"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629975667.4686/events.out.tfevents.1629975667.8e89bd551565.924.161 +3 -0
- model-bin/finetune/base/log/1629976112.672869/events.out.tfevents.1629976112.8e89bd551565.924.163 +3 -0
- model-bin/finetune/base/log/1629976536.284601/events.out.tfevents.1629976536.8e89bd551565.924.165 +3 -0
- model-bin/finetune/base/log/1629976963.4914813/events.out.tfevents.1629976963.8e89bd551565.924.167 +3 -0
- model-bin/finetune/base/log/1629977393.2955256/events.out.tfevents.1629977393.8e89bd551565.924.169 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629975667.8e89bd551565.924.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629976111.8e89bd551565.924.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629976536.8e89bd551565.924.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629976963.8e89bd551565.924.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629977393.8e89bd551565.924.168 +3 -0
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/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:f4e96b3a9a7d55cdd649d6e098224a0ba7192413a6a8708f7a40422b95863d3a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/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 sha256:2dea647f3d6b01e5bbe480a793f8b955cd7c9087775bb765c0df8615a04dc817
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/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 sha256:4a0a4c7a87b8f0ee333b56a1d258e67703ea75ba69bb081514dcb8ceb484c1ca
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/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 sha256:1bcad1740345a45f9d66c9221f63f3f5320f677aee8ad857a6f8952fefe018d3
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/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:f9973f16c8883da62d99f7690ea7b4a1fc2ce70ac02f9a012439717bbf8226fd
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -224454,11 +224454,806 @@
|
|
| 224454 |
"eval_steps_per_second": 0.714,
|
| 224455 |
"eval_wer": 0.186774099883856,
|
| 224456 |
"step": 98319
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 224457 |
}
|
| 224458 |
],
|
| 224459 |
-
"max_steps":
|
| 224460 |
"num_train_epochs": 5000,
|
| 224461 |
-
"total_flos": 2.
|
| 224462 |
"trial_name": null,
|
| 224463 |
"trial_params": null
|
| 224464 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 797.0,
|
| 5 |
+
"global_step": 98940,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 224454 |
"eval_steps_per_second": 0.714,
|
| 224455 |
"eval_wer": 0.186774099883856,
|
| 224456 |
"step": 98319
|
| 224457 |
+
},
|
| 224458 |
+
{
|
| 224459 |
+
"epoch": 786.01,
|
| 224460 |
+
"learning_rate": 8.441025641025642e-06,
|
| 224461 |
+
"loss": 0.4183,
|
| 224462 |
+
"step": 98320
|
| 224463 |
+
},
|
| 224464 |
+
{
|
| 224465 |
+
"epoch": 786.05,
|
| 224466 |
+
"learning_rate": 8.440945512820513e-06,
|
| 224467 |
+
"loss": 0.3092,
|
| 224468 |
+
"step": 98325
|
| 224469 |
+
},
|
| 224470 |
+
{
|
| 224471 |
+
"epoch": 786.09,
|
| 224472 |
+
"learning_rate": 8.440865384615386e-06,
|
| 224473 |
+
"loss": 0.4855,
|
| 224474 |
+
"step": 98330
|
| 224475 |
+
},
|
| 224476 |
+
{
|
| 224477 |
+
"epoch": 786.13,
|
| 224478 |
+
"learning_rate": 8.440785256410256e-06,
|
| 224479 |
+
"loss": 0.3628,
|
| 224480 |
+
"step": 98335
|
| 224481 |
+
},
|
| 224482 |
+
{
|
| 224483 |
+
"epoch": 786.17,
|
| 224484 |
+
"learning_rate": 8.440705128205129e-06,
|
| 224485 |
+
"loss": 0.5092,
|
| 224486 |
+
"step": 98340
|
| 224487 |
+
},
|
| 224488 |
+
{
|
| 224489 |
+
"epoch": 786.21,
|
| 224490 |
+
"learning_rate": 8.440625000000002e-06,
|
| 224491 |
+
"loss": 1.2041,
|
| 224492 |
+
"step": 98345
|
| 224493 |
+
},
|
| 224494 |
