End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 100.0,
-    "eval_loss": 1.2441085577011108,
-    "eval_runtime": 12.8952,
+    "eval_loss": 1.1507658958435059,
+    "eval_runtime": 14.1834,
     "eval_samples": 360,
-    "eval_samples_per_second": 27.917,
-    "eval_steps_per_second": 3.49,
-    "eval_wer": 0.5350842806742454,
-    "train_loss": 0.7377449184876901,
-    "train_runtime": 5722.2661,
+    "eval_samples_per_second": 25.382,
+    "eval_steps_per_second": 3.173,
+    "eval_wer": 0.49078792630341045,
+    "train_loss": 0.8475217450106586,
+    "train_runtime": 6078.829,
     "train_samples": 856,
-    "train_samples_per_second": 14.959,
-    "train_steps_per_second": 0.944
+    "train_samples_per_second": 14.082,
+    "train_steps_per_second": 0.888
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 100.0,
-    "eval_loss": 1.2441085577011108,
-    "eval_runtime": 12.8952,
+    "eval_loss": 1.1507658958435059,
+    "eval_runtime": 14.1834,
     "eval_samples": 360,
-    "eval_samples_per_second": 27.917,
-    "eval_steps_per_second": 3.49,
-    "eval_wer": 0.5350842806742454
+    "eval_samples_per_second": 25.382,
+    "eval_steps_per_second": 3.173,
+    "eval_wer": 0.49078792630341045
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 100.0,
-    "train_loss": 0.7377449184876901,
-    "train_runtime": 5722.2661,
+    "train_loss": 0.8475217450106586,
+    "train_runtime": 6078.829,
     "train_samples": 856,
-    "train_samples_per_second": 14.959,
-    "train_steps_per_second": 0.944
+    "train_samples_per_second": 14.082,
+    "train_steps_per_second": 0.888
 }

trainer_state.json

@@ -9,264 +9,264 @@
   "log_history": [
     {
       "epoch": 3.7,
-      "learning_rate": 5.909999999999999e-05,
-      "loss": 11.4212,
+      "learning_rate": 3.9399999999999995e-05,
+      "loss": 12.8577,
       "step": 200
     },
     {
       "epoch": 7.41,
-      "learning_rate": 0.0001191,
-      "loss": 3.3931,
+      "learning_rate": 7.939999999999999e-05,
+      "loss": 3.5841,
       "step": 400
     },
     {
       "epoch": 9.26,
-      "eval_loss": 2.6012766361236572,
-      "eval_runtime": 13.0763,
-      "eval_samples_per_second": 27.531,
-      "eval_steps_per_second": 3.441,
-      "eval_wer": 0.991375931007448,
+      "eval_loss": 3.2514328956604004,
+      "eval_runtime": 14.446,
+      "eval_samples_per_second": 24.92,
+      "eval_steps_per_second": 3.115,
+      "eval_wer": 0.9941199529596236,
       "step": 500
     },
     {
       "epoch": 11.11,
-      "learning_rate": 0.0001791,
-      "loss": 2.2716,
+      "learning_rate": 0.0001194,
+      "loss": 3.0775,
       "step": 600
     },
     {
       "epoch": 14.81,
-      "learning_rate": 0.00023909999999999998,
-      "loss": 0.5568,
+      "learning_rate": 0.00015939999999999997,
+      "loss": 0.8996,
       "step": 800
     },
     {
       "epoch": 18.52,
-      "learning_rate": 0.0002991,
-      "loss": 0.3457,
+      "learning_rate": 0.00019939999999999997,
+      "loss": 0.3992,
       "step": 1000
     },
     {
       "epoch": 18.52,
-      "eval_loss": 1.0315884351730347,
-      "eval_runtime": 12.6921,
-      "eval_samples_per_second": 28.364,
-      "eval_steps_per_second": 3.546,
-      "eval_wer": 0.6452371618972952,
+      "eval_loss": 0.8790363073348999,
+      "eval_runtime": 14.0346,
+      "eval_samples_per_second": 25.651,
+      "eval_steps_per_second": 3.206,
+      "eval_wer": 0.6107408859270874,
       "step": 1000
     },
     {
       "epoch": 22.22,
-      "learning_rate": 0.00032255581395348835,
-      "loss": 0.2708,
+      "learning_rate": 0.0002394,
+      "loss": 0.2879,
       "step": 1200
     },
     {
       "epoch": 25.93,
-      "learning_rate": 0.00030720697674418605,
-      "loss": 0.2118,
+      "learning_rate": 0.00027939999999999996,
+      "loss": 0.2409,
       "step": 1400
     },
     {
       "epoch": 27.78,
-      "eval_loss": 1.1182368993759155,
