12.62,
-      "learning_rate": 0.0009023452768729642,
-      "loss": 1.283,
+      "epoch": 2.82,
+      "learning_rate": 0.0009643553758325404,
+      "loss": 1.3671,
       "step": 4000
     },
     {
-      "epoch": 12.62,
-      "eval_loss": 0.42783403396606445,
-      "eval_runtime": 141.0912,
-      "eval_samples_per_second": 35.977,
-      "eval_steps_per_second": 2.254,
-      "eval_wer": 0.34039401570137756,
+      "epoch": 2.82,
+      "eval_loss": 0.5062423348426819,
+      "eval_runtime": 423.4242,
+      "eval_samples_per_second": 47.815,
+      "eval_steps_per_second": 2.99,
+      "eval_wer": 0.3827660912712279,
       "step": 4000
     },
     {
-      "epoch": 15.77,
-      "eval_loss": 0.40902915596961975,
-      "eval_runtime": 140.9561,
-      "eval_samples_per_second": 36.011,
-      "eval_steps_per_second": 2.256,
-      "eval_wer": 0.3053939098969465,
+      "epoch": 3.53,
+      "eval_loss": 0.49790534377098083,
+      "eval_runtime": 423.631,
+      "eval_samples_per_second": 47.792,
+      "eval_steps_per_second": 2.988,
+      "eval_wer": 0.3671902299252219,
       "step": 5000
     },
     {
-      "epoch": 18.93,
-      "learning_rate": 0.0008372312703583062,
-      "loss": 1.1777,
+      "epoch": 4.23,
+      "learning_rate": 0.0009405803996194102,
+      "loss": 1.3421,
       "step": 6000
     },
     {
-      "epoch": 18.93,
-      "eval_loss": 0.3892641067504883,
-      "eval_runtime": 139.9163,
-      "eval_samples_per_second": 36.279,
-      "eval_steps_per_second": 2.273,
-      "eval_wer": 0.30056922783926193,
+      "epoch": 4.23,
+      "eval_loss": 0.4905886650085449,
+      "eval_runtime": 422.3048,
+      "eval_samples_per_second": 47.942,
+      "eval_steps_per_second": 2.998,
+      "eval_wer": 0.38159549933608217,
       "step": 6000
     },
     {
-      "epoch": 22.08,
-      "eval_loss": 0.3967570960521698,
-      "eval_runtime": 139.2725,
-      "eval_samples_per_second": 36.447,
-      "eval_steps_per_second": 2.283,
-      "eval_wer": 0.28565080305563195,
+      "epoch": 4.94,
+      "eval_loss": 0.4783899188041687,
+      "eval_runtime": 421.2427,
+      "eval_samples_per_second": 48.063,
+      "eval_steps_per_second": 3.005,
+      "eval_wer": 0.36512859039765183,
       "step": 7000
     },
     {
-      "epoch": 25.24,
-      "learning_rate": 0.0007720846905537459,
-      "loss": 1.0994,
+      "epoch": 5.64,
+      "learning_rate": 0.0009168054234062797,
+      "loss": 1.328,
       "step": 8000
     },
     {
-      "epoch": 25.24,
-      "eval_loss": 0.3892391324043274,
-      "eval_runtime": 138.7844,
-      "eval_samples_per_second": 36.575,
-      "eval_steps_per_second": 2.291,
-      "eval_wer": 0.27509152083289246,
+      "epoch": 5.64,
+      "eval_loss": 0.48102879524230957,
+      "eval_runtime": 420.6976,
+      "eval_samples_per_second": 48.125,
+      "eval_steps_per_second": 3.009,
+      "eval_wer": 0.3669106855825005,
       "step": 8000
     },
     {
-      "epoch": 28.39,
-      "eval_loss": 0.4061281681060791,
-      "eval_runtime": 139.2312,
-      "eval_samples_per_second": 36.457,
-      "eval_steps_per_second": 2.284,
-      "eval_wer": 0.2689760247159151,
+      "epoch": 6.35,
+      "eval_loss": 0.47466185688972473,
+      "eval_runtime": 421.0932,
+      "eval_samples_per_second": 48.08,
+      "eval_steps_per_second": 3.006,
+      "eval_wer": 0.35974736180026556,
       "step": 9000
     },
     {
-      "epoch": 31.54,
-      "learning_rate": 0.000706970684039088,
-      "loss": 1.0323,
+      "epoch": 7.05,
+      "learning_rate": 0.0008930185537583254,
+      "loss": 1.3109,
       "step": 10000
     },
     {
-      "epoch": 31.54,
-      "eval_loss": 0.41136494278907776,
-      "eval_runtime": 139.4432,
-      "eval_samples_per_second": 36.402,
-      "eval_steps_per_second": 2.28,
-      "eval_wer": 0.25065069725120087,
+      "epoch": 7.05,
+      "eval_loss": 0.4812825620174408,
+      "eval_runtime": 417.8509,
+      "eval_samples_per_second": 48.453,
+      "eval_steps_per_second": 3.03,
+      "eval_wer": 0.3808267523935984,
       "step": 10000
     },
     {
-      "epoch": 34.7,
-      "eval_loss": 0.40214526653289795,
-      "eval_runtime": 139.4093,
-      "eval_samples_per_second": 36.411,
