"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629916283.1054225/events.out.tfevents.1629916283.7e498afd5545.7645.115 +3 -0
- model-bin/finetune/base/log/1629916758.8115587/events.out.tfevents.1629916758.7e498afd5545.7645.117 +3 -0
- model-bin/finetune/base/log/1629917235.212877/events.out.tfevents.1629917235.7e498afd5545.7645.119 +3 -0
- model-bin/finetune/base/log/1629917711.204426/events.out.tfevents.1629917711.7e498afd5545.7645.121 +3 -0
- model-bin/finetune/base/log/1629918181.9293232/events.out.tfevents.1629918181.7e498afd5545.7645.123 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629916283.7e498afd5545.7645.114 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629916758.7e498afd5545.7645.116 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629917234.7e498afd5545.7645.118 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629917711.7e498afd5545.7645.120 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629918181.7e498afd5545.7645.122 +3 -0
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/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 sha256:84d3d4dcc8ff892ba962ed1dad272a1bd4c1a58ba3dd060bdf9a8cc9bad703ce
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/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 sha256:42dbfbaee58c4d5da382b4e7f0fcb067d75a3277d73d7f99a3cf4f8d1c4ed695
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/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 sha256:704c31ddfe3810c1417803352304434a53a78b40b0bf6380385ac47452b7641e
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/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 sha256:424b1cd729b3c4e1fd0997337376e31f6c134ee62c0160776a7e0bd89bd9c008
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/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 sha256:93e9242ed52a0bcbcdb87cd3fb121d9c60638c18bdd4ddd660112a5416e3e0b1
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/trainer_state.json
RENAMED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
"epoch": 672.0,
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -205452,11 +205452,800 @@
|
|
| 205452 |
"eval_steps_per_second": 0.633,
|
| 205453 |
"eval_wer": 0.18727738994158713,
|
| 205454 |
"step": 83380
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 205455 |
}
|
| 205456 |
],
|
| 205457 |
-
"max_steps":
|
| 205458 |
"num_train_epochs": 5000,
|
| 205459 |
-
"total_flos": 2.
|
| 205460 |
"trial_name": null,
|
| 205461 |
"trial_params": null
|
| 205462 |
}
|
|
|
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
"epoch": 672.0,
|
| 5 |
+
"global_step": 84001,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 205452 |
"eval_steps_per_second": 0.633,
|
| 205453 |
"eval_wer": 0.18727738994158713,
|
| 205454 |
"step": 83380
|
| 205455 |
+
},
|
| 205456 |
+
{
|
| 205457 |
+
"epoch": 672.04,
|
| 205458 |
+
"learning_rate": 8.669596122778676e-06,
|
| 205459 |
+
"loss": 0.321,
|
| 205460 |
+
"step": 83385
|
| 205461 |
+
},
|
| 205462 |
+
{
|
| 205463 |
+
"epoch": 672.08,
|
| 205464 |
+
"learning_rate": 8.66951534733441e-06,
|
| 205465 |
+
"loss": 0.2795,
|
| 205466 |
+
"step": 83390
|
| 205467 |
+
},
|
| 205468 |
+
{
|
| 205469 |
+
"epoch": 672.12,
|
| 205470 |
+
"learning_rate": 8.669434571890146e-06,
|
| 205471 |
+
"loss": 0.3193,
|
| 205472 |
+
"step": 83395
|
| 205473 |
+
},
|
| 205474 |
+
{
|
| 205475 |
+
"epoch": 672.16,
|
| 205476 |
+
"learning_rate": 8.66935379644588e-06,
|
| 205477 |
+
"loss": 0.5312,
|
| 205478 |
+
"step": 83400
|
| 205479 |
+
},
|
| 205480 |
+
{
|
| 205481 |
+
"epoch": 672.2,
|
| 205482 |
+
"learning_rate": 8.669273021001616e-06,
|
| 205483 |
+
"loss": 1.3542,
|
| 205484 |
+
"step": 83405
|
| 205485 |
