"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629875783.5501642/events.out.tfevents.1629875783.7e498afd5545.905.73 +3 -0
- model-bin/finetune/base/log/1629876453.8615327/events.out.tfevents.1629876453.7e498afd5545.905.75 +3 -0
- model-bin/finetune/base/log/1629877099.8360853/events.out.tfevents.1629877099.7e498afd5545.905.77 +3 -0
- model-bin/finetune/base/log/1629877746.4587185/events.out.tfevents.1629877746.7e498afd5545.905.79 +3 -0
- model-bin/finetune/base/log/1629878403.3794868/events.out.tfevents.1629878403.7e498afd5545.905.81 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629875783.7e498afd5545.905.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629876453.7e498afd5545.905.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629877099.7e498afd5545.905.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629877746.7e498afd5545.905.78 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629878403.7e498afd5545.905.80 +3 -0
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/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 sha256:56aefc71d0980c995e93aee106a5ea7eb2b8637a7232ca5a4dcc1cdc4511335a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/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 sha256:9bd62e5038f5a12dc25600aabe73f692b4ca8f39c6d252fb9e3be2bbc5ed1145
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/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:a523a00f2b4a8aa255a532787f19739af80b95bca833548653c7ca44d7fe6056
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/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 sha256:89a798027e8a65425e8946906994aff28f6a07534f7b67f44af157705b99bf25
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/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:a5817db540e6b590b213245436521e08f3ff794616869a44c9a430a401dc99d6
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -193410,11 +193410,800 @@
|
|
| 193410 |
"eval_steps_per_second": 0.69,
|
| 193411 |
"eval_wer": 0.1899564928840056,
|
| 193412 |
"step": 73916
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193413 |
}
|
| 193414 |
],
|
| 193415 |
"max_steps": 620000,
|
| 193416 |
"num_train_epochs": 5000,
|
| 193417 |
-
"total_flos": 2.
|
| 193418 |
"trial_name": null,
|
| 193419 |
"trial_params": null
|
| 193420 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 600.995983935743,
|
| 5 |
+
"global_step": 74539,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 193410 |
"eval_steps_per_second": 0.69,
|
| 193411 |
"eval_wer": 0.1899564928840056,
|
| 193412 |
"step": 73916
|
| 193413 |
+
},
|
| 193414 |
+
{
|
| 193415 |
+
"epoch": 596.03,
|
| 193416 |
+
"learning_rate": 8.831875000000001e-06,
|
| 193417 |
+
"loss": 0.4505,
|
| 193418 |
+
"step": 73920
|
| 193419 |
+
},
|
| 193420 |
+
{
|
| 193421 |
+
"epoch": 596.07,
|
| 193422 |
+
"learning_rate": 8.831794871794872e-06,
|
| 193423 |
+
"loss": 0.3225,
|
| 193424 |
+
"step": 73925
|
| 193425 |
+
},
|
| 193426 |
+
{
|
| 193427 |
+
"epoch": 596.11,
|
| 193428 |
+
"learning_rate": 8.831714743589744e-06,
|
| 193429 |
+
"loss": 0.3288,
|
| 193430 |
+
"step": 73930
|
| 193431 |
+
},
|
| 193432 |
+
{
|
| 193433 |
+
"epoch": 596.15,
|
| 193434 |
+
"learning_rate": 8.831634615384617e-06,
|
| 193435 |
+
"loss": 0.413,
|
| 193436 |
+
"step": 73935
|
| 193437 |
+
},
|
| 193438 |
+
{
|
| 193439 |
+
"epoch": 596.19,
|
| 193440 |
+
"learning_rate": 8.831554487179488e-06,
|
| 193441 |
+
"loss": 0.8731,
|
| 193442 |
+
"step": 73940
|
| 193443 |
+
},
|
| 193444 |
+
{
|
| 193445 |
+
"epoch": 596.23,
|
| 193446 |
+
"learning_rate": 8.83147435897436e-06,
|
| 193447 |
+
"loss": 0.6002,
|
| 193448 |
+
