End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 79.99,
-    "eval_loss": 0.49252477288246155,
-    "eval_runtime": 31.5756,
+    "eval_loss": 0.4997510015964508,
+    "eval_runtime": 32.8146,
     "eval_samples": 892,
-    "eval_samples_per_second": 28.25,
-    "eval_steps_per_second": 3.547,
-    "eval_wer": 0.5714079630698211,
-    "train_loss": 1.795632479985555,
-    "train_runtime": 8730.4816,
+    "eval_samples_per_second": 27.183,
+    "eval_steps_per_second": 3.413,
+    "eval_wer": 0.5152914021927294,
+    "train_loss": 1.6836539268493653,
+    "train_runtime": 8656.8344,
     "train_samples": 1941,
-    "train_samples_per_second": 17.786,
-    "train_steps_per_second": 0.55
+    "train_samples_per_second": 17.937,
+    "train_steps_per_second": 0.554
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 79.99,
-    "eval_loss": 0.49252477288246155,
-    "eval_runtime": 31.5756,
+    "eval_loss": 0.4997510015964508,
+    "eval_runtime": 32.8146,
     "eval_samples": 892,
-    "eval_samples_per_second": 28.25,
-    "eval_steps_per_second": 3.547,
-    "eval_wer": 0.5714079630698211
+    "eval_samples_per_second": 27.183,
+    "eval_steps_per_second": 3.413,
+    "eval_wer": 0.5152914021927294
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 79.99,
-    "train_loss": 1.795632479985555,
-    "train_runtime": 8730.4816,
+    "train_loss": 1.6836539268493653,
+    "train_runtime": 8656.8344,
     "train_samples": 1941,
-    "train_samples_per_second": 17.786,
-    "train_steps_per_second": 0.55
+    "train_samples_per_second": 17.937,
+    "train_steps_per_second": 0.554
 }

trainer_state.json

@@ -9,381 +9,381 @@
   "log_history": [
     {
       "epoch": 1.66,
-      "learning_rate": 3.7125e-06,
-      "loss": 14.5868,
+      "learning_rate": 4.752e-06,
+      "loss": 14.1408,
       "step": 100
     },
     {
       "epoch": 3.33,
-      "learning_rate": 7.4625e-06,
-      "loss": 6.8756,
+      "learning_rate": 9.552000000000001e-06,
+      "loss": 6.1022,
       "step": 200
     },
     {
       "epoch": 4.99,
-      "learning_rate": 1.1212499999999998e-05,
-      "loss": 4.2978,
+      "learning_rate": 1.4351999999999999e-05,
+      "loss": 4.037,
       "step": 300
     },
     {
       "epoch": 6.66,
-      "learning_rate": 1.49625e-05,
-      "loss": 3.6126,
+      "learning_rate": 1.9152000000000002e-05,
+      "loss": 3.3885,
       "step": 400
     },
     {
       "epoch": 8.33,
-      "learning_rate": 1.8712499999999997e-05,
-      "loss": 3.1674,
+      "learning_rate": 2.3952e-05,
+      "loss": 3.0021,
       "step": 500
     },
     {
       "epoch": 8.33,
-      "eval_loss": 3.0295047760009766,
-      "eval_runtime": 32.2697,
-      "eval_samples_per_second": 27.642,
-      "eval_steps_per_second": 3.471,
+      "eval_loss": 2.90588116645813,
+      "eval_runtime": 32.387,
+      "eval_samples_per_second": 27.542,
+      "eval_steps_per_second": 3.458,
       "eval_wer": 1.0,
       "step": 500
     },
     {
       "epoch": 9.99,
-      "learning_rate": 2.2462499999999997e-05,
-      "loss": 2.8989,
+      "learning_rate": 2.8752e-05,
+      "loss": 2.8318,
       "step": 600
     },
     {
       "epoch": 11.66,
-      "learning_rate": 2.6212499999999997e-05,
-      "loss": 2.8318,
+      "learning_rate": 3.3552e-05,
+      "loss": 2.7909,
       "step": 700
     },
     {
       "epoch": 13.33,
-      "learning_rate": 2.99625e-05,
-      "loss": 2.7744,
+      "learning_rate": 3.8352e-05,
+      "loss": 2.7355,
       "step": 800
     },
     {
       "epoch": 14.99,
-      "learning_rate": 3.37125e-05,
-      "loss": 2.7043,
+      "learning_rate": 4.3152e-05,
+      "loss": 2.6796,
       "step": 900
     },
     {
       "epoch": 16.66,
-      "learning_rate": 3.7462499999999996e-05,