-      "eval_steps_per_second": 2.281,
-      "eval_wer": 0.2508411452271621,
+      "epoch": 7.76,
+      "eval_loss": 0.46314355731010437,
+      "eval_runtime": 421.3095,
+      "eval_samples_per_second": 48.055,
+      "eval_steps_per_second": 3.005,
+      "eval_wer": 0.3560696065413376,
       "step": 11000
     },
     {
-      "epoch": 37.85,
-      "learning_rate": 0.00064185667752443,
-      "loss": 0.9623,
+      "epoch": 8.46,
+      "learning_rate": 0.0008692435775451952,
+      "loss": 1.2873,
       "step": 12000
     },
     {
-      "epoch": 37.85,
-      "eval_loss": 0.40321338176727295,
-      "eval_runtime": 139.9917,
-      "eval_samples_per_second": 36.259,
-      "eval_steps_per_second": 2.272,
-      "eval_wer": 0.2378060393169266,
+      "epoch": 8.46,
+      "eval_loss": 0.4602561295032501,
+      "eval_runtime": 419.2403,
+      "eval_samples_per_second": 48.292,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.3430533230833741,
       "step": 12000
     },
     {
-      "epoch": 41.01,
-      "eval_loss": 0.4147748053073883,
-      "eval_runtime": 139.3612,
-      "eval_samples_per_second": 36.423,
-      "eval_steps_per_second": 2.282,
-      "eval_wer": 0.23744630425122204,
+      "epoch": 9.17,
+      "eval_loss": 0.4578526020050049,
+      "eval_runtime": 418.3262,
+      "eval_samples_per_second": 48.398,
+      "eval_steps_per_second": 3.026,
+      "eval_wer": 0.3532916346355441,
       "step": 13000
     },
     {
-      "epoch": 44.16,
-      "learning_rate": 0.0005767100977198697,
-      "loss": 0.9077,
+      "epoch": 9.87,
+      "learning_rate": 0.0008454567078972407,
+      "loss": 1.2661,
       "step": 14000
     },
     {
-      "epoch": 44.16,
-      "eval_loss": 0.4350396394729614,
-      "eval_runtime": 138.8108,
-      "eval_samples_per_second": 36.568,
-      "eval_steps_per_second": 2.291,
-      "eval_wer": 0.23230420890026873,
+      "epoch": 9.87,
+      "eval_loss": 0.44712555408477783,
+      "eval_runtime": 420.1714,
+      "eval_samples_per_second": 48.185,
+      "eval_steps_per_second": 3.013,
+      "eval_wer": 0.336475295268712,
       "step": 14000
     },
     {
-      "epoch": 47.32,
-      "eval_loss": 0.4514589309692383,
-      "eval_runtime": 138.719,
-      "eval_samples_per_second": 36.592,
-      "eval_steps_per_second": 2.292,
-      "eval_wer": 0.22464396808938358,
+      "epoch": 10.58,
+      "eval_loss": 0.45836716890335083,
+      "eval_runtime": 415.4093,
+      "eval_samples_per_second": 48.737,
+      "eval_steps_per_second": 3.048,
+      "eval_wer": 0.3436910336152072,
       "step": 15000
     },
     {
-      "epoch": 50.47,
-      "learning_rate": 0.0005115960912052118,
-      "loss": 0.8573,
+      "epoch": 11.28,
+      "learning_rate": 0.0008216817316841104,
+      "loss": 1.249,
       "step": 16000
     },
     {
-      "epoch": 50.47,
-      "eval_loss": 0.4473990499973297,
-      "eval_runtime": 140.3605,
-      "eval_samples_per_second": 36.164,
-      "eval_steps_per_second": 2.266,
-      "eval_wer": 0.21797828893074042,
+      "epoch": 11.28,
+      "eval_loss": 0.4460853040218353,
+      "eval_runtime": 419.1988,
+      "eval_samples_per_second": 48.297,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.3454294499965057,
       "step": 16000
     },
     {
-      "epoch": 53.63,
-      "eval_loss": 0.4649062752723694,
-      "eval_runtime": 137.3039,
-      "eval_samples_per_second": 36.969,
-      "eval_steps_per_second": 2.316,
-      "eval_wer": 0.21713185348202382,
+      "epoch": 11.99,
+      "eval_loss": 0.44824984669685364,
+      "eval_runtime": 419.1569,
+      "eval_samples_per_second": 48.302,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.3367024250471731,
       "step": 17000
     },
     {
-      "epoch": 56.78,
-      "learning_rate": 0.00044651465798045605,
-      "loss": 0.8083,
+      "epoch": 12.69,
+      "learning_rate": 0.00079790675547098,
+      "loss": 1.2322,
       "step": 18000
     },
     {
-      "epoch": 56.78,
-      "eval_loss": 0.44551119208335876,
-      "eval_runtime": 139.4699,
-      "eval_samples_per_second": 36.395,
-      "eval_steps_per_second": 2.28,
-      "eval_wer": 0.2102334045749836,
+      "epoch": 12.69,
+      "eval_loss": 0.44639432430267334,
+      "eval_runtime": 417.2494,
