End of training

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 89.98,
-    "eval_loss": 0.7423784136772156,
-    "eval_runtime": 31.8664,
+    "epoch": 79.99,
+    "eval_loss": 0.49252477288246155,
+    "eval_runtime": 31.5756,
     "eval_samples": 892,
-    "eval_samples_per_second": 27.992,
-    "eval_steps_per_second": 3.515,
-    "eval_wer": 0.6190132717830352,
-    "train_loss": 1.793763725845902,
-    "train_runtime": 8224.5964,
+    "eval_samples_per_second": 28.25,
+    "eval_steps_per_second": 3.547,
+    "eval_wer": 0.5714079630698211,
+    "train_loss": 1.795632479985555,
+    "train_runtime": 8730.4816,
     "train_samples": 1941,
-    "train_samples_per_second": 21.24,
-    "train_steps_per_second": 0.328
+    "train_samples_per_second": 17.786,
+    "train_steps_per_second": 0.55
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 89.98,
-    "eval_loss": 0.7423784136772156,
-    "eval_runtime": 31.8664,
+    "epoch": 79.99,
+    "eval_loss": 0.49252477288246155,
+    "eval_runtime": 31.5756,
     "eval_samples": 892,
-    "eval_samples_per_second": 27.992,
-    "eval_steps_per_second": 3.515,
-    "eval_wer": 0.6190132717830352
+    "eval_samples_per_second": 28.25,
+    "eval_steps_per_second": 3.547,
+    "eval_wer": 0.5714079630698211
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 89.98,
-    "train_loss": 1.793763725845902,
-    "train_runtime": 8224.5964,
+    "epoch": 79.99,
+    "train_loss": 1.795632479985555,
+    "train_runtime": 8730.4816,
     "train_samples": 1941,
-    "train_samples_per_second": 21.24,
-    "train_steps_per_second": 0.328
+    "train_samples_per_second": 17.786,
+    "train_steps_per_second": 0.55
 }

trainer_state.json

@@ -1,232 +1,394 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 89.98360655737704,
-  "global_step": 2700,
+  "epoch": 79.98765432098766,
+  "global_step": 4800,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 3.33,
-      "learning_rate": 1.485e-05,
-      "loss": 11.2315,
+      "epoch": 1.66,
+      "learning_rate": 3.7125e-06,
+      "loss": 14.5868,
       "step": 100
     },
     {
-      "epoch": 6.66,
-      "learning_rate": 2.985e-05,
-      "loss": 4.0404,
+      "epoch": 3.33,
+      "learning_rate": 7.4625e-06,
+      "loss": 6.8756,
       "step": 200
     },
     {
-      "epoch": 9.98,
-      "learning_rate": 4.484999999999999e-05,
-      "loss": 3.056,
+      "epoch": 4.99,
+      "learning_rate": 1.1212499999999998e-05,
+      "loss": 4.2978,
       "step": 300
     },
     {
-      "epoch": 13.33,
-      "learning_rate": 5.985e-05,
-      "loss": 2.8167,
+      "epoch": 6.66,
+      "learning_rate": 1.49625e-05,
+      "loss": 3.6126,
       "step": 400
     },
     {
-      "epoch": 16.66,
-      "learning_rate": 7.484999999999999e-05,
-      "loss": 2.721,
+      "epoch": 8.33,
+      "learning_rate": 1.8712499999999997e-05,
+      "loss": 3.1674,
       "step": 500
     },
     {
-      "epoch": 16.66,
-      "eval_loss": 2.6876425743103027,
-      "eval_runtime": 32.088,
-      "eval_samples_per_second": 27.799,
-      "eval_steps_per_second": 3.49,
+      "epoch": 8.33,
+      "eval_loss": 3.0295047760009766,
+      "eval_runtime": 32.2697,
+      "eval_samples_per_second": 27.642,
+      "eval_steps_per_second": 3.471,
       "eval_wer": 1.0,
       "step": 500
     },
     {
-      "epoch": 19.98,
-      "learning_rate": 8.984999999999999e-05,
-      "loss": 2.6493,
+      "epoch": 9.99,
+      "learning_rate": 2.2462499999999997e-05,
+      "loss": 2.8989,
       "step": 600
     },
     {
-      "epoch": 23.33,
-      "learning_rate": 0.00010484999999999999,
-      "loss": 1.9825,
+      "epoch": 11.66,
+      "learning_rate": 2.6212499999999997e-05,
+      "loss": 2.8318,
       "step": 700
     },
     {
-      "epoch": 26.66,
-      "learning_rate": 0.00011985,
-      "loss": 1.4345,
+      "epoch": 13.33,
+      "learning_rate": 2.99625e-05,
+      "loss": 2.7744,
       "step": 800
     },
     {
-      "epoch": 29.98,
-      "learning_rate": 0.00013485,
-      "loss": 1.3429,
+      "epoch": 14.99,
+      "learning_rate": 3.37125e-05,
+      "loss": 2.7043,
       "step": 900
     },
     {
-      "epoch": 33.33,
