"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629968901.3732219/events.out.tfevents.1629968901.8e89bd551565.924.131 +3 -0
- model-bin/finetune/base/log/1629969336.52509/events.out.tfevents.1629969336.8e89bd551565.924.133 +3 -0
- model-bin/finetune/base/log/1629969764.0650191/events.out.tfevents.1629969764.8e89bd551565.924.135 +3 -0
- model-bin/finetune/base/log/1629970202.6078649/events.out.tfevents.1629970202.8e89bd551565.924.137 +3 -0
- model-bin/finetune/base/log/1629970642.6016197/events.out.tfevents.1629970642.8e89bd551565.924.139 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629968901.8e89bd551565.924.130 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629969335.8e89bd551565.924.132 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629969764.8e89bd551565.924.134 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629970202.8e89bd551565.924.136 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629970642.8e89bd551565.924.138 +3 -0
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c10579dbc985ee5912a553b1b2b1fd66bdf101b6be5020bd5700ff86fc949b9a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb5d57ca4f3a500e8647f8c0c45914eeb02139ae9a13e5b84416f054c7eb1874
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5f4d1a1cee938b18782fcec036ae0ba7a7733390d04434995a7aac778bf86e4
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6dbc4cfdd127e6a648679cffb1cc66e95f2661ad2f1bd13a18b931479d8bf841
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5727fff892b707138e7252948e475b0ded1a0009fa8dff119d069a9df135194
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -222075,11 +222075,806 @@
|
|
| 222075 |
"eval_steps_per_second": 0.664,
|
| 222076 |
"eval_wer": 0.18247860174063152,
|
| 222077 |
"step": 96447
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 222078 |
}
|
| 222079 |
],
|
| 222080 |
"max_steps": 625000,
|
| 222081 |
"num_train_epochs": 5000,
|
| 222082 |
-
"total_flos": 2.
|
| 222083 |
"trial_name": null,
|
| 222084 |
"trial_params": null
|
| 222085 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 776.0,
|
| 5 |
+
"global_step": 97072,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 222075 |
"eval_steps_per_second": 0.664,
|
| 222076 |
"eval_wer": 0.18247860174063152,
|
| 222077 |
"step": 96447
|
| 222078 |
+
},
|
| 222079 |
+
{
|
| 222080 |
+
"epoch": 771.02,
|
| 222081 |
+
"learning_rate": 8.470961538461539e-06,
|
| 222082 |
+
"loss": 0.3421,
|
| 222083 |
+
"step": 96450
|
| 222084 |
+
},
|
| 222085 |
+
{
|
| 222086 |
+
"epoch": 771.06,
|
| 222087 |
+
"learning_rate": 8.47088141025641e-06,
|
| 222088 |
+
"loss": 0.2922,
|
| 222089 |
+
"step": 96455
|
| 222090 |
+
},
|
| 222091 |
+
{
|
| 222092 |
+
"epoch": 771.1,
|
| 222093 |
+
"learning_rate": 8.470801282051283e-06,
|
| 222094 |
+
"loss": 0.2784,
|
| 222095 |
+
"step": 96460
|
| 222096 |
+
},
|
| 222097 |
+
{
|
| 222098 |
+
"epoch": 771.14,
|
| 222099 |
+
"learning_rate": 8.470721153846155e-06,
|
| 222100 |
+
"loss": 0.3606,
|
| 222101 |
+
"step": 96465
|
| 222102 |
+
},
|
| 222103 |
+
{
|
| 222104 |
+
"epoch": 771.18,
|
| 222105 |
+
"learning_rate": 8.470641025641026e-06,
|
| 222106 |
+
"loss": 0.6556,
|
| 222107 |
+
"step": 96470
|
| 222108 |
+
},
|
| 222109 |
+
{
|
| 222110 |
+
"epoch": 771.22,
|
| 222111 |
+
"learning_rate": 8.470560897435897e-06,
|
| 222112 |
+
"loss": 0.9834,
|
| 222113 |
+
"step": 96475
|
| 222114 |
+
},
|
| 222115 |
+
{
|
| 222116 |
+
"epoch": 771.26,
|
| 222117 |
+
"learning_rate": 8.47048076923077e-06,
|
| 222118 |
+
"loss": 0.3059,
|
| 222119 |
+
"step": 96480
|
| 222120 |
+
},
|
| 222121 |
+
{
|
| 222122 |
+
"epoch": 771.3,
|
| 222123 |
+
