"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630142682.9499917/events.out.tfevents.1630142682.86bb0ddabf9b.4092.101 +3 -0
- model-bin/finetune/base/log/1630143090.1370292/events.out.tfevents.1630143090.86bb0ddabf9b.4092.103 +3 -0
- model-bin/finetune/base/log/1630143484.717932/events.out.tfevents.1630143484.86bb0ddabf9b.4092.105 +3 -0
- model-bin/finetune/base/log/1630143880.581071/events.out.tfevents.1630143880.86bb0ddabf9b.4092.107 +3 -0
- model-bin/finetune/base/log/1630144272.0782309/events.out.tfevents.1630144272.86bb0ddabf9b.4092.109 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630142682.86bb0ddabf9b.4092.100 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630143090.86bb0ddabf9b.4092.102 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630143484.86bb0ddabf9b.4092.104 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630143880.86bb0ddabf9b.4092.106 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630144272.86bb0ddabf9b.4092.108 +3 -0
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/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:a7c4a3231088e372b641bcbecefb9d8f7b78420423afc4a419e40f17dac6b497
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/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 sha256:193548cc6244643fe0875e36a0b274952b57c6ff3787884ea6e9340b59be33d1
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/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 sha256:5fe129e917b4c2af29771c1725b443c2df423d938257d23f158bd186a2a90ae4
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/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:d391da4854332be30a64a76369bfbbbfa856719cae3d860e11c28d7890a695af
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/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:d851a34b78ff2a34ac3461ac6fc45d251f315b586c0be20f8792ee63881fd06e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1735723097017633,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-123203",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -257388,11 +257388,800 @@
|
|
| 257388 |
"eval_steps_per_second": 0.708,
|
| 257389 |
"eval_wer": 0.17809599194070663,
|
| 257390 |
"step": 124200
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 257391 |
}
|
| 257392 |
],
|
| 257393 |
-
"max_steps":
|
| 257394 |
"num_train_epochs": 5000,
|
| 257395 |
-
"total_flos": 3.
|
| 257396 |
"trial_name": null,
|
| 257397 |
"trial_params": null
|
| 257398 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1735723097017633,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-123203",
|
| 4 |
+
"epoch": 998.0,
|
| 5 |
+
"global_step": 124823,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 257388 |
"eval_steps_per_second": 0.708,
|
| 257389 |
"eval_wer": 0.17809599194070663,
|
| 257390 |
"step": 124200
|
| 257391 |
+
},
|
| 257392 |
+
{
|
| 257393 |
+
"epoch": 1001.04,
|
| 257394 |
+
"learning_rate": 8.026346153846154e-06,
|
| 257395 |
+
"loss": 0.3085,
|
| 257396 |
+
"step": 124205
|
| 257397 |
+
},
|
| 257398 |
+
{
|
| 257399 |
+
"epoch": 1001.08,
|
| 257400 |
+
"learning_rate": 8.026266025641026e-06,
|
| 257401 |
+
"loss": 0.2722,
|
| 257402 |
+
"step": 124210
|
| 257403 |
+
},
|
| 257404 |
+
{
|
| 257405 |
+
"epoch": 1001.12,
|
| 257406 |
+
"learning_rate": 8.026185897435899e-06,
|
| 257407 |
+
"loss": 0.2798,
|
| 257408 |
+
"step": 124215
|
| 257409 |
+
},
|
| 257410 |
+
{
|
| 257411 |
+
"epoch": 1001.16,
|
| 257412 |
+
"learning_rate": 8.02610576923077e-06,
|
| 257413 |
+
"loss": 0.4422,
|
| 257414 |
+
"step": 124220
|
| 257415 |
+
},
|
| 257416 |
+
{
|
| 257417 |
+
"epoch": 1001.2,
|
| 257418 |
+
"learning_rate": 8.026025641025641e-06,
|
| 257419 |
+
"loss": 1.1585,
|
| 257420 |