-      "learning_rate": 0.00014954999999999998,
-      "loss": 1.2944,
+      "epoch": 16.66,
+      "learning_rate": 3.7462499999999996e-05,
+      "loss": 2.6987,
       "step": 1000
     },
     {
-      "epoch": 33.33,
-      "eval_loss": 0.6482492089271545,
-      "eval_runtime": 32.0457,
-      "eval_samples_per_second": 27.835,
-      "eval_steps_per_second": 3.495,
-      "eval_wer": 0.7135025966532026,
+      "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,
       "step": 1000
     },
     {
-      "epoch": 36.66,
-      "learning_rate": 0.00016455,
-      "loss": 1.2347,
+      "epoch": 18.33,
+      "learning_rate": 4.12125e-05,
+      "loss": 2.6565,
       "step": 1100
     },
     {
-      "epoch": 39.98,
-      "learning_rate": 0.00017955,
-      "loss": 1.1858,
+      "epoch": 19.99,
+      "learning_rate": 4.4962499999999995e-05,
+      "loss": 2.4086,
       "step": 1200
     },
     {
-      "epoch": 43.33,
-      "learning_rate": 0.00019454999999999999,
-      "loss": 1.1475,
+      "epoch": 21.66,
+      "learning_rate": 4.871249999999999e-05,
+      "loss": 1.7625,
       "step": 1300
     },
     {
-      "epoch": 46.66,
-      "learning_rate": 0.00020955,
-      "loss": 1.0875,
+      "epoch": 23.33,
+      "learning_rate": 5.2462499999999994e-05,
+      "loss": 1.4648,
       "step": 1400
     },
     {
-      "epoch": 49.98,
-      "learning_rate": 0.00022455,
-      "loss": 1.0515,
+      "epoch": 24.99,
+      "learning_rate": 5.62125e-05,
+      "loss": 1.3454,
       "step": 1500
     },
     {
-      "epoch": 49.98,
-      "eval_loss": 0.6644838452339172,
-      "eval_runtime": 31.726,
-      "eval_samples_per_second": 28.116,
-      "eval_steps_per_second": 3.53,
-      "eval_wer": 0.6754183496826313,
+      "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,
       "step": 1500
     },
     {
-      "epoch": 53.33,
-      "learning_rate": 0.00023954999999999997,
-      "loss": 1.0217,
+      "epoch": 26.66,
+      "learning_rate": 5.9962499999999994e-05,
+      "loss": 1.2913,
       "step": 1600
     },
     {
-      "epoch": 56.66,
-      "learning_rate": 0.00025455,
-      "loss": 0.9991,
+      "epoch": 28.33,
+      "learning_rate": 6.37125e-05,
+      "loss": 1.2416,
       "step": 1700
     },
     {
-      "epoch": 59.98,
-      "learning_rate": 0.00026954999999999997,
-      "loss": 0.9918,
+      "epoch": 29.99,
+      "learning_rate": 6.746249999999999e-05,
+      "loss": 1.1899,
       "step": 1800
     },
     {
-      "epoch": 63.33,
-      "learning_rate": 0.00028455,
-      "loss": 0.9552,
+      "epoch": 31.66,
+      "learning_rate": 7.121249999999999e-05,
+      "loss": 1.1745,
       "step": 1900
     },
     {
-      "epoch": 66.66,
-      "learning_rate": 0.00029955,
-      "loss": 0.9153,
+      "epoch": 33.33,
+      "learning_rate": 7.492499999999999e-05,
+      "loss": 1.1227,
       "step": 2000
     },
     {
-      "epoch": 66.66,
-      "eval_loss": 0.7648739814758301,
-      "eval_runtime": 31.6976,
-      "eval_samples_per_second": 28.141,
-      "eval_steps_per_second": 3.533,
-      "eval_wer": 0.693594922100404,
+      "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,
       "step": 2000
     },
     {
-      "epoch": 69.98,
-      "learning_rate": 0.0002584285714285714,
-      "loss": 0.8827,
+      "epoch": 34.99,
+      "learning_rate": 7.2375e-05,
+      "loss": 1.0795,
       "step": 2100
     },
     {
-      "epoch": 73.33,
-      "learning_rate": 0.00021557142857142855,
-      "loss": 0.8474,
+      "epoch": 36.66,
+      "learning_rate": 6.972321428571428e-05,
+      "loss": 1.0646,
       "step": 2200
     },
     {
-      "epoch": 76.66,
-      "learning_rate": 0.0001727142857142857,
-      "loss": 0.8016,
+      "epoch": 38.33,
+      "learning_rate": 6.704464285714285e-05,
+      "loss": 1.05,
       "step": 2300
     },
     {
-      "epoch": 79.98,
-      "learning_rate": 0.00012985714285714285,
-      "loss": 0.7467,
+      "epoch": 39.99,
+      "learning_rate": 6.436607142857142e-05,
+      "loss": 1.0149,
       "step": 2400
     },
     {
-      "epoch": 83.33,
-      "learning_rate": 8.699999999999999e-05,
-      "loss": 0.7056,
