"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629804951.1741285/events.out.tfevents.1629804951.c435e1c5ee04.920.121 +3 -0
- model-bin/finetune/base/log/1629805585.754903/events.out.tfevents.1629805585.c435e1c5ee04.920.123 +3 -0
- model-bin/finetune/base/log/1629806228.6566694/events.out.tfevents.1629806228.c435e1c5ee04.920.125 +3 -0
- model-bin/finetune/base/log/1629806875.5539/events.out.tfevents.1629806875.c435e1c5ee04.920.127 +3 -0
- model-bin/finetune/base/log/1629807503.9229152/events.out.tfevents.1629807503.c435e1c5ee04.920.129 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629804951.c435e1c5ee04.920.120 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629805585.c435e1c5ee04.920.122 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629806228.c435e1c5ee04.920.124 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629806875.c435e1c5ee04.920.126 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629807503.c435e1c5ee04.920.128 +3 -0
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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:cc046ce6a70a6176f5a408d38dfce4fa4e0319d7db46487046ae174725c79894
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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:e35458acded70096c07c8451d78abdf2eb7c934e6052e5e6ef4c45b08010c1b5
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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:e870c185f28e6591e77ed6c2bfd0d5d76d25d4d6f220035dbb1ea9d6f5306642
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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 sha256:2f34225a8cf98085e730b36c0e1c50a3349560a0730c3fb1dd814102c3307c41
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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 sha256:6d7b74ffa97962271900982c6c3e598caf5ec2df2234088f585123a33eecc0d8
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/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,
|
|
@@ -177423,11 +177423,800 @@
|
|
| 177423 |
"eval_steps_per_second": 0.677,
|
| 177424 |
"eval_wer": 0.19714999265462024,
|
| 177425 |
"step": 61350
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 177426 |
}
|
| 177427 |
],
|
| 177428 |
"max_steps": 620000,
|
| 177429 |
"num_train_epochs": 5000,
|
| 177430 |
-
"total_flos": 1.
|
| 177431 |
"trial_name": null,
|
| 177432 |
"trial_params": null
|
| 177433 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 499.0,
|
| 5 |
+
"global_step": 61973,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 177423 |
"eval_steps_per_second": 0.677,
|
| 177424 |
"eval_wer": 0.19714999265462024,
|
| 177425 |
"step": 61350
|
| 177426 |
+
},
|
| 177427 |
+
{
|
| 177428 |
+
"epoch": 490.04,
|
| 177429 |
+
"learning_rate": 9.033141025641026e-06,
|
| 177430 |
+
"loss": 0.3504,
|
| 177431 |
+
"step": 61355
|
| 177432 |
+
},
|
| 177433 |
+
{
|
| 177434 |
+
"epoch": 490.08,
|
| 177435 |
+
"learning_rate": 9.033060897435897e-06,
|
| 177436 |
+
"loss": 0.3182,
|
| 177437 |
+
"step": 61360
|
| 177438 |
+
},
|
| 177439 |
+
{
|
| 177440 |
+
"epoch": 490.12,
|
| 177441 |
+
"learning_rate": 9.03298076923077e-06,
|
| 177442 |
+
"loss": 0.3591,
|
| 177443 |
+
"step": 61365
|
| 177444 |
+
},
|
| 177445 |
+
{
|
| 177446 |
+
"epoch": 490.16,
|
| 177447 |
+
"learning_rate": 9.032900641025642e-06,
|
| 177448 |
+
"loss": 0.5046,
|
| 177449 |
+
"step": 61370
|
| 177450 |
+
},
|
| 177451 |
+
{
|
| 177452 |
+
"epoch": 490.2,
|
| 177453 |
+
"learning_rate": 9.032820512820513e-06,
|
| 177454 |
+
"loss": 1.1922,
|
| 177455 |
+
"step": 61375
|
| 177456 |
+
},
|
| 177457 |
+
{
|
| 177458 |
+
"epoch": 490.24,
|
| 177459 |
+
"learning_rate": 9.032740384615385e-06,