"step": 73945
|
| 193449 |
+
},
|
| 193450 |
+
{
|
| 193451 |
+
"epoch": 596.27,
|
| 193452 |
+
"learning_rate": 8.83139423076923e-06,
|
| 193453 |
+
"loss": 0.3424,
|
| 193454 |
+
"step": 73950
|
| 193455 |
+
},
|
| 193456 |
+
{
|
| 193457 |
+
"epoch": 596.31,
|
| 193458 |
+
"learning_rate": 8.831314102564104e-06,
|
| 193459 |
+
"loss": 0.3209,
|
| 193460 |
+
"step": 73955
|
| 193461 |
+
},
|
| 193462 |
+
{
|
| 193463 |
+
"epoch": 596.35,
|
| 193464 |
+
"learning_rate": 8.831233974358975e-06,
|
| 193465 |
+
"loss": 0.4331,
|
| 193466 |
+
"step": 73960
|
| 193467 |
+
},
|
| 193468 |
+
{
|
| 193469 |
+
"epoch": 596.39,
|
| 193470 |
+
"learning_rate": 8.831153846153846e-06,
|
| 193471 |
+
"loss": 0.8799,
|
| 193472 |
+
"step": 73965
|
| 193473 |
+
},
|
| 193474 |
+
{
|
| 193475 |
+
"epoch": 596.43,
|
| 193476 |
+
"learning_rate": 8.831073717948718e-06,
|
| 193477 |
+
"loss": 0.7031,
|
| 193478 |
+
"step": 73970
|
| 193479 |
+
},
|
| 193480 |
+
{
|
| 193481 |
+
"epoch": 596.47,
|
| 193482 |
+
"learning_rate": 8.830993589743591e-06,
|
| 193483 |
+
"loss": 0.3835,
|
| 193484 |
+
"step": 73975
|
| 193485 |
+
},
|
| 193486 |
+
{
|
| 193487 |
+
"epoch": 596.51,
|
| 193488 |
+
"learning_rate": 8.830913461538462e-06,
|
| 193489 |
+
"loss": 0.2978,
|
| 193490 |
+
"step": 73980
|
| 193491 |
+
},
|
| 193492 |
+
{
|
| 193493 |
+
"epoch": 596.55,
|
| 193494 |
+
"learning_rate": 8.830833333333334e-06,
|
| 193495 |
+
"loss": 0.3954,
|
| 193496 |
+
"step": 73985
|
| 193497 |
+
},
|
| 193498 |
+
{
|
| 193499 |
+
"epoch": 596.59,
|
| 193500 |
+
"learning_rate": 8.830753205128207e-06,
|
| 193501 |
+
"loss": 0.8978,
|
| 193502 |
+
"step": 73990
|
| 193503 |
+
},
|
| 193504 |
+
{
|
| 193505 |
+
"epoch": 596.63,
|
| 193506 |
+
"learning_rate": 8.830673076923078e-06,
|
| 193507 |
+
"loss": 0.6776,
|
| 193508 |
+
"step": 73995
|
| 193509 |
+
},
|
| 193510 |
+
{
|
| 193511 |
+
"epoch": 596.67,
|
| 193512 |
+
"learning_rate": 8.83059294871795e-06,
|
| 193513 |
+
"loss": 0.3265,
|
| 193514 |
+
"step": 74000
|
| 193515 |
+
},
|
| 193516 |
+
{
|
| 193517 |
+
"epoch": 596.71,
|
| 193518 |
+
"learning_rate": 8.83051282051282e-06,
|
| 193519 |
+
"loss": 0.3834,
|
| 193520 |
+
"step": 74005
|
| 193521 |
+
},
|
| 193522 |
+
{
|
| 193523 |
+
"epoch": 596.76,
|
| 193524 |
+
"learning_rate": 8.830432692307694e-06,
|
| 193525 |
+
"loss": 0.366,
|
| 193526 |
+
"step": 74010
|
| 193527 |
+
},
|
| 193528 |
+
{
|
| 193529 |
+
"epoch": 596.8,
|
| 193530 |
+
"learning_rate": 8.830352564102565e-06,
|
| 193531 |
+
"loss": 0.9071,
|
| 193532 |
+
"step": 74015
|
| 193533 |
+
},
|
| 193534 |
+
{
|
| 193535 |
+
"epoch": 596.84,
|
| 193536 |
+
"learning_rate": 8.830272435897437e-06,
|
| 193537 |
+
"loss": 0.6363,
|
| 193538 |
+
"step": 74020
|
| 193539 |
+
},
|
| 193540 |
+
{
|
| 193541 |
+
"epoch": 596.88,
|
| 193542 |
+
"learning_rate": 8.830192307692308e-06,
|
| 193543 |
+
"loss": 0.3411,
|
| 193544 |
+
"step": 74025
|
| 193545 |
+
},
|
| 193546 |
+
{
|
| 193547 |
+
"epoch": 596.92,
|
| 193548 |
+
"learning_rate": 8.830112179487181e-06,
|
| 193549 |
+
"loss": 0.3157,
|
| 193550 |
+
"step": 74030
|
| 193551 |
+
},
|
| 193552 |
+
{
|
| 193553 |
+
"epoch": 596.96,
|
| 193554 |
+
"learning_rate": 8.830032051282052e-06,
|
| 193555 |
+
"loss": 0.3928,
|
| 193556 |
+
"step": 74035
|
| 193557 |
+
},
|
| 193558 |
+
{
|
| 193559 |
+
"epoch": 597.0,
|
| 193560 |
+
"learning_rate": 8.829951923076924e-06,
|
| 193561 |
+
"loss": 1.1771,
|
| 193562 |
+
"step": 74040
|
| 193563 |
+
},
|
| 193564 |
+
{
|
| 193565 |
+
"epoch": 597.0,
|
| 193566 |
+
"eval_loss": 0.5054441690444946,
|
| 193567 |
+
"eval_runtime": 40.5036,
|
| 193568 |
+
