"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629739443.170523/events.out.tfevents.1629739443.74272264b15c.932.183 +3 -0
- model-bin/finetune/base/log/1629740075.5392973/events.out.tfevents.1629740075.74272264b15c.932.185 +3 -0
- model-bin/finetune/base/log/1629740711.1643536/events.out.tfevents.1629740711.74272264b15c.932.187 +3 -0
- model-bin/finetune/base/log/1629741346.2360308/events.out.tfevents.1629741346.74272264b15c.932.189 +3 -0
- model-bin/finetune/base/log/1629741982.7596643/events.out.tfevents.1629741982.74272264b15c.932.191 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629739443.74272264b15c.932.182 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629740075.74272264b15c.932.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629740711.74272264b15c.932.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629741346.74272264b15c.932.188 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629741982.74272264b15c.932.190 +3 -0
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:959ce62729227bed55822bda8043de7724c3e5c158eb50b761917a9abfa7be57
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/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 sha256:6a461917b442f70797596af1555f318276458e7e17d444d349f44961efe363f1
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/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:e36128eac80a2569e908c8e07343e4c8d0eabf3932ed0ad3ce6ddff419c12340
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/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 sha256:65dadf708547d192dd1a1f85344e17c1a7a89640d06852cd8566bd64ca2db817
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/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:b6011d3763d3a4d577b6b20622f290fc53b7b5fe40a920457f7d6a696ba3dd22
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -162384,11 +162384,800 @@
|
|
| 162384 |
"eval_steps_per_second": 0.689,
|
| 162385 |
"eval_wer": 0.19971912188631827,
|
| 162386 |
"step": 49530
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 162387 |
}
|
| 162388 |
],
|
| 162389 |
"max_steps": 620000,
|
| 162390 |
"num_train_epochs": 5000,
|
| 162391 |
-
"total_flos": 1.
|
| 162392 |
"trial_name": null,
|
| 162393 |
"trial_params": null
|
| 162394 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 403.99598393574297,
|
| 5 |
+
"global_step": 50151,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 162384 |
"eval_steps_per_second": 0.689,
|
| 162385 |
"eval_wer": 0.19971912188631827,
|
| 162386 |
"step": 49530
|
| 162387 |
+
},
|
| 162388 |
+
{
|
| 162389 |
+
"epoch": 399.04,
|
| 162390 |
+
"learning_rate": 9.222483974358976e-06,
|
| 162391 |
+
"loss": 0.3401,
|
| 162392 |
+
"step": 49535
|
| 162393 |
+
},
|
| 162394 |
+
{
|
| 162395 |
+
"epoch": 399.08,
|
| 162396 |
+
"learning_rate": 9.222403846153847e-06,
|
| 162397 |
+
"loss": 0.2955,
|
| 162398 |
+
"step": 49540
|
| 162399 |
+
},
|
| 162400 |
+
{
|
| 162401 |
+
"epoch": 399.12,
|
| 162402 |
+
"learning_rate": 9.222323717948719e-06,
|
| 162403 |
+
"loss": 0.4354,
|
| 162404 |
+
"step": 49545
|
| 162405 |
+
},
|
| 162406 |
+
{
|
| 162407 |
+
"epoch": 399.16,
|
| 162408 |
+
"learning_rate": 9.22224358974359e-06,
|
| 162409 |
+
"loss": 0.5642,
|
| 162410 |
+
"step": 49550
|
| 162411 |
+
},
|
| 162412 |
+
{
|
| 162413 |
+
"epoch": 399.2,
|
| 162414 |
+
"learning_rate": 9.222163461538463e-06,
|
| 162415 |
+
"loss": 1.2955,
|
| 162416 |
+
"step": 49555
|
| 162417 |
+
},
|
| 162418 |
+
{
|
| 162419 |
+
"epoch": 399.24,
|
| 162420 |
+
"learning_rate": 9.222083333333334e-06,
|
| 162421 |
+
"loss": 0.431,
|
| 162422 |
+
"step": 49560
|
| 162423 |
+
},
|
| 162424 |
