"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629904128.5656571/events.out.tfevents.1629904128.7e498afd5545.7645.65 +3 -0
- model-bin/finetune/base/log/1629904601.0615444/events.out.tfevents.1629904601.7e498afd5545.7645.67 +3 -0
- model-bin/finetune/base/log/1629905064.9216402/events.out.tfevents.1629905064.7e498afd5545.7645.69 +3 -0
- model-bin/finetune/base/log/1629905527.7759078/events.out.tfevents.1629905527.7e498afd5545.7645.71 +3 -0
- model-bin/finetune/base/log/1629905996.4959083/events.out.tfevents.1629905996.7e498afd5545.7645.73 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629904128.7e498afd5545.7645.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629904601.7e498afd5545.7645.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629905064.7e498afd5545.7645.68 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629905527.7e498afd5545.7645.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629905996.7e498afd5545.7645.72 +3 -0
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/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 sha256:499aed0d9010da5763c53fa596bd3690ab653855d90b29a7179c034cca89ab46
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/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:d2f4660a08d626636627c24d99c2d12de51bf5a22d32bade54a02d377b1c8589
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/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 sha256:1359678ed57ef76ba4a01755e2169ac00e857b4fb73f1475b304a099e28649dc
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/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 sha256:2e8a586bc9b63765c3fab357431c101314b42a44cf283ddbb669f9978dbc3e9b
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/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:fb6dd82815b3396a90d14cd2e98065c9d2df14b6b8fa8441658e29c12220c22b
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -201489,11 +201489,806 @@
|
|
| 201489 |
"eval_steps_per_second": 0.626,
|
| 201490 |
"eval_wer": 0.18905362543335547,
|
| 201491 |
"step": 80269
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 201492 |
}
|
| 201493 |
],
|
| 201494 |
-
"max_steps":
|
| 201495 |
"num_train_epochs": 5000,
|
| 201496 |
-
"total_flos": 2.
|
| 201497 |
"trial_name": null,
|
| 201498 |
"trial_params": null
|
| 201499 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 651.995983935743,
|
| 5 |
+
"global_step": 80891,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 201489 |
"eval_steps_per_second": 0.626,
|
| 201490 |
"eval_wer": 0.18905362543335547,
|
| 201491 |
"step": 80269
|
| 201492 |
+
},
|
| 201493 |
+
{
|
| 201494 |
+
"epoch": 642.01,
|
| 201495 |
+
"learning_rate": 8.719886914378029e-06,
|
| 201496 |
+
"loss": 0.4176,
|
| 201497 |
+
"step": 80270
|
| 201498 |
+
},
|
| 201499 |
+
{
|
| 201500 |
+
"epoch": 642.05,
|
| 201501 |
+
"learning_rate": 8.719806138933765e-06,
|
| 201502 |
+
"loss": 0.2807,
|
| 201503 |
+
"step": 80275
|
| 201504 |
+
},
|
| 201505 |
+
{
|
| 201506 |
+
"epoch": 642.09,
|
| 201507 |
+
"learning_rate": 8.719725363489499e-06,
|
| 201508 |
+
"loss": 0.3207,
|
| 201509 |
+
"step": 80280
|
| 201510 |
+
},
|
| 201511 |
+
{
|
| 201512 |
+
"epoch": 642.13,
|
| 201513 |
+
"learning_rate": 8.719644588045235e-06,
|
| 201514 |
+
"loss": 0.369,
|
| 201515 |
+
"step": 80285
|
| 201516 |
+
},
|
| 201517 |
+
{
|
| 201518 |
+
"epoch": 642.17,
|
| 201519 |
+
"learning_rate": 8.719563812600969e-06,
|
| 201520 |
+
"loss": 0.614,
|
| 201521 |
+
"step": 80290
|
| 201522 |
+
},
|
| 201523 |
+
{
|
| 201524 |
+
"epoch": 642.21,
|
| 201525 |
+
"learning_rate": 8.719483037156705e-06,
|
| 201526 |
+
"loss": 1.1049,
|
| 201527 |
+
"step": 80295
|