"eval_samples_per_second": 20.739,
|
| 193569 |
+
"eval_steps_per_second": 0.667,
|
| 193570 |
+
"eval_wer": 0.20609675281643472,
|
| 193571 |
+
"step": 74040
|
| 193572 |
+
},
|
| 193573 |
+
{
|
| 193574 |
+
"epoch": 592.04,
|
| 193575 |
+
"learning_rate": 8.829871794871797e-06,
|
| 193576 |
+
"loss": 0.3384,
|
| 193577 |
+
"step": 74045
|
| 193578 |
+
},
|
| 193579 |
+
{
|
| 193580 |
+
"epoch": 592.08,
|
| 193581 |
+
"learning_rate": 8.829791666666666e-06,
|
| 193582 |
+
"loss": 0.3687,
|
| 193583 |
+
"step": 74050
|
| 193584 |
+
},
|
| 193585 |
+
{
|
| 193586 |
+
"epoch": 592.12,
|
| 193587 |
+
"learning_rate": 8.82971153846154e-06,
|
| 193588 |
+
"loss": 0.3776,
|
| 193589 |
+
"step": 74055
|
| 193590 |
+
},
|
| 193591 |
+
{
|
| 193592 |
+
"epoch": 592.16,
|
| 193593 |
+
"learning_rate": 8.82963141025641e-06,
|
| 193594 |
+
"loss": 0.4675,
|
| 193595 |
+
"step": 74060
|
| 193596 |
+
},
|
| 193597 |
+
{
|
| 193598 |
+
"epoch": 592.2,
|
| 193599 |
+
"learning_rate": 8.829551282051282e-06,
|
| 193600 |
+
"loss": 1.0907,
|
| 193601 |
+
"step": 74065
|
| 193602 |
+
},
|
| 193603 |
+
{
|
| 193604 |
+
"epoch": 592.24,
|
| 193605 |
+
"learning_rate": 8.829471153846153e-06,
|
| 193606 |
+
"loss": 0.4606,
|
| 193607 |
+
"step": 74070
|
| 193608 |
+
},
|
| 193609 |
+
{
|
| 193610 |
+
"epoch": 592.28,
|
| 193611 |
+
"learning_rate": 8.829391025641027e-06,
|
| 193612 |
+
"loss": 0.2879,
|
| 193613 |
+
"step": 74075
|
| 193614 |
+
},
|
| 193615 |
+
{
|
| 193616 |
+
"epoch": 592.32,
|
| 193617 |
+
"learning_rate": 8.829310897435898e-06,
|
| 193618 |
+
"loss": 0.3348,
|
| 193619 |
+
"step": 74080
|
| 193620 |
+
},
|
| 193621 |
+
{
|
| 193622 |
+
"epoch": 592.36,
|
| 193623 |
+
"learning_rate": 8.82923076923077e-06,
|
| 193624 |
+
"loss": 0.5071,
|
| 193625 |
+
"step": 74085
|
| 193626 |
+
},
|
| 193627 |
+
{
|
| 193628 |
+
"epoch": 592.4,
|
| 193629 |
+
"learning_rate": 8.829150641025642e-06,
|
| 193630 |
+
"loss": 1.1857,
|
| 193631 |
+
"step": 74090
|
| 193632 |
+
},
|
| 193633 |
+
{
|
| 193634 |
+
"epoch": 592.44,
|
| 193635 |
+
"learning_rate": 8.829070512820514e-06,
|
| 193636 |
+
"loss": 0.3519,
|
| 193637 |
+
"step": 74095
|
| 193638 |
+
},
|
| 193639 |
+
{
|
| 193640 |
+
"epoch": 592.48,
|
| 193641 |
+
"learning_rate": 8.828990384615385e-06,
|
| 193642 |
+
"loss": 0.2731,
|
| 193643 |
+
"step": 74100
|
| 193644 |
+
},
|
| 193645 |
+
{
|
| 193646 |
+
"epoch": 592.52,
|
| 193647 |
+
"learning_rate": 8.828910256410256e-06,
|
| 193648 |
+
"loss": 0.3082,
|
| 193649 |
+
"step": 74105
|
| 193650 |
+
},
|
| 193651 |
+
{
|
| 193652 |
+
"epoch": 592.56,
|
| 193653 |
+
"learning_rate": 8.82883012820513e-06,
|
| 193654 |
+
"loss": 0.4305,
|
| 193655 |
+
"step": 74110
|
| 193656 |
+
},
|
| 193657 |
+
{
|
| 193658 |
+
"epoch": 592.6,
|
| 193659 |
+
"learning_rate": 8.82875e-06,
|
| 193660 |
+
"loss": 1.1372,
|
| 193661 |
+
"step": 74115
|
| 193662 |
+
},
|
| 193663 |
+
{
|
| 193664 |
+
"epoch": 592.64,
|
| 193665 |
+
"learning_rate": 8.828669871794872e-06,
|
| 193666 |
+
"loss": 0.3594,
|
| 193667 |
+
"step": 74120
|
| 193668 |
+
},
|
| 193669 |
+
{
|
| 193670 |
+
"epoch": 592.68,
|
| 193671 |
+
"learning_rate": 8.828589743589744e-06,
|
| 193672 |
+
"loss": 0.3154,
|
| 193673 |
+
"step": 74125
|
| 193674 |
+
},
|
| 193675 |
+
{
|
| 193676 |
+
"epoch": 592.72,
|
| 193677 |
+
"learning_rate": 8.828509615384617e-06,
|
| 193678 |
+
"loss": 0.3338,
|
| 193679 |
+
"step": 74130
|
| 193680 |
+
},
|
| 193681 |
+
{
|
| 193682 |
+
"epoch": 592.76,
|
| 193683 |
+
"learning_rate": 8.828429487179488e-06,
|
| 193684 |
+
"loss": 0.5104,
|
| 193685 |
+
"step": 74135
|
| 193686 |
+
},
|
| 193687 |
+
{
|
| 193688 |
+
"epoch": 592.8,