-      "loss": 2.6987,
+      "learning_rate": 4.7952000000000004e-05,
+      "loss": 2.6604,
       "step": 1000
     },
     {
       "epoch": 16.66,
-      "eval_loss": 2.687849760055542,
-      "eval_runtime": 32.1535,
-      "eval_samples_per_second": 27.742,
-      "eval_steps_per_second": 3.483,
-      "eval_wer": 1.0,
+      "eval_loss": 2.640192985534668,
+      "eval_runtime": 31.7049,
+      "eval_samples_per_second": 28.134,
+      "eval_steps_per_second": 3.533,
+      "eval_wer": 0.9891806116560877,
       "step": 1000
     },
     {
       "epoch": 18.33,
-      "learning_rate": 4.12125e-05,
-      "loss": 2.6565,
+      "learning_rate": 5.2752e-05,
+      "loss": 2.4114,
       "step": 1100
     },
     {
       "epoch": 19.99,
-      "learning_rate": 4.4962499999999995e-05,
-      "loss": 2.4086,
+      "learning_rate": 5.755200000000001e-05,
+      "loss": 1.675,
       "step": 1200
     },
     {
       "epoch": 21.66,
-      "learning_rate": 4.871249999999999e-05,
-      "loss": 1.7625,
+      "learning_rate": 6.2352e-05,
+      "loss": 1.4118,
       "step": 1300
     },
     {
       "epoch": 23.33,
-      "learning_rate": 5.2462499999999994e-05,
-      "loss": 1.4648,
+      "learning_rate": 6.7152e-05,
+      "loss": 1.291,
       "step": 1400
     },
     {
       "epoch": 24.99,
-      "learning_rate": 5.62125e-05,
-      "loss": 1.3454,
+      "learning_rate": 7.1952e-05,
+      "loss": 1.2216,
       "step": 1500
     },
     {
       "epoch": 24.99,
-      "eval_loss": 0.6813644766807556,
-      "eval_runtime": 32.0646,
-      "eval_samples_per_second": 27.819,
-      "eval_steps_per_second": 3.493,
-      "eval_wer": 0.698066935949221,
+      "eval_loss": 0.6051220893859863,
+      "eval_runtime": 31.8239,
+      "eval_samples_per_second": 28.029,
+      "eval_steps_per_second": 3.519,
+      "eval_wer": 0.6850836699365263,
       "step": 1500
     },
     {
       "epoch": 26.66,
-      "learning_rate": 5.9962499999999994e-05,
-      "loss": 1.2913,
+      "learning_rate": 7.6752e-05,
+      "loss": 1.2115,
       "step": 1600
     },
     {
       "epoch": 28.33,
-      "learning_rate": 6.37125e-05,
-      "loss": 1.2416,
+      "learning_rate": 8.1504e-05,
+      "loss": 1.1555,
       "step": 1700
     },
     {
       "epoch": 29.99,
-      "learning_rate": 6.746249999999999e-05,
-      "loss": 1.1899,
+      "learning_rate": 8.6304e-05,
+      "loss": 1.1142,
       "step": 1800
     },
     {
       "epoch": 31.66,
-      "learning_rate": 7.121249999999999e-05,
-      "loss": 1.1745,
+      "learning_rate": 9.1104e-05,
+      "loss": 1.1029,
       "step": 1900
     },
     {
       "epoch": 33.33,
-      "learning_rate": 7.492499999999999e-05,
-      "loss": 1.1227,
+      "learning_rate": 9.590400000000001e-05,
+      "loss": 1.0754,
       "step": 2000
     },
     {
       "epoch": 33.33,
-      "eval_loss": 0.5790585875511169,
-      "eval_runtime": 32.0814,
-      "eval_samples_per_second": 27.804,
-      "eval_steps_per_second": 3.491,
-      "eval_wer": 0.6513271783035199,
+      "eval_loss": 0.5407921075820923,
+      "eval_runtime": 31.9875,
+      "eval_samples_per_second": 27.886,
+      "eval_steps_per_second": 3.501,
+      "eval_wer": 0.6464223889209464,
       "step": 2000
     },
     {
       "epoch": 34.99,
-      "learning_rate": 7.2375e-05,
-      "loss": 1.0795,
+      "learning_rate": 9.264e-05,
+      "loss": 1.05,
       "step": 2100
     },
     {
       "epoch": 36.66,
-      "learning_rate": 6.972321428571428e-05,
-      "loss": 1.0646,
+      "learning_rate": 8.924571428571428e-05,
+      "loss": 1.0239,
       "step": 2200
     },
     {
       "epoch": 38.33,
-      "learning_rate": 6.704464285714285e-05,
-      "loss": 1.05,
+      "learning_rate": 8.581714285714286e-05,