| 201528 |
+
},
|
| 201529 |
+
{
|
| 201530 |
+
"epoch": 642.25,
|
| 201531 |
+
"learning_rate": 8.719402261712439e-06,
|
| 201532 |
+
"loss": 0.327,
|
| 201533 |
+
"step": 80300
|
| 201534 |
+
},
|
| 201535 |
+
{
|
| 201536 |
+
"epoch": 642.29,
|
| 201537 |
+
"learning_rate": 8.719321486268175e-06,
|
| 201538 |
+
"loss": 0.3335,
|
| 201539 |
+
"step": 80305
|
| 201540 |
+
},
|
| 201541 |
+
{
|
| 201542 |
+
"epoch": 642.33,
|
| 201543 |
+
"learning_rate": 8.71924071082391e-06,
|
| 201544 |
+
"loss": 0.392,
|
| 201545 |
+
"step": 80310
|
| 201546 |
+
},
|
| 201547 |
+
{
|
| 201548 |
+
"epoch": 642.37,
|
| 201549 |
+
"learning_rate": 8.719159935379645e-06,
|
| 201550 |
+
"loss": 0.4432,
|
| 201551 |
+
"step": 80315
|
| 201552 |
+
},
|
| 201553 |
+
{
|
| 201554 |
+
"epoch": 642.41,
|
| 201555 |
+
"learning_rate": 8.71907915993538e-06,
|
| 201556 |
+
"loss": 1.3292,
|
| 201557 |
+
"step": 80320
|
| 201558 |
+
},
|
| 201559 |
+
{
|
| 201560 |
+
"epoch": 642.45,
|
| 201561 |
+
"learning_rate": 8.718998384491115e-06,
|
| 201562 |
+
"loss": 0.3665,
|
| 201563 |
+
"step": 80325
|
| 201564 |
+
},
|
| 201565 |
+
{
|
| 201566 |
+
"epoch": 642.49,
|
| 201567 |
+
"learning_rate": 8.71891760904685e-06,
|
| 201568 |
+
"loss": 0.2565,
|
| 201569 |
+
"step": 80330
|
| 201570 |
+
},
|
| 201571 |
+
{
|
| 201572 |
+
"epoch": 642.53,
|
| 201573 |
+
"learning_rate": 8.718836833602585e-06,
|
| 201574 |
+
"loss": 0.4313,
|
| 201575 |
+
"step": 80335
|
| 201576 |
+
},
|
| 201577 |
+
{
|
| 201578 |
+
"epoch": 642.57,
|
| 201579 |
+
"learning_rate": 8.71875605815832e-06,
|
| 201580 |
+
"loss": 0.5453,
|
| 201581 |
+
"step": 80340
|
| 201582 |
+
},
|
| 201583 |
+
{
|
| 201584 |
+
"epoch": 642.61,
|
| 201585 |
+
"learning_rate": 8.718675282714055e-06,
|
| 201586 |
+
"loss": 0.9556,
|
| 201587 |
+
"step": 80345
|
| 201588 |
+
},
|
| 201589 |
+
{
|
| 201590 |
+
"epoch": 642.65,
|
| 201591 |
+
"learning_rate": 8.71859450726979e-06,
|
| 201592 |
+
"loss": 0.2915,
|
| 201593 |
+
"step": 80350
|
| 201594 |
+
},
|
| 201595 |
+
{
|
| 201596 |
+
"epoch": 642.69,
|
| 201597 |
+
"learning_rate": 8.718513731825525e-06,
|
| 201598 |
+
"loss": 0.3673,
|
| 201599 |
+
"step": 80355
|
| 201600 |
+
},
|
| 201601 |
+
{
|
| 201602 |
+
"epoch": 642.73,
|
| 201603 |
+
"learning_rate": 8.71843295638126e-06,
|
| 201604 |
+
"loss": 0.3407,
|
| 201605 |
+
"step": 80360
|
| 201606 |
+
},
|
| 201607 |
+
{
|
| 201608 |
+
"epoch": 642.77,
|
| 201609 |
+
"learning_rate": 8.718352180936995e-06,
|
| 201610 |
+
"loss": 0.4414,
|
| 201611 |
+
"step": 80365
|
| 201612 |
+
},
|
| 201613 |
+
{
|
| 201614 |
+
"epoch": 642.81,
|
| 201615 |
+
"learning_rate": 8.71827140549273e-06,
|
| 201616 |
+
"loss": 1.264,
|
| 201617 |
+
"step": 80370
|
| 201618 |
+
},
|
| 201619 |
+
{
|
| 201620 |
+
"epoch": 642.85,
|
| 201621 |
+
"learning_rate": 8.718190630048466e-06,
|
| 201622 |
+
"loss": 0.3002,
|
| 201623 |
+
"step": 80375
|
| 201624 |
+
},
|
| 201625 |
+
{
|
| 201626 |
+
"epoch": 642.89,
|
| 201627 |
+
"learning_rate": 8.7181098546042e-06,
|
| 201628 |
+
"loss": 0.3777,
|
| 201629 |
+
"step": 80380
|
| 201630 |
+
},
|
| 201631 |
+
{
|
| 201632 |
+
"epoch": 642.93,
|
| 201633 |
+
"learning_rate": 8.718029079159936e-06,
|
| 201634 |
+
"loss": 0.3856,
|
| 201635 |
+
"step": 80385
|
| 201636 |
+
},
|
| 201637 |
+
{
|
| 201638 |
+
"epoch": 642.97,
|
| 201639 |
+
"learning_rate": 8.71794830371567e-06,
|
| 201640 |
+
"loss": 0.5971,
|
| 201641 |
+
"step": 80390
|
| 201642 |
+
},
|
| 201643 |
+
{
|
| 201644 |
+
"epoch": 643.0,
|
| 201645 |
+
"eval_loss": 0.493080735206604,
|
| 201646 |
+
"eval_runtime": 41.096,
|
| 201647 |
+
"eval_samples_per_second": 20.464,
|
| 201648 |
+
"eval_steps_per_second": 0.657,
