"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629788442.126761/events.out.tfevents.1629788442.c435e1c5ee04.920.71 +3 -0
- model-bin/finetune/base/log/1629789087.6950493/events.out.tfevents.1629789087.c435e1c5ee04.920.73 +3 -0
- model-bin/finetune/base/log/1629789729.6492243/events.out.tfevents.1629789729.c435e1c5ee04.920.75 +3 -0
- model-bin/finetune/base/log/1629790377.8418694/events.out.tfevents.1629790377.c435e1c5ee04.920.77 +3 -0
- model-bin/finetune/base/log/1629791034.1306834/events.out.tfevents.1629791034.c435e1c5ee04.920.79 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629788442.c435e1c5ee04.920.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629789087.c435e1c5ee04.920.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629789729.c435e1c5ee04.920.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629790377.c435e1c5ee04.920.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629791034.c435e1c5ee04.920.78 +3 -0
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abb032a26cbde4b62d89539fc2e8e5881b35423dcd963a58f4158155e7e2bf88
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/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:d2c85302de9ded7dd6c08545970e38ead0397e1e232dc1cc0397a355e678b399
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd52303b56dc5bf8bee3e5b43d2bf8c89b3afc6ef4263289063a2c641f525b38
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/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 sha256:9d9cd3fedc5e7eb14c9de984a70f18c3445b012f7a0208035cf4a96e7423bac5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/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:a378270aa4ce26218f87463f54f775fff54af5666c2db0fab97c9dec6dc51967
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -173466,11 +173466,800 @@
|
|
| 173466 |
"eval_steps_per_second": 0.676,
|
| 173467 |
"eval_wer": 0.1968630900029317,
|
| 173468 |
"step": 58240
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 173469 |
}
|
| 173470 |
],
|
| 173471 |
"max_steps": 625000,
|
| 173472 |
"num_train_epochs": 5000,
|
| 173473 |
-
"total_flos": 1.
|
| 173474 |
"trial_name": null,
|
| 173475 |
"trial_params": null
|
| 173476 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 470.0,
|
| 5 |
+
"global_step": 58863,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 173466 |
"eval_steps_per_second": 0.676,
|
| 173467 |
"eval_wer": 0.1968630900029317,
|
| 173468 |
"step": 58240
|
| 173469 |
+
},
|
| 173470 |
+
{
|
| 173471 |
+
"epoch": 465.04,
|
| 173472 |
+
"learning_rate": 9.082964743589744e-06,
|
| 173473 |
+
"loss": 0.3577,
|
| 173474 |
+
"step": 58245
|
| 173475 |
+
},
|
| 173476 |
+
{
|
| 173477 |
+
"epoch": 465.08,
|
| 173478 |
+
"learning_rate": 9.082884615384617e-06,
|
| 173479 |
+
"loss": 0.3515,
|
| 173480 |
+
"step": 58250
|
| 173481 |
+
},
|
| 173482 |
+
{
|
| 173483 |
+
"epoch": 465.12,
|
| 173484 |
+
"learning_rate": 9.082804487179487e-06,
|
| 173485 |
+
"loss": 0.5173,
|
| 173486 |
+
"step": 58255
|
| 173487 |
+
},
|
| 173488 |
+
{
|
| 173489 |
+
"epoch": 465.16,
|
| 173490 |
+
"learning_rate": 9.08272435897436e-06,
|
| 173491 |
+
"loss": 0.5225,
|
| 173492 |
+
"step": 58260
|
| 173493 |
+
},
|
| 173494 |
+
{
|
| 173495 |
+
"epoch": 465.2,
|
| 173496 |
+
"learning_rate": 9.082644230769233e-06,
|
| 173497 |
+
"loss": 1.2218,
|
| 173498 |
+
"step": 58265
|
| 173499 |
+
},
|
| 173500 |
+
{
|
| 173501 |
+
"epoch": 465.24,
|
| 173502 |
+
"learning_rate": 9.082564102564102e-06,
|
| 173503 |
+
"loss": 0.3957,
|
| 173504 |
+
"step": 58270
|
| 173505 |
+
},
|
| 173506 |
+
{
|
| 173507 |
+
"epoch": 465.28,
|
| 173508 |
+
"learning_rate": 9.082483974358975e-06,
