"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630255201.5100508/events.out.tfevents.1630255201.cc93b136ebf5.1086.229 +3 -0
- model-bin/finetune/base/log/1630255633.58935/events.out.tfevents.1630255633.cc93b136ebf5.1086.231 +3 -0
- model-bin/finetune/base/log/1630256177.9503975/events.out.tfevents.1630256177.cc93b136ebf5.1086.233 +3 -0
- model-bin/finetune/base/log/1630256616.2655816/events.out.tfevents.1630256616.cc93b136ebf5.1086.235 +3 -0
- model-bin/finetune/base/log/1630257046.964922/events.out.tfevents.1630257046.cc93b136ebf5.1086.237 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630255201.cc93b136ebf5.1086.228 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630255633.cc93b136ebf5.1086.230 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630256177.cc93b136ebf5.1086.232 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630256616.cc93b136ebf5.1086.234 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630257046.cc93b136ebf5.1086.236 +3 -0
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/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:c6c4e2f164fe612c9447b7fdd6a75bd151c34385c5157ec3091d0d894af2e00d
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/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:bd584b4168381271c31720a7885b9ee125cb30ba21f1597f757e00db3049a185
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/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 sha256:3f6ff2f68c3c8c25a6448582f63dfac8cbe057a7cbc620566ec94b3b6c1eb228
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/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 sha256:39306f1d255c1140e0b97c10ee337a2e56b425eb5149e6f8db137bc9f9496e80
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/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 sha256:2facb014c0362e2448c4cd83ad0a2f81ed3eca6c82dcc98beab730981e6e8e16
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -294426,11 +294426,800 @@
|
|
| 294426 |
"eval_steps_per_second": 0.649,
|
| 294427 |
"eval_wer": 0.18537154524903376,
|
| 294428 |
"step": 153312
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 294429 |
}
|
| 294430 |
],
|
| 294431 |
-
"max_steps":
|
| 294432 |
"num_train_epochs": 5000,
|
| 294433 |
-
"total_flos": 4.
|
| 294434 |
"trial_name": null,
|
| 294435 |
"trial_params": null
|
| 294436 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
+
"epoch": 1231.0,
|
| 5 |
+
"global_step": 153934,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 294426 |
"eval_steps_per_second": 0.649,
|
| 294427 |
"eval_wer": 0.18537154524903376,
|
| 294428 |
"step": 153312
|
| 294429 |
+
},
|
| 294430 |
+
{
|
| 294431 |
+
"epoch": 1236.02,
|
| 294432 |
+
"learning_rate": 7.5403069466882075e-06,
|
| 294433 |
+
"loss": 0.3773,
|
| 294434 |
+
"step": 153315
|
| 294435 |
+
},
|
| 294436 |
+
{
|
| 294437 |
+
"epoch": 1236.06,
|
| 294438 |
+
"learning_rate": 7.5402261712439425e-06,
|
| 294439 |
+
"loss": 0.3049,
|
| 294440 |
+
"step": 153320
|
| 294441 |
+
},
|
| 294442 |
+
{
|
| 294443 |
+
"epoch": 1236.1,
|
| 294444 |
+
"learning_rate": 7.5401453957996775e-06,
|
| 294445 |
+
"loss": 0.2696,
|
| 294446 |
+
"step": 153325
|
| 294447 |
+
},
|
| 294448 |
+
{
|
| 294449 |
+
"epoch": 1236.14,
|
| 294450 |
+
"learning_rate": 7.5400646203554125e-06,
|
| 294451 |
+
"loss": 0.3384,
|
| 294452 |
+
"step": 153330
|
| 294453 |
+
},
|
| 294454 |
+
{
|
| 294455 |
+
"epoch": 1236.18,
|
| 294456 |
+
"learning_rate": 7.5399838449111475e-06,
|
| 294457 |
+
"loss": 0.7549,
|
| 294458 |
+
"step": 153335
|
| 294459 |
+
},
|
| 294460 |
+
{
|
| 294461 |
+
"epoch": 1236.22,
|
| 294462 |
+
"learning_rate": 7.5399030694668825e-06,
|
| 294463 |
+
"loss": 0.9564,
|
| 294464 |
+
