End of training

all_results.json

@@ -1,33 +1,33 @@
 {
     "epoch": 2.33,
     "eval_loss": 1.6191972494125366,
-    "eval_pretrained_loss": 1.715580940246582,
-    "eval_pretrained_runtime": 131.9468,
-    "eval_pretrained_samples_per_second": 2.501,
-    "eval_pretrained_steps_per_second": 0.318,
-    "eval_pretrained_wer": 264.5217946670924,
+    "eval_pretrained_loss": 1.7155802249908447,
+    "eval_pretrained_runtime": 130.3147,
+    "eval_pretrained_samples_per_second": 2.532,
+    "eval_pretrained_steps_per_second": 0.322,
+    "eval_pretrained_wer": 264.42599393262014,
     "eval_runtime": 56.3363,
     "eval_samples_per_second": 0.071,
     "eval_steps_per_second": 0.036,
     "eval_wer": 153.2258064516129,
-    "test_finetuned_loss": 1.0018435716629028,
-    "test_finetuned_runtime": 238.9736,
-    "test_finetuned_samples_per_second": 3.176,
-    "test_finetuned_steps_per_second": 0.398,
-    "test_finetuned_wer": 173.15939719843325,
+    "test_finetuned_loss": 1.0018900632858276,
+    "test_finetuned_runtime": 230.5902,
+    "test_finetuned_samples_per_second": 3.292,
+    "test_finetuned_steps_per_second": 0.412,
+    "test_finetuned_wer": 172.070636659364,
     "test_loss": 1.7568330764770508,
     "test_pretrained_loss": 1.724961757659912,
-    "test_pretrained_runtime": 275.9851,
-    "test_pretrained_samples_per_second": 2.75,
-    "test_pretrained_steps_per_second": 0.344,
+    "test_pretrained_runtime": 272.5213,
+    "test_pretrained_samples_per_second": 2.785,
+    "test_pretrained_steps_per_second": 0.349,
     "test_pretrained_wer": 261.9066587001262,
     "test_runtime": 37.8582,
     "test_samples_per_second": 0.106,
     "test_steps_per_second": 0.053,
     "test_wer": 138.5964912280702,
     "total_flos": 1.7572960198656e+17,
-    "train_loss": 0.8751795228038516,
-    "train_runtime": 2133.32,
-    "train_samples_per_second": 3.36,
+    "train_loss": 0.8751970188958305,
+    "train_runtime": 2136.6213,
+    "train_samples_per_second": 3.355,
     "train_steps_per_second": 0.21
 }

config.json

@@ -139,6 +139,8 @@
     49870,
     50254,
     50258,
+    50358,
+    50359,
     50360,
     50361,
     50362

eval_pretrained_results.json

@@ -1,7 +1,7 @@
 {
-    "eval_pretrained_loss": 1.715580940246582,
-    "eval_pretrained_runtime": 131.9468,
-    "eval_pretrained_samples_per_second": 2.501,
-    "eval_pretrained_steps_per_second": 0.318,
-    "eval_pretrained_wer": 264.5217946670924
+    "eval_pretrained_loss": 1.7155802249908447,
+    "eval_pretrained_runtime": 130.3147,
+    "eval_pretrained_samples_per_second": 2.532,
+    "eval_pretrained_steps_per_second": 0.322,
+    "eval_pretrained_wer": 264.42599393262014
 }

generation_config.json

@@ -211,6 +211,8 @@
     49870,
     50254,
     50258,
+    50358,
+    50359,
     50360,
     50361,
     50362

