"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629821196.2313719/events.out.tfevents.1629821196.c435e1c5ee04.920.171 +3 -0
- model-bin/finetune/base/log/1629821845.184331/events.out.tfevents.1629821845.c435e1c5ee04.920.173 +3 -0
- model-bin/finetune/base/log/1629822489.0445116/events.out.tfevents.1629822489.c435e1c5ee04.920.175 +3 -0
- model-bin/finetune/base/log/1629823140.8357387/events.out.tfevents.1629823140.c435e1c5ee04.920.177 +3 -0
- model-bin/finetune/base/log/1629823804.588709/events.out.tfevents.1629823804.c435e1c5ee04.920.179 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629821196.c435e1c5ee04.920.170 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629821845.c435e1c5ee04.920.172 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629822489.c435e1c5ee04.920.174 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629823140.c435e1c5ee04.920.176 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629823804.c435e1c5ee04.920.178 +3 -0
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/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:cf0b925889f508e2de0a52ec0cee3966578cde3d8f800e52d2d2fb94617a4b4f
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/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 sha256:616c002c06a7cf81cad53c0251b17bf5180be425f7c89cb8b9747af4a9220572
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/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 sha256:89fad723566bb0d6b4935813ef747d3725047bb08076aa0355ba5b8d4fc8931a
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/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:ed72d3338847ccab1a388d7c3792f1105409f238a637a418c2bbf8290dcbc6b6
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/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 sha256:1e654cce9ba8671a9c638939b19ced0d02b07d9d067d5775b5da979951aae605
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -181380,11 +181380,800 @@
|
|
| 181380 |
"eval_steps_per_second": 0.671,
|
| 181381 |
"eval_wer": 0.19014445992520349,
|
| 181382 |
"step": 64461
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 181383 |
}
|
| 181384 |
],
|
| 181385 |
-
"max_steps":
|
| 181386 |
"num_train_epochs": 5000,
|
| 181387 |
-
"total_flos": 1.
|
| 181388 |
"trial_name": null,
|
| 181389 |
"trial_params": null
|
| 181390 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 524.0,
|
| 5 |
+
"global_step": 65082,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 181380 |
"eval_steps_per_second": 0.671,
|
| 181381 |
"eval_wer": 0.19014445992520349,
|
| 181382 |
"step": 64461
|
| 181383 |
+
},
|
| 181384 |
+
{
|
| 181385 |
+
"epoch": 519.03,
|
| 181386 |
+
"learning_rate": 8.983333333333334e-06,
|
| 181387 |
+
"loss": 0.3258,
|
| 181388 |
+
"step": 64465
|
| 181389 |
+
},
|
| 181390 |
+
{
|
| 181391 |
+
"epoch": 519.07,
|
| 181392 |
+
"learning_rate": 8.983253205128206e-06,
|
| 181393 |
+
"loss": 0.3882,
|
| 181394 |
+
"step": 64470
|
| 181395 |
+
},
|
| 181396 |
+
{
|
| 181397 |
+
"epoch": 519.11,
|
| 181398 |
+
"learning_rate": 8.983173076923077e-06,
|
| 181399 |
+
"loss": 0.4378,
|
| 181400 |
+
"step": 64475
|
| 181401 |
+
},
|
| 181402 |
+
{
|
| 181403 |
+
"epoch": 519.15,
|
| 181404 |
+
"learning_rate": 8.98309294871795e-06,
|
| 181405 |
+
"loss": 0.4579,
|
| 181406 |
+
"step": 64480
|
| 181407 |
+
},
|
| 181408 |
+
{
|
| 181409 |
+
"epoch": 519.19,
|
| 181410 |
+
"learning_rate": 8.983012820512821e-06,
|
| 181411 |
+
"loss": 0.9244,
|
| 181412 |
+
"step": 64485
|
| 181413 |
+
},
|
| 181414 |
+
{
|
| 181415 |
+
"epoch": 519.23,
|
| 181416 |
+
"learning_rate": 8.982932692307693e-06,
|
| 181417 |
+
"loss": 0.7609,
|
| 181418 |
+