+
{
|
| 162425 |
+
"epoch": 399.28,
|
| 162426 |
+
"learning_rate": 9.222003205128206e-06,
|
| 162427 |
+
"loss": 0.3745,
|
| 162428 |
+
"step": 49565
|
| 162429 |
+
},
|
| 162430 |
+
{
|
| 162431 |
+
"epoch": 399.32,
|
| 162432 |
+
"learning_rate": 9.221923076923077e-06,
|
| 162433 |
+
"loss": 0.2961,
|
| 162434 |
+
"step": 49570
|
| 162435 |
+
},
|
| 162436 |
+
{
|
| 162437 |
+
"epoch": 399.36,
|
| 162438 |
+
"learning_rate": 9.22184294871795e-06,
|
| 162439 |
+
"loss": 0.5477,
|
| 162440 |
+
"step": 49575
|
| 162441 |
+
},
|
| 162442 |
+
{
|
| 162443 |
+
"epoch": 399.4,
|
| 162444 |
+
"learning_rate": 9.221762820512821e-06,
|
| 162445 |
+
"loss": 1.3279,
|
| 162446 |
+
"step": 49580
|
| 162447 |
+
},
|
| 162448 |
+
{
|
| 162449 |
+
"epoch": 399.44,
|
| 162450 |
+
"learning_rate": 9.221682692307693e-06,
|
| 162451 |
+
"loss": 0.4807,
|
| 162452 |
+
"step": 49585
|
| 162453 |
+
},
|
| 162454 |
+
{
|
| 162455 |
+
"epoch": 399.48,
|
| 162456 |
+
"learning_rate": 9.221602564102566e-06,
|
| 162457 |
+
"loss": 0.3058,
|
| 162458 |
+
"step": 49590
|
| 162459 |
+
},
|
| 162460 |
+
{
|
| 162461 |
+
"epoch": 399.52,
|
| 162462 |
+
"learning_rate": 9.221522435897437e-06,
|
| 162463 |
+
"loss": 0.4265,
|
| 162464 |
+
"step": 49595
|
| 162465 |
+
},
|
| 162466 |
+
{
|
| 162467 |
+
"epoch": 399.56,
|
| 162468 |
+
"learning_rate": 9.221442307692309e-06,
|
| 162469 |
+
"loss": 0.4743,
|
| 162470 |
+
"step": 49600
|
| 162471 |
+
},
|
| 162472 |
+
{
|
| 162473 |
+
"epoch": 399.6,
|
| 162474 |
+
"learning_rate": 9.22136217948718e-06,
|
| 162475 |
+
"loss": 1.4961,
|
| 162476 |
+
"step": 49605
|
| 162477 |
+
},
|
| 162478 |
+
{
|
| 162479 |
+
"epoch": 399.65,
|
| 162480 |
+
"learning_rate": 9.221282051282053e-06,
|
| 162481 |
+
"loss": 0.4391,
|
| 162482 |
+
"step": 49610
|
| 162483 |
+
},
|
| 162484 |
+
{
|
| 162485 |
+
"epoch": 399.69,
|
| 162486 |
+
"learning_rate": 9.221201923076923e-06,
|
| 162487 |
+
"loss": 0.3347,
|
| 162488 |
+
"step": 49615
|
| 162489 |
+
},
|
| 162490 |
+
{
|
| 162491 |
+
"epoch": 399.73,
|
| 162492 |
+
"learning_rate": 9.221121794871796e-06,
|
| 162493 |
+
"loss": 0.3622,
|
| 162494 |
+
"step": 49620
|
| 162495 |
+
},
|
| 162496 |
+
{
|
| 162497 |
+
"epoch": 399.77,
|
| 162498 |
+
"learning_rate": 9.221041666666667e-06,
|
| 162499 |
+
"loss": 0.5091,
|
| 162500 |
+
"step": 49625
|
| 162501 |
+
},
|
| 162502 |
+
{
|
| 162503 |
+
"epoch": 399.81,
|
| 162504 |
+
"learning_rate": 9.220961538461538e-06,
|
| 162505 |
+
"loss": 1.2901,
|
| 162506 |
+
"step": 49630
|
| 162507 |
+
},
|
| 162508 |
+
{
|
| 162509 |
+
"epoch": 399.85,
|
| 162510 |
+
"learning_rate": 9.220881410256411e-06,
|
| 162511 |
+
"loss": 0.3537,
|
| 162512 |
+
"step": 49635
|
| 162513 |
+
},
|
| 162514 |
+
{
|
| 162515 |
+
"epoch": 399.89,
|
| 162516 |
+
"learning_rate": 9.220801282051283e-06,
|
| 162517 |
+
"loss": 0.4095,
|
| 162518 |
+
"step": 49640
|
| 162519 |
+
},
|
| 162520 |
+
{
|
| 162521 |
+
"epoch": 399.93,
|
| 162522 |
+
"learning_rate": 9.220721153846154e-06,
|
| 162523 |
+
"loss": 0.4307,
|
| 162524 |
+
"step": 49645
|
| 162525 |
+
},
|
| 162526 |
+
{
|
| 162527 |
+
"epoch": 399.97,
|
| 162528 |
+
"learning_rate": 9.220641025641026e-06,
|
| 162529 |
+
"loss": 0.5198,
|
| 162530 |
+
"step": 49650
|
| 162531 |
+
},
|
| 162532 |
+
{
|
| 162533 |
+
"epoch": 400.0,
|
| 162534 |
+
"eval_loss": 0.43548643589019775,
|
| 162535 |
+
"eval_runtime": 40.3971,
|
| 162536 |
+
"eval_samples_per_second": 20.769,
|
| 162537 |
+
"eval_steps_per_second": 0.668,
|
| 162538 |
+
"eval_wer": 0.21171731966312707,
|
| 162539 |
+
"step": 49654
|
| 162540 |
+
},
|
| 162541 |
+
{
|
| 162542 |
+
"epoch": 400.01,
|
| 162543 |