+
{
|
| 224495 |
+
"epoch": 786.25,
|
| 224496 |
+
"learning_rate": 8.440544871794872e-06,
|
| 224497 |
+
"loss": 0.3245,
|
| 224498 |
+
"step": 98350
|
| 224499 |
+
},
|
| 224500 |
+
{
|
| 224501 |
+
"epoch": 786.29,
|
| 224502 |
+
"learning_rate": 8.440464743589745e-06,
|
| 224503 |
+
"loss": 0.304,
|
| 224504 |
+
"step": 98355
|
| 224505 |
+
},
|
| 224506 |
+
{
|
| 224507 |
+
"epoch": 786.33,
|
| 224508 |
+
"learning_rate": 8.440384615384616e-06,
|
| 224509 |
+
"loss": 0.314,
|
| 224510 |
+
"step": 98360
|
| 224511 |
+
},
|
| 224512 |
+
{
|
| 224513 |
+
"epoch": 786.37,
|
| 224514 |
+
"learning_rate": 8.440304487179487e-06,
|
| 224515 |
+
"loss": 0.6307,
|
| 224516 |
+
"step": 98365
|
| 224517 |
+
},
|
| 224518 |
+
{
|
| 224519 |
+
"epoch": 786.41,
|
| 224520 |
+
"learning_rate": 8.440224358974359e-06,
|
| 224521 |
+
"loss": 1.3067,
|
| 224522 |
+
"step": 98370
|
| 224523 |
+
},
|
| 224524 |
+
{
|
| 224525 |
+
"epoch": 786.45,
|
| 224526 |
+
"learning_rate": 8.440144230769232e-06,
|
| 224527 |
+
"loss": 0.2868,
|
| 224528 |
+
"step": 98375
|
| 224529 |
+
},
|
| 224530 |
+
{
|
| 224531 |
+
"epoch": 786.49,
|
| 224532 |
+
"learning_rate": 8.440064102564103e-06,
|
| 224533 |
+
"loss": 0.3017,
|
| 224534 |
+
"step": 98380
|
| 224535 |
+
},
|
| 224536 |
+
{
|
| 224537 |
+
"epoch": 786.53,
|
| 224538 |
+
"learning_rate": 8.439983974358974e-06,
|
| 224539 |
+
"loss": 0.4023,
|
| 224540 |
+
"step": 98385
|
| 224541 |
+
},
|
| 224542 |
+
{
|
| 224543 |
+
"epoch": 786.57,
|
| 224544 |
+
"learning_rate": 8.439903846153847e-06,
|
| 224545 |
+
"loss": 0.634,
|
| 224546 |
+
"step": 98390
|
| 224547 |
+
},
|
| 224548 |
+
{
|
| 224549 |
+
"epoch": 786.61,
|
| 224550 |
+
"learning_rate": 8.439823717948719e-06,
|
| 224551 |
+
"loss": 1.2596,
|
| 224552 |
+
"step": 98395
|
| 224553 |
+
},
|
| 224554 |
+
{
|
| 224555 |
+
"epoch": 786.65,
|
| 224556 |
+
"learning_rate": 8.43974358974359e-06,
|
| 224557 |
+
"loss": 0.3525,
|
| 224558 |
+
"step": 98400
|
| 224559 |
+
},
|
| 224560 |
+
{
|
| 224561 |
+
"epoch": 786.69,
|
| 224562 |
+
"learning_rate": 8.439663461538462e-06,
|
| 224563 |
+
"loss": 0.3087,
|
| 224564 |
+
"step": 98405
|
| 224565 |
+
},
|
| 224566 |
+
{
|
| 224567 |
+
"epoch": 786.73,
|
| 224568 |
+
"learning_rate": 8.439583333333335e-06,
|
| 224569 |
+
"loss": 0.3531,
|
| 224570 |
+
"step": 98410
|
| 224571 |
+
},
|
| 224572 |
+
{
|
| 224573 |
+
"epoch": 786.77,
|
| 224574 |
+
"learning_rate": 8.439503205128206e-06,
|
| 224575 |
+
"loss": 0.5268,
|
| 224576 |
+
"step": 98415
|
| 224577 |
+
},
|
| 224578 |
+
{
|
| 224579 |
+
"epoch": 786.81,
|
| 224580 |
+
"learning_rate": 8.439423076923077e-06,
|
| 224581 |
+
"loss": 1.0259,
|
| 224582 |
+
"step": 98420
|
| 224583 |
+
},
|
| 224584 |
+
{
|
| 224585 |
+
"epoch": 786.85,
|
| 224586 |
+
"learning_rate": 8.439342948717949e-06,
|
| 224587 |
+
"loss": 0.3175,
|
| 224588 |
+
"step": 98425
|
| 224589 |
+
},
|
| 224590 |
+
{
|
| 224591 |
+
"epoch": 786.89,
|
| 224592 |
+
"learning_rate": 8.439262820512822e-06,
|
| 224593 |
+
"loss": 0.3156,
|
| 224594 |
+
"step": 98430
|
| 224595 |
+
},
|
| 224596 |
+
{
|
| 224597 |
+
"epoch": 786.93,
|
| 224598 |
+
"learning_rate": 8.439182692307693e-06,
|
| 224599 |
+
"loss": 0.327,
|
| 224600 |
+
"step": 98435
|
| 224601 |
+
},
|
| 224602 |
+
{
|
| 224603 |
+
"epoch": 786.97,
|
| 224604 |
+
"learning_rate": 8.439102564102564e-06,
|
| 224605 |
+
"loss": 0.6287,
|
| 224606 |
+
"step": 98440
|
| 224607 |
+
},
|
| 224608 |
+
{
|
| 224609 |
+
"epoch": 787.0,
|
| 224610 |
+
"eval_loss": 0.3264864385128021,
|
| 224611 |
+
"eval_runtime": 38.17,
|
| 224612 |
+
"eval_samples_per_second": 21.823,
|
| 224613 |
+
"eval_steps_per_second": 0.707,