+
"step": 124225
|
| 257421 |
+
},
|
| 257422 |
+
{
|
| 257423 |
+
"epoch": 1001.24,
|
| 257424 |
+
"learning_rate": 8.025945512820513e-06,
|
| 257425 |
+
"loss": 0.3022,
|
| 257426 |
+
"step": 124230
|
| 257427 |
+
},
|
| 257428 |
+
{
|
| 257429 |
+
"epoch": 1001.28,
|
| 257430 |
+
"learning_rate": 8.025865384615386e-06,
|
| 257431 |
+
"loss": 0.3218,
|
| 257432 |
+
"step": 124235
|
| 257433 |
+
},
|
| 257434 |
+
{
|
| 257435 |
+
"epoch": 1001.32,
|
| 257436 |
+
"learning_rate": 8.025785256410257e-06,
|
| 257437 |
+
"loss": 0.3082,
|
| 257438 |
+
"step": 124240
|
| 257439 |
+
},
|
| 257440 |
+
{
|
| 257441 |
+
"epoch": 1001.36,
|
| 257442 |
+
"learning_rate": 8.025705128205129e-06,
|
| 257443 |
+
"loss": 0.4458,
|
| 257444 |
+
"step": 124245
|
| 257445 |
+
},
|
| 257446 |
+
{
|
| 257447 |
+
"epoch": 1001.4,
|
| 257448 |
+
"learning_rate": 8.025625e-06,
|
| 257449 |
+
"loss": 1.2388,
|
| 257450 |
+
"step": 124250
|
| 257451 |
+
},
|
| 257452 |
+
{
|
| 257453 |
+
"epoch": 1001.44,
|
| 257454 |
+
"learning_rate": 8.025544871794873e-06,
|
| 257455 |
+
"loss": 0.3478,
|
| 257456 |
+
"step": 124255
|
| 257457 |
+
},
|
| 257458 |
+
{
|
| 257459 |
+
"epoch": 1001.48,
|
| 257460 |
+
"learning_rate": 8.025464743589744e-06,
|
| 257461 |
+
"loss": 0.2347,
|
| 257462 |
+
"step": 124260
|
| 257463 |
+
},
|
| 257464 |
+
{
|
| 257465 |
+
"epoch": 1001.52,
|
| 257466 |
+
"learning_rate": 8.025384615384616e-06,
|
| 257467 |
+
"loss": 0.3652,
|
| 257468 |
+
"step": 124265
|
| 257469 |
+
},
|
| 257470 |
+
{
|
| 257471 |
+
"epoch": 1001.56,
|
| 257472 |
+
"learning_rate": 8.025304487179489e-06,
|
| 257473 |
+
"loss": 0.3937,
|
| 257474 |
+
"step": 124270
|
| 257475 |
+
},
|
| 257476 |
+
{
|
| 257477 |
+
"epoch": 1001.6,
|
| 257478 |
+
"learning_rate": 8.02522435897436e-06,
|
| 257479 |
+
"loss": 1.2448,
|
| 257480 |
+
"step": 124275
|
| 257481 |
+
},
|
| 257482 |
+
{
|
| 257483 |
+
"epoch": 1001.64,
|
| 257484 |
+
"learning_rate": 8.025144230769231e-06,
|
| 257485 |
+
"loss": 0.3001,
|
| 257486 |
+
"step": 124280
|
| 257487 |
+
},
|
| 257488 |
+
{
|
| 257489 |
+
"epoch": 1001.68,
|
| 257490 |
+
"learning_rate": 8.025064102564103e-06,
|
| 257491 |
+
"loss": 0.2776,
|
| 257492 |
+
"step": 124285
|
| 257493 |
+
},
|
| 257494 |
+
{
|
| 257495 |
+
"epoch": 1001.72,
|
| 257496 |
+
"learning_rate": 8.024983974358976e-06,
|
| 257497 |
+
"loss": 0.3356,
|
| 257498 |
+
"step": 124290
|
| 257499 |
+
},
|
| 257500 |
+
{
|
| 257501 |
+
"epoch": 1001.76,
|
| 257502 |
+
"learning_rate": 8.024903846153847e-06,
|
| 257503 |
+
"loss": 0.4398,
|
| 257504 |
+
"step": 124295
|
| 257505 |
+
},
|
| 257506 |
+
{
|
| 257507 |
+
"epoch": 1001.8,
|
| 257508 |
+
"learning_rate": 8.024823717948719e-06,
|
| 257509 |
+
"loss": 1.1453,
|
| 257510 |
+
"step": 124300
|
| 257511 |
+
},
|
| 257512 |
+
{
|
| 257513 |
+
"epoch": 1001.84,
|
| 257514 |
+
"learning_rate": 8.02474358974359e-06,
|
| 257515 |
+
"loss": 0.2821,
|
| 257516 |
+
"step": 124305
|
| 257517 |
+
},
|
| 257518 |
+
{
|
| 257519 |
+
"epoch": 1001.88,
|
| 257520 |
+
"learning_rate": 8.024663461538463e-06,
|
| 257521 |
+
"loss": 0.2459,
|
| 257522 |
+
"step": 124310
|
| 257523 |
+
},
|
| 257524 |
+
{
|
| 257525 |
+
"epoch": 1001.92,
|
| 257526 |
+
"learning_rate": 8.024583333333334e-06,
|
| 257527 |
+
"loss": 0.2939,
|
| 257528 |
+
"step": 124315
|
| 257529 |
+
},
|
| 257530 |
+
{
|
| 257531 |
+
"epoch": 1001.96,
|
| 257532 |
+
"learning_rate": 8.024503205128206e-06,
|
| 257533 |
+
"loss": 0.4506,
|
| 257534 |
+
"step": 124320
|
| 257535 |
+
},
|
| 257536 |
+
{
|
| 257537 |
+
"epoch": 1002.0,
|
| 257538 |
+
"eval_loss": 0.3501935303211212,
|
| 257539 |
+
"eval_runtime": 37.1608,
|
| 257540 |
+
"eval_samples_per_second": 22.604,
|