+
},
|
| 205486 |
+
{
|
| 205487 |
+
"epoch": 672.24,
|
| 205488 |
+
"learning_rate": 8.669192245557352e-06,
|
| 205489 |
+
"loss": 0.3859,
|
| 205490 |
+
"step": 83410
|
| 205491 |
+
},
|
| 205492 |
+
{
|
| 205493 |
+
"epoch": 672.28,
|
| 205494 |
+
"learning_rate": 8.669111470113086e-06,
|
| 205495 |
+
"loss": 0.2965,
|
| 205496 |
+
"step": 83415
|
| 205497 |
+
},
|
| 205498 |
+
{
|
| 205499 |
+
"epoch": 672.32,
|
| 205500 |
+
"learning_rate": 8.669030694668822e-06,
|
| 205501 |
+
"loss": 0.3794,
|
| 205502 |
+
"step": 83420
|
| 205503 |
+
},
|
| 205504 |
+
{
|
| 205505 |
+
"epoch": 672.36,
|
| 205506 |
+
"learning_rate": 8.668949919224556e-06,
|
| 205507 |
+
"loss": 0.4591,
|
| 205508 |
+
"step": 83425
|
| 205509 |
+
},
|
| 205510 |
+
{
|
| 205511 |
+
"epoch": 672.4,
|
| 205512 |
+
"learning_rate": 8.668869143780292e-06,
|
| 205513 |
+
"loss": 1.1486,
|
| 205514 |
+
"step": 83430
|
| 205515 |
+
},
|
| 205516 |
+
{
|
| 205517 |
+
"epoch": 672.44,
|
| 205518 |
+
"learning_rate": 8.668788368336026e-06,
|
| 205519 |
+
"loss": 0.3122,
|
| 205520 |
+
"step": 83435
|
| 205521 |
+
},
|
| 205522 |
+
{
|
| 205523 |
+
"epoch": 672.48,
|
| 205524 |
+
"learning_rate": 8.668707592891761e-06,
|
| 205525 |
+
"loss": 0.2827,
|
| 205526 |
+
"step": 83440
|
| 205527 |
+
},
|
| 205528 |
+
{
|
| 205529 |
+
"epoch": 672.52,
|
| 205530 |
+
"learning_rate": 8.668626817447496e-06,
|
| 205531 |
+
"loss": 0.3438,
|
| 205532 |
+
"step": 83445
|
| 205533 |
+
},
|
| 205534 |
+
{
|
| 205535 |
+
"epoch": 672.56,
|
| 205536 |
+
"learning_rate": 8.668546042003231e-06,
|
| 205537 |
+
"loss": 0.4683,
|
| 205538 |
+
"step": 83450
|
| 205539 |
+
},
|
| 205540 |
+
{
|
| 205541 |
+
"epoch": 672.6,
|
| 205542 |
+
"learning_rate": 8.668465266558966e-06,
|
| 205543 |
+
"loss": 1.1439,
|
| 205544 |
+
"step": 83455
|
| 205545 |
+
},
|
| 205546 |
+
{
|
| 205547 |
+
"epoch": 672.65,
|
| 205548 |
+
"learning_rate": 8.668384491114701e-06,
|
| 205549 |
+
"loss": 0.3886,
|
| 205550 |
+
"step": 83460
|
| 205551 |
+
},
|
| 205552 |
+
{
|
| 205553 |
+
"epoch": 672.69,
|
| 205554 |
+
"learning_rate": 8.668303715670436e-06,
|
| 205555 |
+
"loss": 0.3214,
|
| 205556 |
+
"step": 83465
|
| 205557 |
+
},
|
| 205558 |
+
{
|
| 205559 |
+
"epoch": 672.73,
|
| 205560 |
+
"learning_rate": 8.668222940226171e-06,
|
| 205561 |
+
"loss": 0.3617,
|
| 205562 |
+
"step": 83470
|
| 205563 |
+
},
|
| 205564 |
+
{
|
| 205565 |
+
"epoch": 672.77,
|
| 205566 |
+
"learning_rate": 8.668142164781907e-06,
|
| 205567 |
+
"loss": 0.4773,
|
| 205568 |
+
"step": 83475
|
| 205569 |
+
},
|
| 205570 |
+
{
|
| 205571 |
+
"epoch": 672.81,
|
| 205572 |
+
"learning_rate": 8.668061389337641e-06,
|
| 205573 |
+
"loss": 1.2413,
|
| 205574 |
+
"step": 83480
|
| 205575 |
+
},
|
| 205576 |
+
{
|
| 205577 |
+
"epoch": 672.85,
|
| 205578 |
+
"learning_rate": 8.667980613893377e-06,
|
| 205579 |
+
"loss": 0.3372,
|
| 205580 |
+
"step": 83485
|
| 205581 |
+
},
|
| 205582 |
+
{
|
| 205583 |
+
"epoch": 672.89,
|
| 205584 |
+
"learning_rate": 8.667899838449111e-06,
|
| 205585 |
+
"loss": 0.3303,
|
| 205586 |
+
"step": 83490
|
| 205587 |
+
},
|
| 205588 |
+
{
|
| 205589 |
+
"epoch": 672.93,
|
| 205590 |
+
"learning_rate": 8.667819063004847e-06,
|
| 205591 |
+
"loss": 0.3082,
|
| 205592 |
+
"step": 83495
|
| 205593 |
+
},
|
| 205594 |
+
{
|
| 205595 |
+
"epoch": 672.97,
|
| 205596 |
+
"learning_rate": 8.667738287560581e-06,
|
| 205597 |
+
"loss": 0.5084,
|
| 205598 |
+
"step": 83500
|
| 205599 |
+
},
|
| 205600 |
+
{
|
| 205601 |
+
"epoch": 673.0,
|
| 205602 |
+
"eval_loss": 0.3978157043457031,
|
| 205603 |
+
"eval_runtime": 42.1738,
|
| 205604 |
+
"eval_samples_per_second": 19.918,
|
| 205605 |
+
"eval_steps_per_second": 0.64,