|
| 201649 |
+
"eval_wer": 0.19013722756292653,
|
| 201650 |
+
"step": 80394
|
| 201651 |
+
},
|
| 201652 |
+
{
|
| 201653 |
+
"epoch": 643.01,
|
| 201654 |
+
"learning_rate": 8.717867528271406e-06,
|
| 201655 |
+
"loss": 0.4632,
|
| 201656 |
+
"step": 80395
|
| 201657 |
+
},
|
| 201658 |
+
{
|
| 201659 |
+
"epoch": 643.05,
|
| 201660 |
+
"learning_rate": 8.71778675282714e-06,
|
| 201661 |
+
"loss": 0.3292,
|
| 201662 |
+
"step": 80400
|
| 201663 |
+
},
|
| 201664 |
+
{
|
| 201665 |
+
"epoch": 643.09,
|
| 201666 |
+
"learning_rate": 8.717705977382876e-06,
|
| 201667 |
+
"loss": 0.274,
|
| 201668 |
+
"step": 80405
|
| 201669 |
+
},
|
| 201670 |
+
{
|
| 201671 |
+
"epoch": 643.13,
|
| 201672 |
+
"learning_rate": 8.71762520193861e-06,
|
| 201673 |
+
"loss": 0.3407,
|
| 201674 |
+
"step": 80410
|
| 201675 |
+
},
|
| 201676 |
+
{
|
| 201677 |
+
"epoch": 643.17,
|
| 201678 |
+
"learning_rate": 8.717544426494346e-06,
|
| 201679 |
+
"loss": 0.6008,
|
| 201680 |
+
"step": 80415
|
| 201681 |
+
},
|
| 201682 |
+
{
|
| 201683 |
+
"epoch": 643.21,
|
| 201684 |
+
"learning_rate": 8.71746365105008e-06,
|
| 201685 |
+
"loss": 1.0789,
|
| 201686 |
+
"step": 80420
|
| 201687 |
+
},
|
| 201688 |
+
{
|
| 201689 |
+
"epoch": 643.25,
|
| 201690 |
+
"learning_rate": 8.717382875605816e-06,
|
| 201691 |
+
"loss": 0.3242,
|
| 201692 |
+
"step": 80425
|
| 201693 |
+
},
|
| 201694 |
+
{
|
| 201695 |
+
"epoch": 643.29,
|
| 201696 |
+
"learning_rate": 8.717302100161552e-06,
|
| 201697 |
+
"loss": 0.3146,
|
| 201698 |
+
"step": 80430
|
| 201699 |
+
},
|
| 201700 |
+
{
|
| 201701 |
+
"epoch": 643.33,
|
| 201702 |
+
"learning_rate": 8.717221324717286e-06,
|
| 201703 |
+
"loss": 0.3943,
|
| 201704 |
+
"step": 80435
|
| 201705 |
+
},
|
| 201706 |
+
{
|
| 201707 |
+
"epoch": 643.37,
|
| 201708 |
+
"learning_rate": 8.717140549273022e-06,
|
| 201709 |
+
"loss": 0.548,
|
| 201710 |
+
"step": 80440
|
| 201711 |
+
},
|
| 201712 |
+
{
|
| 201713 |
+
"epoch": 643.41,
|
| 201714 |
+
"learning_rate": 8.717059773828756e-06,
|
| 201715 |
+
"loss": 1.0559,
|
| 201716 |
+
"step": 80445
|
| 201717 |
+
},
|
| 201718 |
+
{
|
| 201719 |
+
"epoch": 643.45,
|
| 201720 |
+
"learning_rate": 8.716978998384492e-06,
|
| 201721 |
+
"loss": 0.2766,
|
| 201722 |
+
"step": 80450
|
| 201723 |
+
},
|
| 201724 |
+
{
|
| 201725 |
+
"epoch": 643.49,
|
| 201726 |
+
"learning_rate": 8.716898222940226e-06,
|
| 201727 |
+
"loss": 0.2528,
|
| 201728 |
+
"step": 80455
|
| 201729 |
+
},
|
| 201730 |
+
{
|
| 201731 |
+
"epoch": 643.53,
|
| 201732 |
+
"learning_rate": 8.716817447495962e-06,
|
| 201733 |
+
"loss": 0.3173,
|
| 201734 |
+
"step": 80460
|
| 201735 |
+
},
|
| 201736 |
+
{
|
| 201737 |
+
"epoch": 643.57,
|
| 201738 |
+
"learning_rate": 8.716736672051696e-06,
|
| 201739 |
+
"loss": 0.513,
|
| 201740 |
+
"step": 80465
|
| 201741 |
+
},
|
| 201742 |
+
{
|
| 201743 |
+
"epoch": 643.61,
|
| 201744 |
+
"learning_rate": 8.716655896607432e-06,
|
| 201745 |
+
"loss": 1.274,
|
| 201746 |
+
"step": 80470
|
| 201747 |
+
},
|
| 201748 |
+
{
|
| 201749 |
+
"epoch": 643.65,
|
| 201750 |
+
"learning_rate": 8.716575121163166e-06,
|
| 201751 |
+
"loss": 0.2854,
|
| 201752 |
+
"step": 80475
|
| 201753 |
+
},
|
| 201754 |
+
{
|
| 201755 |
+
"epoch": 643.69,
|
| 201756 |
+
"learning_rate": 8.716494345718902e-06,
|
| 201757 |
+
"loss": 0.3299,
|
| 201758 |
+
"step": 80480
|
| 201759 |
+
},
|
| 201760 |
+
{
|
| 201761 |
+
"epoch": 643.73,
|
| 201762 |
+
"learning_rate": 8.716413570274638e-06,
|
| 201763 |
+
"loss": 0.3727,
|
| 201764 |
+
"step": 80485
|
| 201765 |
+
},
|
| 201766 |
+
{
|
| 201767 |
+
"epoch": 643.77,
|
| 201768 |
+
"learning_rate": 8.716332794830372e-06,
|
| 201769 |
+