| 257541 |
+
"eval_steps_per_second": 0.727,
|
| 257542 |
+
"eval_wer": 0.18019187765995817,
|
| 257543 |
+
"step": 124324
|
| 257544 |
+
},
|
| 257545 |
+
{
|
| 257546 |
+
"epoch": 994.01,
|
| 257547 |
+
"learning_rate": 8.024423076923079e-06,
|
| 257548 |
+
"loss": 0.3624,
|
| 257549 |
+
"step": 124325
|
| 257550 |
+
},
|
| 257551 |
+
{
|
| 257552 |
+
"epoch": 994.05,
|
| 257553 |
+
"learning_rate": 8.024342948717948e-06,
|
| 257554 |
+
"loss": 0.3273,
|
| 257555 |
+
"step": 124330
|
| 257556 |
+
},
|
| 257557 |
+
{
|
| 257558 |
+
"epoch": 994.09,
|
| 257559 |
+
"learning_rate": 8.024262820512821e-06,
|
| 257560 |
+
"loss": 0.2553,
|
| 257561 |
+
"step": 124335
|
| 257562 |
+
},
|
| 257563 |
+
{
|
| 257564 |
+
"epoch": 994.13,
|
| 257565 |
+
"learning_rate": 8.024182692307693e-06,
|
| 257566 |
+
"loss": 0.3694,
|
| 257567 |
+
"step": 124340
|
| 257568 |
+
},
|
| 257569 |
+
{
|
| 257570 |
+
"epoch": 994.17,
|
| 257571 |
+
"learning_rate": 8.024102564102564e-06,
|
| 257572 |
+
"loss": 0.5139,
|
| 257573 |
+
"step": 124345
|
| 257574 |
+
},
|
| 257575 |
+
{
|
| 257576 |
+
"epoch": 994.21,
|
| 257577 |
+
"learning_rate": 8.024022435897436e-06,
|
| 257578 |
+
"loss": 1.2288,
|
| 257579 |
+
"step": 124350
|
| 257580 |
+
},
|
| 257581 |
+
{
|
| 257582 |
+
"epoch": 994.25,
|
| 257583 |
+
"learning_rate": 8.023942307692309e-06,
|
| 257584 |
+
"loss": 0.3068,
|
| 257585 |
+
"step": 124355
|
| 257586 |
+
},
|
| 257587 |
+
{
|
| 257588 |
+
"epoch": 994.29,
|
| 257589 |
+
"learning_rate": 8.02386217948718e-06,
|
| 257590 |
+
"loss": 0.3395,
|
| 257591 |
+
"step": 124360
|
| 257592 |
+
},
|
| 257593 |
+
{
|
| 257594 |
+
"epoch": 994.33,
|
| 257595 |
+
"learning_rate": 8.023782051282051e-06,
|
| 257596 |
+
"loss": 0.295,
|
| 257597 |
+
"step": 124365
|
| 257598 |
+
},
|
| 257599 |
+
{
|
| 257600 |
+
"epoch": 994.37,
|
| 257601 |
+
"learning_rate": 8.023701923076924e-06,
|
| 257602 |
+
"loss": 0.4902,
|
| 257603 |
+
"step": 124370
|
| 257604 |
+
},
|
| 257605 |
+
{
|
| 257606 |
+
"epoch": 994.41,
|
| 257607 |
+
"learning_rate": 8.023621794871796e-06,
|
| 257608 |
+
"loss": 1.1012,
|
| 257609 |
+
"step": 124375
|
| 257610 |
+
},
|
| 257611 |
+
{
|
| 257612 |
+
"epoch": 994.45,
|
| 257613 |
+
"learning_rate": 8.023541666666667e-06,
|
| 257614 |
+
"loss": 0.3274,
|
| 257615 |
+
"step": 124380
|
| 257616 |
+
},
|
| 257617 |
+
{
|
| 257618 |
+
"epoch": 994.49,
|
| 257619 |
+
"learning_rate": 8.023461538461538e-06,
|
| 257620 |
+
"loss": 0.3741,
|
| 257621 |
+
"step": 124385
|
| 257622 |
+
},
|
| 257623 |
+
{
|
| 257624 |
+
"epoch": 994.53,
|
| 257625 |
+
"learning_rate": 8.023381410256412e-06,
|
| 257626 |
+
"loss": 0.3158,
|
| 257627 |
+
"step": 124390
|
| 257628 |
+
},
|
| 257629 |
+
{
|
| 257630 |
+
"epoch": 994.57,
|
| 257631 |
+
"learning_rate": 8.023301282051283e-06,
|
| 257632 |
+
"loss": 0.6262,
|
| 257633 |
+
"step": 124395
|
| 257634 |
+
},
|
| 257635 |
+
{
|
| 257636 |
+
"epoch": 994.61,
|
| 257637 |
+
"learning_rate": 8.023221153846154e-06,
|
| 257638 |
+
"loss": 1.1513,
|
| 257639 |
+
"step": 124400
|
| 257640 |
+
},
|
| 257641 |
+
{
|
| 257642 |
+
"epoch": 994.65,
|
| 257643 |
+
"learning_rate": 8.023141025641026e-06,
|
| 257644 |
+
"loss": 0.3037,
|
| 257645 |
+
"step": 124405
|
| 257646 |
+
},
|
| 257647 |
+
{
|
| 257648 |
+
"epoch": 994.69,
|
| 257649 |
+
"learning_rate": 8.023060897435899e-06,
|
| 257650 |
+
"loss": 0.2719,
|
| 257651 |
+
"step": 124410
|
| 257652 |
+
},
|
| 257653 |
+
{
|
| 257654 |
+
"epoch": 994.73,
|
| 257655 |
+
"learning_rate": 8.02298076923077e-06,
|
| 257656 |
+
"loss": 0.3865,
|
| 257657 |
+
"step": 124415
|
| 257658 |
+
},
|
| 257659 |
+
{
|
| 257660 |
+
"epoch": 994.77,
|
| 257661 |
+
"learning_rate": 8.022900641025641e-06,