|
| 224614 |
+
"eval_wer": 0.19518141402552702,
|
| 224615 |
+
"step": 98444
|
| 224616 |
+
},
|
| 224617 |
+
{
|
| 224618 |
+
"epoch": 793.01,
|
| 224619 |
+
"learning_rate": 8.439022435897438e-06,
|
| 224620 |
+
"loss": 0.3977,
|
| 224621 |
+
"step": 98445
|
| 224622 |
+
},
|
| 224623 |
+
{
|
| 224624 |
+
"epoch": 793.05,
|
| 224625 |
+
"learning_rate": 8.438942307692309e-06,
|
| 224626 |
+
"loss": 0.3203,
|
| 224627 |
+
"step": 98450
|
| 224628 |
+
},
|
| 224629 |
+
{
|
| 224630 |
+
"epoch": 793.09,
|
| 224631 |
+
"learning_rate": 8.43886217948718e-06,
|
| 224632 |
+
"loss": 0.3443,
|
| 224633 |
+
"step": 98455
|
| 224634 |
+
},
|
| 224635 |
+
{
|
| 224636 |
+
"epoch": 793.13,
|
| 224637 |
+
"learning_rate": 8.438782051282052e-06,
|
| 224638 |
+
"loss": 0.3024,
|
| 224639 |
+
"step": 98460
|
| 224640 |
+
},
|
| 224641 |
+
{
|
| 224642 |
+
"epoch": 793.17,
|
| 224643 |
+
"learning_rate": 8.438701923076925e-06,
|
| 224644 |
+
"loss": 0.5617,
|
| 224645 |
+
"step": 98465
|
| 224646 |
+
},
|
| 224647 |
+
{
|
| 224648 |
+
"epoch": 793.21,
|
| 224649 |
+
"learning_rate": 8.438621794871796e-06,
|
| 224650 |
+
"loss": 1.1566,
|
| 224651 |
+
"step": 98470
|
| 224652 |
+
},
|
| 224653 |
+
{
|
| 224654 |
+
"epoch": 793.25,
|
| 224655 |
+
"learning_rate": 8.438541666666667e-06,
|
| 224656 |
+
"loss": 0.3207,
|
| 224657 |
+
"step": 98475
|
| 224658 |
+
},
|
| 224659 |
+
{
|
| 224660 |
+
"epoch": 793.29,
|
| 224661 |
+
"learning_rate": 8.438461538461539e-06,
|
| 224662 |
+
"loss": 0.9967,
|
| 224663 |
+
"step": 98480
|
| 224664 |
+
},
|
| 224665 |
+
{
|
| 224666 |
+
"epoch": 793.33,
|
| 224667 |
+
"learning_rate": 8.438381410256412e-06,
|
| 224668 |
+
"loss": 0.3944,
|
| 224669 |
+
"step": 98485
|
| 224670 |
+
},
|
| 224671 |
+
{
|
| 224672 |
+
"epoch": 793.37,
|
| 224673 |
+
"learning_rate": 8.438301282051283e-06,
|
| 224674 |
+
"loss": 0.6135,
|
| 224675 |
+
"step": 98490
|
| 224676 |
+
},
|
| 224677 |
+
{
|
| 224678 |
+
"epoch": 793.41,
|
| 224679 |
+
"learning_rate": 8.438221153846154e-06,
|
| 224680 |
+
"loss": 1.139,
|
| 224681 |
+
"step": 98495
|
| 224682 |
+
},
|
| 224683 |
+
{
|
| 224684 |
+
"epoch": 793.45,
|
| 224685 |
+
"learning_rate": 8.438141025641028e-06,
|
| 224686 |
+
"loss": 0.3133,
|
| 224687 |
+
"step": 98500
|
| 224688 |
+
},
|
| 224689 |
+
{
|
| 224690 |
+
"epoch": 793.49,
|
| 224691 |
+
"learning_rate": 8.438060897435897e-06,
|
| 224692 |
+
"loss": 0.294,
|
| 224693 |
+
"step": 98505
|
| 224694 |
+
},
|
| 224695 |
+
{
|
| 224696 |
+
"epoch": 793.53,
|
| 224697 |
+
"learning_rate": 8.43798076923077e-06,
|
| 224698 |
+
"loss": 0.3002,
|
| 224699 |
+
"step": 98510
|
| 224700 |
+
},
|
| 224701 |
+
{
|
| 224702 |
+
"epoch": 793.57,
|
| 224703 |
+
"learning_rate": 8.437900641025642e-06,
|
| 224704 |
+
"loss": 0.5191,
|
| 224705 |
+
"step": 98515
|
| 224706 |
+
},
|
| 224707 |
+
{
|
| 224708 |
+
"epoch": 793.61,
|
| 224709 |
+
"learning_rate": 8.437820512820513e-06,
|
| 224710 |
+
"loss": 1.1345,
|
| 224711 |
+
"step": 98520
|
| 224712 |
+
},
|
| 224713 |
+
{
|
| 224714 |
+
"epoch": 793.65,
|
| 224715 |
+
"learning_rate": 8.437740384615384e-06,
|
| 224716 |
+
"loss": 0.3027,
|
| 224717 |
+
"step": 98525
|
| 224718 |
+
},
|
| 224719 |
+
{
|
| 224720 |
+
"epoch": 793.69,
|
| 224721 |
+
"learning_rate": 8.437660256410257e-06,
|
| 224722 |
+
"loss": 0.2889,
|
| 224723 |
+
"step": 98530
|
| 224724 |
+
},
|
| 224725 |
+
{
|
| 224726 |
+
"epoch": 793.73,
|
| 224727 |
+
"learning_rate": 8.437580128205129e-06,
|
| 224728 |
+
"loss": 0.2961,
|
| 224729 |
+
"step": 98535
|
| 224730 |
+
},
|
| 224731 |
+
{
|
| 224732 |
+
"epoch": 793.77,
|
| 224733 |
+
"learning_rate": 8.4375e-06,
|
| 224734 |
+
"loss": 0.5415,
