End of training

all_results.json

@@ -1,12 +1,12 @@
 {
-    "epoch": 528.76,
+    "epoch": 587.82,
     "eval_loss": 0.56689453125,
-    "eval_runtime": 156.403,
-    "eval_samples_per_second": 1.739,
-    "eval_steps_per_second": 0.109,
+    "eval_runtime": 154.2304,
+    "eval_samples_per_second": 1.764,
+    "eval_steps_per_second": 0.11,
     "eval_wer": 9.899702823179792,
-    "train_loss": 0.00025936167769961886,
-    "train_runtime": 28092.3517,
-    "train_samples_per_second": 10.252,
-    "train_steps_per_second": 0.32
+    "train_loss": 1.4810419082641601e-05,
+    "train_runtime": 3866.6425,
+    "train_samples_per_second": 82.759,
+    "train_steps_per_second": 2.586
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 528.76,
+    "epoch": 587.82,
     "eval_loss": 0.56689453125,
-    "eval_runtime": 156.403,
-    "eval_samples_per_second": 1.739,
-    "eval_steps_per_second": 0.109,
+    "eval_runtime": 154.2304,
+    "eval_samples_per_second": 1.764,
+    "eval_steps_per_second": 0.11,
     "eval_wer": 9.899702823179792
 }

runs/Dec23_07-12-50_129-146-176-120/events.out.tfevents.1671783758.129-146-176-120.858176.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00d6b23a4face7948649a31b01be2108827f32cf3dc24574e56cf90543af30e9
+size 358

train_results.json

@@ -1,7 +1,7 @@
 {
-    "epoch": 528.76,
-    "train_loss": 0.00025936167769961886,
-    "train_runtime": 28092.3517,
-    "train_samples_per_second": 10.252,
-    "train_steps_per_second": 0.32
+    "epoch": 587.82,
+    "train_loss": 1.4810419082641601e-05,
+    "train_runtime": 3866.6425,
+    "train_samples_per_second": 82.759,
+    "train_steps_per_second": 2.586
 }