|
| 205606 |
+
"eval_wer": 0.19420624054190386,
|
| 205607 |
+
"step": 83504
|
| 205608 |
+
},
|
| 205609 |
+
{
|
| 205610 |
+
"epoch": 673.01,
|
| 205611 |
+
"learning_rate": 8.667657512116317e-06,
|
| 205612 |
+
"loss": 0.34,
|
| 205613 |
+
"step": 83505
|
| 205614 |
+
},
|
| 205615 |
+
{
|
| 205616 |
+
"epoch": 673.05,
|
| 205617 |
+
"learning_rate": 8.667576736672051e-06,
|
| 205618 |
+
"loss": 0.3097,
|
| 205619 |
+
"step": 83510
|
| 205620 |
+
},
|
| 205621 |
+
{
|
| 205622 |
+
"epoch": 673.09,
|
| 205623 |
+
"learning_rate": 8.667495961227787e-06,
|
| 205624 |
+
"loss": 0.3382,
|
| 205625 |
+
"step": 83515
|
| 205626 |
+
},
|
| 205627 |
+
{
|
| 205628 |
+
"epoch": 673.13,
|
| 205629 |
+
"learning_rate": 8.667415185783521e-06,
|
| 205630 |
+
"loss": 0.3407,
|
| 205631 |
+
"step": 83520
|
| 205632 |
+
},
|
| 205633 |
+
{
|
| 205634 |
+
"epoch": 673.17,
|
| 205635 |
+
"learning_rate": 8.667334410339257e-06,
|
| 205636 |
+
"loss": 0.7185,
|
| 205637 |
+
"step": 83525
|
| 205638 |
+
},
|
| 205639 |
+
{
|
| 205640 |
+
"epoch": 673.21,
|
| 205641 |
+
"learning_rate": 8.667253634894993e-06,
|
| 205642 |
+
"loss": 1.2853,
|
| 205643 |
+
"step": 83530
|
| 205644 |
+
},
|
| 205645 |
+
{
|
| 205646 |
+
"epoch": 673.25,
|
| 205647 |
+
"learning_rate": 8.667172859450727e-06,
|
| 205648 |
+
"loss": 0.4782,
|
| 205649 |
+
"step": 83535
|
| 205650 |
+
},
|
| 205651 |
+
{
|
| 205652 |
+
"epoch": 673.29,
|
| 205653 |
+
"learning_rate": 8.667092084006463e-06,
|
| 205654 |
+
"loss": 0.2928,
|
| 205655 |
+
"step": 83540
|
| 205656 |
+
},
|
| 205657 |
+
{
|
| 205658 |
+
"epoch": 673.33,
|
| 205659 |
+
"learning_rate": 8.667011308562197e-06,
|
| 205660 |
+
"loss": 0.3669,
|
| 205661 |
+
"step": 83545
|
| 205662 |
+
},
|
| 205663 |
+
{
|
| 205664 |
+
"epoch": 673.37,
|
| 205665 |
+
"learning_rate": 8.666930533117933e-06,
|
| 205666 |
+
"loss": 0.5494,
|
| 205667 |
+
"step": 83550
|
| 205668 |
+
},
|
| 205669 |
+
{
|
| 205670 |
+
"epoch": 673.41,
|
| 205671 |
+
"learning_rate": 8.666849757673667e-06,
|
| 205672 |
+
"loss": 1.1097,
|
| 205673 |
+
"step": 83555
|
| 205674 |
+
},
|
| 205675 |
+
{
|
| 205676 |
+
"epoch": 673.45,
|
| 205677 |
+
"learning_rate": 8.666768982229403e-06,
|
| 205678 |
+
"loss": 0.3213,
|
| 205679 |
+
"step": 83560
|
| 205680 |
+
},
|
| 205681 |
+
{
|
| 205682 |
+
"epoch": 673.49,
|
| 205683 |
+
"learning_rate": 8.666688206785137e-06,
|
| 205684 |
+
"loss": 0.3028,
|
| 205685 |
+
"step": 83565
|
| 205686 |
+
},
|
| 205687 |
+
{
|
| 205688 |
+
"epoch": 673.53,
|
| 205689 |
+
"learning_rate": 8.666607431340873e-06,
|
| 205690 |
+
"loss": 0.4287,
|
| 205691 |
+
"step": 83570
|
| 205692 |
+
},
|
| 205693 |
+
{
|
| 205694 |
+
"epoch": 673.57,
|
| 205695 |
+
"learning_rate": 8.666526655896607e-06,
|
| 205696 |
+
"loss": 0.555,
|
| 205697 |
+
"step": 83575
|
| 205698 |
+
},
|
| 205699 |
+
{
|
| 205700 |
+
"epoch": 673.61,
|
| 205701 |
+
"learning_rate": 8.666445880452343e-06,
|
| 205702 |
+
"loss": 1.2275,
|
| 205703 |
+
"step": 83580
|
| 205704 |
+
},
|
| 205705 |
+
{
|
| 205706 |
+
"epoch": 673.65,
|
| 205707 |
+
"learning_rate": 8.666365105008077e-06,
|
| 205708 |
+
"loss": 0.3096,
|
| 205709 |
+
"step": 83585
|
| 205710 |
+
},
|
| 205711 |
+
{
|
| 205712 |
+
"epoch": 673.69,
|
| 205713 |
+
"learning_rate": 8.666284329563813e-06,
|
| 205714 |
+
"loss": 0.2968,
|
| 205715 |
+
"step": 83590
|
| 205716 |
+
},
|
| 205717 |
+
{
|
| 205718 |
+
"epoch": 673.73,
|
| 205719 |
+
"learning_rate": 8.666203554119549e-06,
|
| 205720 |
+
"loss": 0.3168,
|
| 205721 |
+
"step": 83595
|
| 205722 |
+
},
|
| 205723 |
+
{
|
| 205724 |
+
"epoch": 673.77,
|
| 205725 |
+
"learning_rate": 8.666122778675283e-06,
|
| 205726 |
+