"step": 64490
|
| 181419 |
+
},
|
| 181420 |
+
{
|
| 181421 |
+
"epoch": 519.27,
|
| 181422 |
+
"learning_rate": 8.982852564102564e-06,
|
| 181423 |
+
"loss": 0.2902,
|
| 181424 |
+
"step": 64495
|
| 181425 |
+
},
|
| 181426 |
+
{
|
| 181427 |
+
"epoch": 519.31,
|
| 181428 |
+
"learning_rate": 8.982772435897437e-06,
|
| 181429 |
+
"loss": 0.3364,
|
| 181430 |
+
"step": 64500
|
| 181431 |
+
},
|
| 181432 |
+
{
|
| 181433 |
+
"epoch": 519.35,
|
| 181434 |
+
"learning_rate": 8.982692307692309e-06,
|
| 181435 |
+
"loss": 0.4234,
|
| 181436 |
+
"step": 64505
|
| 181437 |
+
},
|
| 181438 |
+
{
|
| 181439 |
+
"epoch": 519.39,
|
| 181440 |
+
"learning_rate": 8.98261217948718e-06,
|
| 181441 |
+
"loss": 0.9376,
|
| 181442 |
+
"step": 64510
|
| 181443 |
+
},
|
| 181444 |
+
{
|
| 181445 |
+
"epoch": 519.43,
|
| 181446 |
+
"learning_rate": 8.982532051282053e-06,
|
| 181447 |
+
"loss": 0.7222,
|
| 181448 |
+
"step": 64515
|
| 181449 |
+
},
|
| 181450 |
+
{
|
| 181451 |
+
"epoch": 519.47,
|
| 181452 |
+
"learning_rate": 8.982451923076924e-06,
|
| 181453 |
+
"loss": 0.3225,
|
| 181454 |
+
"step": 64520
|
| 181455 |
+
},
|
| 181456 |
+
{
|
| 181457 |
+
"epoch": 519.51,
|
| 181458 |
+
"learning_rate": 8.982371794871796e-06,
|
| 181459 |
+
"loss": 0.3611,
|
| 181460 |
+
"step": 64525
|
| 181461 |
+
},
|
| 181462 |
+
{
|
| 181463 |
+
"epoch": 519.55,
|
| 181464 |
+
"learning_rate": 8.982291666666667e-06,
|
| 181465 |
+
"loss": 0.4063,
|
| 181466 |
+
"step": 64530
|
| 181467 |
+
},
|
| 181468 |
+
{
|
| 181469 |
+
"epoch": 519.59,
|
| 181470 |
+
"learning_rate": 8.98221153846154e-06,
|
| 181471 |
+
"loss": 0.9661,
|
| 181472 |
+
"step": 64535
|
| 181473 |
+
},
|
| 181474 |
+
{
|
| 181475 |
+
"epoch": 519.63,
|
| 181476 |
+
"learning_rate": 8.98213141025641e-06,
|
| 181477 |
+
"loss": 0.6798,
|
| 181478 |
+
"step": 64540
|
| 181479 |
+
},
|
| 181480 |
+
{
|
| 181481 |
+
"epoch": 519.67,
|
| 181482 |
+
"learning_rate": 8.982051282051283e-06,
|
| 181483 |
+
"loss": 0.3482,
|
| 181484 |
+
"step": 64545
|
| 181485 |
+
},
|
| 181486 |
+
{
|
| 181487 |
+
"epoch": 519.71,
|
| 181488 |
+
"learning_rate": 8.981971153846154e-06,
|
| 181489 |
+
"loss": 0.3422,
|
| 181490 |
+
"step": 64550
|
| 181491 |
+
},
|
| 181492 |
+
{
|
| 181493 |
+
"epoch": 519.76,
|
| 181494 |
+
"learning_rate": 8.981891025641026e-06,
|
| 181495 |
+
"loss": 0.4464,
|
| 181496 |
+
"step": 64555
|
| 181497 |
+
},
|
| 181498 |
+
{
|
| 181499 |
+
"epoch": 519.8,
|
| 181500 |
+
"learning_rate": 8.981810897435899e-06,
|
| 181501 |
+
"loss": 1.0214,
|
| 181502 |
+
"step": 64560
|
| 181503 |
+
},
|
| 181504 |
+
{
|
| 181505 |
+
"epoch": 519.84,
|
| 181506 |
+
"learning_rate": 8.98173076923077e-06,
|
| 181507 |
+
"loss": 0.728,
|
| 181508 |
+
"step": 64565
|
| 181509 |
+
},
|
| 181510 |
+
{
|
| 181511 |
+
"epoch": 519.88,
|
| 181512 |
+
"learning_rate": 8.981650641025641e-06,
|
| 181513 |
+
"loss": 0.328,
|
| 181514 |
+
"step": 64570
|
| 181515 |
+
},
|
| 181516 |
+
{
|
| 181517 |
+
"epoch": 519.92,
|
| 181518 |
+
"learning_rate": 8.981570512820513e-06,
|
| 181519 |
+
"loss": 0.3455,
|
| 181520 |
+
"step": 64575
|
| 181521 |
+
},
|
| 181522 |
+
{
|
| 181523 |
+
"epoch": 519.96,
|
| 181524 |
+
"learning_rate": 8.981490384615386e-06,
|
| 181525 |
+
"loss": 0.4709,
|
| 181526 |
+
"step": 64580
|
| 181527 |
+
},
|
| 181528 |
+
{
|
| 181529 |
+
"epoch": 520.0,
|
| 181530 |
+
"learning_rate": 8.981410256410257e-06,
|
| 181531 |
+
"loss": 1.2198,
|
| 181532 |
+
"step": 64585
|
| 181533 |
+
},
|
| 181534 |
+
{
|
| 181535 |
+
"epoch": 520.0,
|
| 181536 |
+
"eval_loss": 0.4248497486114502,
|
| 181537 |
+
"eval_runtime": 39.8905,
|
| 181538 |
+
"eval_samples_per_second": 20.982,