"learning_rate": 8.470400641025642e-06,
|
| 222124 |
+
"loss": 0.4721,
|
| 222125 |
+
"step": 96485
|
| 222126 |
+
},
|
| 222127 |
+
{
|
| 222128 |
+
"epoch": 771.34,
|
| 222129 |
+
"learning_rate": 8.470320512820513e-06,
|
| 222130 |
+
"loss": 0.4288,
|
| 222131 |
+
"step": 96490
|
| 222132 |
+
},
|
| 222133 |
+
{
|
| 222134 |
+
"epoch": 771.38,
|
| 222135 |
+
"learning_rate": 8.470240384615386e-06,
|
| 222136 |
+
"loss": 0.7838,
|
| 222137 |
+
"step": 96495
|
| 222138 |
+
},
|
| 222139 |
+
{
|
| 222140 |
+
"epoch": 771.42,
|
| 222141 |
+
"learning_rate": 8.470160256410258e-06,
|
| 222142 |
+
"loss": 0.9725,
|
| 222143 |
+
"step": 96500
|
| 222144 |
+
},
|
| 222145 |
+
{
|
| 222146 |
+
"epoch": 771.46,
|
| 222147 |
+
"learning_rate": 8.470080128205129e-06,
|
| 222148 |
+
"loss": 0.3023,
|
| 222149 |
+
"step": 96505
|
| 222150 |
+
},
|
| 222151 |
+
{
|
| 222152 |
+
"epoch": 771.5,
|
| 222153 |
+
"learning_rate": 8.47e-06,
|
| 222154 |
+
"loss": 0.2835,
|
| 222155 |
+
"step": 96510
|
| 222156 |
+
},
|
| 222157 |
+
{
|
| 222158 |
+
"epoch": 771.54,
|
| 222159 |
+
"learning_rate": 8.469919871794873e-06,
|
| 222160 |
+
"loss": 0.4104,
|
| 222161 |
+
"step": 96515
|
| 222162 |
+
},
|
| 222163 |
+
{
|
| 222164 |
+
"epoch": 771.58,
|
| 222165 |
+
"learning_rate": 8.469839743589745e-06,
|
| 222166 |
+
"loss": 0.6912,
|
| 222167 |
+
"step": 96520
|
| 222168 |
+
},
|
| 222169 |
+
{
|
| 222170 |
+
"epoch": 771.62,
|
| 222171 |
+
"learning_rate": 8.469759615384616e-06,
|
| 222172 |
+
"loss": 0.8181,
|
| 222173 |
+
"step": 96525
|
| 222174 |
+
},
|
| 222175 |
+
{
|
| 222176 |
+
"epoch": 771.66,
|
| 222177 |
+
"learning_rate": 8.469679487179487e-06,
|
| 222178 |
+
"loss": 0.3261,
|
| 222179 |
+
"step": 96530
|
| 222180 |
+
},
|
| 222181 |
+
{
|
| 222182 |
+
"epoch": 771.7,
|
| 222183 |
+
"learning_rate": 8.46959935897436e-06,
|
| 222184 |
+
"loss": 0.2919,
|
| 222185 |
+
"step": 96535
|
| 222186 |
+
},
|
| 222187 |
+
{
|
| 222188 |
+
"epoch": 771.74,
|
| 222189 |
+
"learning_rate": 8.469519230769232e-06,
|
| 222190 |
+
"loss": 0.3659,
|
| 222191 |
+
"step": 96540
|
| 222192 |
+
},
|
| 222193 |
+
{
|
| 222194 |
+
"epoch": 771.78,
|
| 222195 |
+
"learning_rate": 8.469439102564103e-06,
|
| 222196 |
+
"loss": 0.875,
|
| 222197 |
+
"step": 96545
|
| 222198 |
+
},
|
| 222199 |
+
{
|
| 222200 |
+
"epoch": 771.82,
|
| 222201 |
+
"learning_rate": 8.469358974358976e-06,
|
| 222202 |
+
"loss": 0.8602,
|
| 222203 |
+
"step": 96550
|
| 222204 |
+
},
|
| 222205 |
+
{
|
| 222206 |
+
"epoch": 771.86,
|
| 222207 |
+
"learning_rate": 8.469278846153846e-06,
|
| 222208 |
+
"loss": 0.3233,
|
| 222209 |
+
"step": 96555
|
| 222210 |
+
},
|
| 222211 |
+
{
|
| 222212 |
+
"epoch": 771.9,
|
| 222213 |
+
"learning_rate": 8.469198717948719e-06,
|
| 222214 |
+
"loss": 0.2426,
|
| 222215 |
+
"step": 96560
|
| 222216 |
+
},
|
| 222217 |
+
{
|
| 222218 |
+
"epoch": 771.94,
|
| 222219 |
+
"learning_rate": 8.46911858974359e-06,
|
| 222220 |
+
"loss": 0.3388,
|
| 222221 |
+
"step": 96565
|
| 222222 |
+
},
|
| 222223 |
+
{
|
| 222224 |
+
"epoch": 771.98,
|
| 222225 |
+
"learning_rate": 8.469038461538462e-06,
|
| 222226 |
+
"loss": 0.8395,
|
| 222227 |
+
"step": 96570
|
| 222228 |
+
},
|
| 222229 |
+
{
|
| 222230 |
+
"epoch": 772.0,
|
| 222231 |
+
"eval_loss": 0.3809555470943451,
|
| 222232 |
+
"eval_runtime": 40.6701,
|
| 222233 |
+
"eval_samples_per_second": 20.679,
|
| 222234 |
+
"eval_steps_per_second": 0.664,
|
| 222235 |
+
"eval_wer": 0.18068965517241378,
|
| 222236 |
+
"step": 96572
|
| 222237 |
+
},
|
| 222238 |
+
{
|
| 222239 |
+
"epoch": 772.02,
|
| 222240 |
+
"learning_rate": 8.468958333333333e-06,
|
| 222241 |
+