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68bb0c59266f7c595eaaa4c27443c5b5bde3796bdbf17a2e1ef41723e9e02297
+oid sha256:b8451cf9e0f3b232b59003dac8b3033e1d7ef4a8bf8b4bd868cd1aa95b092f45
 size 151098921

test_finetuned_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 2.33,
-    "test_finetuned_loss": 1.0018435716629028,
-    "test_finetuned_runtime": 238.9736,
-    "test_finetuned_samples_per_second": 3.176,
-    "test_finetuned_steps_per_second": 0.398,
-    "test_finetuned_wer": 173.15939719843325
+    "test_finetuned_loss": 1.0018900632858276,
+    "test_finetuned_runtime": 230.5902,
+    "test_finetuned_samples_per_second": 3.292,
+    "test_finetuned_steps_per_second": 0.412,
+    "test_finetuned_wer": 172.070636659364
 }

test_pretrained_results.json

@@ -1,7 +1,7 @@
 {
     "test_pretrained_loss": 1.724961757659912,
-    "test_pretrained_runtime": 275.9851,
-    "test_pretrained_samples_per_second": 2.75,
-    "test_pretrained_steps_per_second": 0.344,
+    "test_pretrained_runtime": 272.5213,
+    "test_pretrained_samples_per_second": 2.785,
+    "test_pretrained_steps_per_second": 0.349,
     "test_pretrained_wer": 261.9066587001262
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 2.33,
     "total_flos": 1.7572960198656e+17,
-    "train_loss": 0.8751795228038516,
-    "train_runtime": 2133.32,
-    "train_samples_per_second": 3.36,
+    "train_loss": 0.8751970188958305,
+    "train_runtime": 2136.6213,
+    "train_samples_per_second": 3.355,
     "train_steps_per_second": 0.21
 }

trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 173.47916334025228,
+  "best_metric": 160.91330033530258,
   "best_model_checkpoint": "./whisper-training-blog/checkpoint-396",
   "epoch": 2.330357142857143,
   "global_step": 448,
@@ -16,7 +16,7 @@
     {
       "epoch": 0.02,
       "learning_rate": 3.3333333333333335e-07,
-      "loss": 1.6808,
+      "loss": 1.6809,
       "step": 8
     },
     {
@@ -34,7 +34,7 @@
     {
       "epoch": 0.04,
       "learning_rate": 1e-06,
-      "loss": 1.609,
+      "loss": 1.6091,
       "step": 20
     },
     {
@@ -46,13 +46,13 @@
     {
       "epoch": 0.06,
       "learning_rate": 1.4444444444444445e-06,
-      "loss": 1.608,
+      "loss": 1.6081,
       "step": 28
     },
     {
       "epoch": 0.07,
       "learning_rate": 1.6666666666666667e-06,
-      "loss": 1.5075,
+      "loss": 1.5076,
       "step": 32
     },
     {
@@ -75,17 +75,17 @@
     },
     {
       "epoch": 0.1,
-      "eval_loss": 1.4919109344482422,
-      "eval_runtime": 118.4455,
-      "eval_samples_per_second": 2.786,
-      "eval_steps_per_second": 0.355,
-      "eval_wer": 245.29778061631805,
+      "eval_loss": 1.4919402599334717,
+      "eval_runtime": 119.3982,
+      "eval_samples_per_second": 2.764,
+      "eval_steps_per_second": 0.352,
+      "eval_wer": 245.34568098355422,
       "step": 44
     },
     {
       "epoch": 0.11,
       "learning_rate": 2.5555555555555557e-06,
-      "loss": 1.3192,
+      "loss": 1.3193,
       "step": 48
     },
     {
@@ -97,7 +97,7 @@
     {
       "epoch": 0.12,
       "learning_rate": 3e-06,
-      "loss": 1.2895,
+      "loss": 1.2896,
       "step": 56
     },
     {
@@ -109,7 +109,7 @@
     {
       "epoch": 0.14,
       "learning_rate": 3.4444444444444444e-06,
-      "loss": 1.3856,
+      "loss": 1.3855,
       "step": 64
     },
     {
@@ -127,52 +127,52 @@
     {
       "epoch": 0.17,
       "learning_rate": 4.111111111111111e-06,
-      "loss": 1.1661,
+      "loss": 1.1663,
       "step": 76
     },
     {
       "epoch": 0.18,
       "learning_rate": 4.333333333333333e-06,
-      "loss": 1.1351,
+      "loss": 1.1352,
       "step": 80
     },
     {
       "epoch": 0.19,
       "learning_rate": 4.555555555555555e-06,
-      "loss": 1.0729,
+      "loss": 1.0731,
       "step": 84
     },
     {
       "epoch": 0.2,
       "learning_rate": 4.777777777777778e-06,
-      "loss": 1.0501,
+      "loss": 1.0502,
       "step": 88
     },
     {
       "epoch": 0.2,
-      "eval_loss": 1.225467324256897,
-      "eval_runtime": 119.277,
-      "eval_samples_per_second": 2.767,
-      "eval_steps_per_second": 0.352,
-      "eval_wer": 219.94251955931662,
+      "eval_loss": 1.2254914045333862,
+      "eval_runtime": 123.9853,
+      "eval_samples_per_second": 2.662,
+      "eval_steps_per_second": 0.339,
+      "eval_wer": 220.1500878173399,
       "step": 88
     },
     {
       "epoch": 0.21,
       "learning_rate": 4.9999999999999996e-06,
-      "loss": 1.1278,
+      "loss": 1.128,
       "step": 92
     },
     {
       "epoch": 0.21,
       "learning_rate": 5.2222222222222226e-06,
-      "loss": 1.0563,
+      "loss": 1.0564,
       "step": 96
     },
     {
       "epoch": 0.22,
       "learning_rate": 5.444444444444445e-06,
-      "loss": 0.9929,
+      "loss": 0.993,
       "step": 100
     },
     {
@@ -196,13 +196,13 @@
     {
       "epoch": 0.26,
       "learning_rate": 6.333333333333333e-06,
-      "loss": 0.9688,
+      "loss": 0.9687,
       "step": 116
     },
     {
       "epoch": 0.27,
       "learning_rate": 6.555555555555556e-06,
-      "loss": 0.9559,
+      "loss": 0.9558,
       "step": 120
     },
     {