+      "eval_samples_per_second": 48.523,
+      "eval_steps_per_second": 3.034,
+      "eval_wer": 0.33347019358445734,
       "step": 18000
     },
     {
-      "epoch": 59.94,
-      "eval_loss": 0.4586869478225708,
-      "eval_runtime": 139.1403,
-      "eval_samples_per_second": 36.481,
-      "eval_steps_per_second": 2.285,
-      "eval_wer": 0.20917536026408787,
+      "epoch": 13.4,
+      "eval_loss": 0.4426889717578888,
+      "eval_runtime": 417.2512,
+      "eval_samples_per_second": 48.522,
+      "eval_steps_per_second": 3.034,
+      "eval_wer": 0.3454469215179258,
       "step": 19000
     },
     {
-      "epoch": 63.09,
-      "learning_rate": 0.00038140065146579803,
-      "loss": 0.769,
+      "epoch": 14.1,
+      "learning_rate": 0.0007741317792578497,
+      "loss": 1.22,
       "step": 20000
     },
     {
-      "epoch": 63.09,
-      "eval_loss": 0.4793929159641266,
-      "eval_runtime": 139.562,
-      "eval_samples_per_second": 36.371,
-      "eval_steps_per_second": 2.279,
-      "eval_wer": 0.20117654527371606,
+      "epoch": 14.1,
+      "eval_loss": 0.44404253363609314,
+      "eval_runtime": 419.2873,
+      "eval_samples_per_second": 48.287,
+      "eval_steps_per_second": 3.019,
+      "eval_wer": 0.33950660423509676,
       "step": 20000
     },
     {
-      "epoch": 66.25,
-      "eval_loss": 0.4844733476638794,
-      "eval_runtime": 138.9678,
-      "eval_samples_per_second": 36.526,
-      "eval_steps_per_second": 2.288,
-      "eval_wer": 0.20073216666313987,
+      "epoch": 14.81,
+      "eval_loss": 0.44594520330429077,
+      "eval_runtime": 418.1805,
+      "eval_samples_per_second": 48.415,
+      "eval_steps_per_second": 3.027,
+      "eval_wer": 0.3378468097001887,
       "step": 21000
     },
     {
-      "epoch": 69.4,
-      "learning_rate": 0.00031628664495114006,
-      "loss": 0.7308,
+      "epoch": 15.51,
+      "learning_rate": 0.0007503449096098953,
+      "loss": 1.2044,
       "step": 22000
     },
     {
-      "epoch": 69.4,
-      "eval_loss": 0.49372631311416626,
-      "eval_runtime": 139.7927,
-      "eval_samples_per_second": 36.311,
-      "eval_steps_per_second": 2.275,
-      "eval_wer": 0.20075332754935776,
+      "epoch": 15.51,
+      "eval_loss": 0.4406070411205292,
+      "eval_runtime": 418.2792,
+      "eval_samples_per_second": 48.403,
+      "eval_steps_per_second": 3.027,
+      "eval_wer": 0.3199035572017611,
       "step": 22000
     },
     {
-      "epoch": 72.55,
-      "eval_loss": 0.4920376241207123,
-      "eval_runtime": 138.7644,
-      "eval_samples_per_second": 36.58,
-      "eval_steps_per_second": 2.292,
-      "eval_wer": 0.1894745751952092,
+      "epoch": 16.22,
+      "eval_loss": 0.4397943317890167,
+      "eval_runtime": 419.2827,
+      "eval_samples_per_second": 48.287,
+      "eval_steps_per_second": 3.019,
+      "eval_wer": 0.31545705500034943,
       "step": 23000
     },
     {
-      "epoch": 75.71,
-      "learning_rate": 0.0002511400651465798,
-      "loss": 0.6927,
+      "epoch": 16.92,
+      "learning_rate": 0.000726581826831589,
+      "loss": 1.1913,
       "step": 24000
     },
     {
-      "epoch": 75.71,
-      "eval_loss": 0.5178954005241394,
-      "eval_runtime": 139.799,
-      "eval_samples_per_second": 36.309,
-      "eval_steps_per_second": 2.275,
-      "eval_wer": 0.19114628520642443,
+      "epoch": 16.92,
+      "eval_loss": 0.42369282245635986,
+      "eval_runtime": 417.1191,
+      "eval_samples_per_second": 48.538,
+      "eval_steps_per_second": 3.035,
+      "eval_wer": 0.314959116639877,
       "step": 24000
     },
     {
-      "epoch": 78.86,
-      "eval_loss": 0.520152747631073,
-      "eval_runtime": 140.3812,
-      "eval_samples_per_second": 36.159,
-      "eval_steps_per_second": 2.265,
-      "eval_wer": 0.18767589986668642,
+      "epoch": 17.63,
+      "eval_loss": 0.4286690652370453,
+      "eval_runtime": 422.4918,
+      "eval_samples_per_second": 47.92,
+      "eval_steps_per_second": 2.997,
+      "eval_wer": 0.3278880424907401,
       "step": 25000
     },
     {
-      "epoch": 82.02,
-      "learning_rate": 0.00018602605863192182,
-      "loss": 0.6622,
+      "epoch": 18.34,
+      "learning_rate": 0.0007028068506184586,
+      "loss": 1.1705,