|
| 224735 |
+
"step": 98540
|
| 224736 |
+
},
|
| 224737 |
+
{
|
| 224738 |
+
"epoch": 793.81,
|
| 224739 |
+
"learning_rate": 8.437419871794873e-06,
|
| 224740 |
+
"loss": 1.1181,
|
| 224741 |
+
"step": 98545
|
| 224742 |
+
},
|
| 224743 |
+
{
|
| 224744 |
+
"epoch": 793.85,
|
| 224745 |
+
"learning_rate": 8.437339743589745e-06,
|
| 224746 |
+
"loss": 0.3743,
|
| 224747 |
+
"step": 98550
|
| 224748 |
+
},
|
| 224749 |
+
{
|
| 224750 |
+
"epoch": 793.89,
|
| 224751 |
+
"learning_rate": 8.437259615384616e-06,
|
| 224752 |
+
"loss": 0.2981,
|
| 224753 |
+
"step": 98555
|
| 224754 |
+
},
|
| 224755 |
+
{
|
| 224756 |
+
"epoch": 793.93,
|
| 224757 |
+
"learning_rate": 8.437179487179487e-06,
|
| 224758 |
+
"loss": 0.3788,
|
| 224759 |
+
"step": 98560
|
| 224760 |
+
},
|
| 224761 |
+
{
|
| 224762 |
+
"epoch": 793.97,
|
| 224763 |
+
"learning_rate": 8.43709935897436e-06,
|
| 224764 |
+
"loss": 0.6078,
|
| 224765 |
+
"step": 98565
|
| 224766 |
+
},
|
| 224767 |
+
{
|
| 224768 |
+
"epoch": 794.0,
|
| 224769 |
+
"eval_loss": 0.32572391629219055,
|
| 224770 |
+
"eval_runtime": 37.9988,
|
| 224771 |
+
"eval_samples_per_second": 21.948,
|
| 224772 |
+
"eval_steps_per_second": 0.711,
|
| 224773 |
+
"eval_wer": 0.18024709516105308,
|
| 224774 |
+
"step": 98568
|
| 224775 |
+
},
|
| 224776 |
+
{
|
| 224777 |
+
"epoch": 794.02,
|
| 224778 |
+
"learning_rate": 8.437019230769232e-06,
|
| 224779 |
+
"loss": 0.3149,
|
| 224780 |
+
"step": 98570
|
| 224781 |
+
},
|
| 224782 |
+
{
|
| 224783 |
+
"epoch": 794.06,
|
| 224784 |
+
"learning_rate": 8.436939102564103e-06,
|
| 224785 |
+
"loss": 0.3213,
|
| 224786 |
+
"step": 98575
|
| 224787 |
+
},
|
| 224788 |
+
{
|
| 224789 |
+
"epoch": 794.1,
|
| 224790 |
+
"learning_rate": 8.436858974358974e-06,
|
| 224791 |
+
"loss": 0.2927,
|
| 224792 |
+
"step": 98580
|
| 224793 |
+
},
|
| 224794 |
+
{
|
| 224795 |
+
"epoch": 794.14,
|
| 224796 |
+
"learning_rate": 8.436778846153847e-06,
|
| 224797 |
+
"loss": 0.3499,
|
| 224798 |
+
"step": 98585
|
| 224799 |
+
},
|
| 224800 |
+
{
|
| 224801 |
+
"epoch": 794.18,
|
| 224802 |
+
"learning_rate": 8.436698717948719e-06,
|
| 224803 |
+
"loss": 0.5836,
|
| 224804 |
+
"step": 98590
|
| 224805 |
+
},
|
| 224806 |
+
{
|
| 224807 |
+
"epoch": 794.22,
|
| 224808 |
+
"learning_rate": 8.43661858974359e-06,
|
| 224809 |
+
"loss": 1.0623,
|
| 224810 |
+
"step": 98595
|
| 224811 |
+
},
|
| 224812 |
+
{
|
| 224813 |
+
"epoch": 794.26,
|
| 224814 |
+
"learning_rate": 8.436538461538463e-06,
|
| 224815 |
+
"loss": 0.2589,
|
| 224816 |
+
"step": 98600
|
| 224817 |
+
},
|
| 224818 |
+
{
|
| 224819 |
+
"epoch": 794.3,
|
| 224820 |
+
"learning_rate": 8.436458333333335e-06,
|
| 224821 |
+
"loss": 0.2452,
|
| 224822 |
+
"step": 98605
|
| 224823 |
+
},
|
| 224824 |
+
{
|
| 224825 |
+
"epoch": 794.34,
|
| 224826 |
+
"learning_rate": 8.436378205128206e-06,
|
| 224827 |
+
"loss": 0.3308,
|
| 224828 |
+
"step": 98610
|
| 224829 |
+
},
|
| 224830 |
+
{
|
| 224831 |
+
"epoch": 794.38,
|
| 224832 |
+
"learning_rate": 8.436298076923077e-06,
|
| 224833 |
+
"loss": 0.6364,
|
| 224834 |
+
"step": 98615
|
| 224835 |
+
},
|
| 224836 |
+
{
|
| 224837 |
+
"epoch": 794.42,
|
| 224838 |
+
"learning_rate": 8.43621794871795e-06,
|
| 224839 |
+
"loss": 1.0428,
|
| 224840 |
+
"step": 98620
|
| 224841 |
+
},
|
| 224842 |
+
{
|
| 224843 |
+
"epoch": 794.46,
|
| 224844 |
+
"learning_rate": 8.43613782051282e-06,
|
| 224845 |
+
"loss": 0.3348,
|
| 224846 |
+
"step": 98625
|
| 224847 |
+
},
|
| 224848 |
+
{
|
| 224849 |
+
"epoch": 794.5,
|
| 224850 |
+
"learning_rate": 8.436057692307693e-06,
|
| 224851 |
+
"loss": 0.2523,
|
| 224852 |
+
"step": 98630
|
| 224853 |
+
},
|
| 224854 |
+
{
|