"loss": 0.5196,
|
| 201770 |
+
"step": 80490
|
| 201771 |
+
},
|
| 201772 |
+
{
|
| 201773 |
+
"epoch": 643.81,
|
| 201774 |
+
"learning_rate": 8.716252019386108e-06,
|
| 201775 |
+
"loss": 1.0925,
|
| 201776 |
+
"step": 80495
|
| 201777 |
+
},
|
| 201778 |
+
{
|
| 201779 |
+
"epoch": 643.85,
|
| 201780 |
+
"learning_rate": 8.716171243941842e-06,
|
| 201781 |
+
"loss": 0.3021,
|
| 201782 |
+
"step": 80500
|
| 201783 |
+
},
|
| 201784 |
+
{
|
| 201785 |
+
"epoch": 643.89,
|
| 201786 |
+
"learning_rate": 8.716090468497578e-06,
|
| 201787 |
+
"loss": 0.2862,
|
| 201788 |
+
"step": 80505
|
| 201789 |
+
},
|
| 201790 |
+
{
|
| 201791 |
+
"epoch": 643.93,
|
| 201792 |
+
"learning_rate": 8.716009693053312e-06,
|
| 201793 |
+
"loss": 0.3624,
|
| 201794 |
+
"step": 80510
|
| 201795 |
+
},
|
| 201796 |
+
{
|
| 201797 |
+
"epoch": 643.97,
|
| 201798 |
+
"learning_rate": 8.715928917609048e-06,
|
| 201799 |
+
"loss": 0.6027,
|
| 201800 |
+
"step": 80515
|
| 201801 |
+
},
|
| 201802 |
+
{
|
| 201803 |
+
"epoch": 644.0,
|
| 201804 |
+
"eval_loss": 0.38939520716667175,
|
| 201805 |
+
"eval_runtime": 42.8774,
|
| 201806 |
+
"eval_samples_per_second": 19.591,
|
| 201807 |
+
"eval_steps_per_second": 0.63,
|
| 201808 |
+
"eval_wer": 0.19194484760522496,
|
| 201809 |
+
"step": 80519
|
| 201810 |
+
},
|
| 201811 |
+
{
|
| 201812 |
+
"epoch": 649.01,
|
| 201813 |
+
"learning_rate": 8.715848142164782e-06,
|
| 201814 |
+
"loss": 0.4317,
|
| 201815 |
+
"step": 80520
|
| 201816 |
+
},
|
| 201817 |
+
{
|
| 201818 |
+
"epoch": 649.05,
|
| 201819 |
+
"learning_rate": 8.715767366720518e-06,
|
| 201820 |
+
"loss": 0.2902,
|
| 201821 |
+
"step": 80525
|
| 201822 |
+
},
|
| 201823 |
+
{
|
| 201824 |
+
"epoch": 649.09,
|
| 201825 |
+
"learning_rate": 8.715686591276252e-06,
|
| 201826 |
+
"loss": 0.346,
|
| 201827 |
+
"step": 80530
|
| 201828 |
+
},
|
| 201829 |
+
{
|
| 201830 |
+
"epoch": 649.13,
|
| 201831 |
+
"learning_rate": 8.715605815831988e-06,
|
| 201832 |
+
"loss": 0.358,
|
| 201833 |
+
"step": 80535
|
| 201834 |
+
},
|
| 201835 |
+
{
|
| 201836 |
+
"epoch": 649.17,
|
| 201837 |
+
"learning_rate": 8.715525040387722e-06,
|
| 201838 |
+
"loss": 0.6588,
|
| 201839 |
+
"step": 80540
|
| 201840 |
+
},
|
| 201841 |
+
{
|
| 201842 |
+
"epoch": 649.21,
|
| 201843 |
+
"learning_rate": 8.715444264943458e-06,
|
| 201844 |
+
"loss": 1.2163,
|
| 201845 |
+
"step": 80545
|
| 201846 |
+
},
|
| 201847 |
+
{
|
| 201848 |
+
"epoch": 649.25,
|
| 201849 |
+
"learning_rate": 8.715363489499193e-06,
|
| 201850 |
+
"loss": 0.3163,
|
| 201851 |
+
"step": 80550
|
| 201852 |
+
},
|
| 201853 |
+
{
|
| 201854 |
+
"epoch": 649.29,
|
| 201855 |
+
"learning_rate": 8.715282714054928e-06,
|
| 201856 |
+
"loss": 0.3852,
|
| 201857 |
+
"step": 80555
|
| 201858 |
+
},
|
| 201859 |
+
{
|
| 201860 |
+
"epoch": 649.33,
|
| 201861 |
+
"learning_rate": 8.715201938610663e-06,
|
| 201862 |
+
"loss": 0.3614,
|
| 201863 |
+
"step": 80560
|
| 201864 |
+
},
|
| 201865 |
+
{
|
| 201866 |
+
"epoch": 649.37,
|
| 201867 |
+
"learning_rate": 8.715121163166398e-06,
|
| 201868 |
+
"loss": 0.5485,
|
| 201869 |
+
"step": 80565
|
| 201870 |
+
},
|
| 201871 |
+
{
|
| 201872 |
+
"epoch": 649.41,
|
| 201873 |
+
"learning_rate": 8.715040387722133e-06,
|
| 201874 |
+
"loss": 1.2827,
|
| 201875 |
+
"step": 80570
|
| 201876 |
+
},
|
| 201877 |
+
{
|
| 201878 |
+
"epoch": 649.45,
|
| 201879 |
+
"learning_rate": 8.714959612277867e-06,
|
| 201880 |
+
"loss": 0.3936,
|
| 201881 |
+
"step": 80575
|
| 201882 |
+
},
|
| 201883 |
+
{
|
| 201884 |
+
"epoch": 649.49,
|
| 201885 |
+
"learning_rate": 8.714878836833603e-06,
|
| 201886 |
+
"loss": 0.3177,
|
| 201887 |
+
"step": 80580
|
| 201888 |
+
},
|
| 201889 |
+
{