|
| 177460 |
+
"loss": 0.3724,
|
| 177461 |
+
"step": 61380
|
| 177462 |
+
},
|
| 177463 |
+
{
|
| 177464 |
+
"epoch": 490.28,
|
| 177465 |
+
"learning_rate": 9.032660256410258e-06,
|
| 177466 |
+
"loss": 0.2758,
|
| 177467 |
+
"step": 61385
|
| 177468 |
+
},
|
| 177469 |
+
{
|
| 177470 |
+
"epoch": 490.32,
|
| 177471 |
+
"learning_rate": 9.032580128205129e-06,
|
| 177472 |
+
"loss": 0.3468,
|
| 177473 |
+
"step": 61390
|
| 177474 |
+
},
|
| 177475 |
+
{
|
| 177476 |
+
"epoch": 490.36,
|
| 177477 |
+
"learning_rate": 9.0325e-06,
|
| 177478 |
+
"loss": 0.5559,
|
| 177479 |
+
"step": 61395
|
| 177480 |
+
},
|
| 177481 |
+
{
|
| 177482 |
+
"epoch": 490.4,
|
| 177483 |
+
"learning_rate": 9.032419871794873e-06,
|
| 177484 |
+
"loss": 1.147,
|
| 177485 |
+
"step": 61400
|
| 177486 |
+
},
|
| 177487 |
+
{
|
| 177488 |
+
"epoch": 490.44,
|
| 177489 |
+
"learning_rate": 9.032339743589745e-06,
|
| 177490 |
+
"loss": 0.3765,
|
| 177491 |
+
"step": 61405
|
| 177492 |
+
},
|
| 177493 |
+
{
|
| 177494 |
+
"epoch": 490.48,
|
| 177495 |
+
"learning_rate": 9.032259615384616e-06,
|
| 177496 |
+
"loss": 0.3325,
|
| 177497 |
+
"step": 61410
|
| 177498 |
+
},
|
| 177499 |
+
{
|
| 177500 |
+
"epoch": 490.52,
|
| 177501 |
+
"learning_rate": 9.032179487179487e-06,
|
| 177502 |
+
"loss": 0.4258,
|
| 177503 |
+
"step": 61415
|
| 177504 |
+
},
|
| 177505 |
+
{
|
| 177506 |
+
"epoch": 490.56,
|
| 177507 |
+
"learning_rate": 9.03209935897436e-06,
|
| 177508 |
+
"loss": 0.5173,
|
| 177509 |
+
"step": 61420
|
| 177510 |
+
},
|
| 177511 |
+
{
|
| 177512 |
+
"epoch": 490.6,
|
| 177513 |
+
"learning_rate": 9.032019230769232e-06,
|
| 177514 |
+
"loss": 1.4811,
|
| 177515 |
+
"step": 61425
|
| 177516 |
+
},
|
| 177517 |
+
{
|
| 177518 |
+
"epoch": 490.64,
|
| 177519 |
+
"learning_rate": 9.031939102564103e-06,
|
| 177520 |
+
"loss": 0.3446,
|
| 177521 |
+
"step": 61430
|
| 177522 |
+
},
|
| 177523 |
+
{
|
| 177524 |
+
"epoch": 490.68,
|
| 177525 |
+
"learning_rate": 9.031858974358975e-06,
|
| 177526 |
+
"loss": 0.3521,
|
| 177527 |
+
"step": 61435
|
| 177528 |
+
},
|
| 177529 |
+
{
|
| 177530 |
+
"epoch": 490.72,
|
| 177531 |
+
"learning_rate": 9.031778846153848e-06,
|
| 177532 |
+
"loss": 0.4047,
|
| 177533 |
+
"step": 61440
|
| 177534 |
+
},
|
| 177535 |
+
{
|
| 177536 |
+
"epoch": 490.76,
|
| 177537 |
+
"learning_rate": 9.031698717948719e-06,
|
| 177538 |
+
"loss": 0.5132,
|
| 177539 |
+
"step": 61445
|
| 177540 |
+
},
|
| 177541 |
+
{
|
| 177542 |
+
"epoch": 490.8,
|
| 177543 |
+
"learning_rate": 9.03161858974359e-06,
|
| 177544 |
+
"loss": 1.3382,
|
| 177545 |
+
"step": 61450
|
| 177546 |
+
},
|
| 177547 |
+
{
|
| 177548 |
+
"epoch": 490.84,
|
| 177549 |
+
"learning_rate": 9.031538461538463e-06,
|
| 177550 |
+
"loss": 0.3674,
|
| 177551 |
+
"step": 61455
|
| 177552 |
+
},
|
| 177553 |
+
{
|
| 177554 |
+
"epoch": 490.88,
|
| 177555 |
+
"learning_rate": 9.031458333333333e-06,
|
| 177556 |
+
"loss": 0.351,
|
| 177557 |
+
"step": 61460
|
| 177558 |
+
},
|
| 177559 |
+
{
|
| 177560 |
+
"epoch": 490.92,
|
| 177561 |
+
"learning_rate": 9.031378205128206e-06,
|
| 177562 |
+
"loss": 0.3609,
|
| 177563 |
+
"step": 61465
|
| 177564 |
+
},
|
| 177565 |
+
{
|
| 177566 |
+
"epoch": 490.96,
|
| 177567 |
+
"learning_rate": 9.031298076923077e-06,
|
| 177568 |
+
"loss": 0.4871,
|
| 177569 |
+
"step": 61470
|
| 177570 |
+
},
|
| 177571 |
+
{
|
| 177572 |
+
"epoch": 491.0,
|
| 177573 |
+
"learning_rate": 9.031217948717949e-06,
|
| 177574 |
+
"loss": 1.4099,
|
| 177575 |
+
"step": 61475
|
| 177576 |
+
},
|
| 177577 |
+
{
|
| 177578 |
+
"epoch": 491.0,
|
| 177579 |
+
"eval_loss": 0.45204436779022217,
|
| 177580 |
+
"eval_runtime": 39.7359,
|