"step": 153340
|
| 294465 |
+
},
|
| 294466 |
+
{
|
| 294467 |
+
"epoch": 1236.27,
|
| 294468 |
+
"learning_rate": 7.5398222940226175e-06,
|
| 294469 |
+
"loss": 0.29,
|
| 294470 |
+
"step": 153345
|
| 294471 |
+
},
|
| 294472 |
+
{
|
| 294473 |
+
"epoch": 1236.31,
|
| 294474 |
+
"learning_rate": 7.539741518578353e-06,
|
| 294475 |
+
"loss": 0.2812,
|
| 294476 |
+
"step": 153350
|
| 294477 |
+
},
|
| 294478 |
+
{
|
| 294479 |
+
"epoch": 1236.35,
|
| 294480 |
+
"learning_rate": 7.5396607431340874e-06,
|
| 294481 |
+
"loss": 0.3392,
|
| 294482 |
+
"step": 153355
|
| 294483 |
+
},
|
| 294484 |
+
{
|
| 294485 |
+
"epoch": 1236.39,
|
| 294486 |
+
"learning_rate": 7.539579967689823e-06,
|
| 294487 |
+
"loss": 0.7306,
|
| 294488 |
+
"step": 153360
|
| 294489 |
+
},
|
| 294490 |
+
{
|
| 294491 |
+
"epoch": 1236.43,
|
| 294492 |
+
"learning_rate": 7.539499192245558e-06,
|
| 294493 |
+
"loss": 0.9617,
|
| 294494 |
+
"step": 153365
|
| 294495 |
+
},
|
| 294496 |
+
{
|
| 294497 |
+
"epoch": 1236.47,
|
| 294498 |
+
"learning_rate": 7.539418416801293e-06,
|
| 294499 |
+
"loss": 0.2778,
|
| 294500 |
+
"step": 153370
|
| 294501 |
+
},
|
| 294502 |
+
{
|
| 294503 |
+
"epoch": 1236.51,
|
| 294504 |
+
"learning_rate": 7.539337641357028e-06,
|
| 294505 |
+
"loss": 0.2961,
|
| 294506 |
+
"step": 153375
|
| 294507 |
+
},
|
| 294508 |
+
{
|
| 294509 |
+
"epoch": 1236.55,
|
| 294510 |
+
"learning_rate": 7.539256865912763e-06,
|
| 294511 |
+
"loss": 0.3302,
|
| 294512 |
+
"step": 153380
|
| 294513 |
+
},
|
| 294514 |
+
{
|
| 294515 |
+
"epoch": 1236.59,
|
| 294516 |
+
"learning_rate": 7.539176090468498e-06,
|
| 294517 |
+
"loss": 0.683,
|
| 294518 |
+
"step": 153385
|
| 294519 |
+
},
|
| 294520 |
+
{
|
| 294521 |
+
"epoch": 1236.63,
|
| 294522 |
+
"learning_rate": 7.539095315024233e-06,
|
| 294523 |
+
"loss": 0.9545,
|
| 294524 |
+
"step": 153390
|
| 294525 |
+
},
|
| 294526 |
+
{
|
| 294527 |
+
"epoch": 1236.67,
|
| 294528 |
+
"learning_rate": 7.539014539579968e-06,
|
| 294529 |
+
"loss": 0.3683,
|
| 294530 |
+
"step": 153395
|
| 294531 |
+
},
|
| 294532 |
+
{
|
| 294533 |
+
"epoch": 1236.71,
|
| 294534 |
+
"learning_rate": 7.538933764135703e-06,
|
| 294535 |
+
"loss": 0.2904,
|
| 294536 |
+
"step": 153400
|
| 294537 |
+
},
|
| 294538 |
+
{
|
| 294539 |
+
"epoch": 1236.75,
|
| 294540 |
+
"learning_rate": 7.538852988691438e-06,
|
| 294541 |
+
"loss": 0.29,
|
| 294542 |
+
"step": 153405
|
| 294543 |
+
},
|
| 294544 |
+
{
|
| 294545 |
+
"epoch": 1236.79,
|
| 294546 |
+
"learning_rate": 7.538772213247173e-06,
|
| 294547 |
+
"loss": 0.745,
|
| 294548 |
+
"step": 153410
|
| 294549 |
+
},
|
| 294550 |
+
{
|
| 294551 |
+
"epoch": 1236.83,
|
| 294552 |
+
"learning_rate": 7.538691437802909e-06,
|
| 294553 |
+
"loss": 0.9307,
|
| 294554 |
+
"step": 153415
|
| 294555 |
+
},
|
| 294556 |
+
{
|
| 294557 |
+
"epoch": 1236.87,
|
| 294558 |
+
"learning_rate": 7.538610662358644e-06,
|
| 294559 |
+
"loss": 0.3326,
|
| 294560 |
+
"step": 153420
|
| 294561 |
+
},
|
| 294562 |
+
{
|
| 294563 |
+
"epoch": 1236.91,
|
| 294564 |
+
"learning_rate": 7.538529886914379e-06,
|
| 294565 |
+
"loss": 0.268,
|
| 294566 |
+
"step": 153425
|
| 294567 |
+
},
|
| 294568 |
+
{
|
| 294569 |
+
"epoch": 1236.95,
|
| 294570 |
+
"learning_rate": 7.538449111470114e-06,
|
| 294571 |
+
"loss": 0.4247,
|
| 294572 |
+
"step": 153430
|
| 294573 |
+
},
|
| 294574 |
+
{
|
| 294575 |
+
"epoch": 1236.99,
|
| 294576 |
+
"learning_rate": 7.538368336025849e-06,
|
| 294577 |
+
"loss": 0.8271,
|
| 294578 |
+
"step": 153435
|
| 294579 |
+
},
|
| 294580 |
+
{
|
| 294581 |
+
"epoch": 1237.0,
|
| 294582 |
+
"eval_loss": 0.3603706955909729,
|
| 294583 |
+
"eval_runtime": 42.9904,
|
| 294584 |
+
"eval_samples_per_second": 19.446,