+
"learning_rate": 9.220560897435899e-06,
|
| 162544 |
+
"loss": 0.696,
|
| 162545 |
+
"step": 49655
|
| 162546 |
+
},
|
| 162547 |
+
{
|
| 162548 |
+
"epoch": 400.05,
|
| 162549 |
+
"learning_rate": 9.22048076923077e-06,
|
| 162550 |
+
"loss": 0.3643,
|
| 162551 |
+
"step": 49660
|
| 162552 |
+
},
|
| 162553 |
+
{
|
| 162554 |
+
"epoch": 400.09,
|
| 162555 |
+
"learning_rate": 9.220400641025641e-06,
|
| 162556 |
+
"loss": 0.3304,
|
| 162557 |
+
"step": 49665
|
| 162558 |
+
},
|
| 162559 |
+
{
|
| 162560 |
+
"epoch": 400.13,
|
| 162561 |
+
"learning_rate": 9.220320512820513e-06,
|
| 162562 |
+
"loss": 0.3745,
|
| 162563 |
+
"step": 49670
|
| 162564 |
+
},
|
| 162565 |
+
{
|
| 162566 |
+
"epoch": 400.17,
|
| 162567 |
+
"learning_rate": 9.220240384615386e-06,
|
| 162568 |
+
"loss": 0.5855,
|
| 162569 |
+
"step": 49675
|
| 162570 |
+
},
|
| 162571 |
+
{
|
| 162572 |
+
"epoch": 400.21,
|
| 162573 |
+
"learning_rate": 9.220160256410257e-06,
|
| 162574 |
+
"loss": 1.3573,
|
| 162575 |
+
"step": 49680
|
| 162576 |
+
},
|
| 162577 |
+
{
|
| 162578 |
+
"epoch": 400.25,
|
| 162579 |
+
"learning_rate": 9.220080128205128e-06,
|
| 162580 |
+
"loss": 0.3411,
|
| 162581 |
+
"step": 49685
|
| 162582 |
+
},
|
| 162583 |
+
{
|
| 162584 |
+
"epoch": 400.29,
|
| 162585 |
+
"learning_rate": 9.220000000000002e-06,
|
| 162586 |
+
"loss": 0.4071,
|
| 162587 |
+
"step": 49690
|
| 162588 |
+
},
|
| 162589 |
+
{
|
| 162590 |
+
"epoch": 400.33,
|
| 162591 |
+
"learning_rate": 9.219919871794873e-06,
|
| 162592 |
+
"loss": 0.3437,
|
| 162593 |
+
"step": 49695
|
| 162594 |
+
},
|
| 162595 |
+
{
|
| 162596 |
+
"epoch": 400.37,
|
| 162597 |
+
"learning_rate": 9.219839743589744e-06,
|
| 162598 |
+
"loss": 0.6714,
|
| 162599 |
+
"step": 49700
|
| 162600 |
+
},
|
| 162601 |
+
{
|
| 162602 |
+
"epoch": 400.41,
|
| 162603 |
+
"learning_rate": 9.219759615384616e-06,
|
| 162604 |
+
"loss": 1.2815,
|
| 162605 |
+
"step": 49705
|
| 162606 |
+
},
|
| 162607 |
+
{
|
| 162608 |
+
"epoch": 400.45,
|
| 162609 |
+
"learning_rate": 9.219695512820513e-06,
|
| 162610 |
+
"loss": 1.3999,
|
| 162611 |
+
"step": 49710
|
| 162612 |
+
},
|
| 162613 |
+
{
|
| 162614 |
+
"epoch": 400.49,
|
| 162615 |
+
"learning_rate": 9.219615384615386e-06,
|
| 162616 |
+
"loss": 0.4497,
|
| 162617 |
+
"step": 49715
|
| 162618 |
+
},
|
| 162619 |
+
{
|
| 162620 |
+
"epoch": 400.53,
|
| 162621 |
+
"learning_rate": 9.219535256410257e-06,
|
| 162622 |
+
"loss": 0.4104,
|
| 162623 |
+
"step": 49720
|
| 162624 |
+
},
|
| 162625 |
+
{
|
| 162626 |
+
"epoch": 400.57,
|
| 162627 |
+
"learning_rate": 9.219455128205129e-06,
|
| 162628 |
+
"loss": 0.5813,
|
| 162629 |
+
"step": 49725
|
| 162630 |
+
},
|
| 162631 |
+
{
|
| 162632 |
+
"epoch": 400.61,
|
| 162633 |
+
"learning_rate": 9.219375e-06,
|
| 162634 |
+
"loss": 1.224,
|
| 162635 |
+
"step": 49730
|
| 162636 |
+
},
|
| 162637 |
+
{
|
| 162638 |
+
"epoch": 400.65,
|
| 162639 |
+
"learning_rate": 9.219294871794873e-06,
|
| 162640 |
+
"loss": 0.3109,
|
| 162641 |
+
"step": 49735
|
| 162642 |
+
},
|
| 162643 |
+
{
|
| 162644 |
+
"epoch": 400.69,
|
| 162645 |
+
"learning_rate": 9.219214743589745e-06,
|
| 162646 |
+
"loss": 0.3625,
|
| 162647 |
+
"step": 49740
|
| 162648 |
+
},
|
| 162649 |
+
{
|
| 162650 |
+
"epoch": 400.73,
|
| 162651 |
+
"learning_rate": 9.219134615384616e-06,
|
| 162652 |
+
"loss": 0.3834,
|
| 162653 |
+
"step": 49745
|
| 162654 |
+
},
|
| 162655 |
+
{
|
| 162656 |
+
"epoch": 400.77,
|
| 162657 |
+
"learning_rate": 9.219054487179489e-06,
|
| 162658 |
+
"loss": 0.5324,
|
| 162659 |
+
"step": 49750
|
| 162660 |
+
},
|
| 162661 |
+
{
|
| 162662 |
+
"epoch": 400.81,
|
| 162663 |
+