|
| 257662 |
+
"loss": 0.5366,
|
| 257663 |
+
"step": 124420
|
| 257664 |
+
},
|
| 257665 |
+
{
|
| 257666 |
+
"epoch": 994.81,
|
| 257667 |
+
"learning_rate": 8.022820512820514e-06,
|
| 257668 |
+
"loss": 1.0406,
|
| 257669 |
+
"step": 124425
|
| 257670 |
+
},
|
| 257671 |
+
{
|
| 257672 |
+
"epoch": 994.85,
|
| 257673 |
+
"learning_rate": 8.022740384615386e-06,
|
| 257674 |
+
"loss": 0.3121,
|
| 257675 |
+
"step": 124430
|
| 257676 |
+
},
|
| 257677 |
+
{
|
| 257678 |
+
"epoch": 994.89,
|
| 257679 |
+
"learning_rate": 8.022660256410257e-06,
|
| 257680 |
+
"loss": 0.4277,
|
| 257681 |
+
"step": 124435
|
| 257682 |
+
},
|
| 257683 |
+
{
|
| 257684 |
+
"epoch": 994.93,
|
| 257685 |
+
"learning_rate": 8.022580128205128e-06,
|
| 257686 |
+
"loss": 0.2867,
|
| 257687 |
+
"step": 124440
|
| 257688 |
+
},
|
| 257689 |
+
{
|
| 257690 |
+
"epoch": 994.97,
|
| 257691 |
+
"learning_rate": 8.022500000000002e-06,
|
| 257692 |
+
"loss": 0.5523,
|
| 257693 |
+
"step": 124445
|
| 257694 |
+
},
|
| 257695 |
+
{
|
| 257696 |
+
"epoch": 995.0,
|
| 257697 |
+
"eval_loss": 0.37554535269737244,
|
| 257698 |
+
"eval_runtime": 37.1448,
|
| 257699 |
+
"eval_samples_per_second": 22.614,
|
| 257700 |
+
"eval_steps_per_second": 0.727,
|
| 257701 |
+
"eval_wer": 0.178762469278589,
|
| 257702 |
+
"step": 124449
|
| 257703 |
+
},
|
| 257704 |
+
{
|
| 257705 |
+
"epoch": 995.01,
|
| 257706 |
+
"learning_rate": 8.022419871794871e-06,
|
| 257707 |
+
"loss": 0.6121,
|
| 257708 |
+
"step": 124450
|
| 257709 |
+
},
|
| 257710 |
+
{
|
| 257711 |
+
"epoch": 995.05,
|
| 257712 |
+
"learning_rate": 8.022339743589744e-06,
|
| 257713 |
+
"loss": 0.2873,
|
| 257714 |
+
"step": 124455
|
| 257715 |
+
},
|
| 257716 |
+
{
|
| 257717 |
+
"epoch": 995.09,
|
| 257718 |
+
"learning_rate": 8.022259615384616e-06,
|
| 257719 |
+
"loss": 0.299,
|
| 257720 |
+
"step": 124460
|
| 257721 |
+
},
|
| 257722 |
+
{
|
| 257723 |
+
"epoch": 995.13,
|
| 257724 |
+
"learning_rate": 8.022179487179487e-06,
|
| 257725 |
+
"loss": 0.2944,
|
| 257726 |
+
"step": 124465
|
| 257727 |
+
},
|
| 257728 |
+
{
|
| 257729 |
+
"epoch": 995.17,
|
| 257730 |
+
"learning_rate": 8.02209935897436e-06,
|
| 257731 |
+
"loss": 0.5003,
|
| 257732 |
+
"step": 124470
|
| 257733 |
+
},
|
| 257734 |
+
{
|
| 257735 |
+
"epoch": 995.21,
|
| 257736 |
+
"learning_rate": 8.022019230769231e-06,
|
| 257737 |
+
"loss": 1.2021,
|
| 257738 |
+
"step": 124475
|
| 257739 |
+
},
|
| 257740 |
+
{
|
| 257741 |
+
"epoch": 995.25,
|
| 257742 |
+
"learning_rate": 8.021939102564103e-06,
|
| 257743 |
+
"loss": 0.2656,
|
| 257744 |
+
"step": 124480
|
| 257745 |
+
},
|
| 257746 |
+
{
|
| 257747 |
+
"epoch": 995.29,
|
| 257748 |
+
"learning_rate": 8.021858974358974e-06,
|
| 257749 |
+
"loss": 0.2815,
|
| 257750 |
+
"step": 124485
|
| 257751 |
+
},
|
| 257752 |
+
{
|
| 257753 |
+
"epoch": 995.33,
|
| 257754 |
+
"learning_rate": 8.021778846153847e-06,
|
| 257755 |
+
"loss": 0.2868,
|
| 257756 |
+
"step": 124490
|
| 257757 |
+
},
|
| 257758 |
+
{
|
| 257759 |
+
"epoch": 995.37,
|
| 257760 |
+
"learning_rate": 8.021698717948719e-06,
|
| 257761 |
+
"loss": 0.4953,
|
| 257762 |
+
"step": 124495
|
| 257763 |
+
},
|
| 257764 |
+
{
|
| 257765 |
+
"epoch": 995.41,
|
| 257766 |
+
"learning_rate": 8.02161858974359e-06,
|
| 257767 |
+
"loss": 1.0536,
|
| 257768 |
+
"step": 124500
|
| 257769 |
+
},
|
| 257770 |
+
{
|
| 257771 |
+
"epoch": 995.45,
|
| 257772 |
+
"learning_rate": 8.021538461538461e-06,
|
| 257773 |
+
"loss": 0.2525,
|
| 257774 |
+
"step": 124505
|
| 257775 |
+
},
|
| 257776 |
+
{
|
| 257777 |
+
"epoch": 995.49,
|
| 257778 |
+
"learning_rate": 8.021458333333334e-06,
|
| 257779 |
+
"loss": 0.281,
|
| 257780 |
+
"step": 124510
|
| 257781 |
+
},
|
| 257782 |