"loss": 0.5648,
|
| 205727 |
+
"step": 83600
|
| 205728 |
+
},
|
| 205729 |
+
{
|
| 205730 |
+
"epoch": 673.81,
|
| 205731 |
+
"learning_rate": 8.666042003231019e-06,
|
| 205732 |
+
"loss": 1.026,
|
| 205733 |
+
"step": 83605
|
| 205734 |
+
},
|
| 205735 |
+
{
|
| 205736 |
+
"epoch": 673.85,
|
| 205737 |
+
"learning_rate": 8.665961227786753e-06,
|
| 205738 |
+
"loss": 0.3263,
|
| 205739 |
+
"step": 83610
|
| 205740 |
+
},
|
| 205741 |
+
{
|
| 205742 |
+
"epoch": 673.89,
|
| 205743 |
+
"learning_rate": 8.665880452342489e-06,
|
| 205744 |
+
"loss": 0.3361,
|
| 205745 |
+
"step": 83615
|
| 205746 |
+
},
|
| 205747 |
+
{
|
| 205748 |
+
"epoch": 673.93,
|
| 205749 |
+
"learning_rate": 8.665799676898223e-06,
|
| 205750 |
+
"loss": 0.3118,
|
| 205751 |
+
"step": 83620
|
| 205752 |
+
},
|
| 205753 |
+
{
|
| 205754 |
+
"epoch": 673.97,
|
| 205755 |
+
"learning_rate": 8.665718901453959e-06,
|
| 205756 |
+
"loss": 0.5449,
|
| 205757 |
+
"step": 83625
|
| 205758 |
+
},
|
| 205759 |
+
{
|
| 205760 |
+
"epoch": 674.0,
|
| 205761 |
+
"eval_loss": 0.4208163619041443,
|
| 205762 |
+
"eval_runtime": 43.5422,
|
| 205763 |
+
"eval_samples_per_second": 19.315,
|
| 205764 |
+
"eval_steps_per_second": 0.62,
|
| 205765 |
+
"eval_wer": 0.19418748634277805,
|
| 205766 |
+
"step": 83628
|
| 205767 |
+
},
|
| 205768 |
+
{
|
| 205769 |
+
"epoch": 674.02,
|
| 205770 |
+
"learning_rate": 8.665638126009693e-06,
|
| 205771 |
+
"loss": 0.354,
|
| 205772 |
+
"step": 83630
|
| 205773 |
+
},
|
| 205774 |
+
{
|
| 205775 |
+
"epoch": 674.06,
|
| 205776 |
+
"learning_rate": 8.665557350565429e-06,
|
| 205777 |
+
"loss": 0.2855,
|
| 205778 |
+
"step": 83635
|
| 205779 |
+
},
|
| 205780 |
+
{
|
| 205781 |
+
"epoch": 674.1,
|
| 205782 |
+
"learning_rate": 8.665476575121163e-06,
|
| 205783 |
+
"loss": 0.3724,
|
| 205784 |
+
"step": 83640
|
| 205785 |
+
},
|
| 205786 |
+
{
|
| 205787 |
+
"epoch": 674.14,
|
| 205788 |
+
"learning_rate": 8.665395799676899e-06,
|
| 205789 |
+
"loss": 0.3313,
|
| 205790 |
+
"step": 83645
|
| 205791 |
+
},
|
| 205792 |
+
{
|
| 205793 |
+
"epoch": 674.18,
|
| 205794 |
+
"learning_rate": 8.665315024232634e-06,
|
| 205795 |
+
"loss": 0.6355,
|
| 205796 |
+
"step": 83650
|
| 205797 |
+
},
|
| 205798 |
+
{
|
| 205799 |
+
"epoch": 674.22,
|
| 205800 |
+
"learning_rate": 8.665234248788368e-06,
|
| 205801 |
+
"loss": 1.0881,
|
| 205802 |
+
"step": 83655
|
| 205803 |
+
},
|
| 205804 |
+
{
|
| 205805 |
+
"epoch": 674.26,
|
| 205806 |
+
"learning_rate": 8.665153473344104e-06,
|
| 205807 |
+
"loss": 0.3134,
|
| 205808 |
+
"step": 83660
|
| 205809 |
+
},
|
| 205810 |
+
{
|
| 205811 |
+
"epoch": 674.3,
|
| 205812 |
+
"learning_rate": 8.665072697899838e-06,
|
| 205813 |
+
"loss": 0.3421,
|
| 205814 |
+
"step": 83665
|
| 205815 |
+
},
|
| 205816 |
+
{
|
| 205817 |
+
"epoch": 674.34,
|
| 205818 |
+
"learning_rate": 8.664991922455574e-06,
|
| 205819 |
+
"loss": 0.3621,
|
| 205820 |
+
"step": 83670
|
| 205821 |
+
},
|
| 205822 |
+
{
|
| 205823 |
+
"epoch": 674.38,
|
| 205824 |
+
"learning_rate": 8.664911147011308e-06,
|
| 205825 |
+
"loss": 1.0302,
|
| 205826 |
+
"step": 83675
|
| 205827 |
+
},
|
| 205828 |
+
{
|
| 205829 |
+
"epoch": 674.42,
|
| 205830 |
+
"learning_rate": 8.664830371567044e-06,
|
| 205831 |
+
"loss": 1.046,
|
| 205832 |
+
"step": 83680
|
| 205833 |
+
},
|
| 205834 |
+
{
|
| 205835 |
+
"epoch": 674.46,
|
| 205836 |
+
"learning_rate": 8.664749596122778e-06,
|
| 205837 |
+
"loss": 0.2705,
|
| 205838 |
+
"step": 83685
|
| 205839 |
+
},
|
| 205840 |
+
{
|
| 205841 |
+
"epoch": 674.5,
|
| 205842 |
+
"learning_rate": 8.664668820678514e-06,
|
| 205843 |
+
"loss": 0.2697,
|
| 205844 |
+
"step": 83690
|
| 205845 |
+
},
|
| 205846 |
+
{
|
| 205847 |
+