"loss": 0.3108,
|
| 222242 |
+
"step": 96575
|
| 222243 |
+
},
|
| 222244 |
+
{
|
| 222245 |
+
"epoch": 772.06,
|
| 222246 |
+
"learning_rate": 8.468878205128206e-06,
|
| 222247 |
+
"loss": 0.3063,
|
| 222248 |
+
"step": 96580
|
| 222249 |
+
},
|
| 222250 |
+
{
|
| 222251 |
+
"epoch": 772.1,
|
| 222252 |
+
"learning_rate": 8.468798076923077e-06,
|
| 222253 |
+
"loss": 0.3162,
|
| 222254 |
+
"step": 96585
|
| 222255 |
+
},
|
| 222256 |
+
{
|
| 222257 |
+
"epoch": 772.14,
|
| 222258 |
+
"learning_rate": 8.468717948717949e-06,
|
| 222259 |
+
"loss": 0.3951,
|
| 222260 |
+
"step": 96590
|
| 222261 |
+
},
|
| 222262 |
+
{
|
| 222263 |
+
"epoch": 772.18,
|
| 222264 |
+
"learning_rate": 8.468637820512822e-06,
|
| 222265 |
+
"loss": 0.7104,
|
| 222266 |
+
"step": 96595
|
| 222267 |
+
},
|
| 222268 |
+
{
|
| 222269 |
+
"epoch": 772.22,
|
| 222270 |
+
"learning_rate": 8.468557692307693e-06,
|
| 222271 |
+
"loss": 0.9571,
|
| 222272 |
+
"step": 96600
|
| 222273 |
+
},
|
| 222274 |
+
{
|
| 222275 |
+
"epoch": 772.26,
|
| 222276 |
+
"learning_rate": 8.468477564102565e-06,
|
| 222277 |
+
"loss": 0.2776,
|
| 222278 |
+
"step": 96605
|
| 222279 |
+
},
|
| 222280 |
+
{
|
| 222281 |
+
"epoch": 772.3,
|
| 222282 |
+
"learning_rate": 8.468397435897436e-06,
|
| 222283 |
+
"loss": 0.2848,
|
| 222284 |
+
"step": 96610
|
| 222285 |
+
},
|
| 222286 |
+
{
|
| 222287 |
+
"epoch": 772.34,
|
| 222288 |
+
"learning_rate": 8.468317307692309e-06,
|
| 222289 |
+
"loss": 0.3861,
|
| 222290 |
+
"step": 96615
|
| 222291 |
+
},
|
| 222292 |
+
{
|
| 222293 |
+
"epoch": 772.38,
|
| 222294 |
+
"learning_rate": 8.46823717948718e-06,
|
| 222295 |
+
"loss": 0.7346,
|
| 222296 |
+
"step": 96620
|
| 222297 |
+
},
|
| 222298 |
+
{
|
| 222299 |
+
"epoch": 772.42,
|
| 222300 |
+
"learning_rate": 8.468157051282052e-06,
|
| 222301 |
+
"loss": 0.8214,
|
| 222302 |
+
"step": 96625
|
| 222303 |
+
},
|
| 222304 |
+
{
|
| 222305 |
+
"epoch": 772.46,
|
| 222306 |
+
"learning_rate": 8.468076923076923e-06,
|
| 222307 |
+
"loss": 0.3034,
|
| 222308 |
+
"step": 96630
|
| 222309 |
+
},
|
| 222310 |
+
{
|
| 222311 |
+
"epoch": 772.5,
|
| 222312 |
+
"learning_rate": 8.467996794871796e-06,
|
| 222313 |
+
"loss": 0.3315,
|
| 222314 |
+
"step": 96635
|
| 222315 |
+
},
|
| 222316 |
+
{
|
| 222317 |
+
"epoch": 772.54,
|
| 222318 |
+
"learning_rate": 8.467916666666667e-06,
|
| 222319 |
+
"loss": 0.4006,
|
| 222320 |
+
"step": 96640
|
| 222321 |
+
},
|
| 222322 |
+
{
|
| 222323 |
+
"epoch": 772.58,
|
| 222324 |
+
"learning_rate": 8.467836538461539e-06,
|
| 222325 |
+
"loss": 0.7839,
|
| 222326 |
+
"step": 96645
|
| 222327 |
+
},
|
| 222328 |
+
{
|
| 222329 |
+
"epoch": 772.62,
|
| 222330 |
+
"learning_rate": 8.467756410256412e-06,
|
| 222331 |
+
"loss": 0.8439,
|
| 222332 |
+
"step": 96650
|
| 222333 |
+
},
|
| 222334 |
+
{
|
| 222335 |
+
"epoch": 772.66,
|
| 222336 |
+
"learning_rate": 8.467676282051283e-06,
|
| 222337 |
+
"loss": 0.2846,
|
| 222338 |
+
"step": 96655
|
| 222339 |
+
},
|
| 222340 |
+
{
|
| 222341 |
+
"epoch": 772.7,
|
| 222342 |
+
"learning_rate": 8.467596153846155e-06,
|
| 222343 |
+
"loss": 0.2912,
|
| 222344 |
+
"step": 96660
|
| 222345 |
+
},
|
| 222346 |
+
{
|
| 222347 |
+
"epoch": 772.74,
|
| 222348 |
+
"learning_rate": 8.467516025641026e-06,
|
| 222349 |
+
"loss": 0.3036,
|
| 222350 |
+
"step": 96665
|
| 222351 |
+
},
|
| 222352 |
+
{
|
| 222353 |
+
"epoch": 772.78,
|
| 222354 |
+
"learning_rate": 8.467435897435899e-06,
|
| 222355 |
+
"loss": 0.7802,
|
| 222356 |
+
"step": 96670
|
| 222357 |
+
},
|
| 222358 |
+
{
|
| 222359 |
+
"epoch": 772.82,
|
| 222360 |
+
"learning_rate": 8.46735576923077e-06,
|
| 222361 |
+
"loss": 0.912,