|
| 294585 |
+
"eval_steps_per_second": 0.628,
|
| 294586 |
+
"eval_wer": 0.1769643109042178,
|
| 294587 |
+
"step": 153436
|
| 294588 |
+
},
|
| 294589 |
+
{
|
| 294590 |
+
"epoch": 1237.03,
|
| 294591 |
+
"learning_rate": 7.538287560581584e-06,
|
| 294592 |
+
"loss": 0.3607,
|
| 294593 |
+
"step": 153440
|
| 294594 |
+
},
|
| 294595 |
+
{
|
| 294596 |
+
"epoch": 1237.07,
|
| 294597 |
+
"learning_rate": 7.538206785137319e-06,
|
| 294598 |
+
"loss": 0.2871,
|
| 294599 |
+
"step": 153445
|
| 294600 |
+
},
|
| 294601 |
+
{
|
| 294602 |
+
"epoch": 1237.11,
|
| 294603 |
+
"learning_rate": 7.538126009693054e-06,
|
| 294604 |
+
"loss": 0.266,
|
| 294605 |
+
"step": 153450
|
| 294606 |
+
},
|
| 294607 |
+
{
|
| 294608 |
+
"epoch": 1237.15,
|
| 294609 |
+
"learning_rate": 7.538045234248789e-06,
|
| 294610 |
+
"loss": 0.375,
|
| 294611 |
+
"step": 153455
|
| 294612 |
+
},
|
| 294613 |
+
{
|
| 294614 |
+
"epoch": 1237.19,
|
| 294615 |
+
"learning_rate": 7.537964458804524e-06,
|
| 294616 |
+
"loss": 0.835,
|
| 294617 |
+
"step": 153460
|
| 294618 |
+
},
|
| 294619 |
+
{
|
| 294620 |
+
"epoch": 1237.23,
|
| 294621 |
+
"learning_rate": 7.537883683360259e-06,
|
| 294622 |
+
"loss": 0.6919,
|
| 294623 |
+
"step": 153465
|
| 294624 |
+
},
|
| 294625 |
+
{
|
| 294626 |
+
"epoch": 1237.27,
|
| 294627 |
+
"learning_rate": 7.537802907915994e-06,
|
| 294628 |
+
"loss": 0.29,
|
| 294629 |
+
"step": 153470
|
| 294630 |
+
},
|
| 294631 |
+
{
|
| 294632 |
+
"epoch": 1237.31,
|
| 294633 |
+
"learning_rate": 7.537722132471729e-06,
|
| 294634 |
+
"loss": 0.2965,
|
| 294635 |
+
"step": 153475
|
| 294636 |
+
},
|
| 294637 |
+
{
|
| 294638 |
+
"epoch": 1237.35,
|
| 294639 |
+
"learning_rate": 7.537641357027465e-06,
|
| 294640 |
+
"loss": 0.3667,
|
| 294641 |
+
"step": 153480
|
| 294642 |
+
},
|
| 294643 |
+
{
|
| 294644 |
+
"epoch": 1237.39,
|
| 294645 |
+
"learning_rate": 7.5375605815832e-06,
|
| 294646 |
+
"loss": 0.8043,
|
| 294647 |
+
"step": 153485
|
| 294648 |
+
},
|
| 294649 |
+
{
|
| 294650 |
+
"epoch": 1237.43,
|
| 294651 |
+
"learning_rate": 7.537479806138935e-06,
|
| 294652 |
+
"loss": 0.6527,
|
| 294653 |
+
"step": 153490
|
| 294654 |
+
},
|
| 294655 |
+
{
|
| 294656 |
+
"epoch": 1237.47,
|
| 294657 |
+
"learning_rate": 7.53739903069467e-06,
|
| 294658 |
+
"loss": 0.2526,
|
| 294659 |
+
"step": 153495
|
| 294660 |
+
},
|
| 294661 |
+
{
|
| 294662 |
+
"epoch": 1237.51,
|
| 294663 |
+
"learning_rate": 7.537318255250405e-06,
|
| 294664 |
+
"loss": 0.3293,
|
| 294665 |
+
"step": 153500
|
| 294666 |
+
},
|
| 294667 |
+
{
|
| 294668 |
+
"epoch": 1237.55,
|
| 294669 |
+
"learning_rate": 7.53723747980614e-06,
|
| 294670 |
+
"loss": 0.4376,
|
| 294671 |
+
"step": 153505
|
| 294672 |
+
},
|
| 294673 |
+
{
|
| 294674 |
+
"epoch": 1237.59,
|
| 294675 |
+
"learning_rate": 7.537156704361875e-06,
|
| 294676 |
+
"loss": 0.7937,
|
| 294677 |
+
"step": 153510
|
| 294678 |
+
},
|
| 294679 |
+
{
|
| 294680 |
+
"epoch": 1237.63,
|
| 294681 |
+
"learning_rate": 7.5370759289176096e-06,
|
| 294682 |
+
"loss": 0.593,
|
| 294683 |
+
"step": 153515
|
| 294684 |
+
},
|
| 294685 |
+
{
|
| 294686 |
+
"epoch": 1237.67,
|
| 294687 |
+
"learning_rate": 7.5369951534733446e-06,
|
| 294688 |
+
"loss": 0.2965,
|
| 294689 |
+
"step": 153520
|
| 294690 |
+
},
|
| 294691 |
+
{
|
| 294692 |
+
"epoch": 1237.71,
|
| 294693 |
+
"learning_rate": 7.5369143780290795e-06,
|
| 294694 |
+
"loss": 0.3655,
|
| 294695 |
+
"step": 153525
|
| 294696 |
+
},
|
| 294697 |
+
{
|
| 294698 |
+
"epoch": 1237.76,
|
| 294699 |
+
"learning_rate": 7.5368336025848145e-06,
|
| 294700 |
+
"loss": 0.3352,
|
| 294701 |
+
"step": 153530
|
| 294702 |
+
},
|
| 294703 |
+
{
|
| 294704 |
+
"epoch": 1237.8,
|
| 294705 |