@@ -225,29 +225,29 @@
     },
     {
       "epoch": 0.29,
-      "eval_loss": 1.120314359664917,
-      "eval_runtime": 111.0003,
-      "eval_samples_per_second": 2.973,
-      "eval_steps_per_second": 0.378,
-      "eval_wer": 205.7799776464953,
+      "eval_loss": 1.1202749013900757,
+      "eval_runtime": 108.7054,
+      "eval_samples_per_second": 3.036,
+      "eval_steps_per_second": 0.386,
+      "eval_wer": 206.24301452977804,
       "step": 132
     },
     {
       "epoch": 0.3,
       "learning_rate": 7.444444444444444e-06,
-      "loss": 0.9171,
+      "loss": 0.917,
       "step": 136
     },
     {
       "epoch": 0.31,
       "learning_rate": 7.428115015974441e-06,
-      "loss": 0.9445,
+      "loss": 0.9444,
       "step": 140
     },
     {
       "epoch": 0.32,
       "learning_rate": 7.332268370607029e-06,
-      "loss": 0.8926,
+      "loss": 0.8927,
       "step": 144
     },
     {
@@ -259,7 +259,7 @@
     {
       "epoch": 1.0,
       "learning_rate": 7.140575079872205e-06,
-      "loss": 1.239,
+      "loss": 1.2391,
       "step": 152
     },
     {
@@ -271,13 +271,13 @@
     {
       "epoch": 1.02,
       "learning_rate": 6.948881789137381e-06,
-      "loss": 1.0117,
+      "loss": 1.0116,
       "step": 160
     },
     {
       "epoch": 1.03,
       "learning_rate": 6.853035143769968e-06,
-      "loss": 0.9081,
+      "loss": 0.9082,
       "step": 164
     },
     {
@@ -289,22 +289,22 @@
     {
       "epoch": 1.05,
       "learning_rate": 6.6613418530351436e-06,
-      "loss": 0.8672,
+      "loss": 0.8671,
       "step": 172
     },
     {
       "epoch": 1.06,
       "learning_rate": 6.565495207667732e-06,
-      "loss": 0.8142,
+      "loss": 0.8141,
       "step": 176
     },
     {
       "epoch": 1.06,
-      "eval_loss": 1.0674585103988647,
-      "eval_runtime": 102.4925,
-      "eval_samples_per_second": 3.22,
-      "eval_steps_per_second": 0.41,
-      "eval_wer": 192.87881207089254,
+      "eval_loss": 1.067484736442566,
+      "eval_runtime": 108.347,
+      "eval_samples_per_second": 3.046,
+      "eval_steps_per_second": 0.388,
+      "eval_wer": 201.96391505668208,
       "step": 176
     },
     {
@@ -328,25 +328,25 @@
     {
       "epoch": 1.09,
       "learning_rate": 6.182108626198084e-06,
-      "loss": 0.821,
+      "loss": 0.8211,
       "step": 192
     },
     {
       "epoch": 1.1,
       "learning_rate": 6.086261980830671e-06,
-      "loss": 0.8518,
+      "loss": 0.8517,
       "step": 196
     },
     {
       "epoch": 1.11,
       "learning_rate": 5.990415335463259e-06,
-      "loss": 0.7813,
+      "loss": 0.7814,
       "step": 200
     },
     {
       "epoch": 1.12,
       "learning_rate": 5.894568690095847e-06,
-      "loss": 0.7541,
+      "loss": 0.7542,
       "step": 204
     },
     {
@@ -358,13 +358,13 @@
     {
       "epoch": 1.14,
       "learning_rate": 5.702875399361023e-06,
-      "loss": 0.8182,
+      "loss": 0.8181,
       "step": 212
     },
     {
       "epoch": 1.15,
       "learning_rate": 5.607028753993611e-06,
-      "loss": 0.7461,
+      "loss": 0.7462,
       "step": 216
     },
     {
@@ -375,11 +375,11 @@
     },
     {
       "epoch": 1.16,
-      "eval_loss": 1.0393497943878174,
-      "eval_runtime": 101.6759,
-      "eval_samples_per_second": 3.246,
-      "eval_steps_per_second": 0.413,
-      "eval_wer": 178.42886795465432,
+      "eval_loss": 1.0393874645233154,
+      "eval_runtime": 100.1655,
+      "eval_samples_per_second": 3.295,
+      "eval_steps_per_second": 0.419,
+      "eval_wer": 178.36500079833945,
       "step": 220
     },
     {
@@ -391,13 +391,13 @@
     {
       "epoch": 1.17,