+      "loss": 1.0023,
       "step": 2300
     },
     {
       "epoch": 39.99,
-      "learning_rate": 6.436607142857142e-05,
-      "loss": 1.0149,
+      "learning_rate": 8.238857142857142e-05,
+      "loss": 0.9695,
       "step": 2400
     },
     {
       "epoch": 41.66,
-      "learning_rate": 6.16875e-05,
-      "loss": 0.9972,
+      "learning_rate": 7.896e-05,
+      "loss": 0.9582,
       "step": 2500
     },
     {
       "epoch": 41.66,
-      "eval_loss": 0.5235142111778259,
-      "eval_runtime": 31.8053,
-      "eval_samples_per_second": 28.046,
-      "eval_steps_per_second": 3.521,
-      "eval_wer": 0.5718407386035776,
+      "eval_loss": 0.5521320700645447,
+      "eval_runtime": 31.6386,
+      "eval_samples_per_second": 28.193,
+      "eval_steps_per_second": 3.54,
+      "eval_wer": 0.5934795152914022,
       "step": 2500
     },
     {
       "epoch": 43.33,
-      "learning_rate": 5.9008928571428565e-05,
-      "loss": 0.9722,
+      "learning_rate": 7.553142857142858e-05,
+      "loss": 0.9312,
       "step": 2600
     },
     {
       "epoch": 44.99,
-      "learning_rate": 5.633035714285714e-05,
-      "loss": 0.9401,
+      "learning_rate": 7.210285714285715e-05,
+      "loss": 0.9136,
       "step": 2700
     },
     {
       "epoch": 46.66,
-      "learning_rate": 5.3651785714285706e-05,
-      "loss": 0.9439,
+      "learning_rate": 6.867428571428571e-05,
+      "loss": 0.9065,
       "step": 2800
     },
     {
       "epoch": 48.33,
-      "learning_rate": 5.0973214285714276e-05,
-      "loss": 0.923,
+      "learning_rate": 6.524571428571428e-05,
+      "loss": 0.8843,
       "step": 2900
     },
     {
       "epoch": 49.99,
-      "learning_rate": 4.829464285714285e-05,
-      "loss": 0.9123,
+      "learning_rate": 6.181714285714286e-05,
+      "loss": 0.8653,
       "step": 3000
     },
     {
       "epoch": 49.99,
-      "eval_loss": 0.5104396939277649,
-      "eval_runtime": 31.7859,
-      "eval_samples_per_second": 28.063,
-      "eval_steps_per_second": 3.524,
-      "eval_wer": 0.5633294864396999,
+      "eval_loss": 0.5156339406967163,
+      "eval_runtime": 31.9186,
+      "eval_samples_per_second": 27.946,
+      "eval_steps_per_second": 3.509,
+      "eval_wer": 0.5549624927870744,
       "step": 3000
     },
     {
       "epoch": 51.66,
-      "learning_rate": 4.561607142857142e-05,
-      "loss": 0.8979,
+      "learning_rate": 5.842285714285714e-05,
+      "loss": 0.8526,
       "step": 3100
     },
     {
       "epoch": 53.33,
-      "learning_rate": 4.29375e-05,
-      "loss": 0.8814,
+      "learning_rate": 5.4994285714285715e-05,
+      "loss": 0.8323,
       "step": 3200
     },
     {
       "epoch": 54.99,
-      "learning_rate": 4.025892857142857e-05,
-      "loss": 0.8548,
+      "learning_rate": 5.156571428571429e-05,
+      "loss": 0.8089,
       "step": 3300
     },
     {
       "epoch": 56.66,
-      "learning_rate": 3.760714285714286e-05,
-      "loss": 0.8515,
+      "learning_rate": 4.813714285714286e-05,
+      "loss": 0.8016,
       "step": 3400
     },
     {
       "epoch": 58.33,
-      "learning_rate": 3.492857142857142e-05,
-      "loss": 0.836,
+      "learning_rate": 4.4708571428571425e-05,
+      "loss": 0.7867,
       "step": 3500
     },
     {
       "epoch": 58.33,
-      "eval_loss": 0.49266090989112854,
-      "eval_runtime": 31.7464,
-      "eval_samples_per_second": 28.098,
-      "eval_steps_per_second": 3.528,
-      "eval_wer": 0.5579919215233698,
+      "eval_loss": 0.543922483921051,
+      "eval_runtime": 31.6944,
+      "eval_samples_per_second": 28.144,
+      "eval_steps_per_second": 3.534,
+      "eval_wer": 0.5605885747259088,
       "step": 3500
     },
     {
       "epoch": 59.99,
-      "learning_rate": 3.225e-05,