| 177581 |
+
"eval_samples_per_second": 21.24,
|
| 177582 |
+
"eval_steps_per_second": 0.679,
|
| 177583 |
+
"eval_wer": 0.20094156329397697,
|
| 177584 |
+
"step": 61475
|
| 177585 |
+
},
|
| 177586 |
+
{
|
| 177587 |
+
"epoch": 495.04,
|
| 177588 |
+
"learning_rate": 9.03113782051282e-06,
|
| 177589 |
+
"loss": 0.3876,
|
| 177590 |
+
"step": 61480
|
| 177591 |
+
},
|
| 177592 |
+
{
|
| 177593 |
+
"epoch": 495.08,
|
| 177594 |
+
"learning_rate": 9.031057692307693e-06,
|
| 177595 |
+
"loss": 0.3144,
|
| 177596 |
+
"step": 61485
|
| 177597 |
+
},
|
| 177598 |
+
{
|
| 177599 |
+
"epoch": 495.12,
|
| 177600 |
+
"learning_rate": 9.030977564102565e-06,
|
| 177601 |
+
"loss": 0.3777,
|
| 177602 |
+
"step": 61490
|
| 177603 |
+
},
|
| 177604 |
+
{
|
| 177605 |
+
"epoch": 495.16,
|
| 177606 |
+
"learning_rate": 9.030897435897436e-06,
|
| 177607 |
+
"loss": 0.5102,
|
| 177608 |
+
"step": 61495
|
| 177609 |
+
},
|
| 177610 |
+
{
|
| 177611 |
+
"epoch": 495.2,
|
| 177612 |
+
"learning_rate": 9.030817307692309e-06,
|
| 177613 |
+
"loss": 1.1972,
|
| 177614 |
+
"step": 61500
|
| 177615 |
+
},
|
| 177616 |
+
{
|
| 177617 |
+
"epoch": 495.24,
|
| 177618 |
+
"learning_rate": 9.03073717948718e-06,
|
| 177619 |
+
"loss": 0.4333,
|
| 177620 |
+
"step": 61505
|
| 177621 |
+
},
|
| 177622 |
+
{
|
| 177623 |
+
"epoch": 495.28,
|
| 177624 |
+
"learning_rate": 9.030657051282052e-06,
|
| 177625 |
+
"loss": 0.3252,
|
| 177626 |
+
"step": 61510
|
| 177627 |
+
},
|
| 177628 |
+
{
|
| 177629 |
+
"epoch": 495.32,
|
| 177630 |
+
"learning_rate": 9.030576923076923e-06,
|
| 177631 |
+
"loss": 0.2979,
|
| 177632 |
+
"step": 61515
|
| 177633 |
+
},
|
| 177634 |
+
{
|
| 177635 |
+
"epoch": 495.36,
|
| 177636 |
+
"learning_rate": 9.030496794871796e-06,
|
| 177637 |
+
"loss": 0.5486,
|
| 177638 |
+
"step": 61520
|
| 177639 |
+
},
|
| 177640 |
+
{
|
| 177641 |
+
"epoch": 495.4,
|
| 177642 |
+
"learning_rate": 9.030416666666668e-06,
|
| 177643 |
+
"loss": 1.2093,
|
| 177644 |
+
"step": 61525
|
| 177645 |
+
},
|
| 177646 |
+
{
|
| 177647 |
+
"epoch": 495.44,
|
| 177648 |
+
"learning_rate": 9.030336538461539e-06,
|
| 177649 |
+
"loss": 0.3834,
|
| 177650 |
+
"step": 61530
|
| 177651 |
+
},
|
| 177652 |
+
{
|
| 177653 |
+
"epoch": 495.48,
|
| 177654 |
+
"learning_rate": 9.03025641025641e-06,
|
| 177655 |
+
"loss": 0.344,
|
| 177656 |
+
"step": 61535
|
| 177657 |
+
},
|
| 177658 |
+
{
|
| 177659 |
+
"epoch": 495.52,
|
| 177660 |
+
"learning_rate": 9.030176282051283e-06,
|
| 177661 |
+
"loss": 0.383,
|
| 177662 |
+
"step": 61540
|
| 177663 |
+
},
|
| 177664 |
+
{
|
| 177665 |
+
"epoch": 495.56,
|
| 177666 |
+
"learning_rate": 9.030096153846155e-06,
|
| 177667 |
+
"loss": 0.5361,
|
| 177668 |
+
"step": 61545
|
| 177669 |
+
},
|
| 177670 |
+
{
|
| 177671 |
+
"epoch": 495.6,
|
| 177672 |
+
"learning_rate": 9.030016025641026e-06,
|
| 177673 |
+
"loss": 1.2365,
|
| 177674 |
+
"step": 61550
|
| 177675 |
+
},
|
| 177676 |
+
{
|
| 177677 |
+
"epoch": 495.64,
|
| 177678 |
+
"learning_rate": 9.029935897435899e-06,
|
| 177679 |
+
"loss": 0.3559,
|
| 177680 |
+
"step": 61555
|
| 177681 |
+
},
|
| 177682 |
+
{
|
| 177683 |
+
"epoch": 495.68,
|
| 177684 |
+
"learning_rate": 9.02985576923077e-06,
|
| 177685 |
+
"loss": 0.3694,
|
| 177686 |
+
"step": 61560
|
| 177687 |
+
},
|
| 177688 |
+
{
|
| 177689 |
+
"epoch": 495.72,
|
| 177690 |
+
"learning_rate": 9.029775641025642e-06,
|
| 177691 |
+
"loss": 0.3872,
|
| 177692 |
+
"step": 61565
|
| 177693 |
+
},
|
| 177694 |
+
{
|
| 177695 |
+
"epoch": 495.76,
|
| 177696 |
+
"learning_rate": 9.029695512820513e-06,
|
| 177697 |
+
"loss": 0.4842,
|
| 177698 |
+
"step": 61570
|
| 177699 |
+
},
|
| 177700 |
+
{
|
| 177701 |