+
"learning_rate": 7.5367528271405495e-06,
|
| 294706 |
+
"loss": 1.0342,
|
| 294707 |
+
"step": 153535
|
| 294708 |
+
},
|
| 294709 |
+
{
|
| 294710 |
+
"epoch": 1237.84,
|
| 294711 |
+
"learning_rate": 7.536672051696285e-06,
|
| 294712 |
+
"loss": 0.7089,
|
| 294713 |
+
"step": 153540
|
| 294714 |
+
},
|
| 294715 |
+
{
|
| 294716 |
+
"epoch": 1237.88,
|
| 294717 |
+
"learning_rate": 7.5365912762520195e-06,
|
| 294718 |
+
"loss": 0.2906,
|
| 294719 |
+
"step": 153545
|
| 294720 |
+
},
|
| 294721 |
+
{
|
| 294722 |
+
"epoch": 1237.92,
|
| 294723 |
+
"learning_rate": 7.536510500807755e-06,
|
| 294724 |
+
"loss": 0.3161,
|
| 294725 |
+
"step": 153550
|
| 294726 |
+
},
|
| 294727 |
+
{
|
| 294728 |
+
"epoch": 1237.96,
|
| 294729 |
+
"learning_rate": 7.53642972536349e-06,
|
| 294730 |
+
"loss": 0.4502,
|
| 294731 |
+
"step": 153555
|
| 294732 |
+
},
|
| 294733 |
+
{
|
| 294734 |
+
"epoch": 1238.0,
|
| 294735 |
+
"learning_rate": 7.536348949919225e-06,
|
| 294736 |
+
"loss": 1.0289,
|
| 294737 |
+
"step": 153560
|
| 294738 |
+
},
|
| 294739 |
+
{
|
| 294740 |
+
"epoch": 1238.0,
|
| 294741 |
+
"eval_loss": 0.5478711724281311,
|
| 294742 |
+
"eval_runtime": 48.9827,
|
| 294743 |
+
"eval_samples_per_second": 17.067,
|
| 294744 |
+
"eval_steps_per_second": 0.551,
|
| 294745 |
+
"eval_wer": 0.1848707647995149,
|
| 294746 |
+
"step": 153560
|
| 294747 |
+
},
|
| 294748 |
+
{
|
| 294749 |
+
"epoch": 1238.04,
|
| 294750 |
+
"learning_rate": 7.53626817447496e-06,
|
| 294751 |
+
"loss": 0.3428,
|
| 294752 |
+
"step": 153565
|
| 294753 |
+
},
|
| 294754 |
+
{
|
| 294755 |
+
"epoch": 1238.08,
|
| 294756 |
+
"learning_rate": 7.536187399030695e-06,
|
| 294757 |
+
"loss": 0.3268,
|
| 294758 |
+
"step": 153570
|
| 294759 |
+
},
|
| 294760 |
+
{
|
| 294761 |
+
"epoch": 1238.12,
|
| 294762 |
+
"learning_rate": 7.53610662358643e-06,
|
| 294763 |
+
"loss": 0.2791,
|
| 294764 |
+
"step": 153575
|
| 294765 |
+
},
|
| 294766 |
+
{
|
| 294767 |
+
"epoch": 1238.16,
|
| 294768 |
+
"learning_rate": 7.536025848142165e-06,
|
| 294769 |
+
"loss": 0.4305,
|
| 294770 |
+
"step": 153580
|
| 294771 |
+
},
|
| 294772 |
+
{
|
| 294773 |
+
"epoch": 1238.2,
|
| 294774 |
+
"learning_rate": 7.5359450726979e-06,
|
| 294775 |
+
"loss": 0.9733,
|
| 294776 |
+
"step": 153585
|
| 294777 |
+
},
|
| 294778 |
+
{
|
| 294779 |
+
"epoch": 1238.24,
|
| 294780 |
+
"learning_rate": 7.535864297253635e-06,
|
| 294781 |
+
"loss": 0.3815,
|
| 294782 |
+
"step": 153590
|
| 294783 |
+
},
|
| 294784 |
+
{
|
| 294785 |
+
"epoch": 1238.28,
|
| 294786 |
+
"learning_rate": 7.535783521809371e-06,
|
| 294787 |
+
"loss": 0.2843,
|
| 294788 |
+
"step": 153595
|
| 294789 |
+
},
|
| 294790 |
+
{
|
| 294791 |
+
"epoch": 1238.32,
|
| 294792 |
+
"learning_rate": 7.535702746365105e-06,
|
| 294793 |
+
"loss": 0.2388,
|
| 294794 |
+
"step": 153600
|
| 294795 |
+
},
|
| 294796 |
+
{
|
| 294797 |
+
"epoch": 1238.36,
|
| 294798 |
+
"learning_rate": 7.535621970920841e-06,
|
| 294799 |
+
"loss": 0.4122,
|
| 294800 |
+
"step": 153605
|
| 294801 |
+
},
|
| 294802 |
+
{
|
| 294803 |
+
"epoch": 1238.4,
|
| 294804 |
+
"learning_rate": 7.535557350565429e-06,
|
| 294805 |
+
"loss": 1.0435,
|
| 294806 |
+
"step": 153610
|
| 294807 |
+
},
|
| 294808 |
+
{
|
| 294809 |
+
"epoch": 1238.44,
|
| 294810 |
+
"learning_rate": 7.535476575121163e-06,
|
| 294811 |
+
"loss": 0.295,
|
| 294812 |
+
"step": 153615
|
| 294813 |
+
},
|
| 294814 |
+
{
|
| 294815 |
+
"epoch": 1238.48,
|
| 294816 |
+
"learning_rate": 7.535395799676899e-06,
|
| 294817 |
+
"loss": 0.2713,
|
| 294818 |
+
"step": 153620
|
| 294819 |
+
},
|
| 294820 |
+
{
|
| 294821 |
+
"epoch": 1238.52,
|
| 294822 |
+
"learning_rate": 7.535315024232633e-06,
|
| 294823 |
+