       "learning_rate": 5.319488817891374e-06,
-      "loss": 0.7662,
+      "loss": 0.7663,
       "step": 228
     },
     {
       "epoch": 1.18,
       "learning_rate": 5.223642172523962e-06,
-      "loss": 0.7253,
+      "loss": 0.7254,
       "step": 232
     },
     {
@@ -415,7 +415,7 @@
     {
       "epoch": 1.21,
       "learning_rate": 4.936102236421725e-06,
-      "loss": 0.7841,
+      "loss": 0.7842,
       "step": 244
     },
     {
@@ -427,13 +427,13 @@
     {
       "epoch": 1.23,
       "learning_rate": 4.744408945686901e-06,
-      "loss": 0.7077,
+      "loss": 0.7075,
       "step": 252
     },
     {
       "epoch": 1.24,
       "learning_rate": 4.648562300319489e-06,
-      "loss": 0.7812,
+      "loss": 0.7811,
       "step": 256
     },
     {
@@ -450,11 +450,11 @@
     },
     {
       "epoch": 1.25,
-      "eval_loss": 1.0302220582962036,
-      "eval_runtime": 114.7389,
-      "eval_samples_per_second": 2.876,
-      "eval_steps_per_second": 0.366,
-      "eval_wer": 216.6054606418649,
+      "eval_loss": 1.030145525932312,
+      "eval_runtime": 114.3297,
+      "eval_samples_per_second": 2.886,
+      "eval_steps_per_second": 0.367,
+      "eval_wer": 221.2996966310075,
       "step": 264
     },
     {
@@ -466,19 +466,19 @@
     {
       "epoch": 1.27,
       "learning_rate": 4.26517571884984e-06,
-      "loss": 0.7187,
+      "loss": 0.7186,
       "step": 272
     },
     {
       "epoch": 1.28,
       "learning_rate": 4.169329073482428e-06,
-      "loss": 0.6699,
+      "loss": 0.6698,
       "step": 276
     },
     {
       "epoch": 1.29,
       "learning_rate": 4.0734824281150155e-06,
-      "loss": 0.5677,
+      "loss": 0.5678,
       "step": 280
     },
     {
@@ -508,7 +508,7 @@
     {
       "epoch": 1.33,
       "learning_rate": 3.5942492012779555e-06,
-      "loss": 0.8766,
+      "loss": 0.8765,
       "step": 300
     },
     {
@@ -520,34 +520,34 @@
     {
       "epoch": 2.02,
       "learning_rate": 3.4025559105431313e-06,
-      "loss": 0.6971,
+      "loss": 0.6972,
       "step": 308
     },
     {
       "epoch": 2.02,
-      "eval_loss": 1.0134836435317993,
-      "eval_runtime": 101.0358,
-      "eval_samples_per_second": 3.266,
-      "eval_steps_per_second": 0.416,
-      "eval_wer": 179.3709085102986,
+      "eval_loss": 1.0134402513504028,
+      "eval_runtime": 102.1008,
+      "eval_samples_per_second": 3.232,
+      "eval_steps_per_second": 0.411,
+      "eval_wer": 176.67252115599553,
       "step": 308
     },
     {
       "epoch": 2.03,
       "learning_rate": 3.306709265175719e-06,
-      "loss": 0.7432,
+      "loss": 0.7431,
       "step": 312
     },
     {
       "epoch": 2.04,
       "learning_rate": 3.2108626198083067e-06,
-      "loss": 0.6264,
+      "loss": 0.6265,
       "step": 316
     },
     {
       "epoch": 2.04,
       "learning_rate": 3.1150159744408946e-06,
-      "loss": 0.6604,
+      "loss": 0.6606,
       "step": 320
     },
     {
@@ -559,13 +559,13 @@
     {
       "epoch": 2.06,
       "learning_rate": 2.9233226837060704e-06,
-      "loss": 0.6539,
+      "loss": 0.654,
       "step": 328
     },
     {
       "epoch": 2.07,
       "learning_rate": 2.8274760383386583e-06,
-      "loss": 0.611,
+      "loss": 0.6111,
       "step": 332
     },
     {
@@ -583,28 +583,28 @@
     {
       "epoch": 2.1,
       "learning_rate": 2.539936102236422e-06,
-      "loss": 0.6642,
+      "loss": 0.6643,
       "step": 344
     },
     {
       "epoch": 2.11,
       "learning_rate": 2.44408945686901e-06,
-      "loss": 0.6147,
+      "loss": 0.6146,
       "step": 348
     },
     {
       "epoch": 2.12,
       "learning_rate": 2.3482428115015974e-06,