"epoch": 674.54,
|
| 205848 |
+
"learning_rate": 8.664588045234248e-06,
|
| 205849 |
+
"loss": 0.3652,
|
| 205850 |
+
"step": 83695
|
| 205851 |
+
},
|
| 205852 |
+
{
|
| 205853 |
+
"epoch": 674.58,
|
| 205854 |
+
"learning_rate": 8.664507269789984e-06,
|
| 205855 |
+
"loss": 0.7068,
|
| 205856 |
+
"step": 83700
|
| 205857 |
+
},
|
| 205858 |
+
{
|
| 205859 |
+
"epoch": 674.62,
|
| 205860 |
+
"learning_rate": 8.66442649434572e-06,
|
| 205861 |
+
"loss": 1.0163,
|
| 205862 |
+
"step": 83705
|
| 205863 |
+
},
|
| 205864 |
+
{
|
| 205865 |
+
"epoch": 674.66,
|
| 205866 |
+
"learning_rate": 8.664345718901454e-06,
|
| 205867 |
+
"loss": 0.2988,
|
| 205868 |
+
"step": 83710
|
| 205869 |
+
},
|
| 205870 |
+
{
|
| 205871 |
+
"epoch": 674.7,
|
| 205872 |
+
"learning_rate": 8.66426494345719e-06,
|
| 205873 |
+
"loss": 0.3276,
|
| 205874 |
+
"step": 83715
|
| 205875 |
+
},
|
| 205876 |
+
{
|
| 205877 |
+
"epoch": 674.74,
|
| 205878 |
+
"learning_rate": 8.664184168012924e-06,
|
| 205879 |
+
"loss": 0.3733,
|
| 205880 |
+
"step": 83720
|
| 205881 |
+
},
|
| 205882 |
+
{
|
| 205883 |
+
"epoch": 674.78,
|
| 205884 |
+
"learning_rate": 8.66410339256866e-06,
|
| 205885 |
+
"loss": 0.5868,
|
| 205886 |
+
"step": 83725
|
| 205887 |
+
},
|
| 205888 |
+
{
|
| 205889 |
+
"epoch": 674.82,
|
| 205890 |
+
"learning_rate": 8.664022617124394e-06,
|
| 205891 |
+
"loss": 1.1381,
|
| 205892 |
+
"step": 83730
|
| 205893 |
+
},
|
| 205894 |
+
{
|
| 205895 |
+
"epoch": 674.86,
|
| 205896 |
+
"learning_rate": 8.66394184168013e-06,
|
| 205897 |
+
"loss": 0.2754,
|
| 205898 |
+
"step": 83735
|
| 205899 |
+
},
|
| 205900 |
+
{
|
| 205901 |
+
"epoch": 674.9,
|
| 205902 |
+
"learning_rate": 8.663861066235864e-06,
|
| 205903 |
+
"loss": 0.3627,
|
| 205904 |
+
"step": 83740
|
| 205905 |
+
},
|
| 205906 |
+
{
|
| 205907 |
+
"epoch": 674.94,
|
| 205908 |
+
"learning_rate": 8.6637802907916e-06,
|
| 205909 |
+
"loss": 0.3453,
|
| 205910 |
+
"step": 83745
|
| 205911 |
+
},
|
| 205912 |
+
{
|
| 205913 |
+
"epoch": 674.98,
|
| 205914 |
+
"learning_rate": 8.663699515347334e-06,
|
| 205915 |
+
"loss": 0.7058,
|
| 205916 |
+
"step": 83750
|
| 205917 |
+
},
|
| 205918 |
+
{
|
| 205919 |
+
"epoch": 675.0,
|
| 205920 |
+
"eval_loss": 0.40985289216041565,
|
| 205921 |
+
"eval_runtime": 42.042,
|
| 205922 |
+
"eval_samples_per_second": 20.004,
|
| 205923 |
+
"eval_steps_per_second": 0.642,
|
| 205924 |
+
"eval_wer": 0.18818096543762175,
|
| 205925 |
+
"step": 83752
|
| 205926 |
+
},
|
| 205927 |
+
{
|
| 205928 |
+
"epoch": 675.02,
|
| 205929 |
+
"learning_rate": 8.66361873990307e-06,
|
| 205930 |
+
"loss": 0.3476,
|
| 205931 |
+
"step": 83755
|
| 205932 |
+
},
|
| 205933 |
+
{
|
| 205934 |
+
"epoch": 675.06,
|
| 205935 |
+
"learning_rate": 8.663537964458804e-06,
|
| 205936 |
+
"loss": 0.3066,
|
| 205937 |
+
"step": 83760
|
| 205938 |
+
},
|
| 205939 |
+
{
|
| 205940 |
+
"epoch": 675.1,
|
| 205941 |
+
"learning_rate": 8.66345718901454e-06,
|
| 205942 |
+
"loss": 0.2835,
|
| 205943 |
+
"step": 83765
|
| 205944 |
+
},
|
| 205945 |
+
{
|
| 205946 |
+
"epoch": 675.14,
|
| 205947 |
+
"learning_rate": 8.663376413570276e-06,
|
| 205948 |
+
"loss": 0.3918,
|
| 205949 |
+
"step": 83770
|
| 205950 |
+
},
|
| 205951 |
+
{
|
| 205952 |
+
"epoch": 675.18,
|
| 205953 |
+
"learning_rate": 8.66329563812601e-06,
|
| 205954 |
+
"loss": 0.6303,
|
| 205955 |
+
"step": 83775
|
| 205956 |
+
},
|
| 205957 |
+
{
|
| 205958 |
+
"epoch": 675.22,
|
| 205959 |
+
"learning_rate": 8.663214862681746e-06,
|
| 205960 |
+
"loss": 0.8455,
|
| 205961 |
+
"step": 83780
|
| 205962 |
+
},
|
| 205963 |
+
{
|
| 205964 |
+
"epoch": 675.27,
|
| 205965 |
+
"learning_rate": 8.66313408723748e-06,
|
| 205966 |
+