       "step": 26000
     },
     {
-      "epoch": 82.02,
-      "eval_loss": 0.5265706181526184,
-      "eval_runtime": 138.1289,
-      "eval_samples_per_second": 36.748,
-      "eval_steps_per_second": 2.302,
-      "eval_wer": 0.18401506655098715,
+      "epoch": 18.34,
+      "eval_loss": 0.4252525866031647,
+      "eval_runtime": 419.581,
+      "eval_samples_per_second": 48.253,
+      "eval_steps_per_second": 3.017,
+      "eval_wer": 0.31029422042071425,
       "step": 26000
     },
     {
-      "epoch": 85.17,
-      "eval_loss": 0.5350863933563232,
-      "eval_runtime": 140.1605,
-      "eval_samples_per_second": 36.216,
-      "eval_steps_per_second": 2.269,
-      "eval_wer": 0.18541168504136954,
+      "epoch": 19.04,
+      "eval_loss": 0.42342108488082886,
+      "eval_runtime": 420.8706,
+      "eval_samples_per_second": 48.105,
+      "eval_steps_per_second": 3.008,
+      "eval_wer": 0.30976133901740166,
       "step": 27000
     },
     {
-      "epoch": 88.33,
-      "learning_rate": 0.00012091205211726384,
-      "loss": 0.6315,
+      "epoch": 19.75,
+      "learning_rate": 0.0006790199809705044,
+      "loss": 1.1564,
       "step": 28000
     },
     {
-      "epoch": 88.33,
-      "eval_loss": 0.5373002290725708,
-      "eval_runtime": 138.125,
-      "eval_samples_per_second": 36.749,
-      "eval_steps_per_second": 2.302,
-      "eval_wer": 0.18113718602535075,
+      "epoch": 19.75,
+      "eval_loss": 0.4174347519874573,
+      "eval_runtime": 419.264,
+      "eval_samples_per_second": 48.289,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.3076472849255713,
       "step": 28000
     },
     {
-      "epoch": 91.48,
-      "eval_loss": 0.5330832600593567,
-      "eval_runtime": 139.0156,
-      "eval_samples_per_second": 36.514,
-      "eval_steps_per_second": 2.288,
-      "eval_wer": 0.17923270626573842,
+      "epoch": 20.45,
+      "eval_loss": 0.42600032687187195,
+      "eval_runtime": 415.7754,
+      "eval_samples_per_second": 48.695,
+      "eval_steps_per_second": 3.045,
+      "eval_wer": 0.31604235096792227,
       "step": 29000
     },
     {
-      "epoch": 94.64,
-      "learning_rate": 5.576547231270358e-05,
-      "loss": 0.6075,
+      "epoch": 21.16,
+      "learning_rate": 0.0006552450047573739,
+      "loss": 1.1461,
       "step": 30000
     },
     {
-      "epoch": 94.64,
-      "eval_loss": 0.538992166519165,
-      "eval_runtime": 138.185,
-      "eval_samples_per_second": 36.733,
-      "eval_steps_per_second": 2.301,
-      "eval_wer": 0.17787840954779185,
+      "epoch": 21.16,
+      "eval_loss": 0.42350149154663086,
+      "eval_runtime": 419.5333,
+      "eval_samples_per_second": 48.258,
+      "eval_steps_per_second": 3.018,
+      "eval_wer": 0.30363757075966175,
       "step": 30000
     },
     {
-      "epoch": 97.79,
-      "eval_loss": 0.5459240078926086,
-      "eval_runtime": 137.8608,
-      "eval_samples_per_second": 36.82,
-      "eval_steps_per_second": 2.307,
-      "eval_wer": 0.17730706561990817,
+      "epoch": 21.86,
+      "eval_loss": 0.43086713552474976,
+      "eval_runtime": 421.0996,
+      "eval_samples_per_second": 48.079,
+      "eval_steps_per_second": 3.006,
+      "eval_wer": 0.30550702355161086,
       "step": 31000
     },
     {
-      "epoch": 100.0,
-      "step": 31700,
-      "total_flos": 3.173184730349909e+20,
-      "train_loss": 0.8928292760036721,
-      "train_runtime": 80759.6589,
-      "train_samples_per_second": 25.144,
-      "train_steps_per_second": 0.393
+      "epoch": 22.57,
+      "learning_rate": 0.0006314581351094196,
+      "loss": 1.1285,
+      "step": 32000
+    },
+    {
+      "epoch": 22.57,
+      "eval_loss": 0.4263738989830017,
+      "eval_runtime": 418.0963,
+      "eval_samples_per_second": 48.424,
+      "eval_steps_per_second": 3.028,
+      "eval_wer": 0.3006324690754071,
+      "step": 32000
+    },
+    {
+      "epoch": 23.27,
+      "eval_loss": 0.420136034488678,
+      "eval_runtime": 418.959,
+      "eval_samples_per_second": 48.325,
+      "eval_steps_per_second": 3.022,
+      "eval_wer": 0.28796561604584525,
+      "step": 33000
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 0.0006076950523311132,
+      "loss": 1.1135,
+      "step": 34000
+    },