|
| 222362 |
+
"step": 96675
|
| 222363 |
+
},
|
| 222364 |
+
{
|
| 222365 |
+
"epoch": 772.86,
|
| 222366 |
+
"learning_rate": 8.467275641025642e-06,
|
| 222367 |
+
"loss": 0.298,
|
| 222368 |
+
"step": 96680
|
| 222369 |
+
},
|
| 222370 |
+
{
|
| 222371 |
+
"epoch": 772.9,
|
| 222372 |
+
"learning_rate": 8.467195512820513e-06,
|
| 222373 |
+
"loss": 0.4736,
|
| 222374 |
+
"step": 96685
|
| 222375 |
+
},
|
| 222376 |
+
{
|
| 222377 |
+
"epoch": 772.94,
|
| 222378 |
+
"learning_rate": 8.467115384615386e-06,
|
| 222379 |
+
"loss": 0.4088,
|
| 222380 |
+
"step": 96690
|
| 222381 |
+
},
|
| 222382 |
+
{
|
| 222383 |
+
"epoch": 772.98,
|
| 222384 |
+
"learning_rate": 8.467035256410257e-06,
|
| 222385 |
+
"loss": 0.9104,
|
| 222386 |
+
"step": 96695
|
| 222387 |
+
},
|
| 222388 |
+
{
|
| 222389 |
+
"epoch": 773.0,
|
| 222390 |
+
"eval_loss": 0.3602166175842285,
|
| 222391 |
+
"eval_runtime": 37.7026,
|
| 222392 |
+
"eval_samples_per_second": 22.28,
|
| 222393 |
+
"eval_steps_per_second": 0.716,
|
| 222394 |
+
"eval_wer": 0.18798860232337256,
|
| 222395 |
+
"step": 96697
|
| 222396 |
+
},
|
| 222397 |
+
{
|
| 222398 |
+
"epoch": 773.02,
|
| 222399 |
+
"learning_rate": 8.466955128205129e-06,
|
| 222400 |
+
"loss": 0.368,
|
| 222401 |
+
"step": 96700
|
| 222402 |
+
},
|
| 222403 |
+
{
|
| 222404 |
+
"epoch": 773.06,
|
| 222405 |
+
"learning_rate": 8.466875000000002e-06,
|
| 222406 |
+
"loss": 0.3052,
|
| 222407 |
+
"step": 96705
|
| 222408 |
+
},
|
| 222409 |
+
{
|
| 222410 |
+
"epoch": 773.1,
|
| 222411 |
+
"learning_rate": 8.466794871794872e-06,
|
| 222412 |
+
"loss": 0.3156,
|
| 222413 |
+
"step": 96710
|
| 222414 |
+
},
|
| 222415 |
+
{
|
| 222416 |
+
"epoch": 773.14,
|
| 222417 |
+
"learning_rate": 8.466714743589745e-06,
|
| 222418 |
+
"loss": 0.3237,
|
| 222419 |
+
"step": 96715
|
| 222420 |
+
},
|
| 222421 |
+
{
|
| 222422 |
+
"epoch": 773.18,
|
| 222423 |
+
"learning_rate": 8.466634615384616e-06,
|
| 222424 |
+
"loss": 0.7653,
|
| 222425 |
+
"step": 96720
|
| 222426 |
+
},
|
| 222427 |
+
{
|
| 222428 |
+
"epoch": 773.22,
|
| 222429 |
+
"learning_rate": 8.466554487179487e-06,
|
| 222430 |
+
"loss": 0.8359,
|
| 222431 |
+
"step": 96725
|
| 222432 |
+
},
|
| 222433 |
+
{
|
| 222434 |
+
"epoch": 773.26,
|
| 222435 |
+
"learning_rate": 8.466474358974359e-06,
|
| 222436 |
+
"loss": 0.5436,
|
| 222437 |
+
"step": 96730
|
| 222438 |
+
},
|
| 222439 |
+
{
|
| 222440 |
+
"epoch": 773.3,
|
| 222441 |
+
"learning_rate": 8.466394230769232e-06,
|
| 222442 |
+
"loss": 0.3419,
|
| 222443 |
+
"step": 96735
|
| 222444 |
+
},
|
| 222445 |
+
{
|
| 222446 |
+
"epoch": 773.34,
|
| 222447 |
+
"learning_rate": 8.466314102564103e-06,
|
| 222448 |
+
"loss": 0.4179,
|
| 222449 |
+
"step": 96740
|
| 222450 |
+
},
|
| 222451 |
+
{
|
| 222452 |
+
"epoch": 773.38,
|
| 222453 |
+
"learning_rate": 8.466233974358974e-06,
|
| 222454 |
+
"loss": 0.7537,
|
| 222455 |
+
"step": 96745
|
| 222456 |
+
},
|
| 222457 |
+
{
|
| 222458 |
+
"epoch": 773.42,
|
| 222459 |
+
"learning_rate": 8.466153846153847e-06,
|
| 222460 |
+
"loss": 1.0606,
|
| 222461 |
+
"step": 96750
|
| 222462 |
+
},
|
| 222463 |
+
{
|
| 222464 |
+
"epoch": 773.46,
|
| 222465 |
+
"learning_rate": 8.466073717948719e-06,
|
| 222466 |
+
"loss": 0.32,
|
| 222467 |
+
"step": 96755
|
| 222468 |
+
},
|
| 222469 |
+
{
|
| 222470 |
+
"epoch": 773.5,
|
| 222471 |
+
"learning_rate": 8.46599358974359e-06,
|
| 222472 |
+
"loss": 0.42,
|
| 222473 |
+
"step": 96760
|
| 222474 |
+
},
|
| 222475 |
+
{
|
| 222476 |
+
"epoch": 773.54,
|
| 222477 |
+
"learning_rate": 8.465913461538462e-06,
|
| 222478 |
+
"loss": 0.3767,
|
| 222479 |
+
"step": 96765
|
| 222480 |
+
},