|
| 181539 |
+
"eval_steps_per_second": 0.677,
|
| 181540 |
+
"eval_wer": 0.2014564910455525,
|
| 181541 |
+
"step": 64585
|
| 181542 |
+
},
|
| 181543 |
+
{
|
| 181544 |
+
"epoch": 516.04,
|
| 181545 |
+
"learning_rate": 8.981330128205128e-06,
|
| 181546 |
+
"loss": 0.3283,
|
| 181547 |
+
"step": 64590
|
| 181548 |
+
},
|
| 181549 |
+
{
|
| 181550 |
+
"epoch": 516.08,
|
| 181551 |
+
"learning_rate": 8.98125e-06,
|
| 181552 |
+
"loss": 0.3273,
|
| 181553 |
+
"step": 64595
|
| 181554 |
+
},
|
| 181555 |
+
{
|
| 181556 |
+
"epoch": 516.12,
|
| 181557 |
+
"learning_rate": 8.981169871794873e-06,
|
| 181558 |
+
"loss": 0.346,
|
| 181559 |
+
"step": 64600
|
| 181560 |
+
},
|
| 181561 |
+
{
|
| 181562 |
+
"epoch": 516.16,
|
| 181563 |
+
"learning_rate": 8.981089743589744e-06,
|
| 181564 |
+
"loss": 0.4777,
|
| 181565 |
+
"step": 64605
|
| 181566 |
+
},
|
| 181567 |
+
{
|
| 181568 |
+
"epoch": 516.2,
|
| 181569 |
+
"learning_rate": 8.981009615384616e-06,
|
| 181570 |
+
"loss": 1.1656,
|
| 181571 |
+
"step": 64610
|
| 181572 |
+
},
|
| 181573 |
+
{
|
| 181574 |
+
"epoch": 516.24,
|
| 181575 |
+
"learning_rate": 8.980929487179489e-06,
|
| 181576 |
+
"loss": 0.328,
|
| 181577 |
+
"step": 64615
|
| 181578 |
+
},
|
| 181579 |
+
{
|
| 181580 |
+
"epoch": 516.28,
|
| 181581 |
+
"learning_rate": 8.98084935897436e-06,
|
| 181582 |
+
"loss": 0.3855,
|
| 181583 |
+
"step": 64620
|
| 181584 |
+
},
|
| 181585 |
+
{
|
| 181586 |
+
"epoch": 516.32,
|
| 181587 |
+
"learning_rate": 8.980769230769231e-06,
|
| 181588 |
+
"loss": 0.326,
|
| 181589 |
+
"step": 64625
|
| 181590 |
+
},
|
| 181591 |
+
{
|
| 181592 |
+
"epoch": 516.36,
|
| 181593 |
+
"learning_rate": 8.980689102564103e-06,
|
| 181594 |
+
"loss": 0.5385,
|
| 181595 |
+
"step": 64630
|
| 181596 |
+
},
|
| 181597 |
+
{
|
| 181598 |
+
"epoch": 516.4,
|
| 181599 |
+
"learning_rate": 8.980608974358976e-06,
|
| 181600 |
+
"loss": 1.2344,
|
| 181601 |
+
"step": 64635
|
| 181602 |
+
},
|
| 181603 |
+
{
|
| 181604 |
+
"epoch": 516.44,
|
| 181605 |
+
"learning_rate": 8.980528846153847e-06,
|
| 181606 |
+
"loss": 0.4577,
|
| 181607 |
+
"step": 64640
|
| 181608 |
+
},
|
| 181609 |
+
{
|
| 181610 |
+
"epoch": 516.48,
|
| 181611 |
+
"learning_rate": 8.980448717948718e-06,
|
| 181612 |
+
"loss": 0.3076,
|
| 181613 |
+
"step": 64645
|
| 181614 |
+
},
|
| 181615 |
+
{
|
| 181616 |
+
"epoch": 516.52,
|
| 181617 |
+
"learning_rate": 8.98036858974359e-06,
|
| 181618 |
+
"loss": 0.3086,
|
| 181619 |
+
"step": 64650
|
| 181620 |
+
},
|
| 181621 |
+
{
|
| 181622 |
+
"epoch": 516.56,
|
| 181623 |
+
"learning_rate": 8.980288461538463e-06,
|
| 181624 |
+
"loss": 0.4801,
|
| 181625 |
+
"step": 64655
|
| 181626 |
+
},
|
| 181627 |
+
{
|
| 181628 |
+
"epoch": 516.6,
|
| 181629 |
+
"learning_rate": 8.980208333333334e-06,
|
| 181630 |
+
"loss": 1.5408,
|
| 181631 |
+
"step": 64660
|
| 181632 |
+
},
|
| 181633 |
+
{
|
| 181634 |
+
"epoch": 516.64,
|
| 181635 |
+
"learning_rate": 8.980128205128206e-06,
|
| 181636 |
+
"loss": 0.3752,
|
| 181637 |
+
"step": 64665
|
| 181638 |
+
},
|
| 181639 |
+
{
|
| 181640 |
+
"epoch": 516.68,
|
| 181641 |
+
"learning_rate": 8.980048076923079e-06,
|
| 181642 |
+
"loss": 0.3062,
|
| 181643 |
+
"step": 64670
|
| 181644 |
+
},
|
| 181645 |
+
{
|
| 181646 |
+
"epoch": 516.72,
|
| 181647 |
+
"learning_rate": 8.97996794871795e-06,
|
| 181648 |
+
"loss": 0.3165,
|
| 181649 |
+
"step": 64675
|
| 181650 |
+
},
|
| 181651 |
+
{
|
| 181652 |
+
"epoch": 516.76,
|
| 181653 |
+
"learning_rate": 8.979887820512821e-06,
|
| 181654 |
+
"loss": 0.4657,
|
| 181655 |
+
"step": 64680
|
| 181656 |
+
},
|
| 181657 |
+
{
|
| 181658 |
+
"epoch": 516.8,
|
| 181659 |
+