+    {
+      "epoch": 23.98,
+      "eval_loss": 0.41308358311653137,
+      "eval_runtime": 419.6417,
+      "eval_samples_per_second": 48.246,
+      "eval_steps_per_second": 3.017,
+      "eval_wer": 0.29749633098050177,
+      "step": 34000
+    },
+    {
+      "epoch": 24.68,
+      "eval_loss": 0.42022430896759033,
+      "eval_runtime": 415.5643,
+      "eval_samples_per_second": 48.719,
+      "eval_steps_per_second": 3.046,
+      "eval_wer": 0.2848556852330701,
+      "step": 35000
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.0005839200761179829,
+      "loss": 1.0968,
+      "step": 36000
+    },
+    {
+      "epoch": 25.39,
+      "eval_loss": 0.41045960783958435,
+      "eval_runtime": 416.7195,
+      "eval_samples_per_second": 48.584,
+      "eval_steps_per_second": 3.038,
+      "eval_wer": 0.2887867775525893,
+      "step": 36000
+    },
+    {
+      "epoch": 26.09,
+      "eval_loss": 0.4209502339363098,
+      "eval_runtime": 419.7061,
+      "eval_samples_per_second": 48.239,
+      "eval_steps_per_second": 3.016,
+      "eval_wer": 0.28344922775875325,
+      "step": 37000
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 0.0005601332064700286,
+      "loss": 1.087,
+      "step": 38000
+    },
+    {
+      "epoch": 26.8,
+      "eval_loss": 0.4122714400291443,
+      "eval_runtime": 417.6093,
+      "eval_samples_per_second": 48.481,
+      "eval_steps_per_second": 3.032,
+      "eval_wer": 0.2843228038297575,
+      "step": 38000
+    },
+    {
+      "epoch": 27.5,
+      "eval_loss": 0.42156872153282166,
+      "eval_runtime": 415.6358,
+      "eval_samples_per_second": 48.711,
+      "eval_steps_per_second": 3.046,
+      "eval_wer": 0.2802694108602977,
+      "step": 39000
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 0.0005363582302568982,
+      "loss": 1.0707,
+      "step": 40000
+    },
+    {
+      "epoch": 28.21,
+      "eval_loss": 0.4161020517349243,
+      "eval_runtime": 421.5813,
+      "eval_samples_per_second": 48.024,
+      "eval_steps_per_second": 3.003,
+      "eval_wer": 0.2786707666503599,
+      "step": 40000
+    },
+    {
+      "epoch": 28.91,
+      "eval_loss": 0.4186325967311859,
+      "eval_runtime": 419.2011,
+      "eval_samples_per_second": 48.297,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.27398839890977705,
+      "step": 41000
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 0.0005125713606089438,
+      "loss": 1.0575,
+      "step": 42000
+    },
+    {
+      "epoch": 29.62,
+      "eval_loss": 0.41177135705947876,
+      "eval_runtime": 420.5227,
+      "eval_samples_per_second": 48.145,
+      "eval_steps_per_second": 3.011,
+      "eval_wer": 0.2844625760011182,
+      "step": 42000
+    },
+    {
+      "epoch": 30.32,
+      "eval_loss": 0.4242798388004303,
+      "eval_runtime": 415.2438,
+      "eval_samples_per_second": 48.757,
+      "eval_steps_per_second": 3.049,
+      "eval_wer": 0.27729925221888324,
+      "step": 43000
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 0.0004888082778306375,
+      "loss": 1.0474,
+      "step": 44000
+    },
+    {
+      "epoch": 31.03,
+      "eval_loss": 0.4220682382583618,
+      "eval_runtime": 420.3236,
+      "eval_samples_per_second": 48.168,
+      "eval_steps_per_second": 3.012,
+      "eval_wer": 0.27072122440422114,
+      "step": 44000
+    },
+    {
+      "epoch": 31.73,
+      "eval_loss": 0.4138353765010834,
+      "eval_runtime": 416.304,
+      "eval_samples_per_second": 48.633,
+      "eval_steps_per_second": 3.041,
+      "eval_wer": 0.2699524774617374,
+      "step": 45000
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 0.00046502140818268316,
+      "loss": 1.0333,
+      "step": 46000
+    },
+    {
+      "epoch": 32.44,
+      "eval_loss": 0.41024765372276306,
+      "eval_runtime": 419.6952,
+      "eval_samples_per_second": 48.24,
+      "eval_steps_per_second": 3.016,
+      "eval_wer": 0.26381997344328745,
+      "step": 46000
+    },
+    {
+      "epoch": 33.15,
+      "eval_loss": 0.4162220358848572,
+      "eval_runtime": 421.7395,
+      "eval_samples_per_second": 48.006,
+      "eval_steps_per_second": 3.002,
+      "eval_wer": 0.26496435809630303,
+      "step": 47000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 0.00044123453853472884,