| 224855 |
+
"epoch": 794.54,
|
| 224856 |
+
"learning_rate": 8.435977564102564e-06,
|
| 224857 |
+
"loss": 0.4014,
|
| 224858 |
+
"step": 98635
|
| 224859 |
+
},
|
| 224860 |
+
{
|
| 224861 |
+
"epoch": 794.58,
|
| 224862 |
+
"learning_rate": 8.435897435897436e-06,
|
| 224863 |
+
"loss": 0.6197,
|
| 224864 |
+
"step": 98640
|
| 224865 |
+
},
|
| 224866 |
+
{
|
| 224867 |
+
"epoch": 794.62,
|
| 224868 |
+
"learning_rate": 8.435817307692309e-06,
|
| 224869 |
+
"loss": 0.9188,
|
| 224870 |
+
"step": 98645
|
| 224871 |
+
},
|
| 224872 |
+
{
|
| 224873 |
+
"epoch": 794.66,
|
| 224874 |
+
"learning_rate": 8.43573717948718e-06,
|
| 224875 |
+
"loss": 0.3583,
|
| 224876 |
+
"step": 98650
|
| 224877 |
+
},
|
| 224878 |
+
{
|
| 224879 |
+
"epoch": 794.7,
|
| 224880 |
+
"learning_rate": 8.435657051282053e-06,
|
| 224881 |
+
"loss": 0.3027,
|
| 224882 |
+
"step": 98655
|
| 224883 |
+
},
|
| 224884 |
+
{
|
| 224885 |
+
"epoch": 794.74,
|
| 224886 |
+
"learning_rate": 8.435576923076923e-06,
|
| 224887 |
+
"loss": 0.3345,
|
| 224888 |
+
"step": 98660
|
| 224889 |
+
},
|
| 224890 |
+
{
|
| 224891 |
+
"epoch": 794.78,
|
| 224892 |
+
"learning_rate": 8.435496794871796e-06,
|
| 224893 |
+
"loss": 0.6405,
|
| 224894 |
+
"step": 98665
|
| 224895 |
+
},
|
| 224896 |
+
{
|
| 224897 |
+
"epoch": 794.82,
|
| 224898 |
+
"learning_rate": 8.435416666666667e-06,
|
| 224899 |
+
"loss": 1.0105,
|
| 224900 |
+
"step": 98670
|
| 224901 |
+
},
|
| 224902 |
+
{
|
| 224903 |
+
"epoch": 794.86,
|
| 224904 |
+
"learning_rate": 8.435336538461539e-06,
|
| 224905 |
+
"loss": 0.2594,
|
| 224906 |
+
"step": 98675
|
| 224907 |
+
},
|
| 224908 |
+
{
|
| 224909 |
+
"epoch": 794.9,
|
| 224910 |
+
"learning_rate": 8.43525641025641e-06,
|
| 224911 |
+
"loss": 0.2814,
|
| 224912 |
+
"step": 98680
|
| 224913 |
+
},
|
| 224914 |
+
{
|
| 224915 |
+
"epoch": 794.94,
|
| 224916 |
+
"learning_rate": 8.435176282051283e-06,
|
| 224917 |
+
"loss": 0.6279,
|
| 224918 |
+
"step": 98685
|
| 224919 |
+
},
|
| 224920 |
+
{
|
| 224921 |
+
"epoch": 794.98,
|
| 224922 |
+
"learning_rate": 8.435096153846154e-06,
|
| 224923 |
+
"loss": 0.7894,
|
| 224924 |
+
"step": 98690
|
| 224925 |
+
},
|
| 224926 |
+
{
|
| 224927 |
+
"epoch": 795.0,
|
| 224928 |
+
"eval_loss": 0.4592739939689636,
|
| 224929 |
+
"eval_runtime": 40.0032,
|
| 224930 |
+
"eval_samples_per_second": 20.823,
|
| 224931 |
+
"eval_steps_per_second": 0.675,
|
| 224932 |
+
"eval_wer": 0.18193024436226524,
|
| 224933 |
+
"step": 98692
|
| 224934 |
+
},
|
| 224935 |
+
{
|
| 224936 |
+
"epoch": 795.02,
|
| 224937 |
+
"learning_rate": 8.435016025641026e-06,
|
| 224938 |
+
"loss": 0.3153,
|
| 224939 |
+
"step": 98695
|
| 224940 |
+
},
|
| 224941 |
+
{
|
| 224942 |
+
"epoch": 795.06,
|
| 224943 |
+
"learning_rate": 8.434935897435899e-06,
|
| 224944 |
+
"loss": 0.2995,
|
| 224945 |
+
"step": 98700
|
| 224946 |
+
},
|
| 224947 |
+
{
|
| 224948 |
+
"epoch": 795.1,
|
| 224949 |
+
"learning_rate": 8.43485576923077e-06,
|
| 224950 |
+
"loss": 0.2886,
|
| 224951 |
+
"step": 98705
|
| 224952 |
+
},
|
| 224953 |
+
{
|
| 224954 |
+
"epoch": 795.15,
|
| 224955 |
+
"learning_rate": 8.434775641025642e-06,
|
| 224956 |
+
"loss": 0.39,
|
| 224957 |
+
"step": 98710
|
| 224958 |
+
},
|
| 224959 |
+
{
|
| 224960 |
+
"epoch": 795.19,
|
| 224961 |
+
"learning_rate": 8.434695512820513e-06,
|
| 224962 |
+
"loss": 0.7975,
|
| 224963 |
+
"step": 98715
|
| 224964 |
+
},
|
| 224965 |
+
{
|
| 224966 |
+
"epoch": 795.23,
|
| 224967 |
+
"learning_rate": 8.434615384615386e-06,
|
| 224968 |
+
"loss": 0.864,
|
| 224969 |
+
"step": 98720
|
| 224970 |
+
},
|
| 224971 |
+
{
|
| 224972 |
+
"epoch": 795.27,
|
| 224973 |
+
"learning_rate": 8.434535256410257e-06,