|
| 201890 |
+
"epoch": 649.53,
|
| 201891 |
+
"learning_rate": 8.714798061389337e-06,
|
| 201892 |
+
"loss": 0.3352,
|
| 201893 |
+
"step": 80585
|
| 201894 |
+
},
|
| 201895 |
+
{
|
| 201896 |
+
"epoch": 649.57,
|
| 201897 |
+
"learning_rate": 8.714717285945073e-06,
|
| 201898 |
+
"loss": 0.5401,
|
| 201899 |
+
"step": 80590
|
| 201900 |
+
},
|
| 201901 |
+
{
|
| 201902 |
+
"epoch": 649.61,
|
| 201903 |
+
"learning_rate": 8.714636510500807e-06,
|
| 201904 |
+
"loss": 1.0671,
|
| 201905 |
+
"step": 80595
|
| 201906 |
+
},
|
| 201907 |
+
{
|
| 201908 |
+
"epoch": 649.65,
|
| 201909 |
+
"learning_rate": 8.714555735056543e-06,
|
| 201910 |
+
"loss": 0.3136,
|
| 201911 |
+
"step": 80600
|
| 201912 |
+
},
|
| 201913 |
+
{
|
| 201914 |
+
"epoch": 649.69,
|
| 201915 |
+
"learning_rate": 8.714474959612279e-06,
|
| 201916 |
+
"loss": 0.3726,
|
| 201917 |
+
"step": 80605
|
| 201918 |
+
},
|
| 201919 |
+
{
|
| 201920 |
+
"epoch": 649.73,
|
| 201921 |
+
"learning_rate": 8.714394184168013e-06,
|
| 201922 |
+
"loss": 0.3793,
|
| 201923 |
+
"step": 80610
|
| 201924 |
+
},
|
| 201925 |
+
{
|
| 201926 |
+
"epoch": 649.77,
|
| 201927 |
+
"learning_rate": 8.714313408723749e-06,
|
| 201928 |
+
"loss": 0.4862,
|
| 201929 |
+
"step": 80615
|
| 201930 |
+
},
|
| 201931 |
+
{
|
| 201932 |
+
"epoch": 649.81,
|
| 201933 |
+
"learning_rate": 8.714232633279483e-06,
|
| 201934 |
+
"loss": 1.2835,
|
| 201935 |
+
"step": 80620
|
| 201936 |
+
},
|
| 201937 |
+
{
|
| 201938 |
+
"epoch": 649.85,
|
| 201939 |
+
"learning_rate": 8.714151857835219e-06,
|
| 201940 |
+
"loss": 0.2867,
|
| 201941 |
+
"step": 80625
|
| 201942 |
+
},
|
| 201943 |
+
{
|
| 201944 |
+
"epoch": 649.9,
|
| 201945 |
+
"learning_rate": 8.714071082390953e-06,
|
| 201946 |
+
"loss": 0.2916,
|
| 201947 |
+
"step": 80630
|
| 201948 |
+
},
|
| 201949 |
+
{
|
| 201950 |
+
"epoch": 649.94,
|
| 201951 |
+
"learning_rate": 8.713990306946689e-06,
|
| 201952 |
+
"loss": 0.4137,
|
| 201953 |
+
"step": 80635
|
| 201954 |
+
},
|
| 201955 |
+
{
|
| 201956 |
+
"epoch": 649.98,
|
| 201957 |
+
"learning_rate": 8.713909531502423e-06,
|
| 201958 |
+
"loss": 0.6571,
|
| 201959 |
+
"step": 80640
|
| 201960 |
+
},
|
| 201961 |
+
{
|
| 201962 |
+
"epoch": 650.0,
|
| 201963 |
+
"eval_loss": 0.4564504027366638,
|
| 201964 |
+
"eval_runtime": 41.7225,
|
| 201965 |
+
"eval_samples_per_second": 20.157,
|
| 201966 |
+
"eval_steps_per_second": 0.647,
|
| 201967 |
+
"eval_wer": 0.19225489487753775,
|
| 201968 |
+
"step": 80643
|
| 201969 |
+
},
|
| 201970 |
+
{
|
| 201971 |
+
"epoch": 650.02,
|
| 201972 |
+
"learning_rate": 8.713828756058159e-06,
|
| 201973 |
+
"loss": 0.3961,
|
| 201974 |
+
"step": 80645
|
| 201975 |
+
},
|
| 201976 |
+
{
|
| 201977 |
+
"epoch": 650.06,
|
| 201978 |
+
"learning_rate": 8.713747980613893e-06,
|
| 201979 |
+
"loss": 0.2932,
|
| 201980 |
+
"step": 80650
|
| 201981 |
+
},
|
| 201982 |
+
{
|
| 201983 |
+
"epoch": 650.1,
|
| 201984 |
+
"learning_rate": 8.713667205169629e-06,
|
| 201985 |
+
"loss": 0.4166,
|
| 201986 |
+
"step": 80655
|
| 201987 |
+
},
|
| 201988 |
+
{
|
| 201989 |
+
"epoch": 650.14,
|
| 201990 |
+
"learning_rate": 8.713586429725365e-06,
|
| 201991 |
+
"loss": 0.3511,
|
| 201992 |
+
"step": 80660
|
| 201993 |
+
},
|
| 201994 |
+
{
|
| 201995 |
+
"epoch": 650.18,
|
| 201996 |
+
"learning_rate": 8.713505654281099e-06,
|
| 201997 |
+
"loss": 0.6076,
|
| 201998 |
+
"step": 80665
|
| 201999 |
+
},
|
| 202000 |
+
{
|
| 202001 |
+
"epoch": 650.22,
|
| 202002 |
+
"learning_rate": 8.713424878836835e-06,
|
| 202003 |
+
"loss": 1.0276,
|
| 202004 |
+
"step": 80670
|
| 202005 |
+
},
|
| 202006 |
+
{
|
| 202007 |
+
"epoch": 650.26,
|
| 202008 |
+
"learning_rate": 8.713344103392569e-06,