|
| 173509 |
+
"loss": 0.4104,
|
| 173510 |
+
"step": 58275
|
| 173511 |
+
},
|
| 173512 |
+
{
|
| 173513 |
+
"epoch": 465.32,
|
| 173514 |
+
"learning_rate": 9.082403846153847e-06,
|
| 173515 |
+
"loss": 0.3723,
|
| 173516 |
+
"step": 58280
|
| 173517 |
+
},
|
| 173518 |
+
{
|
| 173519 |
+
"epoch": 465.36,
|
| 173520 |
+
"learning_rate": 9.082323717948718e-06,
|
| 173521 |
+
"loss": 0.5224,
|
| 173522 |
+
"step": 58285
|
| 173523 |
+
},
|
| 173524 |
+
{
|
| 173525 |
+
"epoch": 465.4,
|
| 173526 |
+
"learning_rate": 9.08224358974359e-06,
|
| 173527 |
+
"loss": 1.3738,
|
| 173528 |
+
"step": 58290
|
| 173529 |
+
},
|
| 173530 |
+
{
|
| 173531 |
+
"epoch": 465.44,
|
| 173532 |
+
"learning_rate": 9.082163461538463e-06,
|
| 173533 |
+
"loss": 0.4635,
|
| 173534 |
+
"step": 58295
|
| 173535 |
+
},
|
| 173536 |
+
{
|
| 173537 |
+
"epoch": 465.48,
|
| 173538 |
+
"learning_rate": 9.082083333333334e-06,
|
| 173539 |
+
"loss": 0.3809,
|
| 173540 |
+
"step": 58300
|
| 173541 |
+
},
|
| 173542 |
+
{
|
| 173543 |
+
"epoch": 465.52,
|
| 173544 |
+
"learning_rate": 9.082003205128205e-06,
|
| 173545 |
+
"loss": 0.3892,
|
| 173546 |
+
"step": 58305
|
| 173547 |
+
},
|
| 173548 |
+
{
|
| 173549 |
+
"epoch": 465.56,
|
| 173550 |
+
"learning_rate": 9.081923076923078e-06,
|
| 173551 |
+
"loss": 0.488,
|
| 173552 |
+
"step": 58310
|
| 173553 |
+
},
|
| 173554 |
+
{
|
| 173555 |
+
"epoch": 465.6,
|
| 173556 |
+
"learning_rate": 9.08184294871795e-06,
|
| 173557 |
+
"loss": 1.1037,
|
| 173558 |
+
"step": 58315
|
| 173559 |
+
},
|
| 173560 |
+
{
|
| 173561 |
+
"epoch": 465.64,
|
| 173562 |
+
"learning_rate": 9.081762820512821e-06,
|
| 173563 |
+
"loss": 0.3588,
|
| 173564 |
+
"step": 58320
|
| 173565 |
+
},
|
| 173566 |
+
{
|
| 173567 |
+
"epoch": 465.68,
|
| 173568 |
+
"learning_rate": 9.081682692307692e-06,
|
| 173569 |
+
"loss": 0.3848,
|
| 173570 |
+
"step": 58325
|
| 173571 |
+
},
|
| 173572 |
+
{
|
| 173573 |
+
"epoch": 465.72,
|
| 173574 |
+
"learning_rate": 9.081602564102566e-06,
|
| 173575 |
+
"loss": 0.3331,
|
| 173576 |
+
"step": 58330
|
| 173577 |
+
},
|
| 173578 |
+
{
|
| 173579 |
+
"epoch": 465.76,
|
| 173580 |
+
"learning_rate": 9.081522435897437e-06,
|
| 173581 |
+
"loss": 0.871,
|
| 173582 |
+
"step": 58335
|
| 173583 |
+
},
|
| 173584 |
+
{
|
| 173585 |
+
"epoch": 465.8,
|
| 173586 |
+
"learning_rate": 9.081442307692308e-06,
|
| 173587 |
+
"loss": 1.4403,
|
| 173588 |
+
"step": 58340
|
| 173589 |
+
},
|
| 173590 |
+
{
|
| 173591 |
+
"epoch": 465.84,
|
| 173592 |
+
"learning_rate": 9.08136217948718e-06,
|
| 173593 |
+
"loss": 0.3826,
|
| 173594 |
+
"step": 58345
|
| 173595 |
+
},
|
| 173596 |
+
{
|
| 173597 |
+
"epoch": 465.88,
|
| 173598 |
+
"learning_rate": 9.081282051282053e-06,
|
| 173599 |
+
"loss": 0.416,
|
| 173600 |
+
"step": 58350
|
| 173601 |
+
},
|
| 173602 |
+
{
|
| 173603 |
+
"epoch": 465.92,
|
| 173604 |
+
"learning_rate": 9.081201923076924e-06,
|
| 173605 |
+
"loss": 0.3328,
|
| 173606 |
+
"step": 58355
|
| 173607 |
+
},
|
| 173608 |
+
{
|
| 173609 |
+
"epoch": 465.96,
|
| 173610 |
+
"learning_rate": 9.081121794871795e-06,
|
| 173611 |
+
"loss": 0.5354,
|
| 173612 |
+
"step": 58360
|
| 173613 |
+
},
|
| 173614 |
+
{
|
| 173615 |
+
"epoch": 466.0,
|
| 173616 |
+
"learning_rate": 9.081041666666668e-06,
|
| 173617 |
+
"loss": 1.529,
|
| 173618 |
+
"step": 58365
|
| 173619 |
+
},
|
| 173620 |
+
{
|
| 173621 |
+
"epoch": 466.0,
|
| 173622 |
+
"eval_loss": 0.405769407749176,
|
| 173623 |
+
"eval_runtime": 40.2672,
|
| 173624 |
+
"eval_samples_per_second": 20.811,
|
| 173625 |
+
"eval_steps_per_second": 0.671,
|
| 173626 |
+
"eval_wer": 0.2038402569905819,