+
"epoch": 495.8,
|
| 177702 |
+
"learning_rate": 9.029615384615386e-06,
|
| 177703 |
+
"loss": 1.4835,
|
| 177704 |
+
"step": 61575
|
| 177705 |
+
},
|
| 177706 |
+
{
|
| 177707 |
+
"epoch": 495.84,
|
| 177708 |
+
"learning_rate": 9.029535256410258e-06,
|
| 177709 |
+
"loss": 0.4173,
|
| 177710 |
+
"step": 61580
|
| 177711 |
+
},
|
| 177712 |
+
{
|
| 177713 |
+
"epoch": 495.88,
|
| 177714 |
+
"learning_rate": 9.029455128205129e-06,
|
| 177715 |
+
"loss": 0.3018,
|
| 177716 |
+
"step": 61585
|
| 177717 |
+
},
|
| 177718 |
+
{
|
| 177719 |
+
"epoch": 495.92,
|
| 177720 |
+
"learning_rate": 9.029375e-06,
|
| 177721 |
+
"loss": 0.3875,
|
| 177722 |
+
"step": 61590
|
| 177723 |
+
},
|
| 177724 |
+
{
|
| 177725 |
+
"epoch": 495.96,
|
| 177726 |
+
"learning_rate": 9.029294871794873e-06,
|
| 177727 |
+
"loss": 0.5493,
|
| 177728 |
+
"step": 61595
|
| 177729 |
+
},
|
| 177730 |
+
{
|
| 177731 |
+
"epoch": 496.0,
|
| 177732 |
+
"eval_loss": 0.3798507750034332,
|
| 177733 |
+
"eval_runtime": 40.0518,
|
| 177734 |
+
"eval_samples_per_second": 21.073,
|
| 177735 |
+
"eval_steps_per_second": 0.674,
|
| 177736 |
+
"eval_wer": 0.2006220839813375,
|
| 177737 |
+
"step": 61599
|
| 177738 |
+
},
|
| 177739 |
+
{
|
| 177740 |
+
"epoch": 492.01,
|
| 177741 |
+
"learning_rate": 9.029214743589745e-06,
|
| 177742 |
+
"loss": 0.4965,
|
| 177743 |
+
"step": 61600
|
| 177744 |
+
},
|
| 177745 |
+
{
|
| 177746 |
+
"epoch": 492.05,
|
| 177747 |
+
"learning_rate": 9.029134615384616e-06,
|
| 177748 |
+
"loss": 0.4589,
|
| 177749 |
+
"step": 61605
|
| 177750 |
+
},
|
| 177751 |
+
{
|
| 177752 |
+
"epoch": 492.09,
|
| 177753 |
+
"learning_rate": 9.029054487179489e-06,
|
| 177754 |
+
"loss": 0.3753,
|
| 177755 |
+
"step": 61610
|
| 177756 |
+
},
|
| 177757 |
+
{
|
| 177758 |
+
"epoch": 492.13,
|
| 177759 |
+
"learning_rate": 9.028974358974359e-06,
|
| 177760 |
+
"loss": 0.3388,
|
| 177761 |
+
"step": 61615
|
| 177762 |
+
},
|
| 177763 |
+
{
|
| 177764 |
+
"epoch": 492.17,
|
| 177765 |
+
"learning_rate": 9.028894230769232e-06,
|
| 177766 |
+
"loss": 0.574,
|
| 177767 |
+
"step": 61620
|
| 177768 |
+
},
|
| 177769 |
+
{
|
| 177770 |
+
"epoch": 492.21,
|
| 177771 |
+
"learning_rate": 9.028814102564103e-06,
|
| 177772 |
+
"loss": 1.2476,
|
| 177773 |
+
"step": 61625
|
| 177774 |
+
},
|
| 177775 |
+
{
|
| 177776 |
+
"epoch": 492.25,
|
| 177777 |
+
"learning_rate": 9.028733974358975e-06,
|
| 177778 |
+
"loss": 0.3201,
|
| 177779 |
+
"step": 61630
|
| 177780 |
+
},
|
| 177781 |
+
{
|
| 177782 |
+
"epoch": 492.29,
|
| 177783 |
+
"learning_rate": 9.028653846153846e-06,
|
| 177784 |
+
"loss": 0.2998,
|
| 177785 |
+
"step": 61635
|
| 177786 |
+
},
|
| 177787 |
+
{
|
| 177788 |
+
"epoch": 492.33,
|
| 177789 |
+
"learning_rate": 9.028573717948719e-06,
|
| 177790 |
+
"loss": 0.3988,
|
| 177791 |
+
"step": 61640
|
| 177792 |
+
},
|
| 177793 |
+
{
|
| 177794 |
+
"epoch": 492.37,
|
| 177795 |
+
"learning_rate": 9.02849358974359e-06,
|
| 177796 |
+
"loss": 0.606,
|
| 177797 |
+
"step": 61645
|
| 177798 |
+
},
|
| 177799 |
+
{
|
| 177800 |
+
"epoch": 492.41,
|
| 177801 |
+
"learning_rate": 9.028413461538462e-06,
|
| 177802 |
+
"loss": 1.2492,
|
| 177803 |
+
"step": 61650
|
| 177804 |
+
},
|
| 177805 |
+
{
|
| 177806 |
+
"epoch": 492.45,
|
| 177807 |
+
"learning_rate": 9.028333333333335e-06,
|
| 177808 |
+
"loss": 0.3176,
|
| 177809 |
+
"step": 61655
|
| 177810 |
+
},
|
| 177811 |
+
{
|
| 177812 |
+
"epoch": 492.49,
|
| 177813 |
+
"learning_rate": 9.028253205128206e-06,
|
| 177814 |
+
"loss": 0.3107,
|
| 177815 |
+
"step": 61660
|
| 177816 |
+
},
|
| 177817 |
+
{
|
| 177818 |
+
"epoch": 492.53,
|
| 177819 |
+
"learning_rate": 9.028173076923077e-06,