|
| 202009 |
+
"loss": 0.3696,
|
| 202010 |
+
"step": 80675
|
| 202011 |
+
},
|
| 202012 |
+
{
|
| 202013 |
+
"epoch": 650.3,
|
| 202014 |
+
"learning_rate": 8.713263327948305e-06,
|
| 202015 |
+
"loss": 0.336,
|
| 202016 |
+
"step": 80680
|
| 202017 |
+
},
|
| 202018 |
+
{
|
| 202019 |
+
"epoch": 650.34,
|
| 202020 |
+
"learning_rate": 8.713182552504039e-06,
|
| 202021 |
+
"loss": 0.3982,
|
| 202022 |
+
"step": 80685
|
| 202023 |
+
},
|
| 202024 |
+
{
|
| 202025 |
+
"epoch": 650.38,
|
| 202026 |
+
"learning_rate": 8.713101777059775e-06,
|
| 202027 |
+
"loss": 0.7392,
|
| 202028 |
+
"step": 80690
|
| 202029 |
+
},
|
| 202030 |
+
{
|
| 202031 |
+
"epoch": 650.42,
|
| 202032 |
+
"learning_rate": 8.713021001615509e-06,
|
| 202033 |
+
"loss": 1.1054,
|
| 202034 |
+
"step": 80695
|
| 202035 |
+
},
|
| 202036 |
+
{
|
| 202037 |
+
"epoch": 650.46,
|
| 202038 |
+
"learning_rate": 8.712940226171245e-06,
|
| 202039 |
+
"loss": 0.3,
|
| 202040 |
+
"step": 80700
|
| 202041 |
+
},
|
| 202042 |
+
{
|
| 202043 |
+
"epoch": 650.5,
|
| 202044 |
+
"learning_rate": 8.712859450726979e-06,
|
| 202045 |
+
"loss": 0.3392,
|
| 202046 |
+
"step": 80705
|
| 202047 |
+
},
|
| 202048 |
+
{
|
| 202049 |
+
"epoch": 650.54,
|
| 202050 |
+
"learning_rate": 8.712778675282715e-06,
|
| 202051 |
+
"loss": 0.3815,
|
| 202052 |
+
"step": 80710
|
| 202053 |
+
},
|
| 202054 |
+
{
|
| 202055 |
+
"epoch": 650.58,
|
| 202056 |
+
"learning_rate": 8.712697899838449e-06,
|
| 202057 |
+
"loss": 0.7745,
|
| 202058 |
+
"step": 80715
|
| 202059 |
+
},
|
| 202060 |
+
{
|
| 202061 |
+
"epoch": 650.62,
|
| 202062 |
+
"learning_rate": 8.712617124394185e-06,
|
| 202063 |
+
"loss": 1.0624,
|
| 202064 |
+
"step": 80720
|
| 202065 |
+
},
|
| 202066 |
+
{
|
| 202067 |
+
"epoch": 650.66,
|
| 202068 |
+
"learning_rate": 8.71253634894992e-06,
|
| 202069 |
+
"loss": 0.3261,
|
| 202070 |
+
"step": 80725
|
| 202071 |
+
},
|
| 202072 |
+
{
|
| 202073 |
+
"epoch": 650.7,
|
| 202074 |
+
"learning_rate": 8.712455573505655e-06,
|
| 202075 |
+
"loss": 0.3151,
|
| 202076 |
+
"step": 80730
|
| 202077 |
+
},
|
| 202078 |
+
{
|
| 202079 |
+
"epoch": 650.74,
|
| 202080 |
+
"learning_rate": 8.71237479806139e-06,
|
| 202081 |
+
"loss": 0.35,
|
| 202082 |
+
"step": 80735
|
| 202083 |
+
},
|
| 202084 |
+
{
|
| 202085 |
+
"epoch": 650.78,
|
| 202086 |
+
"learning_rate": 8.712294022617125e-06,
|
| 202087 |
+
"loss": 0.7034,
|
| 202088 |
+
"step": 80740
|
| 202089 |
+
},
|
| 202090 |
+
{
|
| 202091 |
+
"epoch": 650.82,
|
| 202092 |
+
"learning_rate": 8.71221324717286e-06,
|
| 202093 |
+
"loss": 1.1486,
|
| 202094 |
+
"step": 80745
|
| 202095 |
+
},
|
| 202096 |
+
{
|
| 202097 |
+
"epoch": 650.86,
|
| 202098 |
+
"learning_rate": 8.712132471728595e-06,
|
| 202099 |
+
"loss": 0.3333,
|
| 202100 |
+
"step": 80750
|
| 202101 |
+
},
|
| 202102 |
+
{
|
| 202103 |
+
"epoch": 650.9,
|
| 202104 |
+
"learning_rate": 8.71205169628433e-06,
|
| 202105 |
+
"loss": 0.3279,
|
| 202106 |
+
"step": 80755
|
| 202107 |
+
},
|
| 202108 |
+
{
|
| 202109 |
+
"epoch": 650.94,
|
| 202110 |
+
"learning_rate": 8.711970920840065e-06,
|
| 202111 |
+
"loss": 0.4051,
|
| 202112 |
+
"step": 80760
|
| 202113 |
+
},
|
| 202114 |
+
{
|
| 202115 |
+
"epoch": 650.98,
|
| 202116 |
+
"learning_rate": 8.7118901453958e-06,
|
| 202117 |
+
"loss": 0.6548,
|
| 202118 |
+
"step": 80765
|
| 202119 |
+
},
|
| 202120 |
+
{
|
| 202121 |
+
"epoch": 651.0,
|
| 202122 |
+
"eval_loss": 0.37322500348091125,
|
| 202123 |
+
"eval_runtime": 42.5291,
|
| 202124 |
+
"eval_samples_per_second": 19.775,
|
| 202125 |
+
"eval_steps_per_second": 0.635,
|
| 202126 |
+
"eval_wer": 0.18461987877017455,
|
| 202127 |
+
"step": 80767
|
| 202128 |
+
},