|
| 193689 |
+
"learning_rate": 8.82834935897436e-06,
|
| 193690 |
+
"loss": 1.0787,
|
| 193691 |
+
"step": 74140
|
| 193692 |
+
},
|
| 193693 |
+
{
|
| 193694 |
+
"epoch": 592.84,
|
| 193695 |
+
"learning_rate": 8.828269230769232e-06,
|
| 193696 |
+
"loss": 0.3784,
|
| 193697 |
+
"step": 74145
|
| 193698 |
+
},
|
| 193699 |
+
{
|
| 193700 |
+
"epoch": 592.88,
|
| 193701 |
+
"learning_rate": 8.828189102564104e-06,
|
| 193702 |
+
"loss": 0.2921,
|
| 193703 |
+
"step": 74150
|
| 193704 |
+
},
|
| 193705 |
+
{
|
| 193706 |
+
"epoch": 592.92,
|
| 193707 |
+
"learning_rate": 8.828108974358975e-06,
|
| 193708 |
+
"loss": 0.3931,
|
| 193709 |
+
"step": 74155
|
| 193710 |
+
},
|
| 193711 |
+
{
|
| 193712 |
+
"epoch": 592.96,
|
| 193713 |
+
"learning_rate": 8.828028846153846e-06,
|
| 193714 |
+
"loss": 0.5299,
|
| 193715 |
+
"step": 74160
|
| 193716 |
+
},
|
| 193717 |
+
{
|
| 193718 |
+
"epoch": 593.0,
|
| 193719 |
+
"learning_rate": 8.82794871794872e-06,
|
| 193720 |
+
"loss": 1.2197,
|
| 193721 |
+
"step": 74165
|
| 193722 |
+
},
|
| 193723 |
+
{
|
| 193724 |
+
"epoch": 593.0,
|
| 193725 |
+
"eval_loss": 0.4171510636806488,
|
| 193726 |
+
"eval_runtime": 40.75,
|
| 193727 |
+
"eval_samples_per_second": 20.638,
|
| 193728 |
+
"eval_steps_per_second": 0.663,
|
| 193729 |
+
"eval_wer": 0.20073123521399383,
|
| 193730 |
+
"step": 74165
|
| 193731 |
+
},
|
| 193732 |
+
{
|
| 193733 |
+
"epoch": 593.04,
|
| 193734 |
+
"learning_rate": 8.82786858974359e-06,
|
| 193735 |
+
"loss": 0.3557,
|
| 193736 |
+
"step": 74170
|
| 193737 |
+
},
|
| 193738 |
+
{
|
| 193739 |
+
"epoch": 593.08,
|
| 193740 |
+
"learning_rate": 8.827788461538462e-06,
|
| 193741 |
+
"loss": 0.3218,
|
| 193742 |
+
"step": 74175
|
| 193743 |
+
},
|
| 193744 |
+
{
|
| 193745 |
+
"epoch": 593.12,
|
| 193746 |
+
"learning_rate": 8.827708333333334e-06,
|
| 193747 |
+
"loss": 0.3495,
|
| 193748 |
+
"step": 74180
|
| 193749 |
+
},
|
| 193750 |
+
{
|
| 193751 |
+
"epoch": 593.16,
|
| 193752 |
+
"learning_rate": 8.827628205128207e-06,
|
| 193753 |
+
"loss": 0.5121,
|
| 193754 |
+
"step": 74185
|
| 193755 |
+
},
|
| 193756 |
+
{
|
| 193757 |
+
"epoch": 593.2,
|
| 193758 |
+
"learning_rate": 8.827548076923078e-06,
|
| 193759 |
+
"loss": 1.1842,
|
| 193760 |
+
"step": 74190
|
| 193761 |
+
},
|
| 193762 |
+
{
|
| 193763 |
+
"epoch": 593.24,
|
| 193764 |
+
"learning_rate": 8.82746794871795e-06,
|
| 193765 |
+
"loss": 0.3834,
|
| 193766 |
+
"step": 74195
|
| 193767 |
+
},
|
| 193768 |
+
{
|
| 193769 |
+
"epoch": 593.28,
|
| 193770 |
+
"learning_rate": 8.827387820512822e-06,
|
| 193771 |
+
"loss": 0.3191,
|
| 193772 |
+
"step": 74200
|
| 193773 |
+
},
|
| 193774 |
+
{
|
| 193775 |
+
"epoch": 593.32,
|
| 193776 |
+
"learning_rate": 8.827307692307692e-06,
|
| 193777 |
+
"loss": 0.3646,
|
| 193778 |
+
"step": 74205
|
| 193779 |
+
},
|
| 193780 |
+
{
|
| 193781 |
+
"epoch": 593.36,
|
| 193782 |
+
"learning_rate": 8.827227564102565e-06,
|
| 193783 |
+
"loss": 0.4677,
|
| 193784 |
+
"step": 74210
|
| 193785 |
+
},
|
| 193786 |
+
{
|
| 193787 |
+
"epoch": 593.4,
|
| 193788 |
+
"learning_rate": 8.827147435897436e-06,
|
| 193789 |
+
"loss": 1.2346,
|
| 193790 |
+
"step": 74215
|
| 193791 |
+
},
|
| 193792 |
+
{
|
| 193793 |
+
"epoch": 593.44,
|
| 193794 |
+
"learning_rate": 8.827067307692308e-06,
|
| 193795 |
+
"loss": 0.3663,
|
| 193796 |
+
"step": 74220
|
| 193797 |
+
},
|
| 193798 |
+
{
|
| 193799 |
+
"epoch": 593.48,
|
| 193800 |
+
"learning_rate": 8.82698717948718e-06,
|
| 193801 |
+
"loss": 0.2893,
|
| 193802 |
+
"step": 74225
|
| 193803 |
+
},
|
| 193804 |
+
{
|
| 193805 |
+
"epoch": 593.52,
|
| 193806 |
+
"learning_rate": 8.826907051282052e-06,