|
| 177820 |
+
"loss": 0.312,
|
| 177821 |
+
"step": 61665
|
| 177822 |
+
},
|
| 177823 |
+
{
|
| 177824 |
+
"epoch": 492.57,
|
| 177825 |
+
"learning_rate": 9.028092948717949e-06,
|
| 177826 |
+
"loss": 0.5562,
|
| 177827 |
+
"step": 61670
|
| 177828 |
+
},
|
| 177829 |
+
{
|
| 177830 |
+
"epoch": 492.61,
|
| 177831 |
+
"learning_rate": 9.028012820512822e-06,
|
| 177832 |
+
"loss": 1.3504,
|
| 177833 |
+
"step": 61675
|
| 177834 |
+
},
|
| 177835 |
+
{
|
| 177836 |
+
"epoch": 492.65,
|
| 177837 |
+
"learning_rate": 9.027932692307693e-06,
|
| 177838 |
+
"loss": 0.3623,
|
| 177839 |
+
"step": 61680
|
| 177840 |
+
},
|
| 177841 |
+
{
|
| 177842 |
+
"epoch": 492.69,
|
| 177843 |
+
"learning_rate": 9.027852564102565e-06,
|
| 177844 |
+
"loss": 0.2931,
|
| 177845 |
+
"step": 61685
|
| 177846 |
+
},
|
| 177847 |
+
{
|
| 177848 |
+
"epoch": 492.73,
|
| 177849 |
+
"learning_rate": 9.027772435897436e-06,
|
| 177850 |
+
"loss": 0.4285,
|
| 177851 |
+
"step": 61690
|
| 177852 |
+
},
|
| 177853 |
+
{
|
| 177854 |
+
"epoch": 492.77,
|
| 177855 |
+
"learning_rate": 9.027692307692309e-06,
|
| 177856 |
+
"loss": 0.6349,
|
| 177857 |
+
"step": 61695
|
| 177858 |
+
},
|
| 177859 |
+
{
|
| 177860 |
+
"epoch": 492.81,
|
| 177861 |
+
"learning_rate": 9.02761217948718e-06,
|
| 177862 |
+
"loss": 1.1928,
|
| 177863 |
+
"step": 61700
|
| 177864 |
+
},
|
| 177865 |
+
{
|
| 177866 |
+
"epoch": 492.85,
|
| 177867 |
+
"learning_rate": 9.027532051282052e-06,
|
| 177868 |
+
"loss": 0.4126,
|
| 177869 |
+
"step": 61705
|
| 177870 |
+
},
|
| 177871 |
+
{
|
| 177872 |
+
"epoch": 492.89,
|
| 177873 |
+
"learning_rate": 9.027451923076925e-06,
|
| 177874 |
+
"loss": 0.3497,
|
| 177875 |
+
"step": 61710
|
| 177876 |
+
},
|
| 177877 |
+
{
|
| 177878 |
+
"epoch": 492.93,
|
| 177879 |
+
"learning_rate": 9.027371794871796e-06,
|
| 177880 |
+
"loss": 0.4182,
|
| 177881 |
+
"step": 61715
|
| 177882 |
+
},
|
| 177883 |
+
{
|
| 177884 |
+
"epoch": 492.97,
|
| 177885 |
+
"learning_rate": 9.027291666666667e-06,
|
| 177886 |
+
"loss": 0.7146,
|
| 177887 |
+
"step": 61720
|
| 177888 |
+
},
|
| 177889 |
+
{
|
| 177890 |
+
"epoch": 493.0,
|
| 177891 |
+
"eval_loss": 0.3710833489894867,
|
| 177892 |
+
"eval_runtime": 39.3349,
|
| 177893 |
+
"eval_samples_per_second": 21.457,
|
| 177894 |
+
"eval_steps_per_second": 0.686,
|
| 177895 |
+
"eval_wer": 0.19281236331826798,
|
| 177896 |
+
"step": 61724
|
| 177897 |
+
},
|
| 177898 |
+
{
|
| 177899 |
+
"epoch": 493.01,
|
| 177900 |
+
"learning_rate": 9.027211538461539e-06,
|
| 177901 |
+
"loss": 0.4987,
|
| 177902 |
+
"step": 61725
|
| 177903 |
+
},
|
| 177904 |
+
{
|
| 177905 |
+
"epoch": 493.05,
|
| 177906 |
+
"learning_rate": 9.027131410256412e-06,
|
| 177907 |
+
"loss": 0.3369,
|
| 177908 |
+
"step": 61730
|
| 177909 |
+
},
|
| 177910 |
+
{
|
| 177911 |
+
"epoch": 493.09,
|
| 177912 |
+
"learning_rate": 9.027051282051282e-06,
|
| 177913 |
+
"loss": 0.3258,
|
| 177914 |
+
"step": 61735
|
| 177915 |
+
},
|
| 177916 |
+
{
|
| 177917 |
+
"epoch": 493.13,
|
| 177918 |
+
"learning_rate": 9.026971153846155e-06,
|
| 177919 |
+
"loss": 0.3438,
|
| 177920 |
+
"step": 61740
|
| 177921 |
+
},
|
| 177922 |
+
{
|
| 177923 |
+
"epoch": 493.17,
|
| 177924 |
+
"learning_rate": 9.026891025641028e-06,
|
| 177925 |
+
"loss": 0.9288,
|
| 177926 |
+
"step": 61745
|
| 177927 |
+
},
|
| 177928 |
+
{
|
| 177929 |
+
"epoch": 493.21,
|
| 177930 |
+
"learning_rate": 9.026810897435897e-06,
|
| 177931 |
+
"loss": 1.1201,
|
| 177932 |
+
"step": 61750
|
| 177933 |
+
},
|
| 177934 |
+
{
|
| 177935 |
+
"epoch": 493.25,
|
| 177936 |
+
"learning_rate": 9.02673076923077e-06,
|
| 177937 |
+
"loss": 0.3035,
|
| 177938 |
+
"step": 61755