-      "loss": 0.8045,
+      "learning_rate": 4.128e-05,
+      "loss": 0.7598,
       "step": 3600
     },
     {
       "epoch": 61.66,
-      "learning_rate": 2.9571428571428568e-05,
-      "loss": 0.8036,
+      "learning_rate": 3.785142857142857e-05,
+      "loss": 0.7582,
       "step": 3700
     },
     {
       "epoch": 63.33,
-      "learning_rate": 2.689285714285714e-05,
-      "loss": 0.7924,
+      "learning_rate": 3.442285714285715e-05,
+      "loss": 0.7423,
       "step": 3800
     },
     {
       "epoch": 64.99,
-      "learning_rate": 2.4214285714285712e-05,
-      "loss": 0.7837,
+      "learning_rate": 3.0994285714285715e-05,
+      "loss": 0.7313,
       "step": 3900
     },
     {
       "epoch": 66.66,
-      "learning_rate": 2.1535714285714285e-05,
-      "loss": 0.7725,
+      "learning_rate": 2.7565714285714287e-05,
+      "loss": 0.7265,
       "step": 4000
     },
     {
       "epoch": 66.66,
-      "eval_loss": 0.5077508687973022,
-      "eval_runtime": 31.8976,
-      "eval_samples_per_second": 27.964,
-      "eval_steps_per_second": 3.511,
-      "eval_wer": 0.5778995960761685,
+      "eval_loss": 0.48627379536628723,
+      "eval_runtime": 31.7498,
+      "eval_samples_per_second": 28.095,
+      "eval_steps_per_second": 3.528,
+      "eval_wer": 0.525533756491633,
       "step": 4000
     },
     {
       "epoch": 68.33,
-      "learning_rate": 1.888392857142857e-05,
-      "loss": 0.7655,
+      "learning_rate": 2.413714285714286e-05,
+      "loss": 0.7134,
       "step": 4100
     },
     {
       "epoch": 69.99,
-      "learning_rate": 1.6205357142857143e-05,
-      "loss": 0.741,
+      "learning_rate": 2.0708571428571428e-05,
+      "loss": 0.6867,
       "step": 4200
     },
     {
       "epoch": 71.66,
-      "learning_rate": 1.3526785714285713e-05,
-      "loss": 0.7499,
+      "learning_rate": 1.728e-05,
+      "loss": 0.6987,
       "step": 4300
     },
     {
       "epoch": 73.33,
-      "learning_rate": 1.0848214285714287e-05,
-      "loss": 0.7479,
+      "learning_rate": 1.3851428571428573e-05,
+      "loss": 0.6976,
       "step": 4400
     },
     {
       "epoch": 74.99,
-      "learning_rate": 8.169642857142857e-06,
-      "loss": 0.7297,
+      "learning_rate": 1.0422857142857143e-05,
+      "loss": 0.6699,
       "step": 4500
     },
     {
       "epoch": 74.99,
-      "eval_loss": 0.4939458966255188,
-      "eval_runtime": 31.9903,
-      "eval_samples_per_second": 27.883,
-      "eval_steps_per_second": 3.501,
-      "eval_wer": 0.5737160992498558,
+      "eval_loss": 0.5050373673439026,
+      "eval_runtime": 31.7366,
+      "eval_samples_per_second": 28.106,
+      "eval_steps_per_second": 3.529,
+      "eval_wer": 0.5168782458165032,
       "step": 4500
     },
     {
       "epoch": 76.66,
-      "learning_rate": 5.491071428571429e-06,
-      "loss": 0.7318,
+      "learning_rate": 6.994285714285714e-06,
+      "loss": 0.6777,
       "step": 4600
     },
     {
       "epoch": 78.33,
-      "learning_rate": 2.8124999999999998e-06,
-      "loss": 0.7232,
+      "learning_rate": 3.565714285714286e-06,
+      "loss": 0.6669,
       "step": 4700
     },
     {
       "epoch": 79.99,
-      "learning_rate": 1.3392857142857142e-07,
-      "loss": 0.7195,
+      "learning_rate": 1.3714285714285715e-07,
+      "loss": 0.6602,
       "step": 4800
     },
     {
       "epoch": 79.99,
       "step": 4800,
       "total_flos": 2.017023736432276e+19,
-      "train_loss": 1.795632479985555,
-      "train_runtime": 8730.4816,
-      "train_samples_per_second": 17.786,
-      "train_steps_per_second": 0.55
+      "train_loss": 1.6836539268493653,
+      "train_runtime": 8656.8344,
+      "train_samples_per_second": 17.937,
+      "train_steps_per_second": 0.554
     }
   ],
   "max_steps": 4800,