"learning_rate": 9.21897435897436e-06,
|
| 162664 |
+
"loss": 1.2051,
|
| 162665 |
+
"step": 49755
|
| 162666 |
+
},
|
| 162667 |
+
{
|
| 162668 |
+
"epoch": 400.85,
|
| 162669 |
+
"learning_rate": 9.218894230769232e-06,
|
| 162670 |
+
"loss": 0.3304,
|
| 162671 |
+
"step": 49760
|
| 162672 |
+
},
|
| 162673 |
+
{
|
| 162674 |
+
"epoch": 400.9,
|
| 162675 |
+
"learning_rate": 9.218814102564103e-06,
|
| 162676 |
+
"loss": 0.3533,
|
| 162677 |
+
"step": 49765
|
| 162678 |
+
},
|
| 162679 |
+
{
|
| 162680 |
+
"epoch": 400.94,
|
| 162681 |
+
"learning_rate": 9.218733974358976e-06,
|
| 162682 |
+
"loss": 0.3579,
|
| 162683 |
+
"step": 49770
|
| 162684 |
+
},
|
| 162685 |
+
{
|
| 162686 |
+
"epoch": 400.98,
|
| 162687 |
+
"learning_rate": 9.218653846153847e-06,
|
| 162688 |
+
"loss": 0.7775,
|
| 162689 |
+
"step": 49775
|
| 162690 |
+
},
|
| 162691 |
+
{
|
| 162692 |
+
"epoch": 401.0,
|
| 162693 |
+
"eval_loss": 0.45663779973983765,
|
| 162694 |
+
"eval_runtime": 40.3567,
|
| 162695 |
+
"eval_samples_per_second": 20.79,
|
| 162696 |
+
"eval_steps_per_second": 0.669,
|
| 162697 |
+
"eval_wer": 0.20139291932675565,
|
| 162698 |
+
"step": 49778
|
| 162699 |
+
},
|
| 162700 |
+
{
|
| 162701 |
+
"epoch": 398.02,
|
| 162702 |
+
"learning_rate": 9.218573717948719e-06,
|
| 162703 |
+
"loss": 0.4261,
|
| 162704 |
+
"step": 49780
|
| 162705 |
+
},
|
| 162706 |
+
{
|
| 162707 |
+
"epoch": 398.06,
|
| 162708 |
+
"learning_rate": 9.21849358974359e-06,
|
| 162709 |
+
"loss": 0.3171,
|
| 162710 |
+
"step": 49785
|
| 162711 |
+
},
|
| 162712 |
+
{
|
| 162713 |
+
"epoch": 398.1,
|
| 162714 |
+
"learning_rate": 9.218413461538463e-06,
|
| 162715 |
+
"loss": 0.4089,
|
| 162716 |
+
"step": 49790
|
| 162717 |
+
},
|
| 162718 |
+
{
|
| 162719 |
+
"epoch": 398.14,
|
| 162720 |
+
"learning_rate": 9.218333333333333e-06,
|
| 162721 |
+
"loss": 0.3868,
|
| 162722 |
+
"step": 49795
|
| 162723 |
+
},
|
| 162724 |
+
{
|
| 162725 |
+
"epoch": 398.18,
|
| 162726 |
+
"learning_rate": 9.218253205128206e-06,
|
| 162727 |
+
"loss": 0.6769,
|
| 162728 |
+
"step": 49800
|
| 162729 |
+
},
|
| 162730 |
+
{
|
| 162731 |
+
"epoch": 398.22,
|
| 162732 |
+
"learning_rate": 9.218173076923079e-06,
|
| 162733 |
+
"loss": 1.0966,
|
| 162734 |
+
"step": 49805
|
| 162735 |
+
},
|
| 162736 |
+
{
|
| 162737 |
+
"epoch": 398.25,
|
| 162738 |
+
"learning_rate": 9.218092948717949e-06,
|
| 162739 |
+
"loss": 0.3772,
|
| 162740 |
+
"step": 49810
|
| 162741 |
+
},
|
| 162742 |
+
{
|
| 162743 |
+
"epoch": 398.29,
|
| 162744 |
+
"learning_rate": 9.218012820512822e-06,
|
| 162745 |
+
"loss": 0.3179,
|
| 162746 |
+
"step": 49815
|
| 162747 |
+
},
|
| 162748 |
+
{
|
| 162749 |
+
"epoch": 398.33,
|
| 162750 |
+
"learning_rate": 9.217932692307693e-06,
|
| 162751 |
+
"loss": 0.4016,
|
| 162752 |
+
"step": 49820
|
| 162753 |
+
},
|
| 162754 |
+
{
|
| 162755 |
+
"epoch": 398.37,
|
| 162756 |
+
"learning_rate": 9.217852564102564e-06,
|
| 162757 |
+
"loss": 0.6761,
|
| 162758 |
+
"step": 49825
|
| 162759 |
+
},
|
| 162760 |
+
{
|
| 162761 |
+
"epoch": 398.41,
|
| 162762 |
+
"learning_rate": 9.217772435897436e-06,
|
| 162763 |
+
"loss": 1.0311,
|
| 162764 |
+
"step": 49830
|
| 162765 |
+
},
|
| 162766 |
+
{
|
| 162767 |
+
"epoch": 398.45,
|
| 162768 |
+
"learning_rate": 9.217692307692309e-06,
|
| 162769 |
+
"loss": 0.4023,
|
| 162770 |
+
"step": 49835
|
| 162771 |
+
},
|
| 162772 |
+
{
|
| 162773 |
+
"epoch": 398.49,
|
| 162774 |
+
"learning_rate": 9.21761217948718e-06,
|
| 162775 |
+
"loss": 0.2977,
|
| 162776 |
+
"step": 49840
|
| 162777 |
+
},
|
| 162778 |
+
{
|
| 162779 |
+
"epoch": 398.53,
|
| 162780 |
+
"learning_rate": 9.217532051282052e-06,
|
| 162781 |
+
"loss": 0.3801,