|
| 193807 |
+
"loss": 0.334,
|
| 193808 |
+
"step": 74230
|
| 193809 |
+
},
|
| 193810 |
+
{
|
| 193811 |
+
"epoch": 593.56,
|
| 193812 |
+
"learning_rate": 8.826826923076924e-06,
|
| 193813 |
+
"loss": 0.5513,
|
| 193814 |
+
"step": 74235
|
| 193815 |
+
},
|
| 193816 |
+
{
|
| 193817 |
+
"epoch": 593.6,
|
| 193818 |
+
"learning_rate": 8.826746794871795e-06,
|
| 193819 |
+
"loss": 1.281,
|
| 193820 |
+
"step": 74240
|
| 193821 |
+
},
|
| 193822 |
+
{
|
| 193823 |
+
"epoch": 593.64,
|
| 193824 |
+
"learning_rate": 8.826666666666668e-06,
|
| 193825 |
+
"loss": 0.3202,
|
| 193826 |
+
"step": 74245
|
| 193827 |
+
},
|
| 193828 |
+
{
|
| 193829 |
+
"epoch": 593.68,
|
| 193830 |
+
"learning_rate": 8.82658653846154e-06,
|
| 193831 |
+
"loss": 0.3076,
|
| 193832 |
+
"step": 74250
|
| 193833 |
+
},
|
| 193834 |
+
{
|
| 193835 |
+
"epoch": 593.72,
|
| 193836 |
+
"learning_rate": 8.82650641025641e-06,
|
| 193837 |
+
"loss": 0.3823,
|
| 193838 |
+
"step": 74255
|
| 193839 |
+
},
|
| 193840 |
+
{
|
| 193841 |
+
"epoch": 593.76,
|
| 193842 |
+
"learning_rate": 8.826426282051282e-06,
|
| 193843 |
+
"loss": 0.4498,
|
| 193844 |
+
"step": 74260
|
| 193845 |
+
},
|
| 193846 |
+
{
|
| 193847 |
+
"epoch": 593.8,
|
| 193848 |
+
"learning_rate": 8.826346153846155e-06,
|
| 193849 |
+
"loss": 1.3628,
|
| 193850 |
+
"step": 74265
|
| 193851 |
+
},
|
| 193852 |
+
{
|
| 193853 |
+
"epoch": 593.84,
|
| 193854 |
+
"learning_rate": 8.826266025641026e-06,
|
| 193855 |
+
"loss": 0.3403,
|
| 193856 |
+
"step": 74270
|
| 193857 |
+
},
|
| 193858 |
+
{
|
| 193859 |
+
"epoch": 593.88,
|
| 193860 |
+
"learning_rate": 8.826185897435898e-06,
|
| 193861 |
+
"loss": 0.3188,
|
| 193862 |
+
"step": 74275
|
| 193863 |
+
},
|
| 193864 |
+
{
|
| 193865 |
+
"epoch": 593.92,
|
| 193866 |
+
"learning_rate": 8.82610576923077e-06,
|
| 193867 |
+
"loss": 0.3402,
|
| 193868 |
+
"step": 74280
|
| 193869 |
+
},
|
| 193870 |
+
{
|
| 193871 |
+
"epoch": 593.96,
|
| 193872 |
+
"learning_rate": 8.826025641025642e-06,
|
| 193873 |
+
"loss": 0.486,
|
| 193874 |
+
"step": 74285
|
| 193875 |
+
},
|
| 193876 |
+
{
|
| 193877 |
+
"epoch": 594.0,
|
| 193878 |
+
"learning_rate": 8.825945512820514e-06,
|
| 193879 |
+
"loss": 1.311,
|
| 193880 |
+
"step": 74290
|
| 193881 |
+
},
|
| 193882 |
+
{
|
| 193883 |
+
"epoch": 594.0,
|
| 193884 |
+
"eval_loss": 0.35683706402778625,
|
| 193885 |
+
"eval_runtime": 40.5555,
|
| 193886 |
+
"eval_samples_per_second": 20.737,
|
| 193887 |
+
"eval_steps_per_second": 0.666,
|
| 193888 |
+
"eval_wer": 0.19015243684248193,
|
| 193889 |
+
"step": 74290
|
| 193890 |
+
},
|
| 193891 |
+
{
|
| 193892 |
+
"epoch": 594.04,
|
| 193893 |
+
"learning_rate": 8.825865384615385e-06,
|
| 193894 |
+
"loss": 0.358,
|
| 193895 |
+
"step": 74295
|
| 193896 |
+
},
|
| 193897 |
+
{
|
| 193898 |
+
"epoch": 594.08,
|
| 193899 |
+
"learning_rate": 8.825785256410258e-06,
|
| 193900 |
+
"loss": 0.2652,
|
| 193901 |
+
"step": 74300
|
| 193902 |
+
},
|
| 193903 |
+
{
|
| 193904 |
+
"epoch": 594.12,
|
| 193905 |
+
"learning_rate": 8.82570512820513e-06,
|
| 193906 |
+
"loss": 0.3892,
|
| 193907 |
+
"step": 74305
|
| 193908 |
+
},
|
| 193909 |
+
{
|
| 193910 |
+
"epoch": 594.16,
|
| 193911 |
+
"learning_rate": 8.825625e-06,
|
| 193912 |
+
"loss": 0.4418,
|
| 193913 |
+
"step": 74310
|
| 193914 |
+
},
|
| 193915 |
+
{
|
| 193916 |
+
"epoch": 594.2,
|
| 193917 |
+
"learning_rate": 8.825544871794872e-06,
|
| 193918 |
+
"loss": 1.2462,
|
| 193919 |
+
"step": 74315
|
| 193920 |
+
},
|
| 193921 |
+
{
|
| 193922 |
+
"epoch": 594.24,
|
| 193923 |
+
"learning_rate": 8.825464743589745e-06,
|
| 193924 |
+
"loss": 0.3345,
|
| 193925 |
+
"step": 74320
|