"learning_rate": 8.979807692307693e-06,
|
| 181660 |
+
"loss": 1.1699,
|
| 181661 |
+
"step": 64685
|
| 181662 |
+
},
|
| 181663 |
+
{
|
| 181664 |
+
"epoch": 516.84,
|
| 181665 |
+
"learning_rate": 8.979727564102566e-06,
|
| 181666 |
+
"loss": 0.2932,
|
| 181667 |
+
"step": 64690
|
| 181668 |
+
},
|
| 181669 |
+
{
|
| 181670 |
+
"epoch": 516.88,
|
| 181671 |
+
"learning_rate": 8.979647435897435e-06,
|
| 181672 |
+
"loss": 0.3501,
|
| 181673 |
+
"step": 64695
|
| 181674 |
+
},
|
| 181675 |
+
{
|
| 181676 |
+
"epoch": 516.92,
|
| 181677 |
+
"learning_rate": 8.979567307692308e-06,
|
| 181678 |
+
"loss": 0.3625,
|
| 181679 |
+
"step": 64700
|
| 181680 |
+
},
|
| 181681 |
+
{
|
| 181682 |
+
"epoch": 516.96,
|
| 181683 |
+
"learning_rate": 8.979487179487182e-06,
|
| 181684 |
+
"loss": 0.5031,
|
| 181685 |
+
"step": 64705
|
| 181686 |
+
},
|
| 181687 |
+
{
|
| 181688 |
+
"epoch": 517.0,
|
| 181689 |
+
"learning_rate": 8.979407051282051e-06,
|
| 181690 |
+
"loss": 1.2624,
|
| 181691 |
+
"step": 64710
|
| 181692 |
+
},
|
| 181693 |
+
{
|
| 181694 |
+
"epoch": 517.0,
|
| 181695 |
+
"eval_loss": 0.35672011971473694,
|
| 181696 |
+
"eval_runtime": 40.2132,
|
| 181697 |
+
"eval_samples_per_second": 20.789,
|
| 181698 |
+
"eval_steps_per_second": 0.671,
|
| 181699 |
+
"eval_wer": 0.19060431968584102,
|
| 181700 |
+
"step": 64710
|
| 181701 |
+
},
|
| 181702 |
+
{
|
| 181703 |
+
"epoch": 521.04,
|
| 181704 |
+
"learning_rate": 8.979326923076924e-06,
|
| 181705 |
+
"loss": 0.3821,
|
| 181706 |
+
"step": 64715
|
| 181707 |
+
},
|
| 181708 |
+
{
|
| 181709 |
+
"epoch": 521.08,
|
| 181710 |
+
"learning_rate": 8.979246794871796e-06,
|
| 181711 |
+
"loss": 0.4827,
|
| 181712 |
+
"step": 64720
|
| 181713 |
+
},
|
| 181714 |
+
{
|
| 181715 |
+
"epoch": 521.12,
|
| 181716 |
+
"learning_rate": 8.979166666666667e-06,
|
| 181717 |
+
"loss": 0.3156,
|
| 181718 |
+
"step": 64725
|
| 181719 |
+
},
|
| 181720 |
+
{
|
| 181721 |
+
"epoch": 521.16,
|
| 181722 |
+
"learning_rate": 8.979086538461538e-06,
|
| 181723 |
+
"loss": 0.4768,
|
| 181724 |
+
"step": 64730
|
| 181725 |
+
},
|
| 181726 |
+
{
|
| 181727 |
+
"epoch": 521.2,
|
| 181728 |
+
"learning_rate": 8.979006410256411e-06,
|
| 181729 |
+
"loss": 1.2266,
|
| 181730 |
+
"step": 64735
|
| 181731 |
+
},
|
| 181732 |
+
{
|
| 181733 |
+
"epoch": 521.24,
|
| 181734 |
+
"learning_rate": 8.978926282051283e-06,
|
| 181735 |
+
"loss": 0.3997,
|
| 181736 |
+
"step": 64740
|
| 181737 |
+
},
|
| 181738 |
+
{
|
| 181739 |
+
"epoch": 521.28,
|
| 181740 |
+
"learning_rate": 8.978846153846154e-06,
|
| 181741 |
+
"loss": 0.324,
|
| 181742 |
+
"step": 64745
|
| 181743 |
+
},
|
| 181744 |
+
{
|
| 181745 |
+
"epoch": 521.32,
|
| 181746 |
+
"learning_rate": 8.978766025641025e-06,
|
| 181747 |
+
"loss": 0.3588,
|
| 181748 |
+
"step": 64750
|
| 181749 |
+
},
|
| 181750 |
+
{
|
| 181751 |
+
"epoch": 521.36,
|
| 181752 |
+
"learning_rate": 8.978685897435898e-06,
|
| 181753 |
+
"loss": 0.4545,
|
| 181754 |
+
"step": 64755
|
| 181755 |
+
},
|
| 181756 |
+
{
|
| 181757 |
+
"epoch": 521.4,
|
| 181758 |
+
"learning_rate": 8.97860576923077e-06,
|
| 181759 |
+
"loss": 1.2435,
|
| 181760 |
+
"step": 64760
|
| 181761 |
+
},
|
| 181762 |
+
{
|
| 181763 |
+
"epoch": 521.44,
|
| 181764 |
+
"learning_rate": 8.978525641025641e-06,
|
| 181765 |
+
"loss": 0.3921,
|
| 181766 |
+
"step": 64765
|
| 181767 |
+
},
|
| 181768 |
+
{
|
| 181769 |
+
"epoch": 521.48,
|
| 181770 |
+
"learning_rate": 8.978445512820514e-06,
|
| 181771 |
+
"loss": 0.3271,
|
| 181772 |
+
"step": 64770
|
| 181773 |
+
},
|
| 181774 |
+
{
|
| 181775 |
+
"epoch": 521.52,
|
| 181776 |
+
"learning_rate": 8.978365384615386e-06,
|
| 181777 |
+