-      "eval_runtime": 12.9665,
-      "eval_samples_per_second": 27.764,
-      "eval_steps_per_second": 3.47,
-      "eval_wer": 0.6354370834966679,
+      "eval_loss": 1.0011698007583618,
+      "eval_runtime": 14.3849,
+      "eval_samples_per_second": 25.026,
+      "eval_steps_per_second": 3.128,
+      "eval_wer": 0.6366130929047432,
       "step": 1500
     },
     {
       "epoch": 29.63,
-      "learning_rate": 0.00029185813953488374,
-      "loss": 0.1765,
+      "learning_rate": 0.0002925384615384615,
+      "loss": 0.2132,
       "step": 1600
     },
     {
       "epoch": 33.33,
-      "learning_rate": 0.0002765093023255814,
-      "loss": 0.1404,
+      "learning_rate": 0.00027715384615384614,
+      "loss": 0.1729,
       "step": 1800
     },
     {
       "epoch": 37.04,
-      "learning_rate": 0.00026116046511627907,
-      "loss": 0.1245,
+      "learning_rate": 0.00026176923076923073,
+      "loss": 0.1447,
       "step": 2000
     },
     {
       "epoch": 37.04,
-      "eval_loss": 1.2044304609298706,
-      "eval_runtime": 12.7752,
-      "eval_samples_per_second": 28.18,
-      "eval_steps_per_second": 3.522,
-      "eval_wer": 0.6295570364562917,
+      "eval_loss": 1.0167392492294312,
+      "eval_runtime": 14.4093,
+      "eval_samples_per_second": 24.984,
+      "eval_steps_per_second": 3.123,
+      "eval_wer": 0.6275970207761662,
       "step": 2000
     },
     {
       "epoch": 40.74,
-      "learning_rate": 0.00024581162790697676,
-      "loss": 0.1166,
+      "learning_rate": 0.0002463846153846153,
+      "loss": 0.1311,
       "step": 2200
     },
     {
       "epoch": 44.44,
-      "learning_rate": 0.00023046279069767443,
-      "loss": 0.1068,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.1109,
       "step": 2400
     },
     {
       "epoch": 46.3,
-      "eval_loss": 1.1400295495986938,
-      "eval_runtime": 12.7392,
-      "eval_samples_per_second": 28.259,
-      "eval_steps_per_second": 3.532,
-      "eval_wer": 0.582124656997256,
+      "eval_loss": 1.0637701749801636,
+      "eval_runtime": 13.6702,
+      "eval_samples_per_second": 26.335,
+      "eval_steps_per_second": 3.292,
+      "eval_wer": 0.5652685221481771,
       "step": 2500
     },
     {
       "epoch": 48.15,
-      "learning_rate": 0.0002151139534883721,
-      "loss": 0.0917,
+      "learning_rate": 0.0002156153846153846,
+      "loss": 0.0968,
       "step": 2600
     },
     {
       "epoch": 51.85,
-      "learning_rate": 0.00019976511627906976,
-      "loss": 0.0842,
+      "learning_rate": 0.00020023076923076922,
+      "loss": 0.0865,
       "step": 2800
     },
     {
       "epoch": 55.56,
-      "learning_rate": 0.00018441627906976745,
-      "loss": 0.0806,
+      "learning_rate": 0.00018484615384615385,
+      "loss": 0.0797,
       "step": 3000
     },
     {
       "epoch": 55.56,
-      "eval_loss": 1.2010185718536377,
-      "eval_runtime": 12.8794,
-      "eval_samples_per_second": 27.952,
-      "eval_steps_per_second": 3.494,
-      "eval_wer": 0.5785966287730302,
+      "eval_loss": 1.1447213888168335,
+      "eval_runtime": 13.6804,
+      "eval_samples_per_second": 26.315,
+      "eval_steps_per_second": 3.289,
+      "eval_wer": 0.5715405723245786,
       "step": 3000
     },
     {
       "epoch": 59.26,
-      "learning_rate": 0.00016906744186046509,
-      "loss": 0.0719,
+      "learning_rate": 0.00016946153846153844,
+      "loss": 0.0728,
       "step": 3200
     },
     {
       "epoch": 62.96,
-      "learning_rate": 0.00015371860465116278,
-      "loss": 0.0679,
+      "learning_rate": 0.00015407692307692306,
+      "loss": 0.0636,
       "step": 3400
     },
     {
       "epoch": 64.81,
-      "eval_loss": 1.1644504070281982,
-      "eval_runtime": 12.9878,
-      "eval_samples_per_second": 27.718,
-      "eval_steps_per_second": 3.465,
-      "eval_wer": 0.546452371618973,
+      "eval_loss": 1.1502803564071655,