|
| 173627 |
+
"step": 58365
|
| 173628 |
+
},
|
| 173629 |
+
{
|
| 173630 |
+
"epoch": 470.04,
|
| 173631 |
+
"learning_rate": 9.08096153846154e-06,
|
| 173632 |
+
"loss": 0.3582,
|
| 173633 |
+
"step": 58370
|
| 173634 |
+
},
|
| 173635 |
+
{
|
| 173636 |
+
"epoch": 470.08,
|
| 173637 |
+
"learning_rate": 9.080881410256411e-06,
|
| 173638 |
+
"loss": 0.351,
|
| 173639 |
+
"step": 58375
|
| 173640 |
+
},
|
| 173641 |
+
{
|
| 173642 |
+
"epoch": 470.12,
|
| 173643 |
+
"learning_rate": 9.080801282051282e-06,
|
| 173644 |
+
"loss": 0.3805,
|
| 173645 |
+
"step": 58380
|
| 173646 |
+
},
|
| 173647 |
+
{
|
| 173648 |
+
"epoch": 470.16,
|
| 173649 |
+
"learning_rate": 9.080721153846156e-06,
|
| 173650 |
+
"loss": 0.456,
|
| 173651 |
+
"step": 58385
|
| 173652 |
+
},
|
| 173653 |
+
{
|
| 173654 |
+
"epoch": 470.2,
|
| 173655 |
+
"learning_rate": 9.080641025641025e-06,
|
| 173656 |
+
"loss": 1.3478,
|
| 173657 |
+
"step": 58390
|
| 173658 |
+
},
|
| 173659 |
+
{
|
| 173660 |
+
"epoch": 470.24,
|
| 173661 |
+
"learning_rate": 9.080560897435898e-06,
|
| 173662 |
+
"loss": 0.3678,
|
| 173663 |
+
"step": 58395
|
| 173664 |
+
},
|
| 173665 |
+
{
|
| 173666 |
+
"epoch": 470.28,
|
| 173667 |
+
"learning_rate": 9.08048076923077e-06,
|
| 173668 |
+
"loss": 0.3497,
|
| 173669 |
+
"step": 58400
|
| 173670 |
+
},
|
| 173671 |
+
{
|
| 173672 |
+
"epoch": 470.32,
|
| 173673 |
+
"learning_rate": 9.080400641025641e-06,
|
| 173674 |
+
"loss": 0.3635,
|
| 173675 |
+
"step": 58405
|
| 173676 |
+
},
|
| 173677 |
+
{
|
| 173678 |
+
"epoch": 470.36,
|
| 173679 |
+
"learning_rate": 9.080320512820514e-06,
|
| 173680 |
+
"loss": 0.5171,
|
| 173681 |
+
"step": 58410
|
| 173682 |
+
},
|
| 173683 |
+
{
|
| 173684 |
+
"epoch": 470.4,
|
| 173685 |
+
"learning_rate": 9.080240384615385e-06,
|
| 173686 |
+
"loss": 1.2932,
|
| 173687 |
+
"step": 58415
|
| 173688 |
+
},
|
| 173689 |
+
{
|
| 173690 |
+
"epoch": 470.44,
|
| 173691 |
+
"learning_rate": 9.080160256410257e-06,
|
| 173692 |
+
"loss": 0.3472,
|
| 173693 |
+
"step": 58420
|
| 173694 |
+
},
|
| 173695 |
+
{
|
| 173696 |
+
"epoch": 470.48,
|
| 173697 |
+
"learning_rate": 9.080080128205128e-06,
|
| 173698 |
+
"loss": 0.3133,
|
| 173699 |
+
"step": 58425
|
| 173700 |
+
},
|
| 173701 |
+
{
|
| 173702 |
+
"epoch": 470.52,
|
| 173703 |
+
"learning_rate": 9.080000000000001e-06,
|
| 173704 |
+
"loss": 0.3152,
|
| 173705 |
+
"step": 58430
|
| 173706 |
+
},
|
| 173707 |
+
{
|
| 173708 |
+
"epoch": 470.56,
|
| 173709 |
+
"learning_rate": 9.079919871794873e-06,
|
| 173710 |
+
"loss": 0.4895,
|
| 173711 |
+
"step": 58435
|
| 173712 |
+
},
|
| 173713 |
+
{
|
| 173714 |
+
"epoch": 470.6,
|
| 173715 |
+
"learning_rate": 9.079839743589744e-06,
|
| 173716 |
+
"loss": 1.096,
|
| 173717 |
+
"step": 58440
|
| 173718 |
+
},
|
| 173719 |
+
{
|
| 173720 |
+
"epoch": 470.65,
|
| 173721 |
+
"learning_rate": 9.079759615384615e-06,
|
| 173722 |
+
"loss": 0.4131,
|
| 173723 |
+
"step": 58445
|
| 173724 |
+
},
|
| 173725 |
+
{
|
| 173726 |
+
"epoch": 470.69,
|
| 173727 |
+
"learning_rate": 9.079679487179488e-06,
|
| 173728 |
+
"loss": 0.3759,
|
| 173729 |
+
"step": 58450
|
| 173730 |
+
},
|
| 173731 |
+
{
|
| 173732 |
+
"epoch": 470.73,
|
| 173733 |
+
"learning_rate": 9.07959935897436e-06,
|
| 173734 |
+
"loss": 0.3681,
|
| 173735 |
+
"step": 58455
|
| 173736 |
+
},
|
| 173737 |
+
{
|
| 173738 |
+
"epoch": 470.77,
|
| 173739 |
+
"learning_rate": 9.079519230769231e-06,
|
| 173740 |
+
"loss": 0.5257,
|
| 173741 |
+
"step": 58460
|
| 173742 |
+
},
|
| 173743 |
+
{
|
| 173744 |
+
"epoch": 470.81,
|
| 173745 |
+
"learning_rate": 9.079439102564104e-06,
|
| 173746 |
+
"loss": 1.3265,
|