+
{
|
| 257783 |
+
"epoch": 995.53,
|
| 257784 |
+
"learning_rate": 8.021378205128206e-06,
|
| 257785 |
+
"loss": 0.3611,
|
| 257786 |
+
"step": 124515
|
| 257787 |
+
},
|
| 257788 |
+
{
|
| 257789 |
+
"epoch": 995.57,
|
| 257790 |
+
"learning_rate": 8.021298076923077e-06,
|
| 257791 |
+
"loss": 0.4902,
|
| 257792 |
+
"step": 124520
|
| 257793 |
+
},
|
| 257794 |
+
{
|
| 257795 |
+
"epoch": 995.61,
|
| 257796 |
+
"learning_rate": 8.02121794871795e-06,
|
| 257797 |
+
"loss": 1.1837,
|
| 257798 |
+
"step": 124525
|
| 257799 |
+
},
|
| 257800 |
+
{
|
| 257801 |
+
"epoch": 995.65,
|
| 257802 |
+
"learning_rate": 8.021137820512821e-06,
|
| 257803 |
+
"loss": 0.3066,
|
| 257804 |
+
"step": 124530
|
| 257805 |
+
},
|
| 257806 |
+
{
|
| 257807 |
+
"epoch": 995.69,
|
| 257808 |
+
"learning_rate": 8.021057692307693e-06,
|
| 257809 |
+
"loss": 0.3343,
|
| 257810 |
+
"step": 124535
|
| 257811 |
+
},
|
| 257812 |
+
{
|
| 257813 |
+
"epoch": 995.73,
|
| 257814 |
+
"learning_rate": 8.020977564102564e-06,
|
| 257815 |
+
"loss": 0.3023,
|
| 257816 |
+
"step": 124540
|
| 257817 |
+
},
|
| 257818 |
+
{
|
| 257819 |
+
"epoch": 995.77,
|
| 257820 |
+
"learning_rate": 8.020897435897437e-06,
|
| 257821 |
+
"loss": 0.519,
|
| 257822 |
+
"step": 124545
|
| 257823 |
+
},
|
| 257824 |
+
{
|
| 257825 |
+
"epoch": 995.81,
|
| 257826 |
+
"learning_rate": 8.020817307692309e-06,
|
| 257827 |
+
"loss": 1.0876,
|
| 257828 |
+
"step": 124550
|
| 257829 |
+
},
|
| 257830 |
+
{
|
| 257831 |
+
"epoch": 995.85,
|
| 257832 |
+
"learning_rate": 8.02073717948718e-06,
|
| 257833 |
+
"loss": 0.2318,
|
| 257834 |
+
"step": 124555
|
| 257835 |
+
},
|
| 257836 |
+
{
|
| 257837 |
+
"epoch": 995.89,
|
| 257838 |
+
"learning_rate": 8.020657051282051e-06,
|
| 257839 |
+
"loss": 0.3073,
|
| 257840 |
+
"step": 124560
|
| 257841 |
+
},
|
| 257842 |
+
{
|
| 257843 |
+
"epoch": 995.93,
|
| 257844 |
+
"learning_rate": 8.020576923076924e-06,
|
| 257845 |
+
"loss": 0.3889,
|
| 257846 |
+
"step": 124565
|
| 257847 |
+
},
|
| 257848 |
+
{
|
| 257849 |
+
"epoch": 995.97,
|
| 257850 |
+
"learning_rate": 8.020496794871796e-06,
|
| 257851 |
+
"loss": 0.4879,
|
| 257852 |
+
"step": 124570
|
| 257853 |
+
},
|
| 257854 |
+
{
|
| 257855 |
+
"epoch": 996.0,
|
| 257856 |
+
"eval_loss": 0.44444289803504944,
|
| 257857 |
+
"eval_runtime": 36.0337,
|
| 257858 |
+
"eval_samples_per_second": 23.284,
|
| 257859 |
+
"eval_steps_per_second": 0.749,
|
| 257860 |
+
"eval_wer": 0.17907788254510357,
|
| 257861 |
+
"step": 124574
|
| 257862 |
+
},
|
| 257863 |
+
{
|
| 257864 |
+
"epoch": 1004.01,
|
| 257865 |
+
"learning_rate": 8.020416666666667e-06,
|
| 257866 |
+
"loss": 0.2902,
|
| 257867 |
+
"step": 124575
|
| 257868 |
+
},
|
| 257869 |
+
{
|
| 257870 |
+
"epoch": 1004.05,
|
| 257871 |
+
"learning_rate": 8.02033653846154e-06,
|
| 257872 |
+
"loss": 0.3478,
|
| 257873 |
+
"step": 124580
|
| 257874 |
+
},
|
| 257875 |
+
{
|
| 257876 |
+
"epoch": 1004.09,
|
| 257877 |
+
"learning_rate": 8.020256410256411e-06,
|
| 257878 |
+
"loss": 0.3836,
|
| 257879 |
+
"step": 124585
|
| 257880 |
+
},
|
| 257881 |
+
{
|
| 257882 |
+
"epoch": 1004.13,
|
| 257883 |
+
"learning_rate": 8.020176282051283e-06,
|
| 257884 |
+
"loss": 0.3269,
|
| 257885 |
+
"step": 124590
|
| 257886 |
+
},
|
| 257887 |
+
{
|
| 257888 |
+
"epoch": 1004.17,
|
| 257889 |
+
"learning_rate": 8.020096153846154e-06,
|
| 257890 |
+
"loss": 0.5404,
|
| 257891 |
+
"step": 124595
|
| 257892 |
+
},
|
| 257893 |
+
{
|
| 257894 |
+
"epoch": 1004.21,
|
| 257895 |
+
"learning_rate": 8.020016025641027e-06,
|
| 257896 |
+
"loss": 1.1555,
|
| 257897 |
+
"step": 124600
|
| 257898 |
+
},
|
| 257899 |
+
{
|
| 257900 |
+
"epoch": 1004.25,
|
| 257901 |
+
"learning_rate": 8.019935897435897e-06,