"loss": 0.3303,
|
| 181778 |
+
"step": 64775
|
| 181779 |
+
},
|
| 181780 |
+
{
|
| 181781 |
+
"epoch": 521.56,
|
| 181782 |
+
"learning_rate": 8.978285256410257e-06,
|
| 181783 |
+
"loss": 0.5433,
|
| 181784 |
+
"step": 64780
|
| 181785 |
+
},
|
| 181786 |
+
{
|
| 181787 |
+
"epoch": 521.6,
|
| 181788 |
+
"learning_rate": 8.978205128205128e-06,
|
| 181789 |
+
"loss": 1.2772,
|
| 181790 |
+
"step": 64785
|
| 181791 |
+
},
|
| 181792 |
+
{
|
| 181793 |
+
"epoch": 521.64,
|
| 181794 |
+
"learning_rate": 8.978125000000001e-06,
|
| 181795 |
+
"loss": 0.3909,
|
| 181796 |
+
"step": 64790
|
| 181797 |
+
},
|
| 181798 |
+
{
|
| 181799 |
+
"epoch": 521.68,
|
| 181800 |
+
"learning_rate": 8.978044871794873e-06,
|
| 181801 |
+
"loss": 0.338,
|
| 181802 |
+
"step": 64795
|
| 181803 |
+
},
|
| 181804 |
+
{
|
| 181805 |
+
"epoch": 521.72,
|
| 181806 |
+
"learning_rate": 8.977964743589744e-06,
|
| 181807 |
+
"loss": 0.3753,
|
| 181808 |
+
"step": 64800
|
| 181809 |
+
},
|
| 181810 |
+
{
|
| 181811 |
+
"epoch": 521.76,
|
| 181812 |
+
"learning_rate": 8.977884615384617e-06,
|
| 181813 |
+
"loss": 0.5181,
|
| 181814 |
+
"step": 64805
|
| 181815 |
+
},
|
| 181816 |
+
{
|
| 181817 |
+
"epoch": 521.8,
|
| 181818 |
+
"learning_rate": 8.977804487179489e-06,
|
| 181819 |
+
"loss": 1.3838,
|
| 181820 |
+
"step": 64810
|
| 181821 |
+
},
|
| 181822 |
+
{
|
| 181823 |
+
"epoch": 521.84,
|
| 181824 |
+
"learning_rate": 8.97772435897436e-06,
|
| 181825 |
+
"loss": 0.2974,
|
| 181826 |
+
"step": 64815
|
| 181827 |
+
},
|
| 181828 |
+
{
|
| 181829 |
+
"epoch": 521.88,
|
| 181830 |
+
"learning_rate": 8.977644230769231e-06,
|
| 181831 |
+
"loss": 0.3573,
|
| 181832 |
+
"step": 64820
|
| 181833 |
+
},
|
| 181834 |
+
{
|
| 181835 |
+
"epoch": 521.92,
|
| 181836 |
+
"learning_rate": 8.977564102564104e-06,
|
| 181837 |
+
"loss": 0.3602,
|
| 181838 |
+
"step": 64825
|
| 181839 |
+
},
|
| 181840 |
+
{
|
| 181841 |
+
"epoch": 521.96,
|
| 181842 |
+
"learning_rate": 8.977483974358974e-06,
|
| 181843 |
+
"loss": 0.4902,
|
| 181844 |
+
"step": 64830
|
| 181845 |
+
},
|
| 181846 |
+
{
|
| 181847 |
+
"epoch": 522.0,
|
| 181848 |
+
"eval_loss": 0.36429762840270996,
|
| 181849 |
+
"eval_runtime": 40.4014,
|
| 181850 |
+
"eval_samples_per_second": 20.692,
|
| 181851 |
+
"eval_steps_per_second": 0.668,
|
| 181852 |
+
"eval_wer": 0.1932991605378501,
|
| 181853 |
+
"step": 64834
|
| 181854 |
+
},
|
| 181855 |
+
{
|
| 181856 |
+
"epoch": 522.01,
|
| 181857 |
+
"learning_rate": 8.977403846153847e-06,
|
| 181858 |
+
"loss": 0.4266,
|
| 181859 |
+
"step": 64835
|
| 181860 |
+
},
|
| 181861 |
+
{
|
| 181862 |
+
"epoch": 522.05,
|
| 181863 |
+
"learning_rate": 8.977323717948718e-06,
|
| 181864 |
+
"loss": 0.3325,
|
| 181865 |
+
"step": 64840
|
| 181866 |
+
},
|
| 181867 |
+
{
|
| 181868 |
+
"epoch": 522.09,
|
| 181869 |
+
"learning_rate": 8.97724358974359e-06,
|
| 181870 |
+
"loss": 0.3123,
|
| 181871 |
+
"step": 64845
|
| 181872 |
+
},
|
| 181873 |
+
{
|
| 181874 |
+
"epoch": 522.13,
|
| 181875 |
+
"learning_rate": 8.977163461538461e-06,
|
| 181876 |
+
"loss": 0.3465,
|
| 181877 |
+
"step": 64850
|
| 181878 |
+
},
|
| 181879 |
+
{
|
| 181880 |
+
"epoch": 522.17,
|
| 181881 |
+
"learning_rate": 8.977083333333334e-06,
|
| 181882 |
+
"loss": 0.5419,
|
| 181883 |
+
"step": 64855
|
| 181884 |
+
},
|
| 181885 |
+
{
|
| 181886 |
+
"epoch": 522.21,
|
| 181887 |
+
"learning_rate": 8.977003205128205e-06,
|
| 181888 |
+
"loss": 1.1883,
|
| 181889 |
+
"step": 64860
|
| 181890 |
+
},
|
| 181891 |
+
{
|
| 181892 |
+
"epoch": 522.25,
|
| 181893 |
+
"learning_rate": 8.976923076923077e-06,
|
| 181894 |
+
"loss": 0.3538,
|
| 181895 |
+
"step": 64865
|
| 181896 |
