"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629690100.2524/events.out.tfevents.1629690100.74272264b15c.932.33 +3 -0
- model-bin/finetune/base/log/1629690756.9281442/events.out.tfevents.1629690756.74272264b15c.932.35 +3 -0
- model-bin/finetune/base/log/1629691397.683956/events.out.tfevents.1629691397.74272264b15c.932.37 +3 -0
- model-bin/finetune/base/log/1629692041.121403/events.out.tfevents.1629692041.74272264b15c.932.39 +3 -0
- model-bin/finetune/base/log/1629692696.2855487/events.out.tfevents.1629692696.74272264b15c.932.41 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629690100.74272264b15c.932.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629690756.74272264b15c.932.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629691397.74272264b15c.932.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629692041.74272264b15c.932.38 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629692696.74272264b15c.932.40 +3 -0
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/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 sha256:1fa1944ad9925ff162d9d338483134cd4cf2d971b4a8a64d3b5f2f025022cb3d
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/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:df133ded1d828e21b4f907be9ad2f00fb68d52bd05a8bcda95c36a1dce156558
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/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:ae38b17925b08fc3d462077ac961c70646e9e91b35d247c392276c77591ecb2f
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/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-40195 → checkpoint-40819}/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:a201f0b3f3c23a08f89cc194dc483842c8de33da82f56f1b279c0a7ea8a2eeb1
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -150507,11 +150507,800 @@
|
|
| 150507 |
"eval_steps_per_second": 0.686,
|
| 150508 |
"eval_wer": 0.22157434402332363,
|
| 150509 |
"step": 40195
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 150510 |
}
|
| 150511 |
],
|
| 150512 |
"max_steps": 625000,
|
| 150513 |
"num_train_epochs": 5000,
|
| 150514 |
-
"total_flos": 1.
|
| 150515 |
"trial_name": null,
|
| 150516 |
"trial_params": null
|
| 150517 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
+
"epoch": 326.0,
|
| 5 |
+
"global_step": 40819,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 150507 |
"eval_steps_per_second": 0.686,
|
| 150508 |
"eval_wer": 0.22157434402332363,
|
| 150509 |
"step": 40195
|
| 150510 |
+
},
|
| 150511 |
+
{
|
| 150512 |
+
"epoch": 321.04,
|
| 150513 |
+
"learning_rate": 9.372035256410256e-06,
|
| 150514 |
+
"loss": 0.3373,
|
| 150515 |
+
"step": 40200
|
| 150516 |
+
},
|
| 150517 |
+
{
|
| 150518 |
+
"epoch": 321.08,
|
| 150519 |
+
"learning_rate": 9.37195512820513e-06,
|
| 150520 |
+
"loss": 0.3361,
|
| 150521 |
+
"step": 40205
|
| 150522 |
+
},
|
| 150523 |
+
{
|
| 150524 |
+
"epoch": 321.12,
|
| 150525 |
+
"learning_rate": 9.371875e-06,
|
| 150526 |
+
"loss": 0.4543,
|
| 150527 |
+
"step": 40210
|
| 150528 |
+
},
|
| 150529 |
+
{
|
| 150530 |
+
"epoch": 321.16,
|
| 150531 |
+
"learning_rate": 9.371794871794872e-06,
|
| 150532 |
+
"loss": 0.6221,
|
| 150533 |
+
"step": 40215
|
| 150534 |
+
},
|
| 150535 |
+
{
|
| 150536 |
+
"epoch": 321.2,
|
| 150537 |
+
"learning_rate": 9.371714743589745e-06,
|
| 150538 |
+
"loss": 1.255,
|
| 150539 |
+
"step": 40220
|
| 150540 |
+
},
|
| 150541 |
+
{
|
| 150542 |
+
"epoch": 321.24,
|
| 150543 |
+
"learning_rate": 9.371634615384617e-06,
|
| 150544 |
+
"loss": 0.4445,
|
| 150545 |
+
"step": 40225
|
| 150546 |
+
},
|
| 150547 |
+
{
|
| 150548 |
+
"epoch": 321.28,
|
| 150549 |
+