| 173747 |
+
"step": 58465
|
| 173748 |
+
},
|
| 173749 |
+
{
|
| 173750 |
+
"epoch": 470.85,
|
| 173751 |
+
"learning_rate": 9.079358974358975e-06,
|
| 173752 |
+
"loss": 0.3764,
|
| 173753 |
+
"step": 58470
|
| 173754 |
+
},
|
| 173755 |
+
{
|
| 173756 |
+
"epoch": 470.89,
|
| 173757 |
+
"learning_rate": 9.079278846153847e-06,
|
| 173758 |
+
"loss": 0.31,
|
| 173759 |
+
"step": 58475
|
| 173760 |
+
},
|
| 173761 |
+
{
|
| 173762 |
+
"epoch": 470.93,
|
| 173763 |
+
"learning_rate": 9.079198717948718e-06,
|
| 173764 |
+
"loss": 0.3444,
|
| 173765 |
+
"step": 58480
|
| 173766 |
+
},
|
| 173767 |
+
{
|
| 173768 |
+
"epoch": 470.97,
|
| 173769 |
+
"learning_rate": 9.079118589743591e-06,
|
| 173770 |
+
"loss": 0.6568,
|
| 173771 |
+
"step": 58485
|
| 173772 |
+
},
|
| 173773 |
+
{
|
| 173774 |
+
"epoch": 471.0,
|
| 173775 |
+
"eval_loss": 0.4264553189277649,
|
| 173776 |
+
"eval_runtime": 39.6989,
|
| 173777 |
+
"eval_samples_per_second": 21.109,
|
| 173778 |
+
"eval_steps_per_second": 0.68,
|
| 173779 |
+
"eval_wer": 0.19773474607234198,
|
| 173780 |
+
"step": 58489
|
| 173781 |
+
},
|
| 173782 |
+
{
|
| 173783 |
+
"epoch": 467.01,
|
| 173784 |
+
"learning_rate": 9.079038461538463e-06,
|
| 173785 |
+
"loss": 0.5071,
|
| 173786 |
+
"step": 58490
|
| 173787 |
+
},
|
| 173788 |
+
{
|
| 173789 |
+
"epoch": 467.05,
|
| 173790 |
+
"learning_rate": 9.078958333333334e-06,
|
| 173791 |
+
"loss": 0.3601,
|
| 173792 |
+
"step": 58495
|
| 173793 |
+
},
|
| 173794 |
+
{
|
| 173795 |
+
"epoch": 467.09,
|
| 173796 |
+
"learning_rate": 9.078878205128205e-06,
|
| 173797 |
+
"loss": 0.378,
|
| 173798 |
+
"step": 58500
|
| 173799 |
+
},
|
| 173800 |
+
{
|
| 173801 |
+
"epoch": 467.13,
|
| 173802 |
+
"learning_rate": 9.078798076923078e-06,
|
| 173803 |
+
"loss": 0.427,
|
| 173804 |
+
"step": 58505
|
| 173805 |
+
},
|
| 173806 |
+
{
|
| 173807 |
+
"epoch": 467.17,
|
| 173808 |
+
"learning_rate": 9.07871794871795e-06,
|
| 173809 |
+
"loss": 0.6693,
|
| 173810 |
+
"step": 58510
|
| 173811 |
+
},
|
| 173812 |
+
{
|
| 173813 |
+
"epoch": 467.21,
|
| 173814 |
+
"learning_rate": 9.078637820512821e-06,
|
| 173815 |
+
"loss": 1.5666,
|
| 173816 |
+
"step": 58515
|
| 173817 |
+
},
|
| 173818 |
+
{
|
| 173819 |
+
"epoch": 467.25,
|
| 173820 |
+
"learning_rate": 9.078557692307694e-06,
|
| 173821 |
+
"loss": 0.3476,
|
| 173822 |
+
"step": 58520
|
| 173823 |
+
},
|
| 173824 |
+
{
|
| 173825 |
+
"epoch": 467.29,
|
| 173826 |
+
"learning_rate": 9.078477564102565e-06,
|
| 173827 |
+
"loss": 0.3344,
|
| 173828 |
+
"step": 58525
|
| 173829 |
+
},
|
| 173830 |
+
{
|
| 173831 |
+
"epoch": 467.33,
|
| 173832 |
+
"learning_rate": 9.078397435897437e-06,
|
| 173833 |
+
"loss": 0.3901,
|
| 173834 |
+
"step": 58530
|
| 173835 |
+
},
|
| 173836 |
+
{
|
| 173837 |
+
"epoch": 467.37,
|
| 173838 |
+
"learning_rate": 9.078317307692308e-06,
|
| 173839 |
+
"loss": 0.5791,
|
| 173840 |
+
"step": 58535
|
| 173841 |
+
},
|
| 173842 |
+
{
|
| 173843 |
+
"epoch": 467.41,
|
| 173844 |
+
"learning_rate": 9.078237179487181e-06,
|
| 173845 |
+
"loss": 1.095,
|
| 173846 |
+
"step": 58540
|
| 173847 |
+
},
|
| 173848 |
+
{
|
| 173849 |
+
"epoch": 467.45,
|
| 173850 |
+
"learning_rate": 9.078157051282051e-06,
|
| 173851 |
+
"loss": 0.32,
|
| 173852 |
+
"step": 58545
|
| 173853 |
+
},
|
| 173854 |
+
{
|
| 173855 |
+
"epoch": 467.49,
|
| 173856 |
+
"learning_rate": 9.078076923076924e-06,
|
| 173857 |
+
"loss": 0.2828,
|
| 173858 |
+
"step": 58550
|
| 173859 |
+
},
|
| 173860 |
+
{
|
| 173861 |
+
"epoch": 467.53,
|
| 173862 |
+
"learning_rate": 9.077996794871795e-06,
|
| 173863 |
+
"loss": 0.3248,
|
| 173864 |
+
"step": 58555
|
| 173865 |