+      "loss": 1.0191,
+      "step": 48000
+    },
+    {
+      "epoch": 33.85,
+      "eval_loss": 0.4154505133628845,
+      "eval_runtime": 415.0181,
+      "eval_samples_per_second": 48.783,
+      "eval_steps_per_second": 3.05,
+      "eval_wer": 0.2636452582290866,
+      "step": 48000
+    },
+    {
+      "epoch": 34.56,
+      "eval_loss": 0.41287967562675476,
+      "eval_runtime": 419.9379,
+      "eval_samples_per_second": 48.212,
+      "eval_steps_per_second": 3.015,
+      "eval_wer": 0.2655933328674261,
+      "step": 49000
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 0.0004174595623215985,
+      "loss": 1.0087,
+      "step": 50000
+    },
+    {
+      "epoch": 35.26,
+      "eval_loss": 0.4157230257987976,
+      "eval_runtime": 417.0542,
+      "eval_samples_per_second": 48.545,
+      "eval_steps_per_second": 3.036,
+      "eval_wer": 0.2631909986721644,
+      "step": 50000
+    },
+    {
+      "epoch": 35.97,
+      "eval_loss": 0.40904876589775085,
+      "eval_runtime": 417.7439,
+      "eval_samples_per_second": 48.465,
+      "eval_steps_per_second": 3.031,
+      "eval_wer": 0.26537493884967506,
+      "step": 51000
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 0.0003936845861084681,
+      "loss": 0.9901,
+      "step": 52000
+    },
+    {
+      "epoch": 36.67,
+      "eval_loss": 0.41830816864967346,
+      "eval_runtime": 417.9487,
+      "eval_samples_per_second": 48.441,
+      "eval_steps_per_second": 3.029,
+      "eval_wer": 0.25867461038507233,
+      "step": 52000
+    },
+    {
+      "epoch": 37.38,
+      "eval_loss": 0.4250655770301819,
+      "eval_runtime": 417.46,
+      "eval_samples_per_second": 48.498,
+      "eval_steps_per_second": 3.033,
+      "eval_wer": 0.2648420574463624,
+      "step": 53000
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 0.00036990960989533776,
+      "loss": 0.9795,
+      "step": 54000
+    },
+    {
+      "epoch": 38.08,
+      "eval_loss": 0.4228881299495697,
+      "eval_runtime": 417.6821,
+      "eval_samples_per_second": 48.472,
+      "eval_steps_per_second": 3.031,
+      "eval_wer": 0.25547732196519674,
+      "step": 54000
+    },
+    {
+      "epoch": 38.79,
+      "eval_loss": 0.4176000952720642,
+      "eval_runtime": 418.3393,
+      "eval_samples_per_second": 48.396,
+      "eval_steps_per_second": 3.026,
+      "eval_wer": 0.2545513313299322,
+      "step": 55000
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 0.00034613463368220746,
+      "loss": 0.9644,
+      "step": 56000
+    },
+    {
+      "epoch": 39.49,
+      "eval_loss": 0.4222715497016907,
+      "eval_runtime": 423.2092,
+      "eval_samples_per_second": 47.839,
+      "eval_steps_per_second": 2.991,
+      "eval_wer": 0.25131036410650637,
+      "step": 56000
+    },
+    {
+      "epoch": 40.2,
+      "eval_loss": 0.4243711531162262,
+      "eval_runtime": 419.2699,
+      "eval_samples_per_second": 48.289,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.2530138374449647,
+      "step": 57000
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 0.0003223477640342531,
+      "loss": 0.9534,
+      "step": 58000
+    },
+    {
+      "epoch": 40.9,
+      "eval_loss": 0.4174785912036896,
+      "eval_runtime": 417.6661,
+      "eval_samples_per_second": 48.474,
+      "eval_steps_per_second": 3.031,
+      "eval_wer": 0.2538349989517087,
+      "step": 58000
+    },
+    {
+      "epoch": 41.61,
+      "eval_loss": 0.4212724566459656,
+      "eval_runtime": 425.7723,
+      "eval_samples_per_second": 47.551,
+      "eval_steps_per_second": 2.973,
+      "eval_wer": 0.2505416171640226,
+      "step": 59000
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 0.00029856089438629876,
+      "loss": 0.9397,
+      "step": 60000
+    },
+    {
+      "epoch": 42.31,
+      "eval_loss": 0.4275393486022949,
+      "eval_runtime": 419.6181,
+      "eval_samples_per_second": 48.249,
+      "eval_steps_per_second": 3.017,
+      "eval_wer": 0.2565343490111119,
+      "step": 60000
+    },
+    {
+      "epoch": 43.02,
+      "eval_loss": 0.4315040111541748,
+      "eval_runtime": 420.9924,
+      "eval_samples_per_second": 48.091,
+      "eval_steps_per_second": 3.007,