"loss": 0.3056,
|
| 205967 |
+
"step": 83785
|
| 205968 |
+
},
|
| 205969 |
+
{
|
| 205970 |
+
"epoch": 675.31,
|
| 205971 |
+
"learning_rate": 8.663053311793216e-06,
|
| 205972 |
+
"loss": 0.4009,
|
| 205973 |
+
"step": 83790
|
| 205974 |
+
},
|
| 205975 |
+
{
|
| 205976 |
+
"epoch": 675.35,
|
| 205977 |
+
"learning_rate": 8.66297253634895e-06,
|
| 205978 |
+
"loss": 0.3786,
|
| 205979 |
+
"step": 83795
|
| 205980 |
+
},
|
| 205981 |
+
{
|
| 205982 |
+
"epoch": 675.39,
|
| 205983 |
+
"learning_rate": 8.662891760904686e-06,
|
| 205984 |
+
"loss": 0.7564,
|
| 205985 |
+
"step": 83800
|
| 205986 |
+
},
|
| 205987 |
+
{
|
| 205988 |
+
"epoch": 675.43,
|
| 205989 |
+
"learning_rate": 8.66281098546042e-06,
|
| 205990 |
+
"loss": 0.9055,
|
| 205991 |
+
"step": 83805
|
| 205992 |
+
},
|
| 205993 |
+
{
|
| 205994 |
+
"epoch": 675.47,
|
| 205995 |
+
"learning_rate": 8.662730210016156e-06,
|
| 205996 |
+
"loss": 0.3277,
|
| 205997 |
+
"step": 83810
|
| 205998 |
+
},
|
| 205999 |
+
{
|
| 206000 |
+
"epoch": 675.51,
|
| 206001 |
+
"learning_rate": 8.66264943457189e-06,
|
| 206002 |
+
"loss": 0.4182,
|
| 206003 |
+
"step": 83815
|
| 206004 |
+
},
|
| 206005 |
+
{
|
| 206006 |
+
"epoch": 675.55,
|
| 206007 |
+
"learning_rate": 8.662568659127626e-06,
|
| 206008 |
+
"loss": 0.4124,
|
| 206009 |
+
"step": 83820
|
| 206010 |
+
},
|
| 206011 |
+
{
|
| 206012 |
+
"epoch": 675.59,
|
| 206013 |
+
"learning_rate": 8.662487883683361e-06,
|
| 206014 |
+
"loss": 0.7556,
|
| 206015 |
+
"step": 83825
|
| 206016 |
+
},
|
| 206017 |
+
{
|
| 206018 |
+
"epoch": 675.63,
|
| 206019 |
+
"learning_rate": 8.662407108239096e-06,
|
| 206020 |
+
"loss": 0.9783,
|
| 206021 |
+
"step": 83830
|
| 206022 |
+
},
|
| 206023 |
+
{
|
| 206024 |
+
"epoch": 675.67,
|
| 206025 |
+
"learning_rate": 8.662326332794831e-06,
|
| 206026 |
+
"loss": 0.275,
|
| 206027 |
+
"step": 83835
|
| 206028 |
+
},
|
| 206029 |
+
{
|
| 206030 |
+
"epoch": 675.71,
|
| 206031 |
+
"learning_rate": 8.662245557350566e-06,
|
| 206032 |
+
"loss": 0.3723,
|
| 206033 |
+
"step": 83840
|
| 206034 |
+
},
|
| 206035 |
+
{
|
| 206036 |
+
"epoch": 675.75,
|
| 206037 |
+
"learning_rate": 8.662164781906301e-06,
|
| 206038 |
+
"loss": 0.3922,
|
| 206039 |
+
"step": 83845
|
| 206040 |
+
},
|
| 206041 |
+
{
|
| 206042 |
+
"epoch": 675.79,
|
| 206043 |
+
"learning_rate": 8.662084006462036e-06,
|
| 206044 |
+
"loss": 0.9253,
|
| 206045 |
+
"step": 83850
|
| 206046 |
+
},
|
| 206047 |
+
{
|
| 206048 |
+
"epoch": 675.83,
|
| 206049 |
+
"learning_rate": 8.662003231017771e-06,
|
| 206050 |
+
"loss": 0.8761,
|
| 206051 |
+
"step": 83855
|
| 206052 |
+
},
|
| 206053 |
+
{
|
| 206054 |
+
"epoch": 675.87,
|
| 206055 |
+
"learning_rate": 8.661922455573506e-06,
|
| 206056 |
+
"loss": 0.3139,
|
| 206057 |
+
"step": 83860
|
| 206058 |
+
},
|
| 206059 |
+
{
|
| 206060 |
+
"epoch": 675.91,
|
| 206061 |
+
"learning_rate": 8.661841680129241e-06,
|
| 206062 |
+
"loss": 0.3688,
|
| 206063 |
+
"step": 83865
|
| 206064 |
+
},
|
| 206065 |
+
{
|
| 206066 |
+
"epoch": 675.95,
|
| 206067 |
+
"learning_rate": 8.661760904684976e-06,
|
| 206068 |
+
"loss": 0.3588,
|
| 206069 |
+
"step": 83870
|
| 206070 |
+
},
|
| 206071 |
+
{
|
| 206072 |
+
"epoch": 675.99,
|
| 206073 |
+
"learning_rate": 8.661680129240711e-06,
|
| 206074 |
+
"loss": 0.9164,
|
| 206075 |
+
"step": 83875
|
| 206076 |
+
},
|
| 206077 |
+
{
|
| 206078 |
+
"epoch": 676.0,
|
| 206079 |
+
"eval_loss": 0.42249971628189087,
|
| 206080 |
+
"eval_runtime": 42.0177,
|
| 206081 |
+
"eval_samples_per_second": 20.015,
|
| 206082 |
+
"eval_steps_per_second": 0.643,
|
| 206083 |
+
"eval_wer": 0.18870289801637047,
|
| 206084 |
+
"step": 83876
|
| 206085 |
+
},
|
| 206086 |
+
{
|
| 206087 |