+
},
|
| 173866 |
+
{
|
| 173867 |
+
"epoch": 467.57,
|
| 173868 |
+
"learning_rate": 9.077916666666667e-06,
|
| 173869 |
+
"loss": 0.5747,
|
| 173870 |
+
"step": 58560
|
| 173871 |
+
},
|
| 173872 |
+
{
|
| 173873 |
+
"epoch": 467.61,
|
| 173874 |
+
"learning_rate": 9.07783653846154e-06,
|
| 173875 |
+
"loss": 1.1822,
|
| 173876 |
+
"step": 58565
|
| 173877 |
+
},
|
| 173878 |
+
{
|
| 173879 |
+
"epoch": 467.65,
|
| 173880 |
+
"learning_rate": 9.077756410256411e-06,
|
| 173881 |
+
"loss": 0.3319,
|
| 173882 |
+
"step": 58570
|
| 173883 |
+
},
|
| 173884 |
+
{
|
| 173885 |
+
"epoch": 467.69,
|
| 173886 |
+
"learning_rate": 9.077676282051282e-06,
|
| 173887 |
+
"loss": 0.3283,
|
| 173888 |
+
"step": 58575
|
| 173889 |
+
},
|
| 173890 |
+
{
|
| 173891 |
+
"epoch": 467.73,
|
| 173892 |
+
"learning_rate": 9.077596153846154e-06,
|
| 173893 |
+
"loss": 0.4325,
|
| 173894 |
+
"step": 58580
|
| 173895 |
+
},
|
| 173896 |
+
{
|
| 173897 |
+
"epoch": 467.77,
|
| 173898 |
+
"learning_rate": 9.077516025641027e-06,
|
| 173899 |
+
"loss": 0.5353,
|
| 173900 |
+
"step": 58585
|
| 173901 |
+
},
|
| 173902 |
+
{
|
| 173903 |
+
"epoch": 467.81,
|
| 173904 |
+
"learning_rate": 9.077435897435898e-06,
|
| 173905 |
+
"loss": 1.3401,
|
| 173906 |
+
"step": 58590
|
| 173907 |
+
},
|
| 173908 |
+
{
|
| 173909 |
+
"epoch": 467.85,
|
| 173910 |
+
"learning_rate": 9.07735576923077e-06,
|
| 173911 |
+
"loss": 0.3492,
|
| 173912 |
+
"step": 58595
|
| 173913 |
+
},
|
| 173914 |
+
{
|
| 173915 |
+
"epoch": 467.89,
|
| 173916 |
+
"learning_rate": 9.077275641025641e-06,
|
| 173917 |
+
"loss": 0.363,
|
| 173918 |
+
"step": 58600
|
| 173919 |
+
},
|
| 173920 |
+
{
|
| 173921 |
+
"epoch": 467.93,
|
| 173922 |
+
"learning_rate": 9.077195512820514e-06,
|
| 173923 |
+
"loss": 0.3184,
|
| 173924 |
+
"step": 58605
|
| 173925 |
+
},
|
| 173926 |
+
{
|
| 173927 |
+
"epoch": 467.97,
|
| 173928 |
+
"learning_rate": 9.077115384615385e-06,
|
| 173929 |
+
"loss": 0.5845,
|
| 173930 |
+
"step": 58610
|
| 173931 |
+
},
|
| 173932 |
+
{
|
| 173933 |
+
"epoch": 468.0,
|
| 173934 |
+
"eval_loss": 0.3841117322444916,
|
| 173935 |
+
"eval_runtime": 40.8509,
|
| 173936 |
+
"eval_samples_per_second": 20.514,
|
| 173937 |
+
"eval_steps_per_second": 0.661,
|
| 173938 |
+
"eval_wer": 0.20130728554641597,
|
| 173939 |
+
"step": 58614
|
| 173940 |
+
},
|
| 173941 |
+
{
|
| 173942 |
+
"epoch": 472.01,
|
| 173943 |
+
"learning_rate": 9.077035256410257e-06,
|
| 173944 |
+
"loss": 0.395,
|
| 173945 |
+
"step": 58615
|
| 173946 |
+
},
|
| 173947 |
+
{
|
| 173948 |
+
"epoch": 472.05,
|
| 173949 |
+
"learning_rate": 9.07695512820513e-06,
|
| 173950 |
+
"loss": 0.3279,
|
| 173951 |
+
"step": 58620
|
| 173952 |
+
},
|
| 173953 |
+
{
|
| 173954 |
+
"epoch": 472.09,
|
| 173955 |
+
"learning_rate": 9.076875000000001e-06,
|
| 173956 |
+
"loss": 0.3135,
|
| 173957 |
+
"step": 58625
|
| 173958 |
+
},
|
| 173959 |
+
{
|
| 173960 |
+
"epoch": 472.13,
|
| 173961 |
+
"learning_rate": 9.076794871794872e-06,
|
| 173962 |
+
"loss": 0.4003,
|
| 173963 |
+
"step": 58630
|
| 173964 |
+
},
|
| 173965 |
+
{
|
| 173966 |
+
"epoch": 472.17,
|
| 173967 |
+
"learning_rate": 9.076714743589744e-06,
|
| 173968 |
+
"loss": 0.6409,
|
| 173969 |
+
"step": 58635
|
| 173970 |
+
},
|
| 173971 |
+
{
|
| 173972 |
+
"epoch": 472.21,
|
| 173973 |
+
"learning_rate": 9.076634615384617e-06,
|
| 173974 |
+
"loss": 1.1497,
|
| 173975 |
+
"step": 58640
|
| 173976 |
+
},
|
| 173977 |
+
{
|
| 173978 |
+
"epoch": 472.25,
|
| 173979 |
+
"learning_rate": 9.076554487179488e-06,
|
| 173980 |
+
"loss": 0.347,
|
| 173981 |
+
"step": 58645
|
| 173982 |
+
},
|
| 173983 |
+
{
|
| 173984 |
+