-      "loss": 0.6051,
+      "loss": 0.6052,
       "step": 352
     },
     {
       "epoch": 2.12,
-      "eval_loss": 1.0064767599105835,
-      "eval_runtime": 107.523,
-      "eval_samples_per_second": 3.069,
-      "eval_steps_per_second": 0.391,
-      "eval_wer": 194.63515886955133,
+      "eval_loss": 1.006484866142273,
+      "eval_runtime": 105.4228,
+      "eval_samples_per_second": 3.13,
+      "eval_steps_per_second": 0.398,
+      "eval_wer": 194.7149928149449,
       "step": 352
     },
     {
@@ -628,7 +628,7 @@
     {
       "epoch": 2.15,
       "learning_rate": 1.964856230031949e-06,
-      "loss": 0.6012,
+      "loss": 0.6017,
       "step": 368
     },
     {
@@ -640,19 +640,19 @@
     {
       "epoch": 2.17,
       "learning_rate": 1.7731629392971245e-06,
-      "loss": 0.6288,
+      "loss": 0.6289,
       "step": 376
     },
     {
       "epoch": 2.18,
       "learning_rate": 1.6773162939297124e-06,
-      "loss": 0.6079,
+      "loss": 0.608,
       "step": 380
     },
     {
       "epoch": 2.19,
       "learning_rate": 1.5814696485623003e-06,
-      "loss": 0.5887,
+      "loss": 0.5888,
       "step": 384
     },
     {
@@ -670,22 +670,22 @@
     {
       "epoch": 2.21,
       "learning_rate": 1.2939297124600638e-06,
-      "loss": 0.6048,
+      "loss": 0.6047,
       "step": 396
     },
     {
       "epoch": 2.21,
-      "eval_loss": 1.002966284751892,
-      "eval_runtime": 97.8102,
-      "eval_samples_per_second": 3.374,
-      "eval_steps_per_second": 0.429,
-      "eval_wer": 173.47916334025228,
+      "eval_loss": 1.002995491027832,
+      "eval_runtime": 93.4309,
+      "eval_samples_per_second": 3.532,
+      "eval_steps_per_second": 0.45,
+      "eval_wer": 160.91330033530258,
       "step": 396
     },
     {
       "epoch": 2.22,
       "learning_rate": 1.1980830670926517e-06,
-      "loss": 0.5774,
+      "loss": 0.5772,
       "step": 400
     },
     {
@@ -703,13 +703,13 @@
     {
       "epoch": 2.25,
       "learning_rate": 9.105431309904153e-07,
-      "loss": 0.5618,
+      "loss": 0.5617,
       "step": 412
     },
     {
       "epoch": 2.26,
       "learning_rate": 8.146964856230032e-07,
-      "loss": 0.5729,
+      "loss": 0.5728,
       "step": 416
     },
     {
@@ -733,7 +733,7 @@
     {
       "epoch": 2.29,
       "learning_rate": 4.313099041533546e-07,
-      "loss": 0.5544,
+      "loss": 0.5543,
       "step": 432
     },
     {
@@ -745,16 +745,16 @@
     {
       "epoch": 2.31,
       "learning_rate": 2.3961661341853033e-07,
-      "loss": 0.585,
+      "loss": 0.5849,
       "step": 440
     },
     {
       "epoch": 2.31,
-      "eval_loss": 1.0049320459365845,
-      "eval_runtime": 105.5439,
-      "eval_samples_per_second": 3.127,
-      "eval_steps_per_second": 0.398,
-      "eval_wer": 186.6677311192719,
+      "eval_loss": 1.0050277709960938,
+      "eval_runtime": 101.6428,
+      "eval_samples_per_second": 3.247,
+      "eval_steps_per_second": 0.413,
+      "eval_wer": 180.05748044068338,
       "step": 440
     },
     {
@@ -773,9 +773,9 @@
       "epoch": 2.33,
       "step": 448,
       "total_flos": 1.7572960198656e+17,
-      "train_loss": 0.8751795228038516,
-      "train_runtime": 2133.32,
-      "train_samples_per_second": 3.36,
+      "train_loss": 0.8751970188958305,
+      "train_runtime": 2136.6213,
+      "train_samples_per_second": 3.355,
       "train_steps_per_second": 0.21
     }
   ],

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:098eb422614f2b69cada5c370a7aafa45f3e0897c25ca997abbd0fbb3b31385a
+oid sha256:2fb52f407e48481c35659717bd5c261226233eadb5d658c9543eba6275568644
 size 3707