"loss": 0.2524,
|
| 294824 |
+
"step": 153625
|
| 294825 |
+
},
|
| 294826 |
+
{
|
| 294827 |
+
"epoch": 1238.56,
|
| 294828 |
+
"learning_rate": 7.535234248788369e-06,
|
| 294829 |
+
"loss": 0.4259,
|
| 294830 |
+
"step": 153630
|
| 294831 |
+
},
|
| 294832 |
+
{
|
| 294833 |
+
"epoch": 1238.6,
|
| 294834 |
+
"learning_rate": 7.535153473344104e-06,
|
| 294835 |
+
"loss": 1.1058,
|
| 294836 |
+
"step": 153635
|
| 294837 |
+
},
|
| 294838 |
+
{
|
| 294839 |
+
"epoch": 1238.65,
|
| 294840 |
+
"learning_rate": 7.535072697899839e-06,
|
| 294841 |
+
"loss": 0.2864,
|
| 294842 |
+
"step": 153640
|
| 294843 |
+
},
|
| 294844 |
+
{
|
| 294845 |
+
"epoch": 1238.69,
|
| 294846 |
+
"learning_rate": 7.534991922455574e-06,
|
| 294847 |
+
"loss": 0.2977,
|
| 294848 |
+
"step": 153645
|
| 294849 |
+
},
|
| 294850 |
+
{
|
| 294851 |
+
"epoch": 1238.73,
|
| 294852 |
+
"learning_rate": 7.534911147011309e-06,
|
| 294853 |
+
"loss": 0.3155,
|
| 294854 |
+
"step": 153650
|
| 294855 |
+
},
|
| 294856 |
+
{
|
| 294857 |
+
"epoch": 1238.77,
|
| 294858 |
+
"learning_rate": 7.534830371567045e-06,
|
| 294859 |
+
"loss": 0.3997,
|
| 294860 |
+
"step": 153655
|
| 294861 |
+
},
|
| 294862 |
+
{
|
| 294863 |
+
"epoch": 1238.81,
|
| 294864 |
+
"learning_rate": 7.534749596122779e-06,
|
| 294865 |
+
"loss": 1.1922,
|
| 294866 |
+
"step": 153660
|
| 294867 |
+
},
|
| 294868 |
+
{
|
| 294869 |
+
"epoch": 1238.85,
|
| 294870 |
+
"learning_rate": 7.534668820678515e-06,
|
| 294871 |
+
"loss": 0.3735,
|
| 294872 |
+
"step": 153665
|
| 294873 |
+
},
|
| 294874 |
+
{
|
| 294875 |
+
"epoch": 1238.89,
|
| 294876 |
+
"learning_rate": 7.534588045234249e-06,
|
| 294877 |
+
"loss": 0.2743,
|
| 294878 |
+
"step": 153670
|
| 294879 |
+
},
|
| 294880 |
+
{
|
| 294881 |
+
"epoch": 1238.93,
|
| 294882 |
+
"learning_rate": 7.534507269789985e-06,
|
| 294883 |
+
"loss": 0.2918,
|
| 294884 |
+
"step": 153675
|
| 294885 |
+
},
|
| 294886 |
+
{
|
| 294887 |
+
"epoch": 1238.97,
|
| 294888 |
+
"learning_rate": 7.534426494345719e-06,
|
| 294889 |
+
"loss": 0.444,
|
| 294890 |
+
"step": 153680
|
| 294891 |
+
},
|
| 294892 |
+
{
|
| 294893 |
+
"epoch": 1239.0,
|
| 294894 |
+
"eval_loss": 0.41387853026390076,
|
| 294895 |
+
"eval_runtime": 42.5902,
|
| 294896 |
+
"eval_samples_per_second": 19.84,
|
| 294897 |
+
"eval_steps_per_second": 0.634,
|
| 294898 |
+
"eval_wer": 0.18675968316256086,
|
| 294899 |
+
"step": 153684
|
| 294900 |
+
},
|
| 294901 |
+
{
|
| 294902 |
+
"epoch": 1229.01,
|
| 294903 |
+
"learning_rate": 7.534345718901455e-06,
|
| 294904 |
+
"loss": 0.2827,
|
| 294905 |
+
"step": 153685
|
| 294906 |
+
},
|
| 294907 |
+
{
|
| 294908 |
+
"epoch": 1229.05,
|
| 294909 |
+
"learning_rate": 7.53426494345719e-06,
|
| 294910 |
+
"loss": 0.3284,
|
| 294911 |
+
"step": 153690
|
| 294912 |
+
},
|
| 294913 |
+
{
|
| 294914 |
+
"epoch": 1229.09,
|
| 294915 |
+
"learning_rate": 7.534184168012925e-06,
|
| 294916 |
+
"loss": 0.2527,
|
| 294917 |
+
"step": 153695
|
| 294918 |
+
},
|
| 294919 |
+
{
|
| 294920 |
+
"epoch": 1229.13,
|
| 294921 |
+
"learning_rate": 7.53410339256866e-06,
|
| 294922 |
+
"loss": 0.2886,
|
| 294923 |
+
"step": 153700
|
| 294924 |
+
},
|
| 294925 |
+
{
|
| 294926 |
+
"epoch": 1229.17,
|
| 294927 |
+
"learning_rate": 7.534022617124395e-06,
|
| 294928 |
+
"loss": 0.4841,
|
| 294929 |
+
"step": 153705
|
| 294930 |
+
},
|
| 294931 |
+
{
|
| 294932 |
+
"epoch": 1229.21,
|
| 294933 |
+
"learning_rate": 7.53394184168013e-06,
|
| 294934 |
+
"loss": 1.2124,
|
| 294935 |
+
"step": 153710
|
| 294936 |
+
},
|
| 294937 |
+
{
|
| 294938 |
+
"epoch": 1229.25,
|
| 294939 |
+
"learning_rate": 7.533861066235865e-06,
|
| 294940 |
+
"loss": 0.2612,
|
| 294941 |
+
"step": 153715
|
| 294942 |
+
},