"epoch": 472.29,
|
| 173985 |
+
"learning_rate": 9.07647435897436e-06,
|
| 173986 |
+
"loss": 0.372,
|
| 173987 |
+
"step": 58650
|
| 173988 |
+
},
|
| 173989 |
+
{
|
| 173990 |
+
"epoch": 472.33,
|
| 173991 |
+
"learning_rate": 9.076394230769231e-06,
|
| 173992 |
+
"loss": 0.3361,
|
| 173993 |
+
"step": 58655
|
| 173994 |
+
},
|
| 173995 |
+
{
|
| 173996 |
+
"epoch": 472.37,
|
| 173997 |
+
"learning_rate": 9.076314102564104e-06,
|
| 173998 |
+
"loss": 0.538,
|
| 173999 |
+
"step": 58660
|
| 174000 |
+
},
|
| 174001 |
+
{
|
| 174002 |
+
"epoch": 472.41,
|
| 174003 |
+
"learning_rate": 9.076233974358975e-06,
|
| 174004 |
+
"loss": 1.161,
|
| 174005 |
+
"step": 58665
|
| 174006 |
+
},
|
| 174007 |
+
{
|
| 174008 |
+
"epoch": 472.45,
|
| 174009 |
+
"learning_rate": 9.076153846153847e-06,
|
| 174010 |
+
"loss": 0.3147,
|
| 174011 |
+
"step": 58670
|
| 174012 |
+
},
|
| 174013 |
+
{
|
| 174014 |
+
"epoch": 472.49,
|
| 174015 |
+
"learning_rate": 9.07607371794872e-06,
|
| 174016 |
+
"loss": 0.3359,
|
| 174017 |
+
"step": 58675
|
| 174018 |
+
},
|
| 174019 |
+
{
|
| 174020 |
+
"epoch": 472.53,
|
| 174021 |
+
"learning_rate": 9.07599358974359e-06,
|
| 174022 |
+
"loss": 0.426,
|
| 174023 |
+
"step": 58680
|
| 174024 |
+
},
|
| 174025 |
+
{
|
| 174026 |
+
"epoch": 472.57,
|
| 174027 |
+
"learning_rate": 9.075913461538462e-06,
|
| 174028 |
+
"loss": 0.6531,
|
| 174029 |
+
"step": 58685
|
| 174030 |
+
},
|
| 174031 |
+
{
|
| 174032 |
+
"epoch": 472.61,
|
| 174033 |
+
"learning_rate": 9.075833333333334e-06,
|
| 174034 |
+
"loss": 1.508,
|
| 174035 |
+
"step": 58690
|
| 174036 |
+
},
|
| 174037 |
+
{
|
| 174038 |
+
"epoch": 472.65,
|
| 174039 |
+
"learning_rate": 9.075753205128205e-06,
|
| 174040 |
+
"loss": 0.3804,
|
| 174041 |
+
"step": 58695
|
| 174042 |
+
},
|
| 174043 |
+
{
|
| 174044 |
+
"epoch": 472.69,
|
| 174045 |
+
"learning_rate": 9.075673076923077e-06,
|
| 174046 |
+
"loss": 0.3454,
|
| 174047 |
+
"step": 58700
|
| 174048 |
+
},
|
| 174049 |
+
{
|
| 174050 |
+
"epoch": 472.73,
|
| 174051 |
+
"learning_rate": 9.07559294871795e-06,
|
| 174052 |
+
"loss": 0.3783,
|
| 174053 |
+
"step": 58705
|
| 174054 |
+
},
|
| 174055 |
+
{
|
| 174056 |
+
"epoch": 472.77,
|
| 174057 |
+
"learning_rate": 9.075512820512821e-06,
|
| 174058 |
+
"loss": 0.613,
|
| 174059 |
+
"step": 58710
|
| 174060 |
+
},
|
| 174061 |
+
{
|
| 174062 |
+
"epoch": 472.81,
|
| 174063 |
+
"learning_rate": 9.075432692307692e-06,
|
| 174064 |
+
"loss": 1.2604,
|
| 174065 |
+
"step": 58715
|
| 174066 |
+
},
|
| 174067 |
+
{
|
| 174068 |
+
"epoch": 472.85,
|
| 174069 |
+
"learning_rate": 9.075352564102565e-06,
|
| 174070 |
+
"loss": 0.3477,
|
| 174071 |
+
"step": 58720
|
| 174072 |
+
},
|
| 174073 |
+
{
|
| 174074 |
+
"epoch": 472.89,
|
| 174075 |
+
"learning_rate": 9.075272435897437e-06,
|
| 174076 |
+
"loss": 0.3303,
|
| 174077 |
+
"step": 58725
|
| 174078 |
+
},
|
| 174079 |
+
{
|
| 174080 |
+
"epoch": 472.93,
|
| 174081 |
+
"learning_rate": 9.075192307692308e-06,
|
| 174082 |
+
"loss": 0.3996,
|
| 174083 |
+
"step": 58730
|
| 174084 |
+
},
|
| 174085 |
+
{
|
| 174086 |
+
"epoch": 472.97,
|
| 174087 |
+
"learning_rate": 9.07511217948718e-06,
|
| 174088 |
+
"loss": 0.6128,
|
| 174089 |
+
"step": 58735
|
| 174090 |
+
},
|
| 174091 |
+
{
|
| 174092 |
+
"epoch": 473.0,
|
| 174093 |
+
"eval_loss": 0.39583051204681396,
|
| 174094 |
+
"eval_runtime": 38.5651,
|
| 174095 |
+
"eval_samples_per_second": 21.729,
|
| 174096 |
+
"eval_steps_per_second": 0.7,
|
| 174097 |
+
"eval_wer": 0.18955640423031728,
|
| 174098 |
+
"step": 58738
|
| 174099 |
+
},
|
| 174100 |
+
{
|
| 174101 |
+
"epoch": 469.02,
|
| 174102 |
+