+
},
|
| 181897 |
+
{
|
| 181898 |
+
"epoch": 522.29,
|
| 181899 |
+
"learning_rate": 8.97684294871795e-06,
|
| 181900 |
+
"loss": 0.3029,
|
| 181901 |
+
"step": 64870
|
| 181902 |
+
},
|
| 181903 |
+
{
|
| 181904 |
+
"epoch": 522.33,
|
| 181905 |
+
"learning_rate": 8.976762820512821e-06,
|
| 181906 |
+
"loss": 0.3871,
|
| 181907 |
+
"step": 64875
|
| 181908 |
+
},
|
| 181909 |
+
{
|
| 181910 |
+
"epoch": 522.37,
|
| 181911 |
+
"learning_rate": 8.976682692307693e-06,
|
| 181912 |
+
"loss": 0.772,
|
| 181913 |
+
"step": 64880
|
| 181914 |
+
},
|
| 181915 |
+
{
|
| 181916 |
+
"epoch": 522.41,
|
| 181917 |
+
"learning_rate": 8.976602564102564e-06,
|
| 181918 |
+
"loss": 1.1618,
|
| 181919 |
+
"step": 64885
|
| 181920 |
+
},
|
| 181921 |
+
{
|
| 181922 |
+
"epoch": 522.45,
|
| 181923 |
+
"learning_rate": 8.976522435897437e-06,
|
| 181924 |
+
"loss": 0.4247,
|
| 181925 |
+
"step": 64890
|
| 181926 |
+
},
|
| 181927 |
+
{
|
| 181928 |
+
"epoch": 522.49,
|
| 181929 |
+
"learning_rate": 8.976442307692308e-06,
|
| 181930 |
+
"loss": 0.344,
|
| 181931 |
+
"step": 64895
|
| 181932 |
+
},
|
| 181933 |
+
{
|
| 181934 |
+
"epoch": 522.53,
|
| 181935 |
+
"learning_rate": 8.97636217948718e-06,
|
| 181936 |
+
"loss": 0.4065,
|
| 181937 |
+
"step": 64900
|
| 181938 |
+
},
|
| 181939 |
+
{
|
| 181940 |
+
"epoch": 522.57,
|
| 181941 |
+
"learning_rate": 8.976282051282053e-06,
|
| 181942 |
+
"loss": 0.5122,
|
| 181943 |
+
"step": 64905
|
| 181944 |
+
},
|
| 181945 |
+
{
|
| 181946 |
+
"epoch": 522.61,
|
| 181947 |
+
"learning_rate": 8.976201923076924e-06,
|
| 181948 |
+
"loss": 1.1144,
|
| 181949 |
+
"step": 64910
|
| 181950 |
+
},
|
| 181951 |
+
{
|
| 181952 |
+
"epoch": 522.65,
|
| 181953 |
+
"learning_rate": 8.976121794871796e-06,
|
| 181954 |
+
"loss": 0.3088,
|
| 181955 |
+
"step": 64915
|
| 181956 |
+
},
|
| 181957 |
+
{
|
| 181958 |
+
"epoch": 522.69,
|
| 181959 |
+
"learning_rate": 8.976041666666667e-06,
|
| 181960 |
+
"loss": 0.3312,
|
| 181961 |
+
"step": 64920
|
| 181962 |
+
},
|
| 181963 |
+
{
|
| 181964 |
+
"epoch": 522.73,
|
| 181965 |
+
"learning_rate": 8.97596153846154e-06,
|
| 181966 |
+
"loss": 0.373,
|
| 181967 |
+
"step": 64925
|
| 181968 |
+
},
|
| 181969 |
+
{
|
| 181970 |
+
"epoch": 522.77,
|
| 181971 |
+
"learning_rate": 8.975881410256411e-06,
|
| 181972 |
+
"loss": 0.7653,
|
| 181973 |
+
"step": 64930
|
| 181974 |
+
},
|
| 181975 |
+
{
|
| 181976 |
+
"epoch": 522.81,
|
| 181977 |
+
"learning_rate": 8.975801282051283e-06,
|
| 181978 |
+
"loss": 1.139,
|
| 181979 |
+
"step": 64935
|
| 181980 |
+
},
|
| 181981 |
+
{
|
| 181982 |
+
"epoch": 522.85,
|
| 181983 |
+
"learning_rate": 8.975721153846154e-06,
|
| 181984 |
+
"loss": 0.3019,
|
| 181985 |
+
"step": 64940
|
| 181986 |
+
},
|
| 181987 |
+
{
|
| 181988 |
+
"epoch": 522.9,
|
| 181989 |
+
"learning_rate": 8.975641025641027e-06,
|
| 181990 |
+
"loss": 0.3356,
|
| 181991 |
+
"step": 64945
|
| 181992 |
+
},
|
| 181993 |
+
{
|
| 181994 |
+
"epoch": 522.94,
|
| 181995 |
+
"learning_rate": 8.975560897435898e-06,
|
| 181996 |
+
"loss": 0.3868,
|
| 181997 |
+
"step": 64950
|
| 181998 |
+
},
|
| 181999 |
+
{
|
| 182000 |
+
"epoch": 522.98,
|
| 182001 |
+
"learning_rate": 8.97548076923077e-06,
|
| 182002 |
+
"loss": 0.7313,
|
| 182003 |
+
"step": 64955
|
| 182004 |
+
},
|
| 182005 |
+
{
|
| 182006 |
+
"epoch": 523.0,
|
| 182007 |
+
"eval_loss": 0.43909579515457153,
|
| 182008 |
+
"eval_runtime": 40.8887,
|
| 182009 |
+
"eval_samples_per_second": 20.446,
|
| 182010 |
+
"eval_steps_per_second": 0.66,
|
| 182011 |
+
"eval_wer": 0.20390979645628934,
|
| 182012 |
+
"step": 64958
|
| 182013 |
+
},
|
| 182014 |
+
{
|
| 182015 |
+
"epoch": 523.02,
|