|
| 224974 |
+
"loss": 0.2819,
|
| 224975 |
+
"step": 98725
|
| 224976 |
+
},
|
| 224977 |
+
{
|
| 224978 |
+
"epoch": 795.31,
|
| 224979 |
+
"learning_rate": 8.434455128205129e-06,
|
| 224980 |
+
"loss": 0.3741,
|
| 224981 |
+
"step": 98730
|
| 224982 |
+
},
|
| 224983 |
+
{
|
| 224984 |
+
"epoch": 795.35,
|
| 224985 |
+
"learning_rate": 8.434375e-06,
|
| 224986 |
+
"loss": 0.3993,
|
| 224987 |
+
"step": 98735
|
| 224988 |
+
},
|
| 224989 |
+
{
|
| 224990 |
+
"epoch": 795.39,
|
| 224991 |
+
"learning_rate": 8.434294871794873e-06,
|
| 224992 |
+
"loss": 0.6831,
|
| 224993 |
+
"step": 98740
|
| 224994 |
+
},
|
| 224995 |
+
{
|
| 224996 |
+
"epoch": 795.43,
|
| 224997 |
+
"learning_rate": 8.434214743589744e-06,
|
| 224998 |
+
"loss": 0.9472,
|
| 224999 |
+
"step": 98745
|
| 225000 |
+
},
|
| 225001 |
+
{
|
| 225002 |
+
"epoch": 795.47,
|
| 225003 |
+
"learning_rate": 8.434134615384616e-06,
|
| 225004 |
+
"loss": 0.2791,
|
| 225005 |
+
"step": 98750
|
| 225006 |
+
},
|
| 225007 |
+
{
|
| 225008 |
+
"epoch": 795.51,
|
| 225009 |
+
"learning_rate": 8.434054487179489e-06,
|
| 225010 |
+
"loss": 0.3373,
|
| 225011 |
+
"step": 98755
|
| 225012 |
+
},
|
| 225013 |
+
{
|
| 225014 |
+
"epoch": 795.55,
|
| 225015 |
+
"learning_rate": 8.43397435897436e-06,
|
| 225016 |
+
"loss": 0.3575,
|
| 225017 |
+
"step": 98760
|
| 225018 |
+
},
|
| 225019 |
+
{
|
| 225020 |
+
"epoch": 795.59,
|
| 225021 |
+
"learning_rate": 8.433894230769232e-06,
|
| 225022 |
+
"loss": 0.8391,
|
| 225023 |
+
"step": 98765
|
| 225024 |
+
},
|
| 225025 |
+
{
|
| 225026 |
+
"epoch": 795.63,
|
| 225027 |
+
"learning_rate": 8.433814102564103e-06,
|
| 225028 |
+
"loss": 0.929,
|
| 225029 |
+
"step": 98770
|
| 225030 |
+
},
|
| 225031 |
+
{
|
| 225032 |
+
"epoch": 795.67,
|
| 225033 |
+
"learning_rate": 8.433733974358976e-06,
|
| 225034 |
+
"loss": 0.3038,
|
| 225035 |
+
"step": 98775
|
| 225036 |
+
},
|
| 225037 |
+
{
|
| 225038 |
+
"epoch": 795.71,
|
| 225039 |
+
"learning_rate": 8.433653846153846e-06,
|
| 225040 |
+
"loss": 0.3214,
|
| 225041 |
+
"step": 98780
|
| 225042 |
+
},
|
| 225043 |
+
{
|
| 225044 |
+
"epoch": 795.75,
|
| 225045 |
+
"learning_rate": 8.433573717948719e-06,
|
| 225046 |
+
"loss": 0.3918,
|
| 225047 |
+
"step": 98785
|
| 225048 |
+
},
|
| 225049 |
+
{
|
| 225050 |
+
"epoch": 795.79,
|
| 225051 |
+
"learning_rate": 8.43349358974359e-06,
|
| 225052 |
+
"loss": 0.8046,
|
| 225053 |
+
"step": 98790
|
| 225054 |
+
},
|
| 225055 |
+
{
|
| 225056 |
+
"epoch": 795.83,
|
| 225057 |
+
"learning_rate": 8.433413461538461e-06,
|
| 225058 |
+
"loss": 0.9981,
|
| 225059 |
+
"step": 98795
|
| 225060 |
+
},
|
| 225061 |
+
{
|
| 225062 |
+
"epoch": 795.87,
|
| 225063 |
+
"learning_rate": 8.433333333333334e-06,
|
| 225064 |
+
"loss": 0.2726,
|
| 225065 |
+
"step": 98800
|
| 225066 |
+
},
|
| 225067 |
+
{
|
| 225068 |
+
"epoch": 795.91,
|
| 225069 |
+
"learning_rate": 8.433253205128206e-06,
|
| 225070 |
+
"loss": 0.2719,
|
| 225071 |
+
"step": 98805
|
| 225072 |
+
},
|
| 225073 |
+
{
|
| 225074 |
+
"epoch": 795.95,
|
| 225075 |
+
"learning_rate": 8.433173076923077e-06,
|
| 225076 |
+
"loss": 0.3691,
|
| 225077 |
+
"step": 98810
|
| 225078 |
+
},
|
| 225079 |
+
{
|
| 225080 |
+
"epoch": 795.99,
|
| 225081 |
+
"learning_rate": 8.433092948717949e-06,
|
| 225082 |
+
"loss": 0.9731,
|
| 225083 |
+
"step": 98815
|
| 225084 |
+
},
|
| 225085 |
+
{
|
| 225086 |
+
"epoch": 796.0,
|
| 225087 |
+
"eval_loss": 0.3381003737449646,
|
| 225088 |
+
"eval_runtime": 40.3469,
|
| 225089 |
+
"eval_samples_per_second": 20.646,
|
| 225090 |
+
"eval_steps_per_second": 0.669,
|
| 225091 |
+
"eval_wer": 0.17832343045314625,
|
| 225092 |
+
"step": 98816
|
| 225093 |
+
},
|