|
| 162782 |
+
"step": 49845
|
| 162783 |
+
},
|
| 162784 |
+
{
|
| 162785 |
+
"epoch": 398.57,
|
| 162786 |
+
"learning_rate": 9.217451923076925e-06,
|
| 162787 |
+
"loss": 0.7346,
|
| 162788 |
+
"step": 49850
|
| 162789 |
+
},
|
| 162790 |
+
{
|
| 162791 |
+
"epoch": 398.61,
|
| 162792 |
+
"learning_rate": 9.217371794871796e-06,
|
| 162793 |
+
"loss": 1.0575,
|
| 162794 |
+
"step": 49855
|
| 162795 |
+
},
|
| 162796 |
+
{
|
| 162797 |
+
"epoch": 398.65,
|
| 162798 |
+
"learning_rate": 9.217291666666667e-06,
|
| 162799 |
+
"loss": 0.289,
|
| 162800 |
+
"step": 49860
|
| 162801 |
+
},
|
| 162802 |
+
{
|
| 162803 |
+
"epoch": 398.69,
|
| 162804 |
+
"learning_rate": 9.217211538461539e-06,
|
| 162805 |
+
"loss": 0.3293,
|
| 162806 |
+
"step": 49865
|
| 162807 |
+
},
|
| 162808 |
+
{
|
| 162809 |
+
"epoch": 398.73,
|
| 162810 |
+
"learning_rate": 9.217131410256412e-06,
|
| 162811 |
+
"loss": 0.4006,
|
| 162812 |
+
"step": 49870
|
| 162813 |
+
},
|
| 162814 |
+
{
|
| 162815 |
+
"epoch": 398.77,
|
| 162816 |
+
"learning_rate": 9.217051282051283e-06,
|
| 162817 |
+
"loss": 0.6412,
|
| 162818 |
+
"step": 49875
|
| 162819 |
+
},
|
| 162820 |
+
{
|
| 162821 |
+
"epoch": 398.81,
|
| 162822 |
+
"learning_rate": 9.216971153846154e-06,
|
| 162823 |
+
"loss": 1.2047,
|
| 162824 |
+
"step": 49880
|
| 162825 |
+
},
|
| 162826 |
+
{
|
| 162827 |
+
"epoch": 398.85,
|
| 162828 |
+
"learning_rate": 9.216891025641026e-06,
|
| 162829 |
+
"loss": 0.4079,
|
| 162830 |
+
"step": 49885
|
| 162831 |
+
},
|
| 162832 |
+
{
|
| 162833 |
+
"epoch": 398.89,
|
| 162834 |
+
"learning_rate": 9.216810897435899e-06,
|
| 162835 |
+
"loss": 0.3128,
|
| 162836 |
+
"step": 49890
|
| 162837 |
+
},
|
| 162838 |
+
{
|
| 162839 |
+
"epoch": 398.93,
|
| 162840 |
+
"learning_rate": 9.21673076923077e-06,
|
| 162841 |
+
"loss": 0.4184,
|
| 162842 |
+
"step": 49895
|
| 162843 |
+
},
|
| 162844 |
+
{
|
| 162845 |
+
"epoch": 398.97,
|
| 162846 |
+
"learning_rate": 9.216650641025642e-06,
|
| 162847 |
+
"loss": 0.7058,
|
| 162848 |
+
"step": 49900
|
| 162849 |
+
},
|
| 162850 |
+
{
|
| 162851 |
+
"epoch": 399.0,
|
| 162852 |
+
"eval_loss": 0.4314434826374054,
|
| 162853 |
+
"eval_runtime": 39.886,
|
| 162854 |
+
"eval_samples_per_second": 21.035,
|
| 162855 |
+
"eval_steps_per_second": 0.677,
|
| 162856 |
+
"eval_wer": 0.20628127946616379,
|
| 162857 |
+
"step": 49903
|
| 162858 |
+
},
|
| 162859 |
+
{
|
| 162860 |
+
"epoch": 402.02,
|
| 162861 |
+
"learning_rate": 9.216570512820515e-06,
|
| 162862 |
+
"loss": 0.3856,
|
| 162863 |
+
"step": 49905
|
| 162864 |
+
},
|
| 162865 |
+
{
|
| 162866 |
+
"epoch": 402.06,
|
| 162867 |
+
"learning_rate": 9.216490384615386e-06,
|
| 162868 |
+
"loss": 0.3282,
|
| 162869 |
+
"step": 49910
|
| 162870 |
+
},
|
| 162871 |
+
{
|
| 162872 |
+
"epoch": 402.1,
|
| 162873 |
+
"learning_rate": 9.216410256410257e-06,
|
| 162874 |
+
"loss": 0.3658,
|
| 162875 |
+
"step": 49915
|
| 162876 |
+
},
|
| 162877 |
+
{
|
| 162878 |
+
"epoch": 402.14,
|
| 162879 |
+
"learning_rate": 9.216330128205129e-06,
|
| 162880 |
+
"loss": 0.4127,
|
| 162881 |
+
"step": 49920
|
| 162882 |
+
},
|
| 162883 |
+
{
|
| 162884 |
+
"epoch": 402.18,
|
| 162885 |
+
"learning_rate": 9.216250000000002e-06,
|
| 162886 |
+
"loss": 0.7727,
|
| 162887 |
+
"step": 49925
|
| 162888 |
+
},
|
| 162889 |
+
{
|
| 162890 |
+
"epoch": 402.22,
|
| 162891 |
+
"learning_rate": 9.216169871794871e-06,
|
| 162892 |
+
"loss": 1.1278,
|
| 162893 |
+
"step": 49930
|
| 162894 |
+
},
|
| 162895 |
+
{
|
| 162896 |
+
"epoch": 402.26,
|
| 162897 |
+
"learning_rate": 9.216089743589744e-06,
|
| 162898 |
+
"loss": 0.3621,
|
| 162899 |
+
"step": 49935
|
| 162900 |