| 182016 |
+
"learning_rate": 8.975400641025643e-06,
|
| 182017 |
+
"loss": 0.4407,
|
| 182018 |
+
"step": 64960
|
| 182019 |
+
},
|
| 182020 |
+
{
|
| 182021 |
+
"epoch": 523.06,
|
| 182022 |
+
"learning_rate": 8.975320512820514e-06,
|
| 182023 |
+
"loss": 0.3605,
|
| 182024 |
+
"step": 64965
|
| 182025 |
+
},
|
| 182026 |
+
{
|
| 182027 |
+
"epoch": 523.1,
|
| 182028 |
+
"learning_rate": 8.975240384615386e-06,
|
| 182029 |
+
"loss": 0.2888,
|
| 182030 |
+
"step": 64970
|
| 182031 |
+
},
|
| 182032 |
+
{
|
| 182033 |
+
"epoch": 523.14,
|
| 182034 |
+
"learning_rate": 8.975160256410257e-06,
|
| 182035 |
+
"loss": 0.3612,
|
| 182036 |
+
"step": 64975
|
| 182037 |
+
},
|
| 182038 |
+
{
|
| 182039 |
+
"epoch": 523.18,
|
| 182040 |
+
"learning_rate": 8.97508012820513e-06,
|
| 182041 |
+
"loss": 0.7397,
|
| 182042 |
+
"step": 64980
|
| 182043 |
+
},
|
| 182044 |
+
{
|
| 182045 |
+
"epoch": 523.22,
|
| 182046 |
+
"learning_rate": 8.975e-06,
|
| 182047 |
+
"loss": 0.9723,
|
| 182048 |
+
"step": 64985
|
| 182049 |
+
},
|
| 182050 |
+
{
|
| 182051 |
+
"epoch": 523.26,
|
| 182052 |
+
"learning_rate": 8.974919871794873e-06,
|
| 182053 |
+
"loss": 0.3375,
|
| 182054 |
+
"step": 64990
|
| 182055 |
+
},
|
| 182056 |
+
{
|
| 182057 |
+
"epoch": 523.3,
|
| 182058 |
+
"learning_rate": 8.974839743589744e-06,
|
| 182059 |
+
"loss": 0.3919,
|
| 182060 |
+
"step": 64995
|
| 182061 |
+
},
|
| 182062 |
+
{
|
| 182063 |
+
"epoch": 523.34,
|
| 182064 |
+
"learning_rate": 8.974759615384615e-06,
|
| 182065 |
+
"loss": 0.3596,
|
| 182066 |
+
"step": 65000
|
| 182067 |
+
},
|
| 182068 |
+
{
|
| 182069 |
+
"epoch": 523.38,
|
| 182070 |
+
"learning_rate": 8.974679487179488e-06,
|
| 182071 |
+
"loss": 0.729,
|
| 182072 |
+
"step": 65005
|
| 182073 |
+
},
|
| 182074 |
+
{
|
| 182075 |
+
"epoch": 523.42,
|
| 182076 |
+
"learning_rate": 8.97459935897436e-06,
|
| 182077 |
+
"loss": 1.0572,
|
| 182078 |
+
"step": 65010
|
| 182079 |
+
},
|
| 182080 |
+
{
|
| 182081 |
+
"epoch": 523.46,
|
| 182082 |
+
"learning_rate": 8.974519230769231e-06,
|
| 182083 |
+
"loss": 0.3455,
|
| 182084 |
+
"step": 65015
|
| 182085 |
+
},
|
| 182086 |
+
{
|
| 182087 |
+
"epoch": 523.5,
|
| 182088 |
+
"learning_rate": 8.974439102564103e-06,
|
| 182089 |
+
"loss": 0.3085,
|
| 182090 |
+
"step": 65020
|
| 182091 |
+
},
|
| 182092 |
+
{
|
| 182093 |
+
"epoch": 523.54,
|
| 182094 |
+
"learning_rate": 8.974358974358976e-06,
|
| 182095 |
+
"loss": 0.3663,
|
| 182096 |
+
"step": 65025
|
| 182097 |
+
},
|
| 182098 |
+
{
|
| 182099 |
+
"epoch": 523.58,
|
| 182100 |
+
"learning_rate": 8.974278846153847e-06,
|
| 182101 |
+
"loss": 0.6379,
|
| 182102 |
+
"step": 65030
|
| 182103 |
+
},
|
| 182104 |
+
{
|
| 182105 |
+
"epoch": 523.62,
|
| 182106 |
+
"learning_rate": 8.974198717948718e-06,
|
| 182107 |
+
"loss": 1.0438,
|
| 182108 |
+
"step": 65035
|
| 182109 |
+
},
|
| 182110 |
+
{
|
| 182111 |
+
"epoch": 523.66,
|
| 182112 |
+
"learning_rate": 8.97411858974359e-06,
|
| 182113 |
+
"loss": 0.3742,
|
| 182114 |
+
"step": 65040
|
| 182115 |
+
},
|
| 182116 |
+
{
|
| 182117 |
+
"epoch": 523.7,
|
| 182118 |
+
"learning_rate": 8.974038461538463e-06,
|
| 182119 |
+
"loss": 0.3424,
|
| 182120 |
+
"step": 65045
|
| 182121 |
+
},
|
| 182122 |
+
{
|
| 182123 |
+
"epoch": 523.74,
|
| 182124 |
+
"learning_rate": 8.973958333333334e-06,
|
| 182125 |
+
"loss": 0.3719,
|
| 182126 |
+
"step": 65050
|
| 182127 |
+
},
|
| 182128 |
+
{
|
| 182129 |
+
"epoch": 523.78,
|
| 182130 |
+
"learning_rate": 8.973878205128205e-06,
|
| 182131 |
+
"loss": 0.6451,
|
| 182132 |
+
"step": 65055
|
| 182133 |
+
},
|
| 182134 |
+
{
|
| 182135 |
+
"epoch": 523.82,
|
| 182136 |
+
"learning_rate": 8.973798076923078e-06,