| 225094 |
+
{
|
| 225095 |
+
"epoch": 796.03,
|
| 225096 |
+
"learning_rate": 8.433012820512822e-06,
|
| 225097 |
+
"loss": 0.3434,
|
| 225098 |
+
"step": 98820
|
| 225099 |
+
},
|
| 225100 |
+
{
|
| 225101 |
+
"epoch": 796.07,
|
| 225102 |
+
"learning_rate": 8.432932692307693e-06,
|
| 225103 |
+
"loss": 0.3494,
|
| 225104 |
+
"step": 98825
|
| 225105 |
+
},
|
| 225106 |
+
{
|
| 225107 |
+
"epoch": 796.11,
|
| 225108 |
+
"learning_rate": 8.432852564102564e-06,
|
| 225109 |
+
"loss": 0.3089,
|
| 225110 |
+
"step": 98830
|
| 225111 |
+
},
|
| 225112 |
+
{
|
| 225113 |
+
"epoch": 796.15,
|
| 225114 |
+
"learning_rate": 8.432772435897436e-06,
|
| 225115 |
+
"loss": 0.418,
|
| 225116 |
+
"step": 98835
|
| 225117 |
+
},
|
| 225118 |
+
{
|
| 225119 |
+
"epoch": 796.19,
|
| 225120 |
+
"learning_rate": 8.432692307692309e-06,
|
| 225121 |
+
"loss": 0.9833,
|
| 225122 |
+
"step": 98840
|
| 225123 |
+
},
|
| 225124 |
+
{
|
| 225125 |
+
"epoch": 796.23,
|
| 225126 |
+
"learning_rate": 8.43261217948718e-06,
|
| 225127 |
+
"loss": 0.6769,
|
| 225128 |
+
"step": 98845
|
| 225129 |
+
},
|
| 225130 |
+
{
|
| 225131 |
+
"epoch": 796.27,
|
| 225132 |
+
"learning_rate": 8.432532051282051e-06,
|
| 225133 |
+
"loss": 0.2876,
|
| 225134 |
+
"step": 98850
|
| 225135 |
+
},
|
| 225136 |
+
{
|
| 225137 |
+
"epoch": 796.31,
|
| 225138 |
+
"learning_rate": 8.432451923076924e-06,
|
| 225139 |
+
"loss": 0.3477,
|
| 225140 |
+
"step": 98855
|
| 225141 |
+
},
|
| 225142 |
+
{
|
| 225143 |
+
"epoch": 796.35,
|
| 225144 |
+
"learning_rate": 8.432371794871796e-06,
|
| 225145 |
+
"loss": 0.4172,
|
| 225146 |
+
"step": 98860
|
| 225147 |
+
},
|
| 225148 |
+
{
|
| 225149 |
+
"epoch": 796.4,
|
| 225150 |
+
"learning_rate": 8.432291666666667e-06,
|
| 225151 |
+
"loss": 0.7942,
|
| 225152 |
+
"step": 98865
|
| 225153 |
+
},
|
| 225154 |
+
{
|
| 225155 |
+
"epoch": 796.44,
|
| 225156 |
+
"learning_rate": 8.432211538461539e-06,
|
| 225157 |
+
"loss": 0.6742,
|
| 225158 |
+
"step": 98870
|
| 225159 |
+
},
|
| 225160 |
+
{
|
| 225161 |
+
"epoch": 796.48,
|
| 225162 |
+
"learning_rate": 8.432131410256412e-06,
|
| 225163 |
+
"loss": 0.2992,
|
| 225164 |
+
"step": 98875
|
| 225165 |
+
},
|
| 225166 |
+
{
|
| 225167 |
+
"epoch": 796.52,
|
| 225168 |
+
"learning_rate": 8.432051282051283e-06,
|
| 225169 |
+
"loss": 0.3674,
|
| 225170 |
+
"step": 98880
|
| 225171 |
+
},
|
| 225172 |
+
{
|
| 225173 |
+
"epoch": 796.56,
|
| 225174 |
+
"learning_rate": 8.431971153846154e-06,
|
| 225175 |
+
"loss": 0.4408,
|
| 225176 |
+
"step": 98885
|
| 225177 |
+
},
|
| 225178 |
+
{
|
| 225179 |
+
"epoch": 796.6,
|
| 225180 |
+
"learning_rate": 8.431891025641026e-06,
|
| 225181 |
+
"loss": 1.0107,
|
| 225182 |
+
"step": 98890
|
| 225183 |
+
},
|
| 225184 |
+
{
|
| 225185 |
+
"epoch": 796.64,
|
| 225186 |
+
"learning_rate": 8.431810897435899e-06,
|
| 225187 |
+
"loss": 0.6655,
|
| 225188 |
+
"step": 98895
|
| 225189 |
+
},
|
| 225190 |
+
{
|
| 225191 |
+
"epoch": 796.68,
|
| 225192 |
+
"learning_rate": 8.43173076923077e-06,
|
| 225193 |
+
"loss": 0.2911,
|
| 225194 |
+
"step": 98900
|
| 225195 |
+
},
|
| 225196 |
+
{
|
| 225197 |
+
"epoch": 796.72,
|
| 225198 |
+
"learning_rate": 8.431650641025641e-06,
|
| 225199 |
+
"loss": 0.3383,
|
| 225200 |
+
"step": 98905
|
| 225201 |
+
},
|
| 225202 |
+
{
|
| 225203 |
+
"epoch": 796.76,
|
| 225204 |
+
"learning_rate": 8.431570512820515e-06,
|
| 225205 |
+
"loss": 0.3569,
|
| 225206 |
+
"step": 98910
|
| 225207 |
+
},
|
| 225208 |
+
{
|
| 225209 |
+
"epoch": 796.8,
|
| 225210 |
+
"learning_rate": 8.431490384615386e-06,
|
| 225211 |
+
"loss": 0.8294,
|
| 225212 |
+
"step": 98915
|
| 225213 |
+
},
|
| 225214 |
+
{
|
| 225215 |
+