"learning_rate": 9.371554487179488e-06,
|
| 150550 |
+
"loss": 0.3638,
|
| 150551 |
+
"step": 40230
|
| 150552 |
+
},
|
| 150553 |
+
{
|
| 150554 |
+
"epoch": 321.32,
|
| 150555 |
+
"learning_rate": 9.37147435897436e-06,
|
| 150556 |
+
"loss": 0.4107,
|
| 150557 |
+
"step": 40235
|
| 150558 |
+
},
|
| 150559 |
+
{
|
| 150560 |
+
"epoch": 321.36,
|
| 150561 |
+
"learning_rate": 9.371394230769232e-06,
|
| 150562 |
+
"loss": 0.519,
|
| 150563 |
+
"step": 40240
|
| 150564 |
+
},
|
| 150565 |
+
{
|
| 150566 |
+
"epoch": 321.4,
|
| 150567 |
+
"learning_rate": 9.371314102564104e-06,
|
| 150568 |
+
"loss": 1.4418,
|
| 150569 |
+
"step": 40245
|
| 150570 |
+
},
|
| 150571 |
+
{
|
| 150572 |
+
"epoch": 321.44,
|
| 150573 |
+
"learning_rate": 9.371233974358975e-06,
|
| 150574 |
+
"loss": 0.4568,
|
| 150575 |
+
"step": 40250
|
| 150576 |
+
},
|
| 150577 |
+
{
|
| 150578 |
+
"epoch": 321.48,
|
| 150579 |
+
"learning_rate": 9.371153846153846e-06,
|
| 150580 |
+
"loss": 0.3485,
|
| 150581 |
+
"step": 40255
|
| 150582 |
+
},
|
| 150583 |
+
{
|
| 150584 |
+
"epoch": 321.52,
|
| 150585 |
+
"learning_rate": 9.37107371794872e-06,
|
| 150586 |
+
"loss": 0.3938,
|
| 150587 |
+
"step": 40260
|
| 150588 |
+
},
|
| 150589 |
+
{
|
| 150590 |
+
"epoch": 321.56,
|
| 150591 |
+
"learning_rate": 9.370993589743589e-06,
|
| 150592 |
+
"loss": 0.4886,
|
| 150593 |
+
"step": 40265
|
| 150594 |
+
},
|
| 150595 |
+
{
|
| 150596 |
+
"epoch": 321.6,
|
| 150597 |
+
"learning_rate": 9.370913461538462e-06,
|
| 150598 |
+
"loss": 1.6128,
|
| 150599 |
+
"step": 40270
|
| 150600 |
+
},
|
| 150601 |
+
{
|
| 150602 |
+
"epoch": 321.64,
|
| 150603 |
+
"learning_rate": 9.370833333333335e-06,
|
| 150604 |
+
"loss": 0.3649,
|
| 150605 |
+
"step": 40275
|
| 150606 |
+
},
|
| 150607 |
+
{
|
| 150608 |
+
"epoch": 321.68,
|
| 150609 |
+
"learning_rate": 9.370753205128205e-06,
|
| 150610 |
+
"loss": 0.3021,
|
| 150611 |
+
"step": 40280
|
| 150612 |
+
},
|
| 150613 |
+
{
|
| 150614 |
+
"epoch": 321.72,
|
| 150615 |
+
"learning_rate": 9.370673076923078e-06,
|
| 150616 |
+
"loss": 0.3555,
|
| 150617 |
+
"step": 40285
|
| 150618 |
+
},
|
| 150619 |
+
{
|
| 150620 |
+
"epoch": 321.76,
|
| 150621 |
+
"learning_rate": 9.37059294871795e-06,
|
| 150622 |
+
"loss": 0.5168,
|
| 150623 |
+
"step": 40290
|
| 150624 |
+
},
|
| 150625 |
+
{
|
| 150626 |
+
"epoch": 321.8,
|
| 150627 |
+
"learning_rate": 9.370512820512822e-06,
|
| 150628 |
+
"loss": 1.2727,
|
| 150629 |
+
"step": 40295
|
| 150630 |
+
},
|
| 150631 |
+
{
|
| 150632 |
+
"epoch": 321.84,
|
| 150633 |
+
"learning_rate": 9.370432692307692e-06,
|
| 150634 |
+
"loss": 0.384,
|
| 150635 |
+
"step": 40300
|
| 150636 |
+
},
|
| 150637 |
+
{
|
| 150638 |
+
"epoch": 321.88,
|
| 150639 |
+
"learning_rate": 9.370352564102565e-06,
|
| 150640 |
+
"loss": 0.3779,
|
| 150641 |
+
"step": 40305
|
| 150642 |
+
},
|
| 150643 |
+
{
|
| 150644 |
+
"epoch": 321.92,
|
| 150645 |
+
"learning_rate": 9.370272435897436e-06,
|
| 150646 |
+
"loss": 0.4172,
|
| 150647 |
+
"step": 40310
|
| 150648 |
+
},
|
| 150649 |
+
{
|
| 150650 |
+
"epoch": 321.96,
|
| 150651 |
+
"learning_rate": 9.370192307692308e-06,
|
| 150652 |
+
"loss": 0.5787,
|
| 150653 |
+
"step": 40315
|
| 150654 |
+
},
|
| 150655 |
+
{
|
| 150656 |
+
"epoch": 322.0,
|
| 150657 |
+
"learning_rate": 9.370112179487181e-06,
|
| 150658 |
+
"loss": 1.3272,
|
| 150659 |
+
"step": 40320
|
| 150660 |
+
},
|
| 150661 |
+
{
|
| 150662 |
+
"epoch": 322.0,
|
| 150663 |
+
"eval_loss": 0.47489798069000244,
|
| 150664 |
+
"eval_runtime": 39.8494,
|
| 150665 |
+
"eval_samples_per_second": 21.104,
|
| 150666 |
+
"eval_steps_per_second": 0.678,
|
| 150667 |
+