| 193926 |
+
},
|
| 193927 |
+
{
|
| 193928 |
+
"epoch": 594.28,
|
| 193929 |
+
"learning_rate": 8.825384615384617e-06,
|
| 193930 |
+
"loss": 0.2987,
|
| 193931 |
+
"step": 74325
|
| 193932 |
+
},
|
| 193933 |
+
{
|
| 193934 |
+
"epoch": 594.32,
|
| 193935 |
+
"learning_rate": 8.825304487179488e-06,
|
| 193936 |
+
"loss": 0.2656,
|
| 193937 |
+
"step": 74330
|
| 193938 |
+
},
|
| 193939 |
+
{
|
| 193940 |
+
"epoch": 594.36,
|
| 193941 |
+
"learning_rate": 8.825224358974361e-06,
|
| 193942 |
+
"loss": 0.5242,
|
| 193943 |
+
"step": 74335
|
| 193944 |
+
},
|
| 193945 |
+
{
|
| 193946 |
+
"epoch": 594.4,
|
| 193947 |
+
"learning_rate": 8.825144230769232e-06,
|
| 193948 |
+
"loss": 1.2575,
|
| 193949 |
+
"step": 74340
|
| 193950 |
+
},
|
| 193951 |
+
{
|
| 193952 |
+
"epoch": 594.44,
|
| 193953 |
+
"learning_rate": 8.825064102564104e-06,
|
| 193954 |
+
"loss": 0.3369,
|
| 193955 |
+
"step": 74345
|
| 193956 |
+
},
|
| 193957 |
+
{
|
| 193958 |
+
"epoch": 594.48,
|
| 193959 |
+
"learning_rate": 8.824983974358975e-06,
|
| 193960 |
+
"loss": 0.305,
|
| 193961 |
+
"step": 74350
|
| 193962 |
+
},
|
| 193963 |
+
{
|
| 193964 |
+
"epoch": 594.52,
|
| 193965 |
+
"learning_rate": 8.824903846153848e-06,
|
| 193966 |
+
"loss": 0.3236,
|
| 193967 |
+
"step": 74355
|
| 193968 |
+
},
|
| 193969 |
+
{
|
| 193970 |
+
"epoch": 594.56,
|
| 193971 |
+
"learning_rate": 8.824823717948718e-06,
|
| 193972 |
+
"loss": 0.4855,
|
| 193973 |
+
"step": 74360
|
| 193974 |
+
},
|
| 193975 |
+
{
|
| 193976 |
+
"epoch": 594.6,
|
| 193977 |
+
"learning_rate": 8.82474358974359e-06,
|
| 193978 |
+
"loss": 1.2836,
|
| 193979 |
+
"step": 74365
|
| 193980 |
+
},
|
| 193981 |
+
{
|
| 193982 |
+
"epoch": 594.64,
|
| 193983 |
+
"learning_rate": 8.824663461538462e-06,
|
| 193984 |
+
"loss": 0.3867,
|
| 193985 |
+
"step": 74370
|
| 193986 |
+
},
|
| 193987 |
+
{
|
| 193988 |
+
"epoch": 594.68,
|
| 193989 |
+
"learning_rate": 8.824583333333333e-06,
|
| 193990 |
+
"loss": 0.3239,
|
| 193991 |
+
"step": 74375
|
| 193992 |
+
},
|
| 193993 |
+
{
|
| 193994 |
+
"epoch": 594.72,
|
| 193995 |
+
"learning_rate": 8.824503205128205e-06,
|
| 193996 |
+
"loss": 0.4218,
|
| 193997 |
+
"step": 74380
|
| 193998 |
+
},
|
| 193999 |
+
{
|
| 194000 |
+
"epoch": 594.76,
|
| 194001 |
+
"learning_rate": 8.824423076923078e-06,
|
| 194002 |
+
"loss": 0.5141,
|
| 194003 |
+
"step": 74385
|
| 194004 |
+
},
|
| 194005 |
+
{
|
| 194006 |
+
"epoch": 594.8,
|
| 194007 |
+
"learning_rate": 8.82434294871795e-06,
|
| 194008 |
+
"loss": 1.3074,
|
| 194009 |
+
"step": 74390
|
| 194010 |
+
},
|
| 194011 |
+
{
|
| 194012 |
+
"epoch": 594.84,
|
| 194013 |
+
"learning_rate": 8.82426282051282e-06,
|
| 194014 |
+
"loss": 0.3945,
|
| 194015 |
+
"step": 74395
|
| 194016 |
+
},
|
| 194017 |
+
{
|
| 194018 |
+
"epoch": 594.88,
|
| 194019 |
+
"learning_rate": 8.824182692307694e-06,
|
| 194020 |
+
"loss": 0.3776,
|
| 194021 |
+
"step": 74400
|
| 194022 |
+
},
|
| 194023 |
+
{
|
| 194024 |
+
"epoch": 594.92,
|
| 194025 |
+
"learning_rate": 8.824102564102565e-06,
|
| 194026 |
+
"loss": 0.3246,
|
| 194027 |
+
"step": 74405
|
| 194028 |
+
},
|
| 194029 |
+
{
|
| 194030 |
+
"epoch": 594.96,
|
| 194031 |
+
"learning_rate": 8.824022435897436e-06,
|
| 194032 |
+
"loss": 0.4568,
|
| 194033 |
+
"step": 74410
|
| 194034 |
+
},
|
| 194035 |
+
{
|
| 194036 |
+
"epoch": 595.0,
|
| 194037 |
+
"learning_rate": 8.823942307692308e-06,
|
| 194038 |
+
"loss": 1.5204,
|
| 194039 |
+
"step": 74415
|
| 194040 |
+
},
|
| 194041 |
+
{
|
| 194042 |
+
"epoch": 595.0,
|
| 194043 |
+
"eval_loss": 0.437023401260376,
|
| 194044 |
+
"eval_runtime": 39.7077,
|
| 194045 |
+
"eval_samples_per_second": 21.18,