"epoch": 796.84,
|
| 225216 |
+
"learning_rate": 8.431410256410257e-06,
|
| 225217 |
+
"loss": 0.5272,
|
| 225218 |
+
"step": 98920
|
| 225219 |
+
},
|
| 225220 |
+
{
|
| 225221 |
+
"epoch": 796.88,
|
| 225222 |
+
"learning_rate": 8.431330128205129e-06,
|
| 225223 |
+
"loss": 0.2995,
|
| 225224 |
+
"step": 98925
|
| 225225 |
+
},
|
| 225226 |
+
{
|
| 225227 |
+
"epoch": 796.92,
|
| 225228 |
+
"learning_rate": 8.431250000000002e-06,
|
| 225229 |
+
"loss": 0.2882,
|
| 225230 |
+
"step": 98930
|
| 225231 |
+
},
|
| 225232 |
+
{
|
| 225233 |
+
"epoch": 796.96,
|
| 225234 |
+
"learning_rate": 8.431169871794871e-06,
|
| 225235 |
+
"loss": 0.5004,
|
| 225236 |
+
"step": 98935
|
| 225237 |
+
},
|
| 225238 |
+
{
|
| 225239 |
+
"epoch": 797.0,
|
| 225240 |
+
"learning_rate": 8.431089743589744e-06,
|
| 225241 |
+
"loss": 1.3016,
|
| 225242 |
+
"step": 98940
|
| 225243 |
+
},
|
| 225244 |
+
{
|
| 225245 |
+
"epoch": 797.0,
|
| 225246 |
+
"eval_loss": 0.37198692560195923,
|
| 225247 |
+
"eval_runtime": 37.7449,
|
| 225248 |
+
"eval_samples_per_second": 22.069,
|
| 225249 |
+
"eval_steps_per_second": 0.715,
|
| 225250 |
+
"eval_wer": 0.18095796963250935,
|
| 225251 |
+
"step": 98940
|
| 225252 |
}
|
| 225253 |
],
|
| 225254 |
+
"max_steps": 620000,
|
| 225255 |
"num_train_epochs": 5000,
|
| 225256 |
+
"total_flos": 2.784297494455742e+20,
|
| 225257 |
"trial_name": null,
|
| 225258 |
"trial_params": null
|
| 225259 |
}
|
model-bin/finetune/base/{checkpoint-98319 → checkpoint-98940}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629975667.4686/events.out.tfevents.1629975667.8e89bd551565.924.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74e14a51d9eb04464e74a76e9a293603f38074cde1d6cc2ba2f8834672578d72
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629976112.672869/events.out.tfevents.1629976112.8e89bd551565.924.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a6cc4cfc4a25b27c3af0495db89283dfd1b6372b80666c0caa942da0689caabc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629976536.284601/events.out.tfevents.1629976536.8e89bd551565.924.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77e6dd1559b6cf06cc7920e3ced49716f375ba4cba6ec7979820167d77ac8a10
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629976963.4914813/events.out.tfevents.1629976963.8e89bd551565.924.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9db6417f2531b69f80ec2bd8b6a85ffbc9ce5327dd10a4d77e473be9d20329c3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629977393.2955256/events.out.tfevents.1629977393.8e89bd551565.924.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:737878aac3075ecdd1b84120371c7369f02f1391553b4c19aeb5c6a72b193a83
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629975667.8e89bd551565.924.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:529c18adbe2102f72a30072402320868727d1dda3c616c5a8b4fbaa791c2a052
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629976111.8e89bd551565.924.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d5f3e6de97e32fc1bc0ad0d34dd1430ec0a659e080f4bac1c7885ac2f555be0e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629976536.8e89bd551565.924.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:301d72d794e1122d17c51cb7b2c934011768354c0bd689758e0245d61b15863e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629976963.8e89bd551565.924.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a4c51b66a0ab6777ce10b8c95250d6e5ea61c76a987e8fd1d019489f75c8ea44
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629977393.8e89bd551565.924.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b904acdb317157f09520a2daf77df9def74a81c164f477897d9e23fe17a0f7e9
|
| 3 |
+
size 8622
|