|
| 177939 |
+
},
|
| 177940 |
+
{
|
| 177941 |
+
"epoch": 493.29,
|
| 177942 |
+
"learning_rate": 9.026650641025642e-06,
|
| 177943 |
+
"loss": 0.3562,
|
| 177944 |
+
"step": 61760
|
| 177945 |
+
},
|
| 177946 |
+
{
|
| 177947 |
+
"epoch": 493.33,
|
| 177948 |
+
"learning_rate": 9.026570512820513e-06,
|
| 177949 |
+
"loss": 0.3918,
|
| 177950 |
+
"step": 61765
|
| 177951 |
+
},
|
| 177952 |
+
{
|
| 177953 |
+
"epoch": 493.37,
|
| 177954 |
+
"learning_rate": 9.026490384615384e-06,
|
| 177955 |
+
"loss": 0.5514,
|
| 177956 |
+
"step": 61770
|
| 177957 |
+
},
|
| 177958 |
+
{
|
| 177959 |
+
"epoch": 493.41,
|
| 177960 |
+
"learning_rate": 9.026410256410257e-06,
|
| 177961 |
+
"loss": 1.1283,
|
| 177962 |
+
"step": 61775
|
| 177963 |
+
},
|
| 177964 |
+
{
|
| 177965 |
+
"epoch": 493.45,
|
| 177966 |
+
"learning_rate": 9.026330128205129e-06,
|
| 177967 |
+
"loss": 0.4402,
|
| 177968 |
+
"step": 61780
|
| 177969 |
+
},
|
| 177970 |
+
{
|
| 177971 |
+
"epoch": 493.49,
|
| 177972 |
+
"learning_rate": 9.02625e-06,
|
| 177973 |
+
"loss": 0.308,
|
| 177974 |
+
"step": 61785
|
| 177975 |
+
},
|
| 177976 |
+
{
|
| 177977 |
+
"epoch": 493.53,
|
| 177978 |
+
"learning_rate": 9.026169871794872e-06,
|
| 177979 |
+
"loss": 0.4094,
|
| 177980 |
+
"step": 61790
|
| 177981 |
+
},
|
| 177982 |
+
{
|
| 177983 |
+
"epoch": 493.57,
|
| 177984 |
+
"learning_rate": 9.026089743589745e-06,
|
| 177985 |
+
"loss": 0.6401,
|
| 177986 |
+
"step": 61795
|
| 177987 |
+
},
|
| 177988 |
+
{
|
| 177989 |
+
"epoch": 493.61,
|
| 177990 |
+
"learning_rate": 9.026009615384616e-06,
|
| 177991 |
+
"loss": 1.0911,
|
| 177992 |
+
"step": 61800
|
| 177993 |
+
},
|
| 177994 |
+
{
|
| 177995 |
+
"epoch": 493.65,
|
| 177996 |
+
"learning_rate": 9.025929487179487e-06,
|
| 177997 |
+
"loss": 0.4626,
|
| 177998 |
+
"step": 61805
|
| 177999 |
+
},
|
| 178000 |
+
{
|
| 178001 |
+
"epoch": 493.69,
|
| 178002 |
+
"learning_rate": 9.02584935897436e-06,
|
| 178003 |
+
"loss": 0.308,
|
| 178004 |
+
"step": 61810
|
| 178005 |
+
},
|
| 178006 |
+
{
|
| 178007 |
+
"epoch": 493.73,
|
| 178008 |
+
"learning_rate": 9.025769230769232e-06,
|
| 178009 |
+
"loss": 0.3748,
|
| 178010 |
+
"step": 61815
|
| 178011 |
+
},
|
| 178012 |
+
{
|
| 178013 |
+
"epoch": 493.77,
|
| 178014 |
+
"learning_rate": 9.025689102564103e-06,
|
| 178015 |
+
"loss": 0.5389,
|
| 178016 |
+
"step": 61820
|
| 178017 |
+
},
|
| 178018 |
+
{
|
| 178019 |
+
"epoch": 493.81,
|
| 178020 |
+
"learning_rate": 9.025608974358974e-06,
|
| 178021 |
+
"loss": 1.1674,
|
| 178022 |
+
"step": 61825
|
| 178023 |
+
},
|
| 178024 |
+
{
|
| 178025 |
+
"epoch": 493.85,
|
| 178026 |
+
"learning_rate": 9.025528846153847e-06,
|
| 178027 |
+
"loss": 0.3328,
|
| 178028 |
+
"step": 61830
|
| 178029 |
+
},
|
| 178030 |
+
{
|
| 178031 |
+
"epoch": 493.89,
|
| 178032 |
+
"learning_rate": 9.025448717948719e-06,
|
| 178033 |
+
"loss": 0.343,
|
| 178034 |
+
"step": 61835
|
| 178035 |
+
},
|
| 178036 |
+
{
|
| 178037 |
+
"epoch": 493.93,
|
| 178038 |
+
"learning_rate": 9.02536858974359e-06,
|
| 178039 |
+
"loss": 0.3814,
|
| 178040 |
+
"step": 61840
|
| 178041 |
+
},
|
| 178042 |
+
{
|
| 178043 |
+
"epoch": 493.97,
|
| 178044 |
+
"learning_rate": 9.025288461538463e-06,
|
| 178045 |
+
"loss": 0.6895,
|
| 178046 |
+
"step": 61845
|
| 178047 |
+
},
|
| 178048 |
+
{
|
| 178049 |
+
"epoch": 494.0,
|
| 178050 |
+
"eval_loss": 0.4307675361633301,
|
| 178051 |
+
"eval_runtime": 39.3445,
|
| 178052 |
+
"eval_samples_per_second": 21.452,
|
| 178053 |
+
"eval_steps_per_second": 0.686,
|
| 178054 |
+
"eval_wer": 0.21091885441527447,
|
| 178055 |
+
"step": 61849
|
| 178056 |
+
},
|
| 178057 |
+
{
|
| 178058 |
+
"epoch": 498.01,
|