|
| 294943 |
+
{
|
| 294944 |
+
"epoch": 1229.29,
|
| 294945 |
+
"learning_rate": 7.533780290791601e-06,
|
| 294946 |
+
"loss": 0.2457,
|
| 294947 |
+
"step": 153720
|
| 294948 |
+
},
|
| 294949 |
+
{
|
| 294950 |
+
"epoch": 1229.33,
|
| 294951 |
+
"learning_rate": 7.533699515347335e-06,
|
| 294952 |
+
"loss": 0.2777,
|
| 294953 |
+
"step": 153725
|
| 294954 |
+
},
|
| 294955 |
+
{
|
| 294956 |
+
"epoch": 1229.37,
|
| 294957 |
+
"learning_rate": 7.533618739903071e-06,
|
| 294958 |
+
"loss": 0.4545,
|
| 294959 |
+
"step": 153730
|
| 294960 |
+
},
|
| 294961 |
+
{
|
| 294962 |
+
"epoch": 1229.41,
|
| 294963 |
+
"learning_rate": 7.533537964458805e-06,
|
| 294964 |
+
"loss": 1.1277,
|
| 294965 |
+
"step": 153735
|
| 294966 |
+
},
|
| 294967 |
+
{
|
| 294968 |
+
"epoch": 1229.45,
|
| 294969 |
+
"learning_rate": 7.5334571890145406e-06,
|
| 294970 |
+
"loss": 0.2915,
|
| 294971 |
+
"step": 153740
|
| 294972 |
+
},
|
| 294973 |
+
{
|
| 294974 |
+
"epoch": 1229.49,
|
| 294975 |
+
"learning_rate": 7.533376413570275e-06,
|
| 294976 |
+
"loss": 0.2849,
|
| 294977 |
+
"step": 153745
|
| 294978 |
+
},
|
| 294979 |
+
{
|
| 294980 |
+
"epoch": 1229.53,
|
| 294981 |
+
"learning_rate": 7.5332956381260105e-06,
|
| 294982 |
+
"loss": 0.2834,
|
| 294983 |
+
"step": 153750
|
| 294984 |
+
},
|
| 294985 |
+
{
|
| 294986 |
+
"epoch": 1229.57,
|
| 294987 |
+
"learning_rate": 7.5332148626817455e-06,
|
| 294988 |
+
"loss": 0.4269,
|
| 294989 |
+
"step": 153755
|
| 294990 |
+
},
|
| 294991 |
+
{
|
| 294992 |
+
"epoch": 1229.61,
|
| 294993 |
+
"learning_rate": 7.5331340872374805e-06,
|
| 294994 |
+
"loss": 0.9413,
|
| 294995 |
+
"step": 153760
|
| 294996 |
+
},
|
| 294997 |
+
{
|
| 294998 |
+
"epoch": 1229.65,
|
| 294999 |
+
"learning_rate": 7.5330533117932155e-06,
|
| 295000 |
+
"loss": 0.3305,
|
| 295001 |
+
"step": 153765
|
| 295002 |
+
},
|
| 295003 |
+
{
|
| 295004 |
+
"epoch": 1229.69,
|
| 295005 |
+
"learning_rate": 7.5329725363489505e-06,
|
| 295006 |
+
"loss": 0.2752,
|
| 295007 |
+
"step": 153770
|
| 295008 |
+
},
|
| 295009 |
+
{
|
| 295010 |
+
"epoch": 1229.73,
|
| 295011 |
+
"learning_rate": 7.5328917609046855e-06,
|
| 295012 |
+
"loss": 0.3041,
|
| 295013 |
+
"step": 153775
|
| 295014 |
+
},
|
| 295015 |
+
{
|
| 295016 |
+
"epoch": 1229.77,
|
| 295017 |
+
"learning_rate": 7.5328109854604205e-06,
|
| 295018 |
+
"loss": 0.4273,
|
| 295019 |
+
"step": 153780
|
| 295020 |
+
},
|
| 295021 |
+
{
|
| 295022 |
+
"epoch": 1229.81,
|
| 295023 |
+
"learning_rate": 7.532730210016156e-06,
|
| 295024 |
+
"loss": 1.0783,
|
| 295025 |
+
"step": 153785
|
| 295026 |
+
},
|
| 295027 |
+
{
|
| 295028 |
+
"epoch": 1229.85,
|
| 295029 |
+
"learning_rate": 7.5326494345718904e-06,
|
| 295030 |
+
"loss": 0.2982,
|
| 295031 |
+
"step": 153790
|
| 295032 |
+
},
|
| 295033 |
+
{
|
| 295034 |
+
"epoch": 1229.89,
|
| 295035 |
+
"learning_rate": 7.532568659127626e-06,
|
| 295036 |
+
"loss": 0.2782,
|
| 295037 |
+
"step": 153795
|
| 295038 |
+
},
|
| 295039 |
+
{
|
| 295040 |
+
"epoch": 1229.93,
|
| 295041 |
+
"learning_rate": 7.53248788368336e-06,
|
| 295042 |
+
"loss": 0.3229,
|
| 295043 |
+
"step": 153800
|
| 295044 |
+
},
|
| 295045 |
+
{
|
| 295046 |
+
"epoch": 1229.97,
|
| 295047 |
+
"learning_rate": 7.532407108239096e-06,
|
| 295048 |
+
"loss": 0.4878,
|
| 295049 |
+
"step": 153805
|
| 295050 |
+
},
|
| 295051 |
+
{
|
| 295052 |
+
"epoch": 1230.0,
|
| 295053 |
+
"eval_loss": 0.426218718290329,
|
| 295054 |
+
"eval_runtime": 40.8122,
|
| 295055 |
+
"eval_samples_per_second": 20.68,
|
| 295056 |
+
"eval_steps_per_second": 0.662,
|
| 295057 |
+
"eval_wer": 0.17735651131425145,
|
| 295058 |
+
"step": 153809
|
| 295059 |
+
},
|
| 295060 |
+
{
|
| 295061 |
+
"epoch": 1230.01,
|