|
| 202129 |
+
{
|
| 202130 |
+
"epoch": 651.02,
|
| 202131 |
+
"learning_rate": 8.711809369951535e-06,
|
| 202132 |
+
"loss": 0.3237,
|
| 202133 |
+
"step": 80770
|
| 202134 |
+
},
|
| 202135 |
+
{
|
| 202136 |
+
"epoch": 651.06,
|
| 202137 |
+
"learning_rate": 8.71172859450727e-06,
|
| 202138 |
+
"loss": 0.3577,
|
| 202139 |
+
"step": 80775
|
| 202140 |
+
},
|
| 202141 |
+
{
|
| 202142 |
+
"epoch": 651.1,
|
| 202143 |
+
"learning_rate": 8.711647819063006e-06,
|
| 202144 |
+
"loss": 0.303,
|
| 202145 |
+
"step": 80780
|
| 202146 |
+
},
|
| 202147 |
+
{
|
| 202148 |
+
"epoch": 651.14,
|
| 202149 |
+
"learning_rate": 8.71156704361874e-06,
|
| 202150 |
+
"loss": 0.3532,
|
| 202151 |
+
"step": 80785
|
| 202152 |
+
},
|
| 202153 |
+
{
|
| 202154 |
+
"epoch": 651.18,
|
| 202155 |
+
"learning_rate": 8.711486268174476e-06,
|
| 202156 |
+
"loss": 0.7385,
|
| 202157 |
+
"step": 80790
|
| 202158 |
+
},
|
| 202159 |
+
{
|
| 202160 |
+
"epoch": 651.22,
|
| 202161 |
+
"learning_rate": 8.71140549273021e-06,
|
| 202162 |
+
"loss": 0.8566,
|
| 202163 |
+
"step": 80795
|
| 202164 |
+
},
|
| 202165 |
+
{
|
| 202166 |
+
"epoch": 651.27,
|
| 202167 |
+
"learning_rate": 8.711324717285946e-06,
|
| 202168 |
+
"loss": 0.3091,
|
| 202169 |
+
"step": 80800
|
| 202170 |
+
},
|
| 202171 |
+
{
|
| 202172 |
+
"epoch": 651.31,
|
| 202173 |
+
"learning_rate": 8.71124394184168e-06,
|
| 202174 |
+
"loss": 0.3321,
|
| 202175 |
+
"step": 80805
|
| 202176 |
+
},
|
| 202177 |
+
{
|
| 202178 |
+
"epoch": 651.35,
|
| 202179 |
+
"learning_rate": 8.711163166397416e-06,
|
| 202180 |
+
"loss": 0.3993,
|
| 202181 |
+
"step": 80810
|
| 202182 |
+
},
|
| 202183 |
+
{
|
| 202184 |
+
"epoch": 651.39,
|
| 202185 |
+
"learning_rate": 8.71108239095315e-06,
|
| 202186 |
+
"loss": 0.8393,
|
| 202187 |
+
"step": 80815
|
| 202188 |
+
},
|
| 202189 |
+
{
|
| 202190 |
+
"epoch": 651.43,
|
| 202191 |
+
"learning_rate": 8.711001615508886e-06,
|
| 202192 |
+
"loss": 0.9883,
|
| 202193 |
+
"step": 80820
|
| 202194 |
+
},
|
| 202195 |
+
{
|
| 202196 |
+
"epoch": 651.47,
|
| 202197 |
+
"learning_rate": 8.71092084006462e-06,
|
| 202198 |
+
"loss": 0.3495,
|
| 202199 |
+
"step": 80825
|
| 202200 |
+
},
|
| 202201 |
+
{
|
| 202202 |
+
"epoch": 651.51,
|
| 202203 |
+
"learning_rate": 8.710840064620356e-06,
|
| 202204 |
+
"loss": 0.3111,
|
| 202205 |
+
"step": 80830
|
| 202206 |
+
},
|
| 202207 |
+
{
|
| 202208 |
+
"epoch": 651.55,
|
| 202209 |
+
"learning_rate": 8.710759289176092e-06,
|
| 202210 |
+
"loss": 0.3887,
|
| 202211 |
+
"step": 80835
|
| 202212 |
+
},
|
| 202213 |
+
{
|
| 202214 |
+
"epoch": 651.59,
|
| 202215 |
+
"learning_rate": 8.710678513731826e-06,
|
| 202216 |
+
"loss": 0.6717,
|
| 202217 |
+
"step": 80840
|
| 202218 |
+
},
|
| 202219 |
+
{
|
| 202220 |
+
"epoch": 651.63,
|
| 202221 |
+
"learning_rate": 8.710597738287562e-06,
|
| 202222 |
+
"loss": 0.9654,
|
| 202223 |
+
"step": 80845
|
| 202224 |
+
},
|
| 202225 |
+
{
|
| 202226 |
+
"epoch": 651.67,
|
| 202227 |
+
"learning_rate": 8.710516962843296e-06,
|
| 202228 |
+
"loss": 0.3719,
|
| 202229 |
+
"step": 80850
|
| 202230 |
+
},
|
| 202231 |
+
{
|
| 202232 |
+
"epoch": 651.71,
|
| 202233 |
+
"learning_rate": 8.710436187399032e-06,
|
| 202234 |
+
"loss": 0.2985,
|
| 202235 |
+
"step": 80855
|
| 202236 |
+
},
|
| 202237 |
+
{
|
| 202238 |
+
"epoch": 651.75,
|
| 202239 |
+
"learning_rate": 8.710355411954766e-06,
|
| 202240 |
+
"loss": 0.3987,
|
| 202241 |
+
"step": 80860
|
| 202242 |
+
},
|
| 202243 |
+
{
|
| 202244 |
+
"epoch": 651.79,
|
| 202245 |
+
"learning_rate": 8.710274636510502e-06,
|
| 202246 |
+
"loss": 0.8482,
|
| 202247 |
+
"step": 80865
|
| 202248 |
+
},
|
| 202249 |
+
{
|
| 202250 |
+
"epoch": 651.83,