+      "eval_wer": 0.2528129149486337,
+      "step": 61000
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.00027479781160799237,
+      "loss": 0.9269,
+      "step": 62000
+    },
+    {
+      "epoch": 43.72,
+      "eval_loss": 0.4316493570804596,
+      "eval_runtime": 417.0076,
+      "eval_samples_per_second": 48.551,
+      "eval_steps_per_second": 3.036,
+      "eval_wer": 0.2501048291285205,
+      "step": 62000
+    },
+    {
+      "epoch": 44.43,
+      "eval_loss": 0.42470675706863403,
+      "eval_runtime": 418.3345,
+      "eval_samples_per_second": 48.397,
+      "eval_steps_per_second": 3.026,
+      "eval_wer": 0.24705604864071562,
+      "step": 63000
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 0.00025101094196003804,
+      "loss": 0.9175,
+      "step": 64000
+    },
+    {
+      "epoch": 45.13,
+      "eval_loss": 0.43763282895088196,
+      "eval_runtime": 418.9083,
+      "eval_samples_per_second": 48.33,
+      "eval_steps_per_second": 3.022,
+      "eval_wer": 0.24685512614438465,
+      "step": 64000
+    },
+    {
+      "epoch": 45.84,
+      "eval_loss": 0.4334784150123596,
+      "eval_runtime": 418.8375,
+      "eval_samples_per_second": 48.339,
+      "eval_steps_per_second": 3.023,
+      "eval_wer": 0.24501188063456567,
+      "step": 65000
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 0.00022724785918173168,
+      "loss": 0.9026,
+      "step": 66000
+    },
+    {
+      "epoch": 46.54,
+      "eval_loss": 0.4336349070072174,
+      "eval_runtime": 419.7669,
+      "eval_samples_per_second": 48.232,
+      "eval_steps_per_second": 3.016,
+      "eval_wer": 0.24517786008805648,
+      "step": 66000
+    },
+    {
+      "epoch": 47.25,
+      "eval_loss": 0.4399877190589905,
+      "eval_runtime": 419.3411,
+      "eval_samples_per_second": 48.281,
+      "eval_steps_per_second": 3.019,
+      "eval_wer": 0.24265322524285415,
+      "step": 67000
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 0.00020346098953377738,
+      "loss": 0.8929,
+      "step": 68000
+    },
+    {
+      "epoch": 47.95,
+      "eval_loss": 0.43824830651283264,
+      "eval_runtime": 419.5549,
+      "eval_samples_per_second": 48.256,
+      "eval_steps_per_second": 3.017,
+      "eval_wer": 0.24285414773918512,
+      "step": 68000
+    },
+    {
+      "epoch": 48.66,
+      "eval_loss": 0.43613117933273315,
+      "eval_runtime": 418.6436,
+      "eval_samples_per_second": 48.361,
+      "eval_steps_per_second": 3.024,
+      "eval_wer": 0.24154378363267873,
+      "step": 69000
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 0.00017967411988582303,
+      "loss": 0.8786,
+      "step": 70000
+    },
+    {
+      "epoch": 49.37,
+      "eval_loss": 0.44130945205688477,
+      "eval_runtime": 418.5492,
+      "eval_samples_per_second": 48.372,
+      "eval_steps_per_second": 3.025,
+      "eval_wer": 0.23977915996925012,
+      "step": 70000
+    },
+    {
+      "epoch": 50.07,
+      "eval_loss": 0.43924885988235474,
+      "eval_runtime": 419.6699,
+      "eval_samples_per_second": 48.243,
+      "eval_steps_per_second": 3.017,
+      "eval_wer": 0.2415001048291285,
+      "step": 71000
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 0.00015591103710751666,
+      "loss": 0.8714,
+      "step": 72000
+    },
+    {
+      "epoch": 50.78,
+      "eval_loss": 0.4345008134841919,
+      "eval_runtime": 418.4339,
+      "eval_samples_per_second": 48.385,
+      "eval_steps_per_second": 3.026,
+      "eval_wer": 0.24062652875812426,
+      "step": 72000
+    },
+    {
+      "epoch": 51.48,
+      "eval_loss": 0.44752031564712524,
+      "eval_runtime": 417.7023,
+      "eval_samples_per_second": 48.47,
+      "eval_steps_per_second": 3.031,
+      "eval_wer": 0.24017226920120205,
+      "step": 73000
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 0.00013212416745956231,
+      "loss": 0.8589,
+      "step": 74000
+    },
+    {
+      "epoch": 52.19,
+      "eval_loss": 0.4473401606082916,
+      "eval_runtime": 419.491,
+      "eval_samples_per_second": 48.263,
+      "eval_steps_per_second": 3.018,
+      "eval_wer": 0.23740303305611854,
+      "step": 74000
+    },
+    {
+      "epoch": 52.89,
+      "eval_loss": 0.4457215368747711,
+      "eval_runtime": 416.4935,