+
},
|
| 162901 |
+
{
|
| 162902 |
+
"epoch": 402.3,
|
| 162903 |
+
"learning_rate": 9.216009615384616e-06,
|
| 162904 |
+
"loss": 0.3462,
|
| 162905 |
+
"step": 49940
|
| 162906 |
+
},
|
| 162907 |
+
{
|
| 162908 |
+
"epoch": 402.34,
|
| 162909 |
+
"learning_rate": 9.215929487179487e-06,
|
| 162910 |
+
"loss": 0.4675,
|
| 162911 |
+
"step": 49945
|
| 162912 |
+
},
|
| 162913 |
+
{
|
| 162914 |
+
"epoch": 402.38,
|
| 162915 |
+
"learning_rate": 9.21584935897436e-06,
|
| 162916 |
+
"loss": 0.794,
|
| 162917 |
+
"step": 49950
|
| 162918 |
+
},
|
| 162919 |
+
{
|
| 162920 |
+
"epoch": 402.42,
|
| 162921 |
+
"learning_rate": 9.215769230769232e-06,
|
| 162922 |
+
"loss": 1.0722,
|
| 162923 |
+
"step": 49955
|
| 162924 |
+
},
|
| 162925 |
+
{
|
| 162926 |
+
"epoch": 402.46,
|
| 162927 |
+
"learning_rate": 9.215689102564103e-06,
|
| 162928 |
+
"loss": 0.344,
|
| 162929 |
+
"step": 49960
|
| 162930 |
+
},
|
| 162931 |
+
{
|
| 162932 |
+
"epoch": 402.5,
|
| 162933 |
+
"learning_rate": 9.215608974358974e-06,
|
| 162934 |
+
"loss": 0.3991,
|
| 162935 |
+
"step": 49965
|
| 162936 |
+
},
|
| 162937 |
+
{
|
| 162938 |
+
"epoch": 402.54,
|
| 162939 |
+
"learning_rate": 9.215528846153847e-06,
|
| 162940 |
+
"loss": 0.3788,
|
| 162941 |
+
"step": 49970
|
| 162942 |
+
},
|
| 162943 |
+
{
|
| 162944 |
+
"epoch": 402.58,
|
| 162945 |
+
"learning_rate": 9.215448717948719e-06,
|
| 162946 |
+
"loss": 0.7787,
|
| 162947 |
+
"step": 49975
|
| 162948 |
+
},
|
| 162949 |
+
{
|
| 162950 |
+
"epoch": 402.62,
|
| 162951 |
+
"learning_rate": 9.21536858974359e-06,
|
| 162952 |
+
"loss": 1.2898,
|
| 162953 |
+
"step": 49980
|
| 162954 |
+
},
|
| 162955 |
+
{
|
| 162956 |
+
"epoch": 402.66,
|
| 162957 |
+
"learning_rate": 9.215288461538461e-06,
|
| 162958 |
+
"loss": 0.3289,
|
| 162959 |
+
"step": 49985
|
| 162960 |
+
},
|
| 162961 |
+
{
|
| 162962 |
+
"epoch": 402.7,
|
| 162963 |
+
"learning_rate": 9.215208333333335e-06,
|
| 162964 |
+
"loss": 0.4183,
|
| 162965 |
+
"step": 49990
|
| 162966 |
+
},
|
| 162967 |
+
{
|
| 162968 |
+
"epoch": 402.74,
|
| 162969 |
+
"learning_rate": 9.215128205128206e-06,
|
| 162970 |
+
"loss": 0.3565,
|
| 162971 |
+
"step": 49995
|
| 162972 |
+
},
|
| 162973 |
+
{
|
| 162974 |
+
"epoch": 402.78,
|
| 162975 |
+
"learning_rate": 9.215048076923077e-06,
|
| 162976 |
+
"loss": 0.6545,
|
| 162977 |
+
"step": 50000
|
| 162978 |
+
},
|
| 162979 |
+
{
|
| 162980 |
+
"epoch": 402.82,
|
| 162981 |
+
"learning_rate": 9.21496794871795e-06,
|
| 162982 |
+
"loss": 1.1079,
|
| 162983 |
+
"step": 50005
|
| 162984 |
+
},
|
| 162985 |
+
{
|
| 162986 |
+
"epoch": 402.86,
|
| 162987 |
+
"learning_rate": 9.214887820512822e-06,
|
| 162988 |
+
"loss": 0.3451,
|
| 162989 |
+
"step": 50010
|
| 162990 |
+
},
|
| 162991 |
+
{
|
| 162992 |
+
"epoch": 402.9,
|
| 162993 |
+
"learning_rate": 9.214807692307693e-06,
|
| 162994 |
+
"loss": 0.3267,
|
| 162995 |
+
"step": 50015
|
| 162996 |
+
},
|
| 162997 |
+
{
|
| 162998 |
+
"epoch": 402.94,
|
| 162999 |
+
"learning_rate": 9.214727564102564e-06,
|
| 163000 |
+
"loss": 0.4015,
|
| 163001 |
+
"step": 50020
|
| 163002 |
+
},
|
| 163003 |
+
{
|
| 163004 |
+
"epoch": 402.98,
|
| 163005 |
+
"learning_rate": 9.214647435897437e-06,
|
| 163006 |
+
"loss": 0.8181,
|
| 163007 |
+
"step": 50025
|
| 163008 |
+
},
|
| 163009 |
+
{
|
| 163010 |
+
"epoch": 403.0,
|
| 163011 |
+
"eval_loss": 0.3694767951965332,
|
| 163012 |
+
"eval_runtime": 40.8543,
|
| 163013 |
+
"eval_samples_per_second": 20.536,
|
| 163014 |
+
"eval_steps_per_second": 0.661,
|
| 163015 |
+
"eval_wer": 0.20543960439240783,
|
| 163016 |
+
"step": 50027
|
| 163017 |
+
},
|
| 163018 |
+
{
|
| 163019 |
+
"epoch": 403.02,