| 295062 |
+
"learning_rate": 7.532326332794831e-06,
|
| 295063 |
+
"loss": 0.3619,
|
| 295064 |
+
"step": 153810
|
| 295065 |
+
},
|
| 295066 |
+
{
|
| 295067 |
+
"epoch": 1230.05,
|
| 295068 |
+
"learning_rate": 7.532245557350566e-06,
|
| 295069 |
+
"loss": 0.2662,
|
| 295070 |
+
"step": 153815
|
| 295071 |
+
},
|
| 295072 |
+
{
|
| 295073 |
+
"epoch": 1230.09,
|
| 295074 |
+
"learning_rate": 7.532164781906301e-06,
|
| 295075 |
+
"loss": 0.2751,
|
| 295076 |
+
"step": 153820
|
| 295077 |
+
},
|
| 295078 |
+
{
|
| 295079 |
+
"epoch": 1230.13,
|
| 295080 |
+
"learning_rate": 7.532084006462036e-06,
|
| 295081 |
+
"loss": 0.3694,
|
| 295082 |
+
"step": 153825
|
| 295083 |
+
},
|
| 295084 |
+
{
|
| 295085 |
+
"epoch": 1230.17,
|
| 295086 |
+
"learning_rate": 7.532003231017771e-06,
|
| 295087 |
+
"loss": 0.5156,
|
| 295088 |
+
"step": 153830
|
| 295089 |
+
},
|
| 295090 |
+
{
|
| 295091 |
+
"epoch": 1230.21,
|
| 295092 |
+
"learning_rate": 7.531922455573506e-06,
|
| 295093 |
+
"loss": 1.0791,
|
| 295094 |
+
"step": 153835
|
| 295095 |
+
},
|
| 295096 |
+
{
|
| 295097 |
+
"epoch": 1230.25,
|
| 295098 |
+
"learning_rate": 7.531841680129241e-06,
|
| 295099 |
+
"loss": 0.2508,
|
| 295100 |
+
"step": 153840
|
| 295101 |
+
},
|
| 295102 |
+
{
|
| 295103 |
+
"epoch": 1230.29,
|
| 295104 |
+
"learning_rate": 7.531760904684976e-06,
|
| 295105 |
+
"loss": 0.2831,
|
| 295106 |
+
"step": 153845
|
| 295107 |
+
},
|
| 295108 |
+
{
|
| 295109 |
+
"epoch": 1230.33,
|
| 295110 |
+
"learning_rate": 7.531680129240711e-06,
|
| 295111 |
+
"loss": 0.3186,
|
| 295112 |
+
"step": 153850
|
| 295113 |
+
},
|
| 295114 |
+
{
|
| 295115 |
+
"epoch": 1230.37,
|
| 295116 |
+
"learning_rate": 7.531599353796446e-06,
|
| 295117 |
+
"loss": 0.4961,
|
| 295118 |
+
"step": 153855
|
| 295119 |
+
},
|
| 295120 |
+
{
|
| 295121 |
+
"epoch": 1230.41,
|
| 295122 |
+
"learning_rate": 7.531518578352182e-06,
|
| 295123 |
+
"loss": 1.0887,
|
| 295124 |
+
"step": 153860
|
| 295125 |
+
},
|
| 295126 |
+
{
|
| 295127 |
+
"epoch": 1230.45,
|
| 295128 |
+
"learning_rate": 7.531437802907917e-06,
|
| 295129 |
+
"loss": 0.2721,
|
| 295130 |
+
"step": 153865
|
| 295131 |
+
},
|
| 295132 |
+
{
|
| 295133 |
+
"epoch": 1230.49,
|
| 295134 |
+
"learning_rate": 7.531357027463652e-06,
|
| 295135 |
+
"loss": 0.2654,
|
| 295136 |
+
"step": 153870
|
| 295137 |
+
},
|
| 295138 |
+
{
|
| 295139 |
+
"epoch": 1230.53,
|
| 295140 |
+
"learning_rate": 7.531276252019387e-06,
|
| 295141 |
+
"loss": 0.2596,
|
| 295142 |
+
"step": 153875
|
| 295143 |
+
},
|
| 295144 |
+
{
|
| 295145 |
+
"epoch": 1230.57,
|
| 295146 |
+
"learning_rate": 7.531195476575122e-06,
|
| 295147 |
+
"loss": 0.5072,
|
| 295148 |
+
"step": 153880
|
| 295149 |
+
},
|
| 295150 |
+
{
|
| 295151 |
+
"epoch": 1230.61,
|
| 295152 |
+
"learning_rate": 7.531114701130857e-06,
|
| 295153 |
+
"loss": 1.0273,
|
| 295154 |
+
"step": 153885
|
| 295155 |
+
},
|
| 295156 |
+
{
|
| 295157 |
+
"epoch": 1230.65,
|
| 295158 |
+
"learning_rate": 7.531033925686592e-06,
|
| 295159 |
+
"loss": 0.3047,
|
| 295160 |
+
"step": 153890
|
| 295161 |
+
},
|
| 295162 |
+
{
|
| 295163 |
+
"epoch": 1230.69,
|
| 295164 |
+
"learning_rate": 7.530953150242327e-06,
|
| 295165 |
+
"loss": 0.2919,
|
| 295166 |
+
"step": 153895
|
| 295167 |
+
},
|
| 295168 |
+
{
|
| 295169 |
+
"epoch": 1230.73,
|
| 295170 |
+
"learning_rate": 7.530872374798062e-06,
|
| 295171 |
+
"loss": 0.3372,
|
| 295172 |
+
"step": 153900
|
| 295173 |
+
},
|
| 295174 |
+
{
|
| 295175 |
+
"epoch": 1230.77,
|
| 295176 |
+
"learning_rate": 7.530791599353797e-06,
|
| 295177 |
+
"loss": 0.4598,
|
| 295178 |
+
"step": 153905
|
| 295179 |
+
},
|
| 295180 |
+
{
|
| 295181 |
+
"epoch": 1230.81,
|
| 295182 |
+