|
| 194046 |
+
"eval_steps_per_second": 0.68,
|
| 194047 |
+
"eval_wer": 0.1853523587652883,
|
| 194048 |
+
"step": 74415
|
| 194049 |
+
},
|
| 194050 |
+
{
|
| 194051 |
+
"epoch": 600.04,
|
| 194052 |
+
"learning_rate": 8.82386217948718e-06,
|
| 194053 |
+
"loss": 0.2979,
|
| 194054 |
+
"step": 74420
|
| 194055 |
+
},
|
| 194056 |
+
{
|
| 194057 |
+
"epoch": 600.08,
|
| 194058 |
+
"learning_rate": 8.823782051282052e-06,
|
| 194059 |
+
"loss": 0.2777,
|
| 194060 |
+
"step": 74425
|
| 194061 |
+
},
|
| 194062 |
+
{
|
| 194063 |
+
"epoch": 600.12,
|
| 194064 |
+
"learning_rate": 8.823701923076924e-06,
|
| 194065 |
+
"loss": 0.3034,
|
| 194066 |
+
"step": 74430
|
| 194067 |
+
},
|
| 194068 |
+
{
|
| 194069 |
+
"epoch": 600.16,
|
| 194070 |
+
"learning_rate": 8.823621794871795e-06,
|
| 194071 |
+
"loss": 0.4698,
|
| 194072 |
+
"step": 74435
|
| 194073 |
+
},
|
| 194074 |
+
{
|
| 194075 |
+
"epoch": 600.2,
|
| 194076 |
+
"learning_rate": 8.823541666666668e-06,
|
| 194077 |
+
"loss": 1.2657,
|
| 194078 |
+
"step": 74440
|
| 194079 |
+
},
|
| 194080 |
+
{
|
| 194081 |
+
"epoch": 600.24,
|
| 194082 |
+
"learning_rate": 8.82346153846154e-06,
|
| 194083 |
+
"loss": 0.3598,
|
| 194084 |
+
"step": 74445
|
| 194085 |
+
},
|
| 194086 |
+
{
|
| 194087 |
+
"epoch": 600.28,
|
| 194088 |
+
"learning_rate": 8.82338141025641e-06,
|
| 194089 |
+
"loss": 0.3204,
|
| 194090 |
+
"step": 74450
|
| 194091 |
+
},
|
| 194092 |
+
{
|
| 194093 |
+
"epoch": 600.32,
|
| 194094 |
+
"learning_rate": 8.823301282051284e-06,
|
| 194095 |
+
"loss": 0.3404,
|
| 194096 |
+
"step": 74455
|
| 194097 |
+
},
|
| 194098 |
+
{
|
| 194099 |
+
"epoch": 600.36,
|
| 194100 |
+
"learning_rate": 8.823221153846155e-06,
|
| 194101 |
+
"loss": 0.4718,
|
| 194102 |
+
"step": 74460
|
| 194103 |
+
},
|
| 194104 |
+
{
|
| 194105 |
+
"epoch": 600.4,
|
| 194106 |
+
"learning_rate": 8.823141025641026e-06,
|
| 194107 |
+
"loss": 1.1487,
|
| 194108 |
+
"step": 74465
|
| 194109 |
+
},
|
| 194110 |
+
{
|
| 194111 |
+
"epoch": 600.44,
|
| 194112 |
+
"learning_rate": 8.823060897435898e-06,
|
| 194113 |
+
"loss": 0.337,
|
| 194114 |
+
"step": 74470
|
| 194115 |
+
},
|
| 194116 |
+
{
|
| 194117 |
+
"epoch": 600.48,
|
| 194118 |
+
"learning_rate": 8.82298076923077e-06,
|
| 194119 |
+
"loss": 0.2972,
|
| 194120 |
+
"step": 74475
|
| 194121 |
+
},
|
| 194122 |
+
{
|
| 194123 |
+
"epoch": 600.52,
|
| 194124 |
+
"learning_rate": 8.82290064102564e-06,
|
| 194125 |
+
"loss": 0.3133,
|
| 194126 |
+
"step": 74480
|
| 194127 |
+
},
|
| 194128 |
+
{
|
| 194129 |
+
"epoch": 600.56,
|
| 194130 |
+
"learning_rate": 8.822820512820514e-06,
|
| 194131 |
+
"loss": 0.5256,
|
| 194132 |
+
"step": 74485
|
| 194133 |
+
},
|
| 194134 |
+
{
|
| 194135 |
+
"epoch": 600.6,
|
| 194136 |
+
"learning_rate": 8.822740384615387e-06,
|
| 194137 |
+
"loss": 1.1288,
|
| 194138 |
+
"step": 74490
|
| 194139 |
+
},
|
| 194140 |
+
{
|
| 194141 |
+
"epoch": 600.64,
|
| 194142 |
+
"learning_rate": 8.822660256410256e-06,
|
| 194143 |
+
"loss": 0.3383,
|
| 194144 |
+
"step": 74495
|
| 194145 |
+
},
|
| 194146 |
+
{
|
| 194147 |
+
"epoch": 600.68,
|
| 194148 |
+
"learning_rate": 8.82258012820513e-06,
|
| 194149 |
+
"loss": 0.3663,
|
| 194150 |
+
"step": 74500
|
| 194151 |
+
},
|
| 194152 |
+
{
|
| 194153 |
+
"epoch": 600.72,
|
| 194154 |
+
"learning_rate": 8.8225e-06,
|
| 194155 |
+
"loss": 0.5953,
|
| 194156 |
+
"step": 74505
|
| 194157 |
+
},
|
| 194158 |
+
{
|
| 194159 |
+
"epoch": 600.76,
|
| 194160 |
+
"learning_rate": 8.822419871794872e-06,
|
| 194161 |
+
"loss": 0.4843,
|
| 194162 |
+
"step": 74510
|
| 194163 |
+
},
|
| 194164 |
+
{
|
| 194165 |
+
"epoch": 600.8,
|
| 194166 |
+