| 178059 |
+
"learning_rate": 9.025208333333335e-06,
|
| 178060 |
+
"loss": 0.503,
|
| 178061 |
+
"step": 61850
|
| 178062 |
+
},
|
| 178063 |
+
{
|
| 178064 |
+
"epoch": 498.05,
|
| 178065 |
+
"learning_rate": 9.025128205128206e-06,
|
| 178066 |
+
"loss": 0.359,
|
| 178067 |
+
"step": 61855
|
| 178068 |
+
},
|
| 178069 |
+
{
|
| 178070 |
+
"epoch": 498.09,
|
| 178071 |
+
"learning_rate": 9.025048076923077e-06,
|
| 178072 |
+
"loss": 0.3245,
|
| 178073 |
+
"step": 61860
|
| 178074 |
+
},
|
| 178075 |
+
{
|
| 178076 |
+
"epoch": 498.13,
|
| 178077 |
+
"learning_rate": 9.02496794871795e-06,
|
| 178078 |
+
"loss": 0.4073,
|
| 178079 |
+
"step": 61865
|
| 178080 |
+
},
|
| 178081 |
+
{
|
| 178082 |
+
"epoch": 498.17,
|
| 178083 |
+
"learning_rate": 9.024887820512822e-06,
|
| 178084 |
+
"loss": 0.6474,
|
| 178085 |
+
"step": 61870
|
| 178086 |
+
},
|
| 178087 |
+
{
|
| 178088 |
+
"epoch": 498.21,
|
| 178089 |
+
"learning_rate": 9.024807692307693e-06,
|
| 178090 |
+
"loss": 1.1753,
|
| 178091 |
+
"step": 61875
|
| 178092 |
+
},
|
| 178093 |
+
{
|
| 178094 |
+
"epoch": 498.25,
|
| 178095 |
+
"learning_rate": 9.024727564102564e-06,
|
| 178096 |
+
"loss": 0.3574,
|
| 178097 |
+
"step": 61880
|
| 178098 |
+
},
|
| 178099 |
+
{
|
| 178100 |
+
"epoch": 498.29,
|
| 178101 |
+
"learning_rate": 9.024647435897438e-06,
|
| 178102 |
+
"loss": 0.3138,
|
| 178103 |
+
"step": 61885
|
| 178104 |
+
},
|
| 178105 |
+
{
|
| 178106 |
+
"epoch": 498.33,
|
| 178107 |
+
"learning_rate": 9.024567307692307e-06,
|
| 178108 |
+
"loss": 0.3245,
|
| 178109 |
+
"step": 61890
|
| 178110 |
+
},
|
| 178111 |
+
{
|
| 178112 |
+
"epoch": 498.37,
|
| 178113 |
+
"learning_rate": 9.02448717948718e-06,
|
| 178114 |
+
"loss": 0.6058,
|
| 178115 |
+
"step": 61895
|
| 178116 |
+
},
|
| 178117 |
+
{
|
| 178118 |
+
"epoch": 498.41,
|
| 178119 |
+
"learning_rate": 9.024407051282053e-06,
|
| 178120 |
+
"loss": 1.2926,
|
| 178121 |
+
"step": 61900
|
| 178122 |
+
},
|
| 178123 |
+
{
|
| 178124 |
+
"epoch": 498.45,
|
| 178125 |
+
"learning_rate": 9.024326923076923e-06,
|
| 178126 |
+
"loss": 0.3079,
|
| 178127 |
+
"step": 61905
|
| 178128 |
+
},
|
| 178129 |
+
{
|
| 178130 |
+
"epoch": 498.49,
|
| 178131 |
+
"learning_rate": 9.024246794871796e-06,
|
| 178132 |
+
"loss": 0.3141,
|
| 178133 |
+
"step": 61910
|
| 178134 |
+
},
|
| 178135 |
+
{
|
| 178136 |
+
"epoch": 498.53,
|
| 178137 |
+
"learning_rate": 9.024166666666667e-06,
|
| 178138 |
+
"loss": 0.3497,
|
| 178139 |
+
"step": 61915
|
| 178140 |
+
},
|
| 178141 |
+
{
|
| 178142 |
+
"epoch": 498.57,
|
| 178143 |
+
"learning_rate": 9.024086538461539e-06,
|
| 178144 |
+
"loss": 0.5925,
|
| 178145 |
+
"step": 61920
|
| 178146 |
+
},
|
| 178147 |
+
{
|
| 178148 |
+
"epoch": 498.61,
|
| 178149 |
+
"learning_rate": 9.02400641025641e-06,
|
| 178150 |
+
"loss": 1.2671,
|
| 178151 |
+
"step": 61925
|
| 178152 |
+
},
|
| 178153 |
+
{
|
| 178154 |
+
"epoch": 498.65,
|
| 178155 |
+
"learning_rate": 9.023926282051283e-06,
|
| 178156 |
+
"loss": 0.3398,
|
| 178157 |
+
"step": 61930
|
| 178158 |
+
},
|
| 178159 |
+
{
|
| 178160 |
+
"epoch": 498.69,
|
| 178161 |
+
"learning_rate": 9.023846153846154e-06,
|
| 178162 |
+
"loss": 0.3201,
|
| 178163 |
+
"step": 61935
|
| 178164 |
+
},
|
| 178165 |
+
{
|
| 178166 |
+
"epoch": 498.73,
|
| 178167 |
+
"learning_rate": 9.023766025641026e-06,
|
| 178168 |
+
"loss": 0.349,
|
| 178169 |
+
"step": 61940
|
| 178170 |
+
},
|
| 178171 |
+
{
|
| 178172 |
+
"epoch": 498.77,
|
| 178173 |
+
"learning_rate": 9.023685897435897e-06,
|
| 178174 |
+
"loss": 0.5431,
|
| 178175 |
+
"step": 61945
|
| 178176 |
+
},
|
| 178177 |
+
{
|
| 178178 |
+
"epoch": 498.81,
|
| 178179 |
+