|
| 257902 |
+
"loss": 0.3135,
|
| 257903 |
+
"step": 124605
|
| 257904 |
+
},
|
| 257905 |
+
{
|
| 257906 |
+
"epoch": 1004.29,
|
| 257907 |
+
"learning_rate": 8.01985576923077e-06,
|
| 257908 |
+
"loss": 0.2687,
|
| 257909 |
+
"step": 124610
|
| 257910 |
+
},
|
| 257911 |
+
{
|
| 257912 |
+
"epoch": 1004.33,
|
| 257913 |
+
"learning_rate": 8.019775641025643e-06,
|
| 257914 |
+
"loss": 0.3204,
|
| 257915 |
+
"step": 124615
|
| 257916 |
+
},
|
| 257917 |
+
{
|
| 257918 |
+
"epoch": 1004.37,
|
| 257919 |
+
"learning_rate": 8.019695512820513e-06,
|
| 257920 |
+
"loss": 0.6159,
|
| 257921 |
+
"step": 124620
|
| 257922 |
+
},
|
| 257923 |
+
{
|
| 257924 |
+
"epoch": 1004.41,
|
| 257925 |
+
"learning_rate": 8.019615384615386e-06,
|
| 257926 |
+
"loss": 1.1154,
|
| 257927 |
+
"step": 124625
|
| 257928 |
+
},
|
| 257929 |
+
{
|
| 257930 |
+
"epoch": 1004.45,
|
| 257931 |
+
"learning_rate": 8.019535256410257e-06,
|
| 257932 |
+
"loss": 0.3203,
|
| 257933 |
+
"step": 124630
|
| 257934 |
+
},
|
| 257935 |
+
{
|
| 257936 |
+
"epoch": 1004.49,
|
| 257937 |
+
"learning_rate": 8.019455128205128e-06,
|
| 257938 |
+
"loss": 0.2448,
|
| 257939 |
+
"step": 124635
|
| 257940 |
+
},
|
| 257941 |
+
{
|
| 257942 |
+
"epoch": 1004.53,
|
| 257943 |
+
"learning_rate": 8.019375e-06,
|
| 257944 |
+
"loss": 0.3757,
|
| 257945 |
+
"step": 124640
|
| 257946 |
+
},
|
| 257947 |
+
{
|
| 257948 |
+
"epoch": 1004.57,
|
| 257949 |
+
"learning_rate": 8.019294871794873e-06,
|
| 257950 |
+
"loss": 0.5318,
|
| 257951 |
+
"step": 124645
|
| 257952 |
+
},
|
| 257953 |
+
{
|
| 257954 |
+
"epoch": 1004.61,
|
| 257955 |
+
"learning_rate": 8.019214743589744e-06,
|
| 257956 |
+
"loss": 1.0301,
|
| 257957 |
+
"step": 124650
|
| 257958 |
+
},
|
| 257959 |
+
{
|
| 257960 |
+
"epoch": 1004.65,
|
| 257961 |
+
"learning_rate": 8.019134615384616e-06,
|
| 257962 |
+
"loss": 0.3138,
|
| 257963 |
+
"step": 124655
|
| 257964 |
+
},
|
| 257965 |
+
{
|
| 257966 |
+
"epoch": 1004.69,
|
| 257967 |
+
"learning_rate": 8.019054487179487e-06,
|
| 257968 |
+
"loss": 0.2722,
|
| 257969 |
+
"step": 124660
|
| 257970 |
+
},
|
| 257971 |
+
{
|
| 257972 |
+
"epoch": 1004.73,
|
| 257973 |
+
"learning_rate": 8.01897435897436e-06,
|
| 257974 |
+
"loss": 0.3035,
|
| 257975 |
+
"step": 124665
|
| 257976 |
+
},
|
| 257977 |
+
{
|
| 257978 |
+
"epoch": 1004.77,
|
| 257979 |
+
"learning_rate": 8.018894230769231e-06,
|
| 257980 |
+
"loss": 0.4899,
|
| 257981 |
+
"step": 124670
|
| 257982 |
+
},
|
| 257983 |
+
{
|
| 257984 |
+
"epoch": 1004.81,
|
| 257985 |
+
"learning_rate": 8.018814102564103e-06,
|
| 257986 |
+
"loss": 1.1714,
|
| 257987 |
+
"step": 124675
|
| 257988 |
+
},
|
| 257989 |
+
{
|
| 257990 |
+
"epoch": 1004.85,
|
| 257991 |
+
"learning_rate": 8.018733974358976e-06,
|
| 257992 |
+
"loss": 0.2925,
|
| 257993 |
+
"step": 124680
|
| 257994 |
+
},
|
| 257995 |
+
{
|
| 257996 |
+
"epoch": 1004.9,
|
| 257997 |
+
"learning_rate": 8.018653846153847e-06,
|
| 257998 |
+
"loss": 0.2628,
|
| 257999 |
+
"step": 124685
|
| 258000 |
+
},
|
| 258001 |
+
{
|
| 258002 |
+
"epoch": 1004.94,
|
| 258003 |
+
"learning_rate": 8.018573717948718e-06,
|
| 258004 |
+
"loss": 0.3557,
|
| 258005 |
+
"step": 124690
|
| 258006 |
+
},
|
| 258007 |
+
{
|
| 258008 |
+
"epoch": 1004.98,
|
| 258009 |
+
"learning_rate": 8.01849358974359e-06,
|
| 258010 |
+
"loss": 0.6265,
|
| 258011 |
+
"step": 124695
|
| 258012 |
+
},
|
| 258013 |
+
{
|
| 258014 |
+
"epoch": 1005.0,
|
| 258015 |
+
"eval_loss": 0.33506569266319275,
|
| 258016 |
+
"eval_runtime": 36.6106,
|
| 258017 |
+
"eval_samples_per_second": 22.917,
|
| 258018 |
+
"eval_steps_per_second": 0.737,
|
| 258019 |
+
"eval_wer": 0.18211167363709738,
|
| 258020 |
+
"step": 124698
|
| 258021 |
+