|
| 163020 |
+
"learning_rate": 9.214567307692309e-06,
|
| 163021 |
+
"loss": 0.3587,
|
| 163022 |
+
"step": 50030
|
| 163023 |
+
},
|
| 163024 |
+
{
|
| 163025 |
+
"epoch": 403.06,
|
| 163026 |
+
"learning_rate": 9.21448717948718e-06,
|
| 163027 |
+
"loss": 0.358,
|
| 163028 |
+
"step": 50035
|
| 163029 |
+
},
|
| 163030 |
+
{
|
| 163031 |
+
"epoch": 403.1,
|
| 163032 |
+
"learning_rate": 9.214407051282051e-06,
|
| 163033 |
+
"loss": 0.335,
|
| 163034 |
+
"step": 50040
|
| 163035 |
+
},
|
| 163036 |
+
{
|
| 163037 |
+
"epoch": 403.14,
|
| 163038 |
+
"learning_rate": 9.214326923076925e-06,
|
| 163039 |
+
"loss": 0.3689,
|
| 163040 |
+
"step": 50045
|
| 163041 |
+
},
|
| 163042 |
+
{
|
| 163043 |
+
"epoch": 403.18,
|
| 163044 |
+
"learning_rate": 9.214246794871796e-06,
|
| 163045 |
+
"loss": 0.777,
|
| 163046 |
+
"step": 50050
|
| 163047 |
+
},
|
| 163048 |
+
{
|
| 163049 |
+
"epoch": 403.22,
|
| 163050 |
+
"learning_rate": 9.214166666666667e-06,
|
| 163051 |
+
"loss": 0.8766,
|
| 163052 |
+
"step": 50055
|
| 163053 |
+
},
|
| 163054 |
+
{
|
| 163055 |
+
"epoch": 403.27,
|
| 163056 |
+
"learning_rate": 9.21408653846154e-06,
|
| 163057 |
+
"loss": 0.3625,
|
| 163058 |
+
"step": 50060
|
| 163059 |
+
},
|
| 163060 |
+
{
|
| 163061 |
+
"epoch": 403.31,
|
| 163062 |
+
"learning_rate": 9.214006410256412e-06,
|
| 163063 |
+
"loss": 0.3834,
|
| 163064 |
+
"step": 50065
|
| 163065 |
+
},
|
| 163066 |
+
{
|
| 163067 |
+
"epoch": 403.35,
|
| 163068 |
+
"learning_rate": 9.213926282051283e-06,
|
| 163069 |
+
"loss": 0.4218,
|
| 163070 |
+
"step": 50070
|
| 163071 |
+
},
|
| 163072 |
+
{
|
| 163073 |
+
"epoch": 403.39,
|
| 163074 |
+
"learning_rate": 9.213846153846154e-06,
|
| 163075 |
+
"loss": 1.0129,
|
| 163076 |
+
"step": 50075
|
| 163077 |
+
},
|
| 163078 |
+
{
|
| 163079 |
+
"epoch": 403.43,
|
| 163080 |
+
"learning_rate": 9.213766025641027e-06,
|
| 163081 |
+
"loss": 0.8793,
|
| 163082 |
+
"step": 50080
|
| 163083 |
+
},
|
| 163084 |
+
{
|
| 163085 |
+
"epoch": 403.47,
|
| 163086 |
+
"learning_rate": 9.213685897435897e-06,
|
| 163087 |
+
"loss": 0.4697,
|
| 163088 |
+
"step": 50085
|
| 163089 |
+
},
|
| 163090 |
+
{
|
| 163091 |
+
"epoch": 403.51,
|
| 163092 |
+
"learning_rate": 9.21360576923077e-06,
|
| 163093 |
+
"loss": 0.3166,
|
| 163094 |
+
"step": 50090
|
| 163095 |
+
},
|
| 163096 |
+
{
|
| 163097 |
+
"epoch": 403.55,
|
| 163098 |
+
"learning_rate": 9.213525641025642e-06,
|
| 163099 |
+
"loss": 0.4842,
|
| 163100 |
+
"step": 50095
|
| 163101 |
+
},
|
| 163102 |
+
{
|
| 163103 |
+
"epoch": 403.59,
|
| 163104 |
+
"learning_rate": 9.213445512820513e-06,
|
| 163105 |
+
"loss": 0.7842,
|
| 163106 |
+
"step": 50100
|
| 163107 |
+
},
|
| 163108 |
+
{
|
| 163109 |
+
"epoch": 403.63,
|
| 163110 |
+
"learning_rate": 9.213365384615386e-06,
|
| 163111 |
+
"loss": 0.928,
|
| 163112 |
+
"step": 50105
|
| 163113 |
+
},
|
| 163114 |
+
{
|
| 163115 |
+
"epoch": 403.67,
|
| 163116 |
+
"learning_rate": 9.213285256410257e-06,
|
| 163117 |
+
"loss": 0.3819,
|
| 163118 |
+
"step": 50110
|
| 163119 |
+
},
|
| 163120 |
+
{
|
| 163121 |
+
"epoch": 403.71,
|
| 163122 |
+
"learning_rate": 9.213205128205129e-06,
|
| 163123 |
+
"loss": 0.3871,
|
| 163124 |
+
"step": 50115
|
| 163125 |
+
},
|
| 163126 |
+
{
|
| 163127 |
+
"epoch": 403.75,
|
| 163128 |
+
"learning_rate": 9.213125e-06,
|
| 163129 |
+
"loss": 0.5323,
|
| 163130 |
+
"step": 50120
|
| 163131 |
+
},
|
| 163132 |
+
{
|
| 163133 |
+
"epoch": 403.79,
|
| 163134 |
+
"learning_rate": 9.213044871794873e-06,
|
| 163135 |
+
"loss": 0.825,
|
| 163136 |
+
"step": 50125
|
| 163137 |
+
},
|
| 163138 |
+
{
|
| 163139 |
+
"epoch": 403.83,
|
| 163140 |
+