|
| 202251 |
+
"learning_rate": 8.710193861066236e-06,
|
| 202252 |
+
"loss": 0.8805,
|
| 202253 |
+
"step": 80870
|
| 202254 |
+
},
|
| 202255 |
+
{
|
| 202256 |
+
"epoch": 651.87,
|
| 202257 |
+
"learning_rate": 8.710113085621972e-06,
|
| 202258 |
+
"loss": 0.2904,
|
| 202259 |
+
"step": 80875
|
| 202260 |
+
},
|
| 202261 |
+
{
|
| 202262 |
+
"epoch": 651.91,
|
| 202263 |
+
"learning_rate": 8.710032310177706e-06,
|
| 202264 |
+
"loss": 0.3254,
|
| 202265 |
+
"step": 80880
|
| 202266 |
+
},
|
| 202267 |
+
{
|
| 202268 |
+
"epoch": 651.95,
|
| 202269 |
+
"learning_rate": 8.709951534733442e-06,
|
| 202270 |
+
"loss": 0.4833,
|
| 202271 |
+
"step": 80885
|
| 202272 |
+
},
|
| 202273 |
+
{
|
| 202274 |
+
"epoch": 651.99,
|
| 202275 |
+
"learning_rate": 8.709870759289178e-06,
|
| 202276 |
+
"loss": 0.8421,
|
| 202277 |
+
"step": 80890
|
| 202278 |
+
},
|
| 202279 |
+
{
|
| 202280 |
+
"epoch": 652.0,
|
| 202281 |
+
"eval_loss": 0.4140271544456482,
|
| 202282 |
+
"eval_runtime": 42.7669,
|
| 202283 |
+
"eval_samples_per_second": 19.665,
|
| 202284 |
+
"eval_steps_per_second": 0.631,
|
| 202285 |
+
"eval_wer": 0.19266525144572139,
|
| 202286 |
+
"step": 80891
|
| 202287 |
}
|
| 202288 |
],
|
| 202289 |
+
"max_steps": 620000,
|
| 202290 |
"num_train_epochs": 5000,
|
| 202291 |
+
"total_flos": 2.276386500512849e+20,
|
| 202292 |
"trial_name": null,
|
| 202293 |
"trial_params": null
|
| 202294 |
}
|
model-bin/finetune/base/{checkpoint-80269 → checkpoint-80891}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629904128.5656571/events.out.tfevents.1629904128.7e498afd5545.7645.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:807060dd41c3169495b1e22f76cdb8737f06b98c46114a0759447886e27f200a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629904601.0615444/events.out.tfevents.1629904601.7e498afd5545.7645.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bffb35c16cae0781fdd54b961b0ca24b32955827b3d412703ba15425873aac3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629905064.9216402/events.out.tfevents.1629905064.7e498afd5545.7645.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c318e0ec29741b86863d7d0587626020ffe179765e364ef63f2b913bdc9686b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629905527.7759078/events.out.tfevents.1629905527.7e498afd5545.7645.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16f11a624cab250697a2b34acc4089cb93135b82a6be8b62b6afd2ecafc3ef72
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629905996.4959083/events.out.tfevents.1629905996.7e498afd5545.7645.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72686a7940c11667371b7801acaac5d86f3ce96ab29da3803f18b703853c2200
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629904128.7e498afd5545.7645.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5889f431c237f270bd245751e9d85810688e97f44d112e3715b553bdcbee2d2a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629904601.7e498afd5545.7645.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:175b6cd71599f79a269596e913f046cdbfa74bccad3cd822ae0f774cb5e7a779
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629905064.7e498afd5545.7645.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a0354dcb77e0dd9b9dc16a1a824a4ec5b4ebfcf7e697b889efae36b88cf8ffb
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629905527.7e498afd5545.7645.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:274058007081c38d153a280cb49c8c59516d7e14d2e8bb9a3a33da7129bc5425
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629905996.7e498afd5545.7645.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:185b3e42b764871d1a4fde3c95e21fe832306f6d5af8fcaa06896db11a31e1b8
|
| 3 |
+
size 8622
|