},
|
| 258022 |
+
{
|
| 258023 |
+
"epoch": 997.02,
|
| 258024 |
+
"learning_rate": 8.018413461538463e-06,
|
| 258025 |
+
"loss": 0.394,
|
| 258026 |
+
"step": 124700
|
| 258027 |
+
},
|
| 258028 |
+
{
|
| 258029 |
+
"epoch": 997.06,
|
| 258030 |
+
"learning_rate": 8.018333333333334e-06,
|
| 258031 |
+
"loss": 0.3194,
|
| 258032 |
+
"step": 124705
|
| 258033 |
+
},
|
| 258034 |
+
{
|
| 258035 |
+
"epoch": 997.1,
|
| 258036 |
+
"learning_rate": 8.018253205128206e-06,
|
| 258037 |
+
"loss": 0.2794,
|
| 258038 |
+
"step": 124710
|
| 258039 |
+
},
|
| 258040 |
+
{
|
| 258041 |
+
"epoch": 997.14,
|
| 258042 |
+
"learning_rate": 8.018173076923079e-06,
|
| 258043 |
+
"loss": 0.3031,
|
| 258044 |
+
"step": 124715
|
| 258045 |
+
},
|
| 258046 |
+
{
|
| 258047 |
+
"epoch": 997.18,
|
| 258048 |
+
"learning_rate": 8.01809294871795e-06,
|
| 258049 |
+
"loss": 0.5738,
|
| 258050 |
+
"step": 124720
|
| 258051 |
+
},
|
| 258052 |
+
{
|
| 258053 |
+
"epoch": 997.22,
|
| 258054 |
+
"learning_rate": 8.018012820512821e-06,
|
| 258055 |
+
"loss": 0.9485,
|
| 258056 |
+
"step": 124725
|
| 258057 |
+
},
|
| 258058 |
+
{
|
| 258059 |
+
"epoch": 997.26,
|
| 258060 |
+
"learning_rate": 8.017932692307693e-06,
|
| 258061 |
+
"loss": 0.3145,
|
| 258062 |
+
"step": 124730
|
| 258063 |
+
},
|
| 258064 |
+
{
|
| 258065 |
+
"epoch": 997.3,
|
| 258066 |
+
"learning_rate": 8.017852564102566e-06,
|
| 258067 |
+
"loss": 0.2922,
|
| 258068 |
+
"step": 124735
|
| 258069 |
+
},
|
| 258070 |
+
{
|
| 258071 |
+
"epoch": 997.34,
|
| 258072 |
+
"learning_rate": 8.017772435897435e-06,
|
| 258073 |
+
"loss": 0.3783,
|
| 258074 |
+
"step": 124740
|
| 258075 |
+
},
|
| 258076 |
+
{
|
| 258077 |
+
"epoch": 997.38,
|
| 258078 |
+
"learning_rate": 8.017692307692308e-06,
|
| 258079 |
+
"loss": 0.6634,
|
| 258080 |
+
"step": 124745
|
| 258081 |
+
},
|
| 258082 |
+
{
|
| 258083 |
+
"epoch": 997.42,
|
| 258084 |
+
"learning_rate": 8.01761217948718e-06,
|
| 258085 |
+
"loss": 0.9275,
|
| 258086 |
+
"step": 124750
|
| 258087 |
+
},
|
| 258088 |
+
{
|
| 258089 |
+
"epoch": 997.46,
|
| 258090 |
+
"learning_rate": 8.017532051282053e-06,
|
| 258091 |
+
"loss": 0.3074,
|
| 258092 |
+
"step": 124755
|
| 258093 |
+
},
|
| 258094 |
+
{
|
| 258095 |
+
"epoch": 997.5,
|
| 258096 |
+
"learning_rate": 8.017451923076923e-06,
|
| 258097 |
+
"loss": 0.2433,
|
| 258098 |
+
"step": 124760
|
| 258099 |
+
},
|
| 258100 |
+
{
|
| 258101 |
+
"epoch": 997.54,
|
| 258102 |
+
"learning_rate": 8.017371794871796e-06,
|
| 258103 |
+
"loss": 0.3525,
|
| 258104 |
+
"step": 124765
|
| 258105 |
+
},
|
| 258106 |
+
{
|
| 258107 |
+
"epoch": 997.58,
|
| 258108 |
+
"learning_rate": 8.017291666666669e-06,
|
| 258109 |
+
"loss": 0.5824,
|
| 258110 |
+
"step": 124770
|
| 258111 |
+
},
|
| 258112 |
+
{
|
| 258113 |
+
"epoch": 997.62,
|
| 258114 |
+
"learning_rate": 8.017211538461538e-06,
|
| 258115 |
+
"loss": 0.9745,
|
| 258116 |
+
"step": 124775
|
| 258117 |
+
},
|
| 258118 |
+
{
|
| 258119 |
+
"epoch": 997.66,
|
| 258120 |
+
"learning_rate": 8.017131410256411e-06,
|
| 258121 |
+
"loss": 0.2424,
|
| 258122 |
+
"step": 124780
|
| 258123 |
+
},
|
| 258124 |
+
{
|
| 258125 |
+
"epoch": 997.7,
|
| 258126 |
+
"learning_rate": 8.017051282051283e-06,
|
| 258127 |
+
"loss": 0.3636,
|
| 258128 |
+
"step": 124785
|
| 258129 |
+
},
|
| 258130 |
+
{
|
| 258131 |
+
"epoch": 997.74,
|
| 258132 |
+
"learning_rate": 8.016971153846154e-06,
|
| 258133 |
+
"loss": 0.3245,
|
| 258134 |
+
"step": 124790
|
| 258135 |
+
},
|
| 258136 |
+
{
|
| 258137 |
+
"epoch": 997.78,
|
| 258138 |
+
"learning_rate": 8.016891025641025e-06,
|
| 258139 |
+
"loss": 0.6432,
|
| 258140 |
+
"step": 124795
|
| 258141 |
+
},
|
| 258142 |
+
{
|
| 258143 |
+
"epoch": 997.82,
|