+
"epoch": 671.03,
|
| 206088 |
+
"learning_rate": 8.661599353796447e-06,
|
| 206089 |
+
"loss": 0.3709,
|
| 206090 |
+
"step": 83880
|
| 206091 |
+
},
|
| 206092 |
+
{
|
| 206093 |
+
"epoch": 671.07,
|
| 206094 |
+
"learning_rate": 8.661518578352181e-06,
|
| 206095 |
+
"loss": 0.292,
|
| 206096 |
+
"step": 83885
|
| 206097 |
+
},
|
| 206098 |
+
{
|
| 206099 |
+
"epoch": 671.11,
|
| 206100 |
+
"learning_rate": 8.661437802907917e-06,
|
| 206101 |
+
"loss": 0.341,
|
| 206102 |
+
"step": 83890
|
| 206103 |
+
},
|
| 206104 |
+
{
|
| 206105 |
+
"epoch": 671.15,
|
| 206106 |
+
"learning_rate": 8.661357027463651e-06,
|
| 206107 |
+
"loss": 0.3583,
|
| 206108 |
+
"step": 83895
|
| 206109 |
+
},
|
| 206110 |
+
{
|
| 206111 |
+
"epoch": 671.19,
|
| 206112 |
+
"learning_rate": 8.661276252019387e-06,
|
| 206113 |
+
"loss": 0.9241,
|
| 206114 |
+
"step": 83900
|
| 206115 |
+
},
|
| 206116 |
+
{
|
| 206117 |
+
"epoch": 671.23,
|
| 206118 |
+
"learning_rate": 8.661195476575121e-06,
|
| 206119 |
+
"loss": 0.7109,
|
| 206120 |
+
"step": 83905
|
| 206121 |
+
},
|
| 206122 |
+
{
|
| 206123 |
+
"epoch": 671.27,
|
| 206124 |
+
"learning_rate": 8.661114701130857e-06,
|
| 206125 |
+
"loss": 0.2876,
|
| 206126 |
+
"step": 83910
|
| 206127 |
+
},
|
| 206128 |
+
{
|
| 206129 |
+
"epoch": 671.31,
|
| 206130 |
+
"learning_rate": 8.661033925686591e-06,
|
| 206131 |
+
"loss": 0.3005,
|
| 206132 |
+
"step": 83915
|
| 206133 |
+
},
|
| 206134 |
+
{
|
| 206135 |
+
"epoch": 671.35,
|
| 206136 |
+
"learning_rate": 8.660953150242327e-06,
|
| 206137 |
+
"loss": 0.3872,
|
| 206138 |
+
"step": 83920
|
| 206139 |
+
},
|
| 206140 |
+
{
|
| 206141 |
+
"epoch": 671.39,
|
| 206142 |
+
"learning_rate": 8.660872374798061e-06,
|
| 206143 |
+
"loss": 0.9051,
|
| 206144 |
+
"step": 83925
|
| 206145 |
+
},
|
| 206146 |
+
{
|
| 206147 |
+
"epoch": 671.43,
|
| 206148 |
+
"learning_rate": 8.660791599353797e-06,
|
| 206149 |
+
"loss": 0.5334,
|
| 206150 |
+
"step": 83930
|
| 206151 |
+
},
|
| 206152 |
+
{
|
| 206153 |
+
"epoch": 671.47,
|
| 206154 |
+
"learning_rate": 8.660710823909531e-06,
|
| 206155 |
+
"loss": 0.37,
|
| 206156 |
+
"step": 83935
|
| 206157 |
+
},
|
| 206158 |
+
{
|
| 206159 |
+
"epoch": 671.51,
|
| 206160 |
+
"learning_rate": 8.660630048465267e-06,
|
| 206161 |
+
"loss": 0.3576,
|
| 206162 |
+
"step": 83940
|
| 206163 |
+
},
|
| 206164 |
+
{
|
| 206165 |
+
"epoch": 671.55,
|
| 206166 |
+
"learning_rate": 8.660549273021003e-06,
|
| 206167 |
+
"loss": 0.3855,
|
| 206168 |
+
"step": 83945
|
| 206169 |
+
},
|
| 206170 |
+
{
|
| 206171 |
+
"epoch": 671.59,
|
| 206172 |
+
"learning_rate": 8.660468497576737e-06,
|
| 206173 |
+
"loss": 0.8962,
|
| 206174 |
+
"step": 83950
|
| 206175 |
+
},
|
| 206176 |
+
{
|
| 206177 |
+
"epoch": 671.63,
|
| 206178 |
+
"learning_rate": 8.660387722132473e-06,
|
| 206179 |
+
"loss": 0.6493,
|
| 206180 |
+
"step": 83955
|
| 206181 |
+
},
|
| 206182 |
+
{
|
| 206183 |
+
"epoch": 671.67,
|
| 206184 |
+
"learning_rate": 8.660306946688207e-06,
|
| 206185 |
+
"loss": 0.3845,
|
| 206186 |
+
"step": 83960
|
| 206187 |
+
},
|
| 206188 |
+
{
|
| 206189 |
+
"epoch": 671.71,
|
| 206190 |
+
"learning_rate": 8.660226171243943e-06,
|
| 206191 |
+
"loss": 0.3092,
|
| 206192 |
+
"step": 83965
|
| 206193 |
+
},
|
| 206194 |
+
{
|
| 206195 |
+
"epoch": 671.75,
|
| 206196 |
+
"learning_rate": 8.660145395799677e-06,
|
| 206197 |
+
"loss": 0.4117,
|
| 206198 |
+
"step": 83970
|
| 206199 |
+
},
|
| 206200 |
+
{
|
| 206201 |
+
"epoch": 671.79,
|
| 206202 |
+
"learning_rate": 8.660064620355413e-06,
|
| 206203 |
+
"loss": 1.0803,
|
| 206204 |
+
"step": 83975
|
| 206205 |
+
},
|
| 206206 |
+
{
|
| 206207 |
+
"epoch": 671.83,
|
| 206208 |
+