+      "eval_samples_per_second": 48.611,
+      "eval_steps_per_second": 3.04,
+      "eval_wer": 0.23568208819624012,
+      "step": 75000
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 0.00010834919124643197,
+      "loss": 0.8493,
+      "step": 76000
+    },
+    {
+      "epoch": 53.6,
+      "eval_loss": 0.44615164399147034,
+      "eval_runtime": 416.8896,
+      "eval_samples_per_second": 48.564,
+      "eval_steps_per_second": 3.037,
+      "eval_wer": 0.2365556642672444,
+      "step": 76000
+    },
+    {
+      "epoch": 54.3,
+      "eval_loss": 0.44939640164375305,
+      "eval_runtime": 416.8835,
+      "eval_samples_per_second": 48.565,
+      "eval_steps_per_second": 3.037,
+      "eval_wer": 0.2355947305891397,
+      "step": 77000
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 8.456232159847763e-05,
+      "loss": 0.8395,
+      "step": 78000
+    },
+    {
+      "epoch": 55.01,
+      "eval_loss": 0.44722679257392883,
+      "eval_runtime": 419.1529,
+      "eval_samples_per_second": 48.302,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.23519288559647775,
+      "step": 78000
+    },
+    {
+      "epoch": 55.71,
+      "eval_loss": 0.44897979497909546,
+      "eval_runtime": 419.2062,
+      "eval_samples_per_second": 48.296,
+      "eval_steps_per_second": 3.02,
+      "eval_wer": 0.23388252148997135,
+      "step": 79000
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 6.0799238820171265e-05,
+      "loss": 0.8295,
+      "step": 80000
+    },
+    {
+      "epoch": 56.42,
+      "eval_loss": 0.4489339590072632,
+      "eval_runtime": 420.6011,
+      "eval_samples_per_second": 48.136,
+      "eval_steps_per_second": 3.01,
+      "eval_wer": 0.23176846739814103,
+      "step": 80000
+    },
+    {
+      "epoch": 57.12,
+      "eval_loss": 0.4468826949596405,
+      "eval_runtime": 417.4618,
+      "eval_samples_per_second": 48.498,
+      "eval_steps_per_second": 3.033,
+      "eval_wer": 0.23203927598015237,
+      "step": 81000
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 3.7012369172216936e-05,
+      "loss": 0.8225,
+      "step": 82000
+    },
+    {
+      "epoch": 57.83,
+      "eval_loss": 0.4478228688240051,
+      "eval_runtime": 417.4954,
+      "eval_samples_per_second": 48.494,
+      "eval_steps_per_second": 3.032,
+      "eval_wer": 0.23214410510867287,
+      "step": 82000
+    },
+    {
+      "epoch": 58.53,
+      "eval_loss": 0.4525238573551178,
+      "eval_runtime": 415.49,
+      "eval_samples_per_second": 48.728,
+      "eval_steps_per_second": 3.047,
+      "eval_wer": 0.2326071004263051,
+      "step": 83000
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 1.3225499524262608e-05,
+      "loss": 0.816,
+      "step": 84000
+    },
+    {
+      "epoch": 59.24,
+      "eval_loss": 0.4532177150249481,
+      "eval_runtime": 419.041,
+      "eval_samples_per_second": 48.315,
+      "eval_steps_per_second": 3.021,
+      "eval_wer": 0.2315588091411,
+      "step": 84000
+    },
+    {
+      "epoch": 59.94,
+      "eval_loss": 0.45018914341926575,
+      "eval_runtime": 418.7584,
+      "eval_samples_per_second": 48.348,
+      "eval_steps_per_second": 3.023,
+      "eval_wer": 0.23179467468027115,
+      "step": 85000
+    },
+    {
+      "epoch": 60.0,
+      "step": 85080,
+      "total_flos": 4.6401496923493315e+20,
+      "train_loss": 1.0617500136590194,
+      "train_runtime": 191647.3893,
+      "train_samples_per_second": 28.42,
+      "train_steps_per_second": 0.444
     }
   ],
-  "max_steps": 31700,
-  "num_train_epochs": 100,
-  "total_flos": 3.173184730349909e+20,
+  "max_steps": 85080,
+  "num_train_epochs": 60,
+  "total_flos": 4.6401496923493315e+20,
   "trial_name": null,
   "trial_params": null
 }

validation_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 100.0,
-    "eval_loss": 0.5438345074653625,
-    "eval_runtime": 140.268,
-    "eval_samples": 5076,
-    "eval_samples_per_second": 36.188,
-    "eval_steps_per_second": 2.267,
-    "eval_wer": 0.177349387392344
+    "epoch": 60.0,
+    "eval_loss": 0.4502493739128113,
+    "eval_runtime": 418.5824,
+    "eval_samples": 20246,
+    "eval_samples_per_second": 48.368,
+    "eval_steps_per_second": 3.024,
+    "eval_wer": 0.23182961772311134
 }