| 258144 |
+
"learning_rate": 8.016810897435898e-06,
|
| 258145 |
+
"loss": 1.0522,
|
| 258146 |
+
"step": 124800
|
| 258147 |
+
},
|
| 258148 |
+
{
|
| 258149 |
+
"epoch": 997.86,
|
| 258150 |
+
"learning_rate": 8.01673076923077e-06,
|
| 258151 |
+
"loss": 0.3268,
|
| 258152 |
+
"step": 124805
|
| 258153 |
+
},
|
| 258154 |
+
{
|
| 258155 |
+
"epoch": 997.9,
|
| 258156 |
+
"learning_rate": 8.016650641025641e-06,
|
| 258157 |
+
"loss": 0.3003,
|
| 258158 |
+
"step": 124810
|
| 258159 |
+
},
|
| 258160 |
+
{
|
| 258161 |
+
"epoch": 997.94,
|
| 258162 |
+
"learning_rate": 8.016570512820514e-06,
|
| 258163 |
+
"loss": 0.3971,
|
| 258164 |
+
"step": 124815
|
| 258165 |
+
},
|
| 258166 |
+
{
|
| 258167 |
+
"epoch": 997.98,
|
| 258168 |
+
"learning_rate": 8.016490384615386e-06,
|
| 258169 |
+
"loss": 0.8373,
|
| 258170 |
+
"step": 124820
|
| 258171 |
+
},
|
| 258172 |
+
{
|
| 258173 |
+
"epoch": 998.0,
|
| 258174 |
+
"eval_loss": 0.33471718430519104,
|
| 258175 |
+
"eval_runtime": 36.9243,
|
| 258176 |
+
"eval_samples_per_second": 22.722,
|
| 258177 |
+
"eval_steps_per_second": 0.731,
|
| 258178 |
+
"eval_wer": 0.17628693821244465,
|
| 258179 |
+
"step": 124823
|
| 258180 |
}
|
| 258181 |
],
|
| 258182 |
+
"max_steps": 625000,
|
| 258183 |
"num_train_epochs": 5000,
|
| 258184 |
+
"total_flos": 3.512665253182511e+20,
|
| 258185 |
"trial_name": null,
|
| 258186 |
"trial_params": null
|
| 258187 |
}
|
model-bin/finetune/base/{checkpoint-124200 → checkpoint-124823}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630142682.9499917/events.out.tfevents.1630142682.86bb0ddabf9b.4092.101
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f319ac7ae2ab5392d7d00d15f3297ab59cbbbde61f8372c9631341eb423e509b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630143090.1370292/events.out.tfevents.1630143090.86bb0ddabf9b.4092.103
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96d67f24f891edaea8c43021b11dcb066ef39aa23900a660406dc0a1ebc30789
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630143484.717932/events.out.tfevents.1630143484.86bb0ddabf9b.4092.105
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0eaefe4e5eed208a079ae3ff3abe458b31b7f71f2a07828ade0b82115cdd6787
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630143880.581071/events.out.tfevents.1630143880.86bb0ddabf9b.4092.107
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb41b7a0f67bccf6d2e417f78c3f3143c76d1caec9c72e30429360a4f68bdc2a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630144272.0782309/events.out.tfevents.1630144272.86bb0ddabf9b.4092.109
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6852a3bf4a5e074ca636924bfe3024fa46fedd2198833fdf9864169003d58bf9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630142682.86bb0ddabf9b.4092.100
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59e3dc53c8e5997181cc0efbd0fa3405bd99a0884444c36c005b77f007ff1385
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630143090.86bb0ddabf9b.4092.102
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b39ecb1cb5e78cfc8d8e73b4ff8b300d03798d14d03c92f43df8a014da04aa88
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630143484.86bb0ddabf9b.4092.104
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e08cb9815d5defd31de4030a1abe503a138b41bfe938b123d576eea642c365b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630143880.86bb0ddabf9b.4092.106
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:100eeb8e9f781bca8ebe3fbcfefa62ac55794bf91f5fc7b50011baa413b9f4d5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630144272.86bb0ddabf9b.4092.108
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c94aeaad7cee77a479f87513f320d9957cb957e4428cc9088cbb31fecbf7e3fd
|
| 3 |
+
size 8622
|