"eval_wer": 0.2066909090909091,
|
| 150668 |
+
"step": 40320
|
| 150669 |
+
},
|
| 150670 |
+
{
|
| 150671 |
+
"epoch": 322.04,
|
| 150672 |
+
"learning_rate": 9.370032051282052e-06,
|
| 150673 |
+
"loss": 0.4442,
|
| 150674 |
+
"step": 40325
|
| 150675 |
+
},
|
| 150676 |
+
{
|
| 150677 |
+
"epoch": 322.08,
|
| 150678 |
+
"learning_rate": 9.369951923076924e-06,
|
| 150679 |
+
"loss": 0.4147,
|
| 150680 |
+
"step": 40330
|
| 150681 |
+
},
|
| 150682 |
+
{
|
| 150683 |
+
"epoch": 322.12,
|
| 150684 |
+
"learning_rate": 9.369871794871795e-06,
|
| 150685 |
+
"loss": 0.3458,
|
| 150686 |
+
"step": 40335
|
| 150687 |
+
},
|
| 150688 |
+
{
|
| 150689 |
+
"epoch": 322.16,
|
| 150690 |
+
"learning_rate": 9.369791666666668e-06,
|
| 150691 |
+
"loss": 0.5212,
|
| 150692 |
+
"step": 40340
|
| 150693 |
+
},
|
| 150694 |
+
{
|
| 150695 |
+
"epoch": 322.2,
|
| 150696 |
+
"learning_rate": 9.36971153846154e-06,
|
| 150697 |
+
"loss": 1.2348,
|
| 150698 |
+
"step": 40345
|
| 150699 |
+
},
|
| 150700 |
+
{
|
| 150701 |
+
"epoch": 322.24,
|
| 150702 |
+
"learning_rate": 9.36963141025641e-06,
|
| 150703 |
+
"loss": 0.3912,
|
| 150704 |
+
"step": 40350
|
| 150705 |
+
},
|
| 150706 |
+
{
|
| 150707 |
+
"epoch": 322.28,
|
| 150708 |
+
"learning_rate": 9.369551282051282e-06,
|
| 150709 |
+
"loss": 0.3632,
|
| 150710 |
+
"step": 40355
|
| 150711 |
+
},
|
| 150712 |
+
{
|
| 150713 |
+
"epoch": 322.32,
|
| 150714 |
+
"learning_rate": 9.369471153846155e-06,
|
| 150715 |
+
"loss": 0.4197,
|
| 150716 |
+
"step": 40360
|
| 150717 |
+
},
|
| 150718 |
+
{
|
| 150719 |
+
"epoch": 322.36,
|
| 150720 |
+
"learning_rate": 9.369391025641026e-06,
|
| 150721 |
+
"loss": 0.5759,
|
| 150722 |
+
"step": 40365
|
| 150723 |
+
},
|
| 150724 |
+
{
|
| 150725 |
+
"epoch": 322.4,
|
| 150726 |
+
"learning_rate": 9.369310897435898e-06,
|
| 150727 |
+
"loss": 1.417,
|
| 150728 |
+
"step": 40370
|
| 150729 |
+
},
|
| 150730 |
+
{
|
| 150731 |
+
"epoch": 322.44,
|
| 150732 |
+
"learning_rate": 9.369230769230771e-06,
|
| 150733 |
+
"loss": 0.4237,
|
| 150734 |
+
"step": 40375
|
| 150735 |
+
},
|
| 150736 |
+
{
|
| 150737 |
+
"epoch": 322.48,
|
| 150738 |
+
"learning_rate": 9.369150641025642e-06,
|
| 150739 |
+
"loss": 0.3722,
|
| 150740 |
+
"step": 40380
|
| 150741 |
+
},
|
| 150742 |
+
{
|
| 150743 |
+
"epoch": 322.52,
|
| 150744 |
+
"learning_rate": 9.369070512820514e-06,
|
| 150745 |
+
"loss": 0.442,
|
| 150746 |
+
"step": 40385
|
| 150747 |
+
},
|
| 150748 |
+
{
|
| 150749 |
+
"epoch": 322.56,
|
| 150750 |
+
"learning_rate": 9.368990384615385e-06,
|
| 150751 |
+
"loss": 0.5427,
|
| 150752 |
+
"step": 40390
|
| 150753 |
+
},
|
| 150754 |
+
{
|
| 150755 |
+
"epoch": 322.6,
|
| 150756 |
+
"learning_rate": 9.368910256410258e-06,
|
| 150757 |
+
"loss": 1.5102,
|
| 150758 |
+
"step": 40395
|
| 150759 |
+
},
|
| 150760 |
+
{
|
| 150761 |
+
"epoch": 322.64,
|
| 150762 |
+
"learning_rate": 9.36883012820513e-06,
|
| 150763 |
+
"loss": 0.4065,
|
| 150764 |
+
"step": 40400
|
| 150765 |
+
},
|
| 150766 |
+
{
|
| 150767 |
+
"epoch": 322.68,
|
| 150768 |
+
"learning_rate": 9.36875e-06,
|
| 150769 |
+
"loss": 0.3247,
|
| 150770 |
+
"step": 40405
|
| 150771 |
+
},
|
| 150772 |
+
{
|
| 150773 |
+
"epoch": 322.72,
|
| 150774 |
+
"learning_rate": 9.368669871794872e-06,
|
| 150775 |
+
"loss": 0.5111,
|
| 150776 |
+
"step": 40410
|
| 150777 |
+
},
|
| 150778 |
+
{
|
| 150779 |
+
"epoch": 322.76,
|
| 150780 |
+
"learning_rate": 9.368589743589745e-06,
|
| 150781 |
+
"loss": 0.566,
|
| 150782 |
+
"step": 40415
|
| 150783 |
+
},
|
| 150784 |
+
{
|
| 150785 |
+
"epoch": 322.8,
|
| 150786 |
+
"learning_rate": 9.368509615384616e-06,
|
| 150787 |
+