|
| 182137 |
+
"loss": 1.085,
|
| 182138 |
+
"step": 65060
|
| 182139 |
+
},
|
| 182140 |
+
{
|
| 182141 |
+
"epoch": 523.86,
|
| 182142 |
+
"learning_rate": 8.97371794871795e-06,
|
| 182143 |
+
"loss": 0.3687,
|
| 182144 |
+
"step": 65065
|
| 182145 |
+
},
|
| 182146 |
+
{
|
| 182147 |
+
"epoch": 523.9,
|
| 182148 |
+
"learning_rate": 8.973637820512821e-06,
|
| 182149 |
+
"loss": 0.3705,
|
| 182150 |
+
"step": 65070
|
| 182151 |
+
},
|
| 182152 |
+
{
|
| 182153 |
+
"epoch": 523.94,
|
| 182154 |
+
"learning_rate": 8.973557692307693e-06,
|
| 182155 |
+
"loss": 0.466,
|
| 182156 |
+
"step": 65075
|
| 182157 |
+
},
|
| 182158 |
+
{
|
| 182159 |
+
"epoch": 523.98,
|
| 182160 |
+
"learning_rate": 8.973477564102566e-06,
|
| 182161 |
+
"loss": 0.8072,
|
| 182162 |
+
"step": 65080
|
| 182163 |
+
},
|
| 182164 |
+
{
|
| 182165 |
+
"epoch": 524.0,
|
| 182166 |
+
"eval_loss": 0.40178999304771423,
|
| 182167 |
+
"eval_runtime": 40.3217,
|
| 182168 |
+
"eval_samples_per_second": 20.758,
|
| 182169 |
+
"eval_steps_per_second": 0.67,
|
| 182170 |
+
"eval_wer": 0.19669648009232546,
|
| 182171 |
+
"step": 65082
|
| 182172 |
}
|
| 182173 |
],
|
| 182174 |
+
"max_steps": 620000,
|
| 182175 |
"num_train_epochs": 5000,
|
| 182176 |
+
"total_flos": 1.8314457593924465e+20,
|
| 182177 |
"trial_name": null,
|
| 182178 |
"trial_params": null
|
| 182179 |
}
|
model-bin/finetune/base/{checkpoint-64461 → checkpoint-65082}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629821196.2313719/events.out.tfevents.1629821196.c435e1c5ee04.920.171
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0c33f545b92d38ede164b6fc4532c7315546277b501e057840a159d93182050
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629821845.184331/events.out.tfevents.1629821845.c435e1c5ee04.920.173
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86800141394780e5838a4943b504737d21308a52d4d13ce96049c5f82b165a42
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629822489.0445116/events.out.tfevents.1629822489.c435e1c5ee04.920.175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae0a6131a49f386d6c4bba8958f2601dc69aeab5e86652508037d2a9fe74bda7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629823140.8357387/events.out.tfevents.1629823140.c435e1c5ee04.920.177
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a94e6c19b48bea2f745903b3400c3833b53b5c7cc7469d1241f52240b4cc8d51
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629823804.588709/events.out.tfevents.1629823804.c435e1c5ee04.920.179
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1df747c578c67b4e7da20237a10beede714da5d26448c1ab6e4a9780466a3064
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629821196.c435e1c5ee04.920.170
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ade45d0c76eea8365079028091f7949afa8bb5cf30236ebd8bd0550d518ff6d2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629821845.c435e1c5ee04.920.172
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:64073bbf805317986b362afbe54f98607820be52feca11a35426964430509a40
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629822489.c435e1c5ee04.920.174
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f38a157470aa6c75bdf924633a006b7a9d39efc1ecc9f68a36911c12b10daff
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629823140.c435e1c5ee04.920.176
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b8d127dc7177ab974911d49cf0876e5dd4a96a4b95a6fe2545023d23487e3e2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629823804.c435e1c5ee04.920.178
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec9762489c3fc6975a61f9a91dde32292205440a42aa95450f19baf4bb9c24b1
|
| 3 |
+
size 8622
|