trainer_state.json

@@ -1,8 +1,8 @@
 {
   "best_metric": 9.899702823179792,
   "best_model_checkpoint": "./checkpoint-8000",
-  "epoch": 528.7647058823529,
-  "global_step": 9000,
+  "epoch": 587.8235294117648,
+  "global_step": 10000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -2249,18 +2249,267 @@
       "step": 9000
     },
     {
-      "epoch": 528.76,
-      "step": 9000,
-      "total_flos": 2.7722894647818505e+20,
-      "train_loss": 0.00025936167769961886,
-      "train_runtime": 28092.3517,
-      "train_samples_per_second": 10.252,
-      "train_steps_per_second": 0.32
+      "epoch": 530.47,
+      "learning_rate": 3.145263157894737e-07,
+      "loss": 0.0001,
+      "step": 9025
+    },
+    {
+      "epoch": 531.94,
+      "learning_rate": 3.0663157894736844e-07,
+      "loss": 0.0002,
+      "step": 9050
+    },
+    {
+      "epoch": 533.41,
+      "learning_rate": 2.987368421052632e-07,
+      "loss": 0.0002,
+      "step": 9075
+    },
+    {
+      "epoch": 534.88,
+      "learning_rate": 2.908421052631579e-07,
+      "loss": 0.0001,
+      "step": 9100
+    },
+    {
+      "epoch": 536.35,
+      "learning_rate": 2.829473684210526e-07,
+      "loss": 0.0001,
+      "step": 9125
+    },
+    {
+      "epoch": 537.82,
+      "learning_rate": 2.7505263157894737e-07,
+      "loss": 0.0003,
+      "step": 9150
+    },
+    {
+      "epoch": 539.29,
+      "learning_rate": 2.671578947368421e-07,
+      "loss": 0.0002,
+      "step": 9175
+    },
+    {
+      "epoch": 540.76,
+      "learning_rate": 2.5926315789473686e-07,
+      "loss": 0.0002,
+      "step": 9200
+    },
+    {
+      "epoch": 542.24,
+      "learning_rate": 2.513684210526316e-07,
+      "loss": 0.0001,
+      "step": 9225
+    },
+    {
+      "epoch": 543.71,
+      "learning_rate": 2.4347368421052635e-07,
+      "loss": 0.0002,
+      "step": 9250
+    },
+    {
+      "epoch": 545.18,
+      "learning_rate": 2.3557894736842104e-07,
+      "loss": 0.0001,
+      "step": 9275
+    },
+    {
+      "epoch": 546.65,
+      "learning_rate": 2.276842105263158e-07,
+      "loss": 0.0002,
+      "step": 9300
+    },
+    {
+      "epoch": 548.12,
+      "learning_rate": 2.197894736842105e-07,
+      "loss": 0.0002,
+      "step": 9325
+    },
+    {
+      "epoch": 549.59,
+      "learning_rate": 2.1189473684210525e-07,
+      "loss": 0.0001,
+      "step": 9350
+    },
+    {
+      "epoch": 551.06,
+      "learning_rate": 2.0400000000000003e-07,
+      "loss": 0.0001,
+      "step": 9375
+    },
+    {
+      "epoch": 552.53,
+      "learning_rate": 1.9610526315789472e-07,
+      "loss": 0.0001,
+      "step": 9400
+    },
+    {
+      "epoch": 554.0,
+      "learning_rate": 1.882105263157895e-07,
+      "loss": 0.0002,
+      "step": 9425
+    },
+    {
+      "epoch": 555.47,
+      "learning_rate": 1.803157894736842e-07,
+      "loss": 0.0002,
+      "step": 9450
+    },
+    {
+      "epoch": 556.94,
+      "learning_rate": 1.7242105263157896e-07,
+      "loss": 0.0001,
+      "step": 9475
+    },
+    {
+      "epoch": 558.41,
+      "learning_rate": 1.6452631578947368e-07,
+      "loss": 0.0002,
+      "step": 9500
+    },
+    {
+      "epoch": 559.88,
+      "learning_rate": 1.5663157894736842e-07,
+      "loss": 0.0001,
+      "step": 9525
+    },
+    {
+      "epoch": 561.35,
+      "learning_rate": 1.4873684210526317e-07,
+      "loss": 0.0001,
+      "step": 9550
+    },
+    {
+      "epoch": 562.82,
+      "learning_rate": 1.408421052631579e-07,
+      "loss": 0.0001,
+      "step": 9575
+    },
+    {
+      "epoch": 564.29,
+      "learning_rate": 1.3294736842105263e-07,
+      "loss": 0.0002,
+      "step": 9600
+    },
+    {
+      "epoch": 565.76,
+      "learning_rate": 1.2505263157894738e-07,
+      "loss": 0.0001,
+      "step": 9625
+    },
+    {
+      "epoch": 567.24,
+      "learning_rate": 1.1715789473684211e-07,
+      "loss": 0.0002,
+      "step": 9650
+    },
+    {
+      "epoch": 568.71,
+      "learning_rate": 1.0926315789473684e-07,
+      "loss": 0.0001,
+      "step": 9675
+    },
+    {
+      "epoch": 570.18,
+      "learning_rate": 1.0136842105263159e-07,
+      "loss": 0.0001,
+      "step": 9700
+    },
+    {
+      "epoch": 571.65,
+      "learning_rate": 9.347368421052632e-08,
+      "loss": 0.0001,
+      "step": 9725
+    },
+    {
+      "epoch": 573.12,
+      "learning_rate": 8.557894736842105e-08,
+      "loss": 0.0001,
+      "step": 9750
+    },
+    {
+      "epoch": 574.59,
+      "learning_rate": 7.768421052631578e-08,
+      "loss": 0.0002,
+      "step": 9775
+    },
+    {
+      "epoch": 576.06,
+      "learning_rate": 6.978947368421053e-08,
+      "loss": 0.0001,
+      "step": 9800
+    },
+    {
+      "epoch": 577.53,
+      "learning_rate": 6.189473684210526e-08,
+      "loss": 0.0002,
+      "step": 9825
+    },
+    {
+      "epoch": 579.0,
+      "learning_rate": 5.4e-08,
+      "loss": 0.0001,
+      "step": 9850
+    },
+    {
+      "epoch": 580.47,
+      "learning_rate": 4.610526315789473e-08,
+      "loss": 0.0001,
+      "step": 9875
+    },
+    {
+      "epoch": 581.94,
+      "learning_rate": 3.821052631578947e-08,
+      "loss": 0.0002,
+      "step": 9900
+    },
+    {
+      "epoch": 583.41,
+      "learning_rate": 3.031578947368421e-08,
+      "loss": 0.0001,
+      "step": 9925
+    },
+    {
+      "epoch": 584.88,
+      "learning_rate": 2.2421052631578946e-08,
+      "loss": 0.0002,
+      "step": 9950
+    },
+    {
+      "epoch": 586.35,
+      "learning_rate": 1.4526315789473685e-08,
+      "loss": 0.0001,
+      "step": 9975
+    },
+    {
+      "epoch": 587.82,
+      "learning_rate": 6.631578947368421e-09,
+      "loss": 0.0001,
+      "step": 10000
+    },
+    {
+      "epoch": 587.82,
+      "eval_loss": 0.5673828125,
+      "eval_runtime": 153.8197,
+      "eval_samples_per_second": 1.768,
+      "eval_steps_per_second": 0.111,
+      "eval_wer": 9.964710252600298,
+      "step": 10000
+    },
+    {
+      "epoch": 587.82,
+      "step": 10000,
+      "total_flos": 3.080532552348504e+20,
+      "train_loss": 1.4810419082641601e-05,
+      "train_runtime": 3866.6425,
+      "train_samples_per_second": 82.759,
+      "train_steps_per_second": 2.586
     }
   ],
-  "max_steps": 9000,
-  "num_train_epochs": 530,
-  "total_flos": 2.7722894647818505e+20,
+  "max_steps": 10000,
+  "num_train_epochs": 589,
+  "total_flos": 3.080532552348504e+20,
   "trial_name": null,
   "trial_params": null
 }