|
| 222481 |
+
{
|
| 222482 |
+
"epoch": 773.58,
|
| 222483 |
+
"learning_rate": 8.465833333333335e-06,
|
| 222484 |
+
"loss": 0.8556,
|
| 222485 |
+
"step": 96770
|
| 222486 |
+
},
|
| 222487 |
+
{
|
| 222488 |
+
"epoch": 773.62,
|
| 222489 |
+
"learning_rate": 8.465753205128206e-06,
|
| 222490 |
+
"loss": 0.8406,
|
| 222491 |
+
"step": 96775
|
| 222492 |
+
},
|
| 222493 |
+
{
|
| 222494 |
+
"epoch": 773.66,
|
| 222495 |
+
"learning_rate": 8.465673076923077e-06,
|
| 222496 |
+
"loss": 0.2392,
|
| 222497 |
+
"step": 96780
|
| 222498 |
+
},
|
| 222499 |
+
{
|
| 222500 |
+
"epoch": 773.7,
|
| 222501 |
+
"learning_rate": 8.465592948717949e-06,
|
| 222502 |
+
"loss": 0.288,
|
| 222503 |
+
"step": 96785
|
| 222504 |
+
},
|
| 222505 |
+
{
|
| 222506 |
+
"epoch": 773.74,
|
| 222507 |
+
"learning_rate": 8.465512820512822e-06,
|
| 222508 |
+
"loss": 0.3576,
|
| 222509 |
+
"step": 96790
|
| 222510 |
+
},
|
| 222511 |
+
{
|
| 222512 |
+
"epoch": 773.78,
|
| 222513 |
+
"learning_rate": 8.465432692307693e-06,
|
| 222514 |
+
"loss": 0.8087,
|
| 222515 |
+
"step": 96795
|
| 222516 |
+
},
|
| 222517 |
+
{
|
| 222518 |
+
"epoch": 773.82,
|
| 222519 |
+
"learning_rate": 8.465352564102564e-06,
|
| 222520 |
+
"loss": 0.8085,
|
| 222521 |
+
"step": 96800
|
| 222522 |
+
},
|
| 222523 |
+
{
|
| 222524 |
+
"epoch": 773.86,
|
| 222525 |
+
"learning_rate": 8.465272435897437e-06,
|
| 222526 |
+
"loss": 0.3603,
|
| 222527 |
+
"step": 96805
|
| 222528 |
+
},
|
| 222529 |
+
{
|
| 222530 |
+
"epoch": 773.9,
|
| 222531 |
+
"learning_rate": 8.465192307692309e-06,
|
| 222532 |
+
"loss": 0.2768,
|
| 222533 |
+
"step": 96810
|
| 222534 |
+
},
|
| 222535 |
+
{
|
| 222536 |
+
"epoch": 773.94,
|
| 222537 |
+
"learning_rate": 8.46511217948718e-06,
|
| 222538 |
+
"loss": 0.3357,
|
| 222539 |
+
"step": 96815
|
| 222540 |
+
},
|
| 222541 |
+
{
|
| 222542 |
+
"epoch": 773.98,
|
| 222543 |
+
"learning_rate": 8.465032051282052e-06,
|
| 222544 |
+
"loss": 0.7603,
|
| 222545 |
+
"step": 96820
|
| 222546 |
+
},
|
| 222547 |
+
{
|
| 222548 |
+
"epoch": 774.0,
|
| 222549 |
+
"eval_loss": 0.41661757230758667,
|
| 222550 |
+
"eval_runtime": 37.2578,
|
| 222551 |
+
"eval_samples_per_second": 22.546,
|
| 222552 |
+
"eval_steps_per_second": 0.725,
|
| 222553 |
+
"eval_wer": 0.19196902990739334,
|
| 222554 |
+
"step": 96822
|
| 222555 |
+
},
|
| 222556 |
+
{
|
| 222557 |
+
"epoch": 774.02,
|
| 222558 |
+
"learning_rate": 8.464951923076925e-06,
|
| 222559 |
+
"loss": 0.3927,
|
| 222560 |
+
"step": 96825
|
| 222561 |
+
},
|
| 222562 |
+
{
|
| 222563 |
+
"epoch": 774.06,
|
| 222564 |
+
"learning_rate": 8.464871794871794e-06,
|
| 222565 |
+
"loss": 0.256,
|
| 222566 |
+
"step": 96830
|
| 222567 |
+
},
|
| 222568 |
+
{
|
| 222569 |
+
"epoch": 774.1,
|
| 222570 |
+
"learning_rate": 8.464791666666667e-06,
|
| 222571 |
+
"loss": 0.2987,
|
| 222572 |
+
"step": 96835
|
| 222573 |
+
},
|
| 222574 |
+
{
|
| 222575 |
+
"epoch": 774.14,
|
| 222576 |
+
"learning_rate": 8.46471153846154e-06,
|
| 222577 |
+
"loss": 0.3521,
|
| 222578 |
+
"step": 96840
|
| 222579 |
+
},
|
| 222580 |
+
{
|
| 222581 |
+
"epoch": 774.18,
|
| 222582 |
+
"learning_rate": 8.46463141025641e-06,
|
| 222583 |
+
"loss": 0.8093,
|
| 222584 |
+
"step": 96845
|
| 222585 |
+
},
|
| 222586 |
+
{
|
| 222587 |
+
"epoch": 774.22,
|
| 222588 |
+
"learning_rate": 8.464551282051283e-06,
|
| 222589 |
+
"loss": 0.9202,
|
| 222590 |
+
"step": 96850
|
| 222591 |
+
},
|
| 222592 |
+
{
|
| 222593 |
+
"epoch": 774.26,
|
| 222594 |
+
"learning_rate": 8.464471153846154e-06,
|
| 222595 |
+
"loss": 0.3084,
|
| 222596 |
+
"step": 96855
|
| 222597 |
+
},
|
| 222598 |
+
{
|
| 222599 |
+
"epoch": 774.3,