"learning_rate": 9.02360576923077e-06,
|
| 178180 |
+
"loss": 1.1888,
|
| 178181 |
+
"step": 61950
|
| 178182 |
+
},
|
| 178183 |
+
{
|
| 178184 |
+
"epoch": 498.85,
|
| 178185 |
+
"learning_rate": 9.023525641025642e-06,
|
| 178186 |
+
"loss": 0.3208,
|
| 178187 |
+
"step": 61955
|
| 178188 |
+
},
|
| 178189 |
+
{
|
| 178190 |
+
"epoch": 498.9,
|
| 178191 |
+
"learning_rate": 9.023445512820513e-06,
|
| 178192 |
+
"loss": 0.3231,
|
| 178193 |
+
"step": 61960
|
| 178194 |
+
},
|
| 178195 |
+
{
|
| 178196 |
+
"epoch": 498.94,
|
| 178197 |
+
"learning_rate": 9.023365384615386e-06,
|
| 178198 |
+
"loss": 0.3823,
|
| 178199 |
+
"step": 61965
|
| 178200 |
+
},
|
| 178201 |
+
{
|
| 178202 |
+
"epoch": 498.98,
|
| 178203 |
+
"learning_rate": 9.023285256410257e-06,
|
| 178204 |
+
"loss": 0.557,
|
| 178205 |
+
"step": 61970
|
| 178206 |
+
},
|
| 178207 |
+
{
|
| 178208 |
+
"epoch": 499.0,
|
| 178209 |
+
"eval_loss": 0.3884941339492798,
|
| 178210 |
+
"eval_runtime": 40.4848,
|
| 178211 |
+
"eval_samples_per_second": 20.847,
|
| 178212 |
+
"eval_steps_per_second": 0.667,
|
| 178213 |
+
"eval_wer": 0.1950281359071159,
|
| 178214 |
+
"step": 61973
|
| 178215 |
}
|
| 178216 |
],
|
| 178217 |
"max_steps": 620000,
|
| 178218 |
"num_train_epochs": 5000,
|
| 178219 |
+
"total_flos": 1.7440618623953663e+20,
|
| 178220 |
"trial_name": null,
|
| 178221 |
"trial_params": null
|
| 178222 |
}
|
model-bin/finetune/base/{checkpoint-61350 → checkpoint-61973}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629804951.1741285/events.out.tfevents.1629804951.c435e1c5ee04.920.121
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:94129def0cfa49d59b4d96e053f0aef4b96e7d78d940752bc7b251dc325a0911
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629805585.754903/events.out.tfevents.1629805585.c435e1c5ee04.920.123
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e7d4a441057de61f646bf20398aea83730ae96c2b686ee1acfa14c3ae8a8cda1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629806228.6566694/events.out.tfevents.1629806228.c435e1c5ee04.920.125
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44c50eb8b72825a7cd545f0e182831c56765fa44a859355441318c4302f73d9c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629806875.5539/events.out.tfevents.1629806875.c435e1c5ee04.920.127
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b0230bd87e40e31808f116061c339a91a1f531d4e690884691851e8ce2d9ce9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629807503.9229152/events.out.tfevents.1629807503.c435e1c5ee04.920.129
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1de29f6094e4f53699e982b10b0dd7c4142ac47dd138bdd5f7965bef648e5d6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629804951.c435e1c5ee04.920.120
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c8824fe371ea4bfc115fdb0ed6b0f9b247fb59bc85e6bd463bcf9086d4ee304
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629805585.c435e1c5ee04.920.122
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd11a12f9478a150340846003aee7b3b05e672d8307eb1874f7136068156d0f9
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629806228.c435e1c5ee04.920.124
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91e92cfec06856824f902f366e32aaaf63621547bfbf4e5e2a5ec816cfbc8635
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629806875.c435e1c5ee04.920.126
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30b1276085a49c1a16cf86bfd39e74ed2fb6258f093c3eb5a2a9d25910bc5d30
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629807503.c435e1c5ee04.920.128
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:889ee32c9352cd80a53944e80bf99e34d3ab32a7a47a5bf1dc0d8e07011f5191
|
| 3 |
+
size 8622
|