"learning_rate": 9.212964743589744e-06,
|
| 163141 |
+
"loss": 0.9118,
|
| 163142 |
+
"step": 50130
|
| 163143 |
+
},
|
| 163144 |
+
{
|
| 163145 |
+
"epoch": 403.87,
|
| 163146 |
+
"learning_rate": 9.212884615384616e-06,
|
| 163147 |
+
"loss": 0.3248,
|
| 163148 |
+
"step": 50135
|
| 163149 |
+
},
|
| 163150 |
+
{
|
| 163151 |
+
"epoch": 403.91,
|
| 163152 |
+
"learning_rate": 9.212804487179487e-06,
|
| 163153 |
+
"loss": 0.3549,
|
| 163154 |
+
"step": 50140
|
| 163155 |
+
},
|
| 163156 |
+
{
|
| 163157 |
+
"epoch": 403.95,
|
| 163158 |
+
"learning_rate": 9.21272435897436e-06,
|
| 163159 |
+
"loss": 0.3968,
|
| 163160 |
+
"step": 50145
|
| 163161 |
+
},
|
| 163162 |
+
{
|
| 163163 |
+
"epoch": 403.99,
|
| 163164 |
+
"learning_rate": 9.212644230769232e-06,
|
| 163165 |
+
"loss": 0.9281,
|
| 163166 |
+
"step": 50150
|
| 163167 |
+
},
|
| 163168 |
+
{
|
| 163169 |
+
"epoch": 404.0,
|
| 163170 |
+
"eval_loss": 0.42465662956237793,
|
| 163171 |
+
"eval_runtime": 39.9556,
|
| 163172 |
+
"eval_samples_per_second": 21.023,
|
| 163173 |
+
"eval_steps_per_second": 0.676,
|
| 163174 |
+
"eval_wer": 0.2024080463989428,
|
| 163175 |
+
"step": 50151
|
| 163176 |
}
|
| 163177 |
],
|
| 163178 |
"max_steps": 620000,
|
| 163179 |
"num_train_epochs": 5000,
|
| 163180 |
+
"total_flos": 1.4111387032874315e+20,
|
| 163181 |
"trial_name": null,
|
| 163182 |
"trial_params": null
|
| 163183 |
}
|
model-bin/finetune/base/{checkpoint-49530 → checkpoint-50151}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629739443.170523/events.out.tfevents.1629739443.74272264b15c.932.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05614eb8b16fee60b725c40faf15d89c81b1692dcfb630b0e8ccaa59f3002563
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629740075.5392973/events.out.tfevents.1629740075.74272264b15c.932.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db5b2ec03ca5c04c382e539caba6df5c92a095e64987efdde3bd12bbb0082c71
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629740711.1643536/events.out.tfevents.1629740711.74272264b15c.932.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:deafe982df91982bd4bf2777889d67f14e611cd54f84d90eee9aff760da28f17
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629741346.2360308/events.out.tfevents.1629741346.74272264b15c.932.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fbfabc8af4985357ab68ce92edbfa6cb43e1508c4077d47849a2f855b9900595
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629741982.7596643/events.out.tfevents.1629741982.74272264b15c.932.191
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb34af6d054f7518c8dfa444fb93b520f61f46399855ce053b2812d332199977
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629739443.74272264b15c.932.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22bf85d81fb13298ebe4bbeeeaa1b178b88d9bc2c95369be72ee972a62d41379
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629740075.74272264b15c.932.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f798284ffd604ed3dd7e039c63332e089a39f40591769b416f24209811e4ac0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629740711.74272264b15c.932.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5e9e0f1617165e3111f1c3da411c90483f4034eb0027e90f2e580f746624c31
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629741346.74272264b15c.932.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7cac169c2a24ba1edd324a8bdd71713f622fcf8c2cade09a614222c8a70cedb2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629741982.74272264b15c.932.190
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:603c48cf6fd19d1aabdb928ea1f91fa54c59eb0649b8bea3d8f8b476208258c8
|
| 3 |
+
size 8622
|