"loss": 1.3414,
|
| 150788 |
+
"step": 40420
|
| 150789 |
+
},
|
| 150790 |
+
{
|
| 150791 |
+
"epoch": 322.84,
|
| 150792 |
+
"learning_rate": 9.368429487179488e-06,
|
| 150793 |
+
"loss": 0.5129,
|
| 150794 |
+
"step": 40425
|
| 150795 |
+
},
|
| 150796 |
+
{
|
| 150797 |
+
"epoch": 322.88,
|
| 150798 |
+
"learning_rate": 9.368349358974361e-06,
|
| 150799 |
+
"loss": 0.3569,
|
| 150800 |
+
"step": 40430
|
| 150801 |
+
},
|
| 150802 |
+
{
|
| 150803 |
+
"epoch": 322.92,
|
| 150804 |
+
"learning_rate": 9.36826923076923e-06,
|
| 150805 |
+
"loss": 0.3942,
|
| 150806 |
+
"step": 40435
|
| 150807 |
+
},
|
| 150808 |
+
{
|
| 150809 |
+
"epoch": 322.96,
|
| 150810 |
+
"learning_rate": 9.368189102564104e-06,
|
| 150811 |
+
"loss": 0.4821,
|
| 150812 |
+
"step": 40440
|
| 150813 |
+
},
|
| 150814 |
+
{
|
| 150815 |
+
"epoch": 323.0,
|
| 150816 |
+
"learning_rate": 9.368108974358975e-06,
|
| 150817 |
+
"loss": 1.1952,
|
| 150818 |
+
"step": 40445
|
| 150819 |
+
},
|
| 150820 |
+
{
|
| 150821 |
+
"epoch": 323.0,
|
| 150822 |
+
"eval_loss": 0.4279903471469879,
|
| 150823 |
+
"eval_runtime": 36.3482,
|
| 150824 |
+
"eval_samples_per_second": 23.11,
|
| 150825 |
+
"eval_steps_per_second": 0.743,
|
| 150826 |
+
"eval_wer": 0.2130316331163917,
|
| 150827 |
+
"step": 40445
|
| 150828 |
+
},
|
| 150829 |
+
{
|
| 150830 |
+
"epoch": 326.04,
|
| 150831 |
+
"learning_rate": 9.368028846153846e-06,
|
| 150832 |
+
"loss": 0.3865,
|
| 150833 |
+
"step": 40450
|
| 150834 |
+
},
|
| 150835 |
+
{
|
| 150836 |
+
"epoch": 326.08,
|
| 150837 |
+
"learning_rate": 9.367948717948718e-06,
|
| 150838 |
+
"loss": 0.3995,
|
| 150839 |
+
"step": 40455
|
| 150840 |
+
},
|
| 150841 |
+
{
|
| 150842 |
+
"epoch": 326.12,
|
| 150843 |
+
"learning_rate": 9.36786858974359e-06,
|
| 150844 |
+
"loss": 0.33,
|
| 150845 |
+
"step": 40460
|
| 150846 |
+
},
|
| 150847 |
+
{
|
| 150848 |
+
"epoch": 326.16,
|
| 150849 |
+
"learning_rate": 9.367788461538462e-06,
|
| 150850 |
+
"loss": 0.5698,
|
| 150851 |
+
"step": 40465
|
| 150852 |
+
},
|
| 150853 |
+
{
|
| 150854 |
+
"epoch": 326.2,
|
| 150855 |
+
"learning_rate": 9.367708333333333e-06,
|
| 150856 |
+
"loss": 1.3174,
|
| 150857 |
+
"step": 40470
|
| 150858 |
+
},
|
| 150859 |
+
{
|
| 150860 |
+
"epoch": 326.24,
|
| 150861 |
+
"learning_rate": 9.367628205128207e-06,
|
| 150862 |
+
"loss": 0.3983,
|
| 150863 |
+
"step": 40475
|
| 150864 |
+
},
|
| 150865 |
+
{
|
| 150866 |
+
"epoch": 326.28,
|
| 150867 |
+
"learning_rate": 9.367548076923078e-06,
|
| 150868 |
+
"loss": 0.4105,
|
| 150869 |
+
"step": 40480
|
| 150870 |
+
},
|
| 150871 |
+
{
|
| 150872 |
+
"epoch": 326.32,
|
| 150873 |
+
"learning_rate": 9.36746794871795e-06,
|
| 150874 |
+
"loss": 0.3333,
|
| 150875 |
+
"step": 40485
|
| 150876 |
+
},
|
| 150877 |
+
{
|
| 150878 |
+
"epoch": 326.36,
|
| 150879 |
+
"learning_rate": 9.36738782051282e-06,
|
| 150880 |
+
"loss": 0.4608,
|
| 150881 |
+
"step": 40490
|
| 150882 |
+
},
|
| 150883 |
+
{
|
| 150884 |
+
"epoch": 326.4,
|
| 150885 |
+
"learning_rate": 9.367307692307694e-06,
|
| 150886 |
+
"loss": 1.283,
|
| 150887 |
+
"step": 40495
|
| 150888 |
+
},
|
| 150889 |
+
{
|
| 150890 |
+
"epoch": 326.44,
|
| 150891 |
+
"learning_rate": 9.367227564102565e-06,
|
| 150892 |
+
"loss": 0.3786,
|
| 150893 |
+
"step": 40500
|
| 150894 |
+
},
|
| 150895 |
+
{
|
| 150896 |
+
"epoch": 326.48,
|
| 150897 |
+
"learning_rate": 9.367147435897436e-06,
|
| 150898 |
+
"loss": 0.3512,
|
| 150899 |
+
"step": 40505
|
| 150900 |
+
},
|
| 150901 |
+
{
|
| 150902 |
+
"epoch": 326.52,
|
| 150903 |
+
"learning_rate": 9.367067307692308e-06,
|
| 150904 |
+
"loss": 0.3313,
|
| 150905 |
+
"step": 40510
|