"learning_rate": 9.075032051282052e-06,
|
| 174103 |
+
"loss": 0.3805,
|
| 174104 |
+
"step": 58740
|
| 174105 |
+
},
|
| 174106 |
+
{
|
| 174107 |
+
"epoch": 469.06,
|
| 174108 |
+
"learning_rate": 9.074951923076924e-06,
|
| 174109 |
+
"loss": 0.2926,
|
| 174110 |
+
"step": 58745
|
| 174111 |
+
},
|
| 174112 |
+
{
|
| 174113 |
+
"epoch": 469.1,
|
| 174114 |
+
"learning_rate": 9.074871794871795e-06,
|
| 174115 |
+
"loss": 0.4002,
|
| 174116 |
+
"step": 58750
|
| 174117 |
+
},
|
| 174118 |
+
{
|
| 174119 |
+
"epoch": 469.14,
|
| 174120 |
+
"learning_rate": 9.074791666666667e-06,
|
| 174121 |
+
"loss": 0.4061,
|
| 174122 |
+
"step": 58755
|
| 174123 |
+
},
|
| 174124 |
+
{
|
| 174125 |
+
"epoch": 469.18,
|
| 174126 |
+
"learning_rate": 9.07471153846154e-06,
|
| 174127 |
+
"loss": 0.6746,
|
| 174128 |
+
"step": 58760
|
| 174129 |
+
},
|
| 174130 |
+
{
|
| 174131 |
+
"epoch": 469.22,
|
| 174132 |
+
"learning_rate": 9.074631410256411e-06,
|
| 174133 |
+
"loss": 1.0583,
|
| 174134 |
+
"step": 58765
|
| 174135 |
+
},
|
| 174136 |
+
{
|
| 174137 |
+
"epoch": 469.26,
|
| 174138 |
+
"learning_rate": 9.074551282051282e-06,
|
| 174139 |
+
"loss": 0.3149,
|
| 174140 |
+
"step": 58770
|
| 174141 |
+
},
|
| 174142 |
+
{
|
| 174143 |
+
"epoch": 469.3,
|
| 174144 |
+
"learning_rate": 9.074471153846155e-06,
|
| 174145 |
+
"loss": 0.3262,
|
| 174146 |
+
"step": 58775
|
| 174147 |
+
},
|
| 174148 |
+
{
|
| 174149 |
+
"epoch": 469.34,
|
| 174150 |
+
"learning_rate": 9.074391025641027e-06,
|
| 174151 |
+
"loss": 0.4636,
|
| 174152 |
+
"step": 58780
|
| 174153 |
+
},
|
| 174154 |
+
{
|
| 174155 |
+
"epoch": 469.38,
|
| 174156 |
+
"learning_rate": 9.074310897435898e-06,
|
| 174157 |
+
"loss": 0.7496,
|
| 174158 |
+
"step": 58785
|
| 174159 |
+
},
|
| 174160 |
+
{
|
| 174161 |
+
"epoch": 469.42,
|
| 174162 |
+
"learning_rate": 9.07423076923077e-06,
|
| 174163 |
+
"loss": 0.9887,
|
| 174164 |
+
"step": 58790
|
| 174165 |
+
},
|
| 174166 |
+
{
|
| 174167 |
+
"epoch": 469.46,
|
| 174168 |
+
"learning_rate": 9.074150641025643e-06,
|
| 174169 |
+
"loss": 0.3095,
|
| 174170 |
+
"step": 58795
|
| 174171 |
+
},
|
| 174172 |
+
{
|
| 174173 |
+
"epoch": 469.5,
|
| 174174 |
+
"learning_rate": 9.074070512820514e-06,
|
| 174175 |
+
"loss": 0.3382,
|
| 174176 |
+
"step": 58800
|
| 174177 |
+
},
|
| 174178 |
+
{
|
| 174179 |
+
"epoch": 469.54,
|
| 174180 |
+
"learning_rate": 9.073990384615385e-06,
|
| 174181 |
+
"loss": 0.3929,
|
| 174182 |
+
"step": 58805
|
| 174183 |
+
},
|
| 174184 |
+
{
|
| 174185 |
+
"epoch": 469.58,
|
| 174186 |
+
"learning_rate": 9.073910256410258e-06,
|
| 174187 |
+
"loss": 0.6524,
|
| 174188 |
+
"step": 58810
|
| 174189 |
+
},
|
| 174190 |
+
{
|
| 174191 |
+
"epoch": 469.62,
|
| 174192 |
+
"learning_rate": 9.07383012820513e-06,
|
| 174193 |
+
"loss": 1.1419,
|
| 174194 |
+
"step": 58815
|
| 174195 |
+
},
|
| 174196 |
+
{
|
| 174197 |
+
"epoch": 469.66,
|
| 174198 |
+
"learning_rate": 9.073750000000001e-06,
|
| 174199 |
+
"loss": 0.3383,
|
| 174200 |
+
"step": 58820
|
| 174201 |
+
},
|
| 174202 |
+
{
|
| 174203 |
+
"epoch": 469.7,
|
| 174204 |
+
"learning_rate": 9.073669871794872e-06,
|
| 174205 |
+
"loss": 0.398,
|
| 174206 |
+
"step": 58825
|
| 174207 |
+
},
|
| 174208 |
+
{
|
| 174209 |
+
"epoch": 469.74,
|
| 174210 |
+
"learning_rate": 9.073589743589745e-06,
|
| 174211 |
+
"loss": 0.3604,
|
| 174212 |
+
"step": 58830
|
| 174213 |
+
},
|
| 174214 |
+
{
|
| 174215 |
+
"epoch": 469.78,
|
| 174216 |
+
"learning_rate": 9.073509615384615e-06,
|
| 174217 |
+
"loss": 0.7135,
|
| 174218 |
+
"step": 58835
|
| 174219 |
+
},
|
| 174220 |
+
{
|
| 174221 |
+
"epoch": 469.82,
|
| 174222 |
+