|
| 222600 |
+
"learning_rate": 8.464391025641028e-06,
|
| 222601 |
+
"loss": 0.2711,
|
| 222602 |
+
"step": 96860
|
| 222603 |
+
},
|
| 222604 |
+
{
|
| 222605 |
+
"epoch": 774.34,
|
| 222606 |
+
"learning_rate": 8.464310897435897e-06,
|
| 222607 |
+
"loss": 0.4106,
|
| 222608 |
+
"step": 96865
|
| 222609 |
+
},
|
| 222610 |
+
{
|
| 222611 |
+
"epoch": 774.38,
|
| 222612 |
+
"learning_rate": 8.46423076923077e-06,
|
| 222613 |
+
"loss": 0.7044,
|
| 222614 |
+
"step": 96870
|
| 222615 |
+
},
|
| 222616 |
+
{
|
| 222617 |
+
"epoch": 774.42,
|
| 222618 |
+
"learning_rate": 8.464150641025642e-06,
|
| 222619 |
+
"loss": 0.7626,
|
| 222620 |
+
"step": 96875
|
| 222621 |
+
},
|
| 222622 |
+
{
|
| 222623 |
+
"epoch": 774.46,
|
| 222624 |
+
"learning_rate": 8.464070512820513e-06,
|
| 222625 |
+
"loss": 0.2576,
|
| 222626 |
+
"step": 96880
|
| 222627 |
+
},
|
| 222628 |
+
{
|
| 222629 |
+
"epoch": 774.5,
|
| 222630 |
+
"learning_rate": 8.463990384615384e-06,
|
| 222631 |
+
"loss": 0.3461,
|
| 222632 |
+
"step": 96885
|
| 222633 |
+
},
|
| 222634 |
+
{
|
| 222635 |
+
"epoch": 774.54,
|
| 222636 |
+
"learning_rate": 8.463910256410257e-06,
|
| 222637 |
+
"loss": 0.3534,
|
| 222638 |
+
"step": 96890
|
| 222639 |
+
},
|
| 222640 |
+
{
|
| 222641 |
+
"epoch": 774.58,
|
| 222642 |
+
"learning_rate": 8.463830128205129e-06,
|
| 222643 |
+
"loss": 0.8297,
|
| 222644 |
+
"step": 96895
|
| 222645 |
+
},
|
| 222646 |
+
{
|
| 222647 |
+
"epoch": 774.62,
|
| 222648 |
+
"learning_rate": 8.46375e-06,
|
| 222649 |
+
"loss": 0.9096,
|
| 222650 |
+
"step": 96900
|
| 222651 |
+
},
|
| 222652 |
+
{
|
| 222653 |
+
"epoch": 774.66,
|
| 222654 |
+
"learning_rate": 8.463669871794873e-06,
|
| 222655 |
+
"loss": 0.3551,
|
| 222656 |
+
"step": 96905
|
| 222657 |
+
},
|
| 222658 |
+
{
|
| 222659 |
+
"epoch": 774.7,
|
| 222660 |
+
"learning_rate": 8.463589743589744e-06,
|
| 222661 |
+
"loss": 0.3067,
|
| 222662 |
+
"step": 96910
|
| 222663 |
+
},
|
| 222664 |
+
{
|
| 222665 |
+
"epoch": 774.74,
|
| 222666 |
+
"learning_rate": 8.463509615384616e-06,
|
| 222667 |
+
"loss": 0.3873,
|
| 222668 |
+
"step": 96915
|
| 222669 |
+
},
|
| 222670 |
+
{
|
| 222671 |
+
"epoch": 774.78,
|
| 222672 |
+
"learning_rate": 8.463429487179487e-06,
|
| 222673 |
+
"loss": 0.7954,
|
| 222674 |
+
"step": 96920
|
| 222675 |
+
},
|
| 222676 |
+
{
|
| 222677 |
+
"epoch": 774.82,
|
| 222678 |
+
"learning_rate": 8.46334935897436e-06,
|
| 222679 |
+
"loss": 0.8324,
|
| 222680 |
+
"step": 96925
|
| 222681 |
+
},
|
| 222682 |
+
{
|
| 222683 |
+
"epoch": 774.86,
|
| 222684 |
+
"learning_rate": 8.463269230769232e-06,
|
| 222685 |
+
"loss": 0.2722,
|
| 222686 |
+
"step": 96930
|
| 222687 |
+
},
|
| 222688 |
+
{
|
| 222689 |
+
"epoch": 774.9,
|
| 222690 |
+
"learning_rate": 8.463189102564103e-06,
|
| 222691 |
+
"loss": 0.319,
|
| 222692 |
+
"step": 96935
|
| 222693 |
+
},
|
| 222694 |
+
{
|
| 222695 |
+
"epoch": 774.94,
|
| 222696 |
+
"learning_rate": 8.463108974358976e-06,
|
| 222697 |
+
"loss": 0.3622,
|
| 222698 |
+
"step": 96940
|
| 222699 |
+
},
|
| 222700 |
+
{
|
| 222701 |
+
"epoch": 774.98,
|
| 222702 |
+
"learning_rate": 8.463028846153847e-06,
|
| 222703 |
+
"loss": 0.7881,
|
| 222704 |
+
"step": 96945
|
| 222705 |
+
},
|
| 222706 |
+
{
|
| 222707 |
+
"epoch": 775.0,
|
| 222708 |
+
"eval_loss": 0.36674395203590393,
|
| 222709 |
+
"eval_runtime": 38.2199,
|
| 222710 |
+
"eval_samples_per_second": 21.978,
|
| 222711 |
+
"eval_steps_per_second": 0.706,
|
| 222712 |
+
"eval_wer": 0.18024691358024691,
|
| 222713 |
+
"step": 96947
|
| 222714 |
+
},
|
| 222715 |
+
{
|
| 222716 |
+
"epoch": 775.02,
|
| 222717 |
+
"learning_rate": 8.462964743589745e-06,