+      "epoch": 41.66,
+      "learning_rate": 6.16875e-05,
+      "loss": 0.9972,
       "step": 2500
     },
     {
-      "epoch": 83.33,
-      "eval_loss": 0.8047966361045837,
-      "eval_runtime": 31.4528,
-      "eval_samples_per_second": 28.36,
-      "eval_steps_per_second": 3.561,
-      "eval_wer": 0.6755626081938835,
+      "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,
       "step": 2500
     },
     {
-      "epoch": 86.66,
-      "learning_rate": 4.414285714285714e-05,
-      "loss": 0.6637,
+      "epoch": 43.33,
+      "learning_rate": 5.9008928571428565e-05,
+      "loss": 0.9722,
       "step": 2600
     },
     {
-      "epoch": 89.98,
-      "learning_rate": 1.2857142857142856e-06,
-      "loss": 0.6249,
+      "epoch": 44.99,
+      "learning_rate": 5.633035714285714e-05,
+      "loss": 0.9401,
       "step": 2700
     },
     {
-      "epoch": 89.98,
-      "step": 2700,
-      "total_flos": 2.311805909026184e+19,
-      "train_loss": 1.793763725845902,
-      "train_runtime": 8224.5964,
-      "train_samples_per_second": 21.24,
-      "train_steps_per_second": 0.328
+      "epoch": 46.66,
+      "learning_rate": 5.3651785714285706e-05,
+      "loss": 0.9439,
+      "step": 2800
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 5.0973214285714276e-05,
+      "loss": 0.923,
+      "step": 2900
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 4.829464285714285e-05,
+      "loss": 0.9123,
+      "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,
+      "step": 3000
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 4.561607142857142e-05,
+      "loss": 0.8979,
+      "step": 3100
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.29375e-05,
+      "loss": 0.8814,
+      "step": 3200
+    },
+    {
+      "epoch": 54.99,
+      "learning_rate": 4.025892857142857e-05,
+      "loss": 0.8548,
+      "step": 3300
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 3.760714285714286e-05,
+      "loss": 0.8515,
+      "step": 3400
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 3.492857142857142e-05,
+      "loss": 0.836,
+      "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,
+      "step": 3500
+    },
+    {
+      "epoch": 59.99,
+      "learning_rate": 3.225e-05,
+      "loss": 0.8045,
+      "step": 3600
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 2.9571428571428568e-05,
+      "loss": 0.8036,
+      "step": 3700
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 2.689285714285714e-05,
+      "loss": 0.7924,
+      "step": 3800
+    },
+    {
+      "epoch": 64.99,
+      "learning_rate": 2.4214285714285712e-05,
+      "loss": 0.7837,
+      "step": 3900
+    },
+    {
+      "epoch": 66.66,
+      "learning_rate": 2.1535714285714285e-05,
+      "loss": 0.7725,
+      "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,
+      "step": 4000
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 1.888392857142857e-05,
+      "loss": 0.7655,
+      "step": 4100
+    },
+    {
+      "epoch": 69.99,
+      "learning_rate": 1.6205357142857143e-05,
+      "loss": 0.741,
+      "step": 4200
+    },
+    {
+      "epoch": 71.66,
+      "learning_rate": 1.3526785714285713e-05,
+      "loss": 0.7499,
+      "step": 4300
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 1.0848214285714287e-05,
+      "loss": 0.7479,
+      "step": 4400
+    },
+    {
+      "epoch": 74.99,
+      "learning_rate": 8.169642857142857e-06,
+      "loss": 0.7297,
+      "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,
+      "step": 4500
+    },
+    {
+      "epoch": 76.66,
+      "learning_rate": 5.491071428571429e-06,
+      "loss": 0.7318,
+      "step": 4600
+    },
+    {
+      "epoch": 78.33,
+      "learning_rate": 2.8124999999999998e-06,
+      "loss": 0.7232,
+      "step": 4700
+    },
+    {
+      "epoch": 79.99,
+      "learning_rate": 1.3392857142857142e-07,
+      "loss": 0.7195,
+      "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
     }
   ],
-  "max_steps": 2700,
-  "num_train_epochs": 90,
-  "total_flos": 2.311805909026184e+19,
+  "max_steps": 4800,
+  "num_train_epochs": 80,
+  "total_flos": 2.017023736432276e+19,
   "trial_name": null,
   "trial_params": null
 }