"learning_rate": 8.659983844911147e-06,
|
| 206209 |
+
"loss": 0.7499,
|
| 206210 |
+
"step": 83980
|
| 206211 |
+
},
|
| 206212 |
+
{
|
| 206213 |
+
"epoch": 671.87,
|
| 206214 |
+
"learning_rate": 8.659903069466883e-06,
|
| 206215 |
+
"loss": 0.2696,
|
| 206216 |
+
"step": 83985
|
| 206217 |
+
},
|
| 206218 |
+
{
|
| 206219 |
+
"epoch": 671.91,
|
| 206220 |
+
"learning_rate": 8.659822294022617e-06,
|
| 206221 |
+
"loss": 0.3393,
|
| 206222 |
+
"step": 83990
|
| 206223 |
+
},
|
| 206224 |
+
{
|
| 206225 |
+
"epoch": 671.95,
|
| 206226 |
+
"learning_rate": 8.659741518578353e-06,
|
| 206227 |
+
"loss": 0.4359,
|
| 206228 |
+
"step": 83995
|
| 206229 |
+
},
|
| 206230 |
+
{
|
| 206231 |
+
"epoch": 671.99,
|
| 206232 |
+
"learning_rate": 8.659660743134089e-06,
|
| 206233 |
+
"loss": 1.0414,
|
| 206234 |
+
"step": 84000
|
| 206235 |
+
},
|
| 206236 |
+
{
|
| 206237 |
+
"epoch": 672.0,
|
| 206238 |
+
"eval_loss": 0.4100021421909332,
|
| 206239 |
+
"eval_runtime": 42.3055,
|
| 206240 |
+
"eval_samples_per_second": 19.879,
|
| 206241 |
+
"eval_steps_per_second": 0.638,
|
| 206242 |
+
"eval_wer": 0.19464897572355472,
|
| 206243 |
+
"step": 84001
|
| 206244 |
}
|
| 206245 |
],
|
| 206246 |
+
"max_steps": 625000,
|
| 206247 |
"num_train_epochs": 5000,
|
| 206248 |
+
"total_flos": 2.363888951443245e+20,
|
| 206249 |
"trial_name": null,
|
| 206250 |
"trial_params": null
|
| 206251 |
}
|
model-bin/finetune/base/{checkpoint-83380 → checkpoint-84001}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629916283.1054225/events.out.tfevents.1629916283.7e498afd5545.7645.115
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4017f29ee22f6d554ccea65f389dbc239e7865d321fb16e8b2eb80cb9862f747
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629916758.8115587/events.out.tfevents.1629916758.7e498afd5545.7645.117
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:490701c12971975b015f858101a7065cfbfab0aacc00f5c0f09ce1222878e52e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629917235.212877/events.out.tfevents.1629917235.7e498afd5545.7645.119
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:421705e54cf92c173a646a4067b7df89593e076b1a273916c4b7df0ec1019766
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629917711.204426/events.out.tfevents.1629917711.7e498afd5545.7645.121
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b496af4975b93736b29ef364e94bbf9e95f6ec1013efa32ef5664a8fee9152d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629918181.9293232/events.out.tfevents.1629918181.7e498afd5545.7645.123
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89d538b5f68eeba69ec775fe5e5367369790485dcfde69ff83f0fe057db25c07
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629916283.7e498afd5545.7645.114
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb1fca8760c1e259bc5016e176390a7dfb65011177b00e3f28275fc8ebd3f727
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629916758.7e498afd5545.7645.116
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:93e2956b8878ac50cc505e8039e1b27b1781395b8932e9acfadac033f14e4b13
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629917234.7e498afd5545.7645.118
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:47913bf9a87d6fb79be43c7034d16d6d8f6166a2995d60e8522158ae25810df0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629917711.7e498afd5545.7645.120
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f986100179df08c47e0c10a7a2337be2bedacfc29aad22688b123ba6fbd3e096
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629918181.7e498afd5545.7645.122
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f10ea9d8de1ad81361d32b308a5b6c0e45fef4a3213893977f01ed7d7ebff8c4
|
| 3 |
+
size 8622
|