|
| 222718 |
+
"loss": 0.2903,
|
| 222719 |
+
"step": 96950
|
| 222720 |
+
},
|
| 222721 |
+
{
|
| 222722 |
+
"epoch": 775.06,
|
| 222723 |
+
"learning_rate": 8.462884615384616e-06,
|
| 222724 |
+
"loss": 0.2675,
|
| 222725 |
+
"step": 96955
|
| 222726 |
+
},
|
| 222727 |
+
{
|
| 222728 |
+
"epoch": 775.1,
|
| 222729 |
+
"learning_rate": 8.462804487179488e-06,
|
| 222730 |
+
"loss": 0.3137,
|
| 222731 |
+
"step": 96960
|
| 222732 |
+
},
|
| 222733 |
+
{
|
| 222734 |
+
"epoch": 775.14,
|
| 222735 |
+
"learning_rate": 8.46272435897436e-06,
|
| 222736 |
+
"loss": 0.3445,
|
| 222737 |
+
"step": 96965
|
| 222738 |
+
},
|
| 222739 |
+
{
|
| 222740 |
+
"epoch": 775.18,
|
| 222741 |
+
"learning_rate": 8.462644230769232e-06,
|
| 222742 |
+
"loss": 0.7379,
|
| 222743 |
+
"step": 96970
|
| 222744 |
+
},
|
| 222745 |
+
{
|
| 222746 |
+
"epoch": 775.22,
|
| 222747 |
+
"learning_rate": 8.462564102564103e-06,
|
| 222748 |
+
"loss": 0.9077,
|
| 222749 |
+
"step": 96975
|
| 222750 |
+
},
|
| 222751 |
+
{
|
| 222752 |
+
"epoch": 775.26,
|
| 222753 |
+
"learning_rate": 8.462483974358975e-06,
|
| 222754 |
+
"loss": 0.3001,
|
| 222755 |
+
"step": 96980
|
| 222756 |
+
},
|
| 222757 |
+
{
|
| 222758 |
+
"epoch": 775.3,
|
| 222759 |
+
"learning_rate": 8.462403846153848e-06,
|
| 222760 |
+
"loss": 0.2626,
|
| 222761 |
+
"step": 96985
|
| 222762 |
+
},
|
| 222763 |
+
{
|
| 222764 |
+
"epoch": 775.34,
|
| 222765 |
+
"learning_rate": 8.462323717948719e-06,
|
| 222766 |
+
"loss": 0.3309,
|
| 222767 |
+
"step": 96990
|
| 222768 |
+
},
|
| 222769 |
+
{
|
| 222770 |
+
"epoch": 775.38,
|
| 222771 |
+
"learning_rate": 8.46224358974359e-06,
|
| 222772 |
+
"loss": 0.6284,
|
| 222773 |
+
"step": 96995
|
| 222774 |
+
},
|
| 222775 |
+
{
|
| 222776 |
+
"epoch": 775.42,
|
| 222777 |
+
"learning_rate": 8.462163461538462e-06,
|
| 222778 |
+
"loss": 0.9079,
|
| 222779 |
+
"step": 97000
|
| 222780 |
+
},
|
| 222781 |
+
{
|
| 222782 |
+
"epoch": 775.46,
|
| 222783 |
+
"learning_rate": 8.462083333333335e-06,
|
| 222784 |
+
"loss": 0.2994,
|
| 222785 |
+
"step": 97005
|
| 222786 |
+
},
|
| 222787 |
+
{
|
| 222788 |
+
"epoch": 775.5,
|
| 222789 |
+
"learning_rate": 8.462003205128204e-06,
|
| 222790 |
+
"loss": 0.3336,
|
| 222791 |
+
"step": 97010
|
| 222792 |
+
},
|
| 222793 |
+
{
|
| 222794 |
+
"epoch": 775.54,
|
| 222795 |
+
"learning_rate": 8.461923076923078e-06,
|
| 222796 |
+
"loss": 0.3821,
|
| 222797 |
+
"step": 97015
|
| 222798 |
+
},
|
| 222799 |
+
{
|
| 222800 |
+
"epoch": 775.58,
|
| 222801 |
+
"learning_rate": 8.46184294871795e-06,
|
| 222802 |
+
"loss": 0.6663,
|
| 222803 |
+
"step": 97020
|
| 222804 |
+
},
|
| 222805 |
+
{
|
| 222806 |
+
"epoch": 775.62,
|
| 222807 |
+
"learning_rate": 8.46176282051282e-06,
|
| 222808 |
+
"loss": 0.8284,
|
| 222809 |
+
"step": 97025
|
| 222810 |
+
},
|
| 222811 |
+
{
|
| 222812 |
+
"epoch": 775.66,
|
| 222813 |
+
"learning_rate": 8.461682692307693e-06,
|
| 222814 |
+
"loss": 0.2903,
|
| 222815 |
+
"step": 97030
|
| 222816 |
+
},
|
| 222817 |
+
{
|
| 222818 |
+
"epoch": 775.7,
|
| 222819 |
+
"learning_rate": 8.461602564102565e-06,
|
| 222820 |
+
"loss": 0.3274,
|
| 222821 |
+
"step": 97035
|
| 222822 |
+
},
|
| 222823 |
+
{
|
| 222824 |
+
"epoch": 775.74,
|
| 222825 |
+
"learning_rate": 8.461522435897436e-06,
|
| 222826 |
+
"loss": 0.436,
|
| 222827 |
+
"step": 97040
|
| 222828 |
+
},
|
| 222829 |
+
{
|
| 222830 |
+
"epoch": 775.78,
|
| 222831 |
+
"learning_rate": 8.461442307692307e-06,
|
| 222832 |
+
"loss": 0.7548,
|
| 222833 |
+
"step": 97045
|
| 222834 |
+
},
|
| 222835 |
+
{
|
| 222836 |
+
"epoch": 775.82,
|
| 222837 |
+
"learning_rate": 8.46136217948718e-06,
|