"learning_rate": 7.530710823909532e-06,
|
| 295183 |
+
"loss": 1.2106,
|
| 295184 |
+
"step": 153910
|
| 295185 |
+
},
|
| 295186 |
+
{
|
| 295187 |
+
"epoch": 1230.85,
|
| 295188 |
+
"learning_rate": 7.530630048465267e-06,
|
| 295189 |
+
"loss": 0.3627,
|
| 295190 |
+
"step": 153915
|
| 295191 |
+
},
|
| 295192 |
+
{
|
| 295193 |
+
"epoch": 1230.89,
|
| 295194 |
+
"learning_rate": 7.530549273021002e-06,
|
| 295195 |
+
"loss": 0.2599,
|
| 295196 |
+
"step": 153920
|
| 295197 |
+
},
|
| 295198 |
+
{
|
| 295199 |
+
"epoch": 1230.93,
|
| 295200 |
+
"learning_rate": 7.530468497576738e-06,
|
| 295201 |
+
"loss": 0.2899,
|
| 295202 |
+
"step": 153925
|
| 295203 |
+
},
|
| 295204 |
+
{
|
| 295205 |
+
"epoch": 1230.97,
|
| 295206 |
+
"learning_rate": 7.530387722132473e-06,
|
| 295207 |
+
"loss": 0.4721,
|
| 295208 |
+
"step": 153930
|
| 295209 |
+
},
|
| 295210 |
+
{
|
| 295211 |
+
"epoch": 1231.0,
|
| 295212 |
+
"eval_loss": 0.3832179605960846,
|
| 295213 |
+
"eval_runtime": 42.084,
|
| 295214 |
+
"eval_samples_per_second": 20.079,
|
| 295215 |
+
"eval_steps_per_second": 0.642,
|
| 295216 |
+
"eval_wer": 0.17948529411764705,
|
| 295217 |
+
"step": 153934
|
| 295218 |
}
|
| 295219 |
],
|
| 295220 |
+
"max_steps": 625000,
|
| 295221 |
"num_train_epochs": 5000,
|
| 295222 |
+
"total_flos": 4.331791340637063e+20,
|
| 295223 |
"trial_name": null,
|
| 295224 |
"trial_params": null
|
| 295225 |
}
|
model-bin/finetune/base/{checkpoint-153312 → checkpoint-153934}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630255201.5100508/events.out.tfevents.1630255201.cc93b136ebf5.1086.229
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3c705e91bc7083cacdbb6e3ea2d47851baed071b6f5c8e4577088faed2362e2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630255633.58935/events.out.tfevents.1630255633.cc93b136ebf5.1086.231
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b63fccc5857491b99b5739495ed2be5087930f24a10b687fbf40415f783d523b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630256177.9503975/events.out.tfevents.1630256177.cc93b136ebf5.1086.233
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a630c947a6f333b8473ee3f23138bc64d6785f9d00986278b06a30f8020b218
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630256616.2655816/events.out.tfevents.1630256616.cc93b136ebf5.1086.235
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f50a9bac3409292271e60f77ab2e8fd4ac623bb8b7a62fc7b0b2b25a90957a55
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630257046.964922/events.out.tfevents.1630257046.cc93b136ebf5.1086.237
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd08c75684eb37b19f78400983c8e53fa797ee0de9523400f2e5ef29575155e4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630255201.cc93b136ebf5.1086.228
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:782bdf8f3cfb090f13eab643e64ba25bab0b9d06fcc958f9f11eae9c08acc88c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630255633.cc93b136ebf5.1086.230
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abbb3352bfa1783fb894e721926ff0819978e7ef7567481e0f4e8ca1934056bf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630256177.cc93b136ebf5.1086.232
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:42cb727f7eb722f684c8f8c3d603d4a89ef672d7fa8e4dfd675ffd8227a59de7
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630256616.cc93b136ebf5.1086.234
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e2b66ef6a795218838308e9f89ad071f07bd164a9141a9f4ea4bf195f96b68e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630257046.cc93b136ebf5.1086.236
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7d4dc274a35f05f7eb0c9f9e0ec2c3f3b907459fdfc1f66f38050fc947f4a3b
|
| 3 |
+
size 8622
|