| 150906 |
+
},
|
| 150907 |
+
{
|
| 150908 |
+
"epoch": 326.56,
|
| 150909 |
+
"learning_rate": 9.36698717948718e-06,
|
| 150910 |
+
"loss": 0.6023,
|
| 150911 |
+
"step": 40515
|
| 150912 |
+
},
|
| 150913 |
+
{
|
| 150914 |
+
"epoch": 326.6,
|
| 150915 |
+
"learning_rate": 9.366907051282052e-06,
|
| 150916 |
+
"loss": 1.3057,
|
| 150917 |
+
"step": 40520
|
| 150918 |
+
},
|
| 150919 |
+
{
|
| 150920 |
+
"epoch": 326.64,
|
| 150921 |
+
"learning_rate": 9.366826923076923e-06,
|
| 150922 |
+
"loss": 0.3721,
|
| 150923 |
+
"step": 40525
|
| 150924 |
+
},
|
| 150925 |
+
{
|
| 150926 |
+
"epoch": 326.68,
|
| 150927 |
+
"learning_rate": 9.366746794871797e-06,
|
| 150928 |
+
"loss": 0.4084,
|
| 150929 |
+
"step": 40530
|
| 150930 |
+
},
|
| 150931 |
+
{
|
| 150932 |
+
"epoch": 326.72,
|
| 150933 |
+
"learning_rate": 9.366666666666668e-06,
|
| 150934 |
+
"loss": 0.4063,
|
| 150935 |
+
"step": 40535
|
| 150936 |
+
},
|
| 150937 |
+
{
|
| 150938 |
+
"epoch": 326.76,
|
| 150939 |
+
"learning_rate": 9.36658653846154e-06,
|
| 150940 |
+
"loss": 0.4393,
|
| 150941 |
+
"step": 40540
|
| 150942 |
+
},
|
| 150943 |
+
{
|
| 150944 |
+
"epoch": 326.8,
|
| 150945 |
+
"learning_rate": 9.36650641025641e-06,
|
| 150946 |
+
"loss": 1.2195,
|
| 150947 |
+
"step": 40545
|
| 150948 |
+
},
|
| 150949 |
+
{
|
| 150950 |
+
"epoch": 326.84,
|
| 150951 |
+
"learning_rate": 9.366426282051284e-06,
|
| 150952 |
+
"loss": 0.382,
|
| 150953 |
+
"step": 40550
|
| 150954 |
+
},
|
| 150955 |
+
{
|
| 150956 |
+
"epoch": 326.88,
|
| 150957 |
+
"learning_rate": 9.366346153846155e-06,
|
| 150958 |
+
"loss": 0.3907,
|
| 150959 |
+
"step": 40555
|
| 150960 |
+
},
|
| 150961 |
+
{
|
| 150962 |
+
"epoch": 326.92,
|
| 150963 |
+
"learning_rate": 9.366266025641026e-06,
|
| 150964 |
+
"loss": 0.4608,
|
| 150965 |
+
"step": 40560
|
| 150966 |
+
},
|
| 150967 |
+
{
|
| 150968 |
+
"epoch": 326.96,
|
| 150969 |
+
"learning_rate": 9.366185897435898e-06,
|
| 150970 |
+
"loss": 0.4996,
|
| 150971 |
+
"step": 40565
|
| 150972 |
+
},
|
| 150973 |
+
{
|
| 150974 |
+
"epoch": 327.0,
|
| 150975 |
+
"eval_loss": 0.41771993041038513,
|
| 150976 |
+
"eval_runtime": 38.7873,
|
| 150977 |
+
"eval_samples_per_second": 21.657,
|
| 150978 |
+
"eval_steps_per_second": 0.696,
|
| 150979 |
+
"eval_wer": 0.21109467455621303,
|
| 150980 |
+
"step": 40569
|
| 150981 |
+
},
|
| 150982 |
+
{
|
| 150983 |
+
"epoch": 324.01,
|
| 150984 |
+
"learning_rate": 9.36610576923077e-06,
|
| 150985 |
+
"loss": 0.4814,
|
| 150986 |
+
"step": 40570
|
| 150987 |
+
},
|
| 150988 |
+
{
|
| 150989 |
+
"epoch": 324.05,
|
| 150990 |
+
"learning_rate": 9.366025641025642e-06,
|
| 150991 |
+
"loss": 0.3825,
|
| 150992 |
+
"step": 40575
|
| 150993 |
+
},
|
| 150994 |
+
{
|
| 150995 |
+
"epoch": 324.09,
|
| 150996 |
+
"learning_rate": 9.365945512820514e-06,
|
| 150997 |
+
"loss": 0.3723,
|
| 150998 |
+
"step": 40580
|
| 150999 |
+
},
|
| 151000 |
+
{
|
| 151001 |
+
"epoch": 324.13,
|
| 151002 |
+
"learning_rate": 9.365865384615387e-06,
|
| 151003 |
+
"loss": 0.4287,
|
| 151004 |
+
"step": 40585
|
| 151005 |
+
},
|
| 151006 |
+
{
|
| 151007 |
+
"epoch": 324.17,
|
| 151008 |
+
"learning_rate": 9.365785256410256e-06,
|
| 151009 |
+
"loss": 0.6168,
|
| 151010 |
+
"step": 40590
|
| 151011 |
+
},
|
| 151012 |
+
{
|
| 151013 |
+
"epoch": 324.21,
|
| 151014 |
+
"learning_rate": 9.36570512820513e-06,
|
| 151015 |
+
"loss": 1.4175,
|
| 151016 |
+
"step": 40595
|
| 151017 |
+
},
|
| 151018 |
+
{
|
| 151019 |
+
"epoch": 324.25,
|
| 151020 |
+
"learning_rate": 9.365625e-06,
|
| 151021 |
+
"loss": 0.3484,
|
| 151022 |
+
"step": 40600
|
| 151023 |
+
},
|
| 151024 |
+
{
|
| 151025 |
+