+      "eval_runtime": 14.7332,
+      "eval_samples_per_second": 24.435,
+      "eval_steps_per_second": 3.054,
+      "eval_wer": 0.5315562524500196,
       "step": 3500
     },
     {
       "epoch": 66.67,
-      "learning_rate": 0.00013836976744186047,
-      "loss": 0.0582,
+      "learning_rate": 0.0001386923076923077,
+      "loss": 0.0604,
       "step": 3600
     },
     {
       "epoch": 70.37,
-      "learning_rate": 0.00012302093023255814,
-      "loss": 0.0566,
+      "learning_rate": 0.0001233076923076923,
+      "loss": 0.0499,
       "step": 3800
     },
     {
       "epoch": 74.07,
-      "learning_rate": 0.00010767209302325581,
-      "loss": 0.0506,
+      "learning_rate": 0.0001079230769230769,
+      "loss": 0.0466,
       "step": 4000
     },
     {
       "epoch": 74.07,
-      "eval_loss": 1.2748003005981445,
-      "eval_runtime": 12.9412,
-      "eval_samples_per_second": 27.818,
-      "eval_steps_per_second": 3.477,
-      "eval_wer": 0.5507644061152489,
+      "eval_loss": 1.2227113246917725,
+      "eval_runtime": 14.6209,
+      "eval_samples_per_second": 24.622,
+      "eval_steps_per_second": 3.078,
+      "eval_wer": 0.5386123088984712,
       "step": 4000
     },
     {
       "epoch": 77.78,
-      "learning_rate": 9.232325581395348e-05,
-      "loss": 0.0443,
+      "learning_rate": 9.253846153846153e-05,
+      "loss": 0.0426,
       "step": 4200
     },
     {
       "epoch": 81.48,
-      "learning_rate": 7.697441860465116e-05,
-      "loss": 0.0397,
+      "learning_rate": 7.715384615384615e-05,
+      "loss": 0.0372,
       "step": 4400
     },
     {
       "epoch": 83.33,
-      "eval_loss": 1.1869040727615356,
-      "eval_runtime": 12.9633,
-      "eval_samples_per_second": 27.771,
-      "eval_steps_per_second": 3.471,
-      "eval_wer": 0.5393963151705213,
+      "eval_loss": 1.121378779411316,
+      "eval_runtime": 13.6605,
+      "eval_samples_per_second": 26.353,
+      "eval_steps_per_second": 3.294,
+      "eval_wer": 0.5225401803214426,
       "step": 4500
     },
     {
       "epoch": 85.19,
-      "learning_rate": 6.162558139534884e-05,
-      "loss": 0.0353,
+      "learning_rate": 6.176923076923076e-05,
+      "loss": 0.0349,
       "step": 4600
     },
     {
       "epoch": 88.89,
-      "learning_rate": 4.6276744186046515e-05,
-      "loss": 0.0282,
+      "learning_rate": 4.6384615384615385e-05,
+      "loss": 0.029,
       "step": 4800
     },
     {
       "epoch": 92.59,
-      "learning_rate": 3.092790697674419e-05,
-      "loss": 0.0276,
+      "learning_rate": 3.0999999999999995e-05,
+      "loss": 0.0239,
       "step": 5000
     },
     {
       "epoch": 92.59,
-      "eval_loss": 1.2239253520965576,
-      "eval_runtime": 12.8775,
-      "eval_samples_per_second": 27.956,
-      "eval_steps_per_second": 3.494,
-      "eval_wer": 0.5390043120344963,
+      "eval_loss": 1.1375247240066528,
+      "eval_runtime": 14.3296,
+      "eval_samples_per_second": 25.123,
+      "eval_steps_per_second": 3.14,
+      "eval_wer": 0.4998039984319875,
       "step": 5000
     },
     {
       "epoch": 96.3,
-      "learning_rate": 1.5579069767441862e-05,
-      "loss": 0.0247,
+      "learning_rate": 1.5615384615384614e-05,
+      "loss": 0.0208,
       "step": 5200
     },
     {
       "epoch": 100.0,
-      "learning_rate": 2.302325581395349e-07,
-      "loss": 0.0218,
+      "learning_rate": 2.3076923076923075e-07,
+      "loss": 0.0188,
       "step": 5400
     },
     {
       "epoch": 100.0,
       "step": 5400,
       "total_flos": 1.3559072525710848e+19,
-      "train_loss": 0.7377449184876901,
-      "train_runtime": 5722.2661,
-      "train_samples_per_second": 14.959,
-      "train_steps_per_second": 0.944
+      "train_loss": 0.8475217450106586,
+      "train_runtime": 6078.829,
+      "train_samples_per_second": 14.082,
+      "train_steps_per_second": 0.888
     }
   ],
   "max_steps": 5400,