"learning_rate": 8.822339743589743e-06,
|
| 194167 |
+
"loss": 1.1774,
|
| 194168 |
+
"step": 74515
|
| 194169 |
+
},
|
| 194170 |
+
{
|
| 194171 |
+
"epoch": 600.84,
|
| 194172 |
+
"learning_rate": 8.822259615384616e-06,
|
| 194173 |
+
"loss": 0.3648,
|
| 194174 |
+
"step": 74520
|
| 194175 |
+
},
|
| 194176 |
+
{
|
| 194177 |
+
"epoch": 600.88,
|
| 194178 |
+
"learning_rate": 8.822179487179488e-06,
|
| 194179 |
+
"loss": 0.312,
|
| 194180 |
+
"step": 74525
|
| 194181 |
+
},
|
| 194182 |
+
{
|
| 194183 |
+
"epoch": 600.92,
|
| 194184 |
+
"learning_rate": 8.822099358974359e-06,
|
| 194185 |
+
"loss": 0.3538,
|
| 194186 |
+
"step": 74530
|
| 194187 |
+
},
|
| 194188 |
+
{
|
| 194189 |
+
"epoch": 600.96,
|
| 194190 |
+
"learning_rate": 8.82201923076923e-06,
|
| 194191 |
+
"loss": 0.4753,
|
| 194192 |
+
"step": 74535
|
| 194193 |
+
},
|
| 194194 |
+
{
|
| 194195 |
+
"epoch": 601.0,
|
| 194196 |
+
"eval_loss": 0.4208657741546631,
|
| 194197 |
+
"eval_runtime": 40.5395,
|
| 194198 |
+
"eval_samples_per_second": 20.745,
|
| 194199 |
+
"eval_steps_per_second": 0.666,
|
| 194200 |
+
"eval_wer": 0.18913761467889909,
|
| 194201 |
+
"step": 74539
|
| 194202 |
}
|
| 194203 |
],
|
| 194204 |
"max_steps": 620000,
|
| 194205 |
"num_train_epochs": 5000,
|
| 194206 |
+
"total_flos": 2.097692339430875e+20,
|
| 194207 |
"trial_name": null,
|
| 194208 |
"trial_params": null
|
| 194209 |
}
|
model-bin/finetune/base/{checkpoint-73916 → checkpoint-74539}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629875783.5501642/events.out.tfevents.1629875783.7e498afd5545.905.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:56464a81e08d7f1a82ade83b736f0756eb87693e0ddad64d889d382198e563fd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629876453.8615327/events.out.tfevents.1629876453.7e498afd5545.905.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be4bfbaec62c29f516a7e9a9d3917649f8fe892178caed42353cf30289db3e17
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629877099.8360853/events.out.tfevents.1629877099.7e498afd5545.905.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:645beaf58ccf8e567b5e37c2c291607674188f19f50f3e7ee43210e6407638dd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629877746.4587185/events.out.tfevents.1629877746.7e498afd5545.905.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f60338304f297ccf1a1c0d050fa5246f8e8acbd98bdea98b1eef75c28f7d1170
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629878403.3794868/events.out.tfevents.1629878403.7e498afd5545.905.81
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df1922471be3de7f9dd9e4f2a4714803af3f9197b17624b7f01328af60d370dc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629875783.7e498afd5545.905.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77271c3153a042764056f6227bdbbc65ec6f3766b6800dfffce708dfdd9b7f0c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629876453.7e498afd5545.905.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30d0a63e279858ca5f738732cff0f1f32c7986714dad24fba8428eb4bd16a122
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629877099.7e498afd5545.905.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:709a794ebfbf29355911f836c4c1181d57b07ab751c07e59e543d2171a1932a6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629877746.7e498afd5545.905.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:32769bd5c161b9f9e3573e58a8963e15f99507303208c12bcc0e1745c936edc3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629878403.7e498afd5545.905.80
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9de9c1b83cc90e427f15e980270464cf53514719155c0e57b3b2b6f8df1b48f1
|
| 3 |
+
size 8462
|