"epoch": 324.29,
|
| 151026 |
+
"learning_rate": 9.365544871794872e-06,
|
| 151027 |
+
"loss": 0.3285,
|
| 151028 |
+
"step": 40605
|
| 151029 |
+
},
|
| 151030 |
+
{
|
| 151031 |
+
"epoch": 324.33,
|
| 151032 |
+
"learning_rate": 9.365464743589743e-06,
|
| 151033 |
+
"loss": 0.4078,
|
| 151034 |
+
"step": 40610
|
| 151035 |
+
},
|
| 151036 |
+
{
|
| 151037 |
+
"epoch": 324.37,
|
| 151038 |
+
"learning_rate": 9.365384615384616e-06,
|
| 151039 |
+
"loss": 0.6791,
|
| 151040 |
+
"step": 40615
|
| 151041 |
+
},
|
| 151042 |
+
{
|
| 151043 |
+
"epoch": 324.41,
|
| 151044 |
+
"learning_rate": 9.365304487179488e-06,
|
| 151045 |
+
"loss": 1.389,
|
| 151046 |
+
"step": 40620
|
| 151047 |
+
},
|
| 151048 |
+
{
|
| 151049 |
+
"epoch": 324.45,
|
| 151050 |
+
"learning_rate": 9.365224358974359e-06,
|
| 151051 |
+
"loss": 0.3931,
|
| 151052 |
+
"step": 40625
|
| 151053 |
+
},
|
| 151054 |
+
{
|
| 151055 |
+
"epoch": 324.49,
|
| 151056 |
+
"learning_rate": 9.365144230769232e-06,
|
| 151057 |
+
"loss": 0.4694,
|
| 151058 |
+
"step": 40630
|
| 151059 |
+
},
|
| 151060 |
+
{
|
| 151061 |
+
"epoch": 324.53,
|
| 151062 |
+
"learning_rate": 9.365064102564104e-06,
|
| 151063 |
+
"loss": 0.4064,
|
| 151064 |
+
"step": 40635
|
| 151065 |
+
},
|
| 151066 |
+
{
|
| 151067 |
+
"epoch": 324.57,
|
| 151068 |
+
"learning_rate": 9.364983974358975e-06,
|
| 151069 |
+
"loss": 1.0477,
|
| 151070 |
+
"step": 40640
|
| 151071 |
+
},
|
| 151072 |
+
{
|
| 151073 |
+
"epoch": 324.61,
|
| 151074 |
+
"learning_rate": 9.364903846153846e-06,
|
| 151075 |
+
"loss": 1.2383,
|
| 151076 |
+
"step": 40645
|
| 151077 |
+
},
|
| 151078 |
+
{
|
| 151079 |
+
"epoch": 324.65,
|
| 151080 |
+
"learning_rate": 9.36482371794872e-06,
|
| 151081 |
+
"loss": 0.3461,
|
| 151082 |
+
"step": 40650
|
| 151083 |
+
},
|
| 151084 |
+
{
|
| 151085 |
+
"epoch": 324.69,
|
| 151086 |
+
"learning_rate": 9.36474358974359e-06,
|
| 151087 |
+
"loss": 0.3728,
|
| 151088 |
+
"step": 40655
|
| 151089 |
+
},
|
| 151090 |
+
{
|
| 151091 |
+
"epoch": 324.73,
|
| 151092 |
+
"learning_rate": 9.364663461538462e-06,
|
| 151093 |
+
"loss": 0.4339,
|
| 151094 |
+
"step": 40660
|
| 151095 |
+
},
|
| 151096 |
+
{
|
| 151097 |
+
"epoch": 324.77,
|
| 151098 |
+
"learning_rate": 9.364583333333333e-06,
|
| 151099 |
+
"loss": 0.7665,
|
| 151100 |
+
"step": 40665
|
| 151101 |
+
},
|
| 151102 |
+
{
|
| 151103 |
+
"epoch": 324.81,
|
| 151104 |
+
"learning_rate": 9.364503205128206e-06,
|
| 151105 |
+
"loss": 1.3851,
|
| 151106 |
+
"step": 40670
|
| 151107 |
+
},
|
| 151108 |
+
{
|
| 151109 |
+
"epoch": 324.85,
|
| 151110 |
+
"learning_rate": 9.364423076923078e-06,
|
| 151111 |
+
"loss": 0.3449,
|
| 151112 |
+
"step": 40675
|
| 151113 |
+
},
|
| 151114 |
+
{
|
| 151115 |
+
"epoch": 324.89,
|
| 151116 |
+
"learning_rate": 9.36434294871795e-06,
|
| 151117 |
+
"loss": 0.3454,
|
| 151118 |
+
"step": 40680
|
| 151119 |
+
},
|
| 151120 |
+
{
|
| 151121 |
+
"epoch": 324.93,
|
| 151122 |
+
"learning_rate": 9.364262820512822e-06,
|
| 151123 |
+
"loss": 0.4433,
|
| 151124 |
+
"step": 40685
|
| 151125 |
+
},
|
| 151126 |
+
{
|
| 151127 |
+
"epoch": 324.97,
|
| 151128 |
+
"learning_rate": 9.364182692307694e-06,
|
| 151129 |
+
"loss": 0.6765,
|
| 151130 |
+
"step": 40690
|
| 151131 |
+
},
|
| 151132 |
+
{
|
| 151133 |
+
"epoch": 325.0,
|
| 151134 |
+
"eval_loss": 0.43804189562797546,
|
| 151135 |
+
"eval_runtime": 38.5123,
|
| 151136 |
+
"eval_samples_per_second": 21.811,
|
| 151137 |
+
"eval_steps_per_second": 0.701,
|
| 151138 |
+
"eval_wer": 0.2098334424321769,
|
| 151139 |
+
"step": 40694
|
| 151140 |
+
},
|
| 151141 |
+
{
|
| 151142 |
+
"epoch": 325.01,
|
| 151143 |
+