| 222838 |
+
"loss": 0.7949,
|
| 222839 |
+
"step": 97050
|
| 222840 |
+
},
|
| 222841 |
+
{
|
| 222842 |
+
"epoch": 775.86,
|
| 222843 |
+
"learning_rate": 8.461282051282052e-06,
|
| 222844 |
+
"loss": 0.3488,
|
| 222845 |
+
"step": 97055
|
| 222846 |
+
},
|
| 222847 |
+
{
|
| 222848 |
+
"epoch": 775.9,
|
| 222849 |
+
"learning_rate": 8.461201923076923e-06,
|
| 222850 |
+
"loss": 0.3829,
|
| 222851 |
+
"step": 97060
|
| 222852 |
+
},
|
| 222853 |
+
{
|
| 222854 |
+
"epoch": 775.94,
|
| 222855 |
+
"learning_rate": 8.461121794871796e-06,
|
| 222856 |
+
"loss": 0.3449,
|
| 222857 |
+
"step": 97065
|
| 222858 |
+
},
|
| 222859 |
+
{
|
| 222860 |
+
"epoch": 775.98,
|
| 222861 |
+
"learning_rate": 8.461041666666668e-06,
|
| 222862 |
+
"loss": 0.8363,
|
| 222863 |
+
"step": 97070
|
| 222864 |
+
},
|
| 222865 |
+
{
|
| 222866 |
+
"epoch": 776.0,
|
| 222867 |
+
"eval_loss": 0.3837716281414032,
|
| 222868 |
+
"eval_runtime": 40.1434,
|
| 222869 |
+
"eval_samples_per_second": 20.925,
|
| 222870 |
+
"eval_steps_per_second": 0.673,
|
| 222871 |
+
"eval_wer": 0.1876936697653829,
|
| 222872 |
+
"step": 97072
|
| 222873 |
}
|
| 222874 |
],
|
| 222875 |
"max_steps": 625000,
|
| 222876 |
"num_train_epochs": 5000,
|
| 222877 |
+
"total_flos": 2.7317098092751012e+20,
|
| 222878 |
"trial_name": null,
|
| 222879 |
"trial_params": null
|
| 222880 |
}
|
model-bin/finetune/base/{checkpoint-96447 → checkpoint-97072}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629968901.3732219/events.out.tfevents.1629968901.8e89bd551565.924.131
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6052c2890684047c9df5572be1eb785b2d3a908195c0ac599abd87ea00265e94
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629969336.52509/events.out.tfevents.1629969336.8e89bd551565.924.133
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c35b38d0a95741f72a7fdaad798f97296da00cb1ea06d355bbfdfd7e088a025
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629969764.0650191/events.out.tfevents.1629969764.8e89bd551565.924.135
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca2643b63e16bc4b73bedb21a3b86a43a7eee60675de06a29d93e0b447507a9b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629970202.6078649/events.out.tfevents.1629970202.8e89bd551565.924.137
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:813d2ce5fa683c6aca4d6224d578b7a52a922d145c51666565c88488be1837eb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629970642.6016197/events.out.tfevents.1629970642.8e89bd551565.924.139
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db18d2d70769da428e5f463184bb72837c8f41c6664a9b26c5e5b302ca6362bd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629968901.8e89bd551565.924.130
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44ea26cdc86e3d89a768b648f3fad7e7aba3116f7fcca318394ddee9c8956cd9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629969335.8e89bd551565.924.132
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:343834868c88df0f32950910a72bb3788ef6d458e5d5dc93c8a0aa2869aca8be
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629969764.8e89bd551565.924.134
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9255954b2592d06e44a64b40eae3ad1aa5d21d30ac9ce4b7449812466b9cd358
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629970202.8e89bd551565.924.136
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ece59c664d74c5fbb24a378dab5ef71cc5e06da207631f8ab5684d1dc072dd20
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629970642.8e89bd551565.924.138
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:27963b41d321aafc08862d48045695e41421fd56013e8ba9fe89c27932cb4d9c
|
| 3 |
+
size 8622
|