"learning_rate": 9.073429487179488e-06,
|
| 174223 |
+
"loss": 1.2744,
|
| 174224 |
+
"step": 58840
|
| 174225 |
+
},
|
| 174226 |
+
{
|
| 174227 |
+
"epoch": 469.86,
|
| 174228 |
+
"learning_rate": 9.07334935897436e-06,
|
| 174229 |
+
"loss": 0.351,
|
| 174230 |
+
"step": 58845
|
| 174231 |
+
},
|
| 174232 |
+
{
|
| 174233 |
+
"epoch": 469.9,
|
| 174234 |
+
"learning_rate": 9.073269230769231e-06,
|
| 174235 |
+
"loss": 0.3801,
|
| 174236 |
+
"step": 58850
|
| 174237 |
+
},
|
| 174238 |
+
{
|
| 174239 |
+
"epoch": 469.94,
|
| 174240 |
+
"learning_rate": 9.073189102564102e-06,
|
| 174241 |
+
"loss": 0.372,
|
| 174242 |
+
"step": 58855
|
| 174243 |
+
},
|
| 174244 |
+
{
|
| 174245 |
+
"epoch": 469.98,
|
| 174246 |
+
"learning_rate": 9.073108974358975e-06,
|
| 174247 |
+
"loss": 0.7963,
|
| 174248 |
+
"step": 58860
|
| 174249 |
+
},
|
| 174250 |
+
{
|
| 174251 |
+
"epoch": 470.0,
|
| 174252 |
+
"eval_loss": 0.41615360975265503,
|
| 174253 |
+
"eval_runtime": 39.8592,
|
| 174254 |
+
"eval_samples_per_second": 21.024,
|
| 174255 |
+
"eval_steps_per_second": 0.677,
|
| 174256 |
+
"eval_wer": 0.19737984944991313,
|
| 174257 |
+
"step": 58863
|
| 174258 |
}
|
| 174259 |
],
|
| 174260 |
"max_steps": 625000,
|
| 174261 |
"num_train_epochs": 5000,
|
| 174262 |
+
"total_flos": 1.656479841453496e+20,
|
| 174263 |
"trial_name": null,
|
| 174264 |
"trial_params": null
|
| 174265 |
}
|
model-bin/finetune/base/{checkpoint-58240 → checkpoint-58863}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629788442.126761/events.out.tfevents.1629788442.c435e1c5ee04.920.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:604c72456103c048ed1693e5c48b46865ca4fea2d95ce2a32c3d17509cf2ba26
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629789087.6950493/events.out.tfevents.1629789087.c435e1c5ee04.920.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5180d2b36f470c893631930ac5adca1c645d82258244fe95434f4e498c26e4db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629789729.6492243/events.out.tfevents.1629789729.c435e1c5ee04.920.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:99ea0f398e53dd81171f7d57f07e05be7476c85d3ca419e3f98e954337368f48
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629790377.8418694/events.out.tfevents.1629790377.c435e1c5ee04.920.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b596940341ed38c06524c37906bbb076cfc0432abecb3799d62914a952caadb5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629791034.1306834/events.out.tfevents.1629791034.c435e1c5ee04.920.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:955ccd10408e0fdcfecfab845910d4dd18e96bb30e648211e0d888e69c852e7c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629788442.c435e1c5ee04.920.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22499354f72e96e46369d0e780671621e3b5e3f983268f0d379381a43c72d156
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629789087.c435e1c5ee04.920.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4c486c4efc6b61b3f00d9db9cbff1c2cbe905fe172ebb850ac5f2fd27d2b850
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629789729.c435e1c5ee04.920.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b15cf706c251a6c6f585b6a14b9dc4b5a93669409ff6f686376fa453db3a85b2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629790377.c435e1c5ee04.920.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3defc3aac724851bdd10b9cd6305b5aa096a37939112160e5446f442fafa6d6c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629791034.c435e1c5ee04.920.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:587fb13106e4ac50c2068eb3c57e604e6fbe32a14a832b92beabcfdfa87c4c6c
|
| 3 |
+
size 8622
|