"learning_rate": 9.364118589743591e-06,
|
| 151144 |
+
"loss": 0.6051,
|
| 151145 |
+
"step": 40695
|
| 151146 |
+
},
|
| 151147 |
+
{
|
| 151148 |
+
"epoch": 325.05,
|
| 151149 |
+
"learning_rate": 9.364038461538462e-06,
|
| 151150 |
+
"loss": 0.3775,
|
| 151151 |
+
"step": 40700
|
| 151152 |
+
},
|
| 151153 |
+
{
|
| 151154 |
+
"epoch": 325.09,
|
| 151155 |
+
"learning_rate": 9.363958333333334e-06,
|
| 151156 |
+
"loss": 0.3355,
|
| 151157 |
+
"step": 40705
|
| 151158 |
+
},
|
| 151159 |
+
{
|
| 151160 |
+
"epoch": 325.13,
|
| 151161 |
+
"learning_rate": 9.363878205128207e-06,
|
| 151162 |
+
"loss": 0.4757,
|
| 151163 |
+
"step": 40710
|
| 151164 |
+
},
|
| 151165 |
+
{
|
| 151166 |
+
"epoch": 325.17,
|
| 151167 |
+
"learning_rate": 9.363798076923078e-06,
|
| 151168 |
+
"loss": 0.5973,
|
| 151169 |
+
"step": 40715
|
| 151170 |
+
},
|
| 151171 |
+
{
|
| 151172 |
+
"epoch": 325.21,
|
| 151173 |
+
"learning_rate": 9.36371794871795e-06,
|
| 151174 |
+
"loss": 1.25,
|
| 151175 |
+
"step": 40720
|
| 151176 |
+
},
|
| 151177 |
+
{
|
| 151178 |
+
"epoch": 325.25,
|
| 151179 |
+
"learning_rate": 9.36363782051282e-06,
|
| 151180 |
+
"loss": 0.3833,
|
| 151181 |
+
"step": 40725
|
| 151182 |
+
},
|
| 151183 |
+
{
|
| 151184 |
+
"epoch": 325.29,
|
| 151185 |
+
"learning_rate": 9.363557692307694e-06,
|
| 151186 |
+
"loss": 0.3853,
|
| 151187 |
+
"step": 40730
|
| 151188 |
+
},
|
| 151189 |
+
{
|
| 151190 |
+
"epoch": 325.33,
|
| 151191 |
+
"learning_rate": 9.363477564102564e-06,
|
| 151192 |
+
"loss": 0.4051,
|
| 151193 |
+
"step": 40735
|
| 151194 |
+
},
|
| 151195 |
+
{
|
| 151196 |
+
"epoch": 325.37,
|
| 151197 |
+
"learning_rate": 9.363397435897437e-06,
|
| 151198 |
+
"loss": 0.6165,
|
| 151199 |
+
"step": 40740
|
| 151200 |
+
},
|
| 151201 |
+
{
|
| 151202 |
+
"epoch": 325.41,
|
| 151203 |
+
"learning_rate": 9.36331730769231e-06,
|
| 151204 |
+
"loss": 1.2113,
|
| 151205 |
+
"step": 40745
|
| 151206 |
+
},
|
| 151207 |
+
{
|
| 151208 |
+
"epoch": 325.45,
|
| 151209 |
+
"learning_rate": 9.36323717948718e-06,
|
| 151210 |
+
"loss": 0.3662,
|
| 151211 |
+
"step": 40750
|
| 151212 |
+
},
|
| 151213 |
+
{
|
| 151214 |
+
"epoch": 325.49,
|
| 151215 |
+
"learning_rate": 9.363157051282052e-06,
|
| 151216 |
+
"loss": 0.6469,
|
| 151217 |
+
"step": 40755
|
| 151218 |
+
},
|
| 151219 |
+
{
|
| 151220 |
+
"epoch": 325.53,
|
| 151221 |
+
"learning_rate": 9.363076923076924e-06,
|
| 151222 |
+
"loss": 0.4419,
|
| 151223 |
+
"step": 40760
|
| 151224 |
+
},
|
| 151225 |
+
{
|
| 151226 |
+
"epoch": 325.57,
|
| 151227 |
+
"learning_rate": 9.362996794871795e-06,
|
| 151228 |
+
"loss": 0.599,
|
| 151229 |
+
"step": 40765
|
| 151230 |
+
},
|
| 151231 |
+
{
|
| 151232 |
+
"epoch": 325.61,
|
| 151233 |
+
"learning_rate": 9.362916666666666e-06,
|
| 151234 |
+
"loss": 1.1019,
|
| 151235 |
+
"step": 40770
|
| 151236 |
+
},
|
| 151237 |
+
{
|
| 151238 |
+
"epoch": 325.65,
|
| 151239 |
+
"learning_rate": 9.36283653846154e-06,
|
| 151240 |
+
"loss": 0.3234,
|
| 151241 |
+
"step": 40775
|
| 151242 |
+
},
|
| 151243 |
+
{
|
| 151244 |
+
"epoch": 325.69,
|
| 151245 |
+
"learning_rate": 9.362756410256411e-06,
|
| 151246 |
+
"loss": 0.3723,
|
| 151247 |
+
"step": 40780
|
| 151248 |
+
},
|
| 151249 |
+
{
|
| 151250 |
+
"epoch": 325.73,
|
| 151251 |
+
"learning_rate": 9.362676282051282e-06,
|
| 151252 |
+
"loss": 0.4181,
|
| 151253 |
+
"step": 40785
|
| 151254 |
+
},
|
| 151255 |
+
{
|
| 151256 |
+
"epoch": 325.77,
|
| 151257 |
+
"learning_rate": 9.362596153846155e-06,
|
| 151258 |
+
"loss": 0.5771,
|
| 151259 |
+
"step": 40790
|
| 151260 |
+
},
|
| 151261 |
+
{
|
| 151262 |
+
"epoch": 325.81,
|
| 151263 |
+
"learning_rate": 9.362516025641027e-06,
|
| 151264 |
+
"loss": 1.3528,
|
| 151265 |
+
"step": 40795
|
| 151266 |
+
},
|
| 151267 |
+
{
|
| 151268 |
+
"epoch": 325.85,
|
| 151269 |
+
"learning_rate": 9.362435897435898e-06,
|
| 151270 |
+
"loss": 0.3821,
|
| 151271 |
+
"step": 40800
|
| 151272 |
+
},
|
| 151273 |
+
{
|
| 151274 |
+
"epoch": 325.89,
|
| 151275 |
+
"learning_rate": 9.36235576923077e-06,
|
| 151276 |
+
"loss": 0.3939,
|
| 151277 |
+
"step": 40805
|
| 151278 |
+
},
|
| 151279 |
+
{
|
| 151280 |
+
"epoch": 325.93,
|
| 151281 |
+
"learning_rate": 9.362275641025642e-06,
|
| 151282 |
+
"loss": 0.4103,
|
| 151283 |
+
"step": 40810
|
| 151284 |
+
},
|
| 151285 |
+
{
|
| 151286 |
+
"epoch": 325.97,
|
| 151287 |
+
"learning_rate": 9.362195512820514e-06,
|
| 151288 |
+
"loss": 0.65,
|
| 151289 |
+
"step": 40815
|
| 151290 |
+
},
|
| 151291 |
+
{
|
| 151292 |
+
"epoch": 326.0,
|
| 151293 |
+
"eval_loss": 0.40566712617874146,
|
| 151294 |
+
"eval_runtime": 39.1923,
|
| 151295 |
+
"eval_samples_per_second": 21.433,
|
| 151296 |
+
"eval_steps_per_second": 0.689,
|
| 151297 |
+
"eval_wer": 0.2033972763215698,
|
| 151298 |
+
"step": 40819
|
| 151299 |
}
|
| 151300 |
],
|
| 151301 |
"max_steps": 625000,
|
| 151302 |
"num_train_epochs": 5000,
|
| 151303 |
+
"total_flos": 1.148630537488186e+20,
|
| 151304 |
"trial_name": null,
|
| 151305 |
"trial_params": null
|
| 151306 |
}
|
model-bin/finetune/base/{checkpoint-40195 → checkpoint-40819}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629690100.2524/events.out.tfevents.1629690100.74272264b15c.932.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b70b15a46d9247feabb13ca61798324d2ae04bac5174f12a13d4125844671b7e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629690756.9281442/events.out.tfevents.1629690756.74272264b15c.932.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebc14044e348ba7c624ee179ef864ee5f680d94a09e4d7936c67ab8ddbd23d42
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629691397.683956/events.out.tfevents.1629691397.74272264b15c.932.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6f8b69455a2ee36e125c7a1a70d868d0c44083361da4401fbed91f153193e9fd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629692041.121403/events.out.tfevents.1629692041.74272264b15c.932.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:33fd68a1471a0351ab66f7270ff6eb4d76ac7f5cf49c91bcbe676a31c1fbf085
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629692696.2855487/events.out.tfevents.1629692696.74272264b15c.932.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daf6eab4f5fa93b92f1b9e6fd46bb8809d46b55c83de315aaaaa40220a999c79
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629690100.74272264b15c.932.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15a96cb23f2536b0ddeac7e1c9525a0fbd6226ed2ab4d874029eb6e627675e00
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629690756.74272264b15c.932.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:18b4bd2f848ffc95040fc1414b54c03eb5a0f2aab7925c210c0e7e36bf3315ec
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629691397.74272264b15c.932.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:237ea131a3535a3f42bcbb6d315b26104412c9438f79b9058763968539b53d81
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629692041.74272264b15c.932.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a458e324b756078292edd471beff1699fb5f1f9495558c9e6f388b1e8dc3072
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629692696.74272264b15c.932.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bfa10e446a27602aeae2f2ab88405868e5e614e02be522d65b758837e039229a
|
| 3 |
+
size 8622
|