"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629944194.7486594/events.out.tfevents.1629944194.8e89bd551565.924.21 +3 -0
- model-bin/finetune/base/log/1629944627.279278/events.out.tfevents.1629944628.8e89bd551565.924.23 +3 -0
- model-bin/finetune/base/log/1629945066.2757561/events.out.tfevents.1629945066.8e89bd551565.924.25 +3 -0
- model-bin/finetune/base/log/1629945511.106437/events.out.tfevents.1629945511.8e89bd551565.924.27 +3 -0
- model-bin/finetune/base/log/1629945940.0520022/events.out.tfevents.1629945941.8e89bd551565.924.29 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629944194.8e89bd551565.924.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629944627.8e89bd551565.924.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629945065.8e89bd551565.924.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629945511.8e89bd551565.924.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629945940.8e89bd551565.924.28 +3 -0
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/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 sha256:8572d131ce8d59b8860c47164157d951991c9c0ef7b171cad275b160c84bd30b
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/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:a46bfb9ac87cdbf1e5989ab2bf749ab9a164a555e55d2c2eec7b5e09680b9616
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0b05680653568416b05354e34ad62d4f0ac27a86e0a07c5052a36b53751ea2d
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/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 sha256:43cf4a0bc3cccd678f9d1d7aac0abd075111f7eb824da38827242e3d76f31e4d
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/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 sha256:4919a489fdfa86cdcf94ef6754b2bfe5e67807bc39be182ad4fe93c03c95ae2a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -213366,11 +213366,800 @@
|
|
| 213366 |
"eval_steps_per_second": 0.708,
|
| 213367 |
"eval_wer": 0.18980540226546616,
|
| 213368 |
"step": 89603
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 213369 |
}
|
| 213370 |
],
|
| 213371 |
"max_steps": 625000,
|
| 213372 |
"num_train_epochs": 5000,
|
| 213373 |
-
"total_flos": 2.
|
| 213374 |
"trial_name": null,
|
| 213375 |
"trial_params": null
|
| 213376 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 721.0,
|
| 5 |
+
"global_step": 90224,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 213366 |
"eval_steps_per_second": 0.708,
|
| 213367 |
"eval_wer": 0.18980540226546616,
|
| 213368 |
"step": 89603
|
| 213369 |
+
},
|
| 213370 |
+
{
|
| 213371 |
+
"epoch": 722.02,
|
| 213372 |
+
"learning_rate": 8.580625e-06,
|
| 213373 |
+
"loss": 0.3116,
|
| 213374 |
+
"step": 89605
|
| 213375 |
+
},
|
| 213376 |
+
{
|
| 213377 |
+
"epoch": 722.06,
|
| 213378 |
+
"learning_rate": 8.580544871794872e-06,
|
| 213379 |
+
"loss": 0.3186,
|
| 213380 |
+
"step": 89610
|
| 213381 |
+
},
|
| 213382 |
+
{
|
| 213383 |
+
"epoch": 722.1,
|
| 213384 |
+
"learning_rate": 8.580464743589745e-06,
|
| 213385 |
+
"loss": 0.3267,
|
| 213386 |
+
"step": 89615
|
| 213387 |
+
},
|
| 213388 |
+
{
|
| 213389 |
+
"epoch": 722.14,
|
| 213390 |
+
"learning_rate": 8.580384615384616e-06,
|
| 213391 |
+
"loss": 0.3022,
|
| 213392 |
+
"step": 89620
|
| 213393 |
+
},
|
| 213394 |
+
{
|
| 213395 |
+
"epoch": 722.18,
|
| 213396 |
+
"learning_rate": 8.580304487179488e-06,
|
| 213397 |
+
"loss": 0.5975,
|
| 213398 |
+
"step": 89625
|
| 213399 |
+
},
|
| 213400 |
+
{
|
| 213401 |
+
"epoch": 722.22,
|
| 213402 |
+
"learning_rate": 8.580224358974359e-06,
|
| 213403 |
+
"loss": 1.3322,
|
| 213404 |
+
"step": 89630
|
| 213405 |
+
},
|
| 213406 |
+
{
|
| 213407 |
+
"epoch": 722.26,
|
| 213408 |
+
"learning_rate": 8.580144230769232e-06,
|
| 213409 |
+
"loss": 0.3857,
|
| 213410 |
+
"step": 89635
|
| 213411 |
+
},
|
| 213412 |
+
{
|
| 213413 |
+
"epoch": 722.3,
|
| 213414 |
+
"learning_rate": 8.580064102564103e-06,
|
| 213415 |
+
"loss": 0.3251,
|
| 213416 |
+
"step": 89640
|
| 213417 |
+
},
|
| 213418 |
+
{
|
| 213419 |
+
"epoch": 722.34,
|
| 213420 |
+
"learning_rate": 8.579983974358975e-06,
|
| 213421 |
+
"loss": 0.3432,
|
| 213422 |
+
"step": 89645
|
| 213423 |
+
},
|
| 213424 |
+
{
|
| 213425 |
+
"epoch": 722.38,
|
| 213426 |
+
"learning_rate": 8.579903846153846e-06,
|
| 213427 |
+
"loss": 0.545,
|
| 213428 |
+
"step": 89650
|
| 213429 |
+
},
|
| 213430 |
+
{
|
| 213431 |
+
"epoch": 722.42,
|
| 213432 |
+
"learning_rate": 8.57982371794872e-06,
|
| 213433 |
+
"loss": 0.9188,
|
| 213434 |
+
"step": 89655
|
| 213435 |
+
},
|
| 213436 |
+
{
|
| 213437 |
+
"epoch": 722.46,
|
| 213438 |
+
"learning_rate": 8.57974358974359e-06,
|
| 213439 |
+
"loss": 0.3483,
|
| 213440 |
+
"step": 89660
|
| 213441 |
+
},
|
| 213442 |
+
{
|
| 213443 |
+
"epoch": 722.5,
|
| 213444 |
+
"learning_rate": 8.579663461538462e-06,
|
| 213445 |
+
"loss": 0.351,
|
| 213446 |
+
"step": 89665
|
| 213447 |
+
},
|
| 213448 |
+
{
|
| 213449 |
+
"epoch": 722.54,
|
| 213450 |
+
"learning_rate": 8.579583333333335e-06,
|
| 213451 |
+
"loss": 0.4175,
|
| 213452 |
+
"step": 89670
|
| 213453 |
+
},
|
| 213454 |
+
{
|
| 213455 |
+
"epoch": 722.58,
|
| 213456 |
+
"learning_rate": 8.579503205128206e-06,
|
| 213457 |
+
"loss": 0.6306,
|
| 213458 |
+
"step": 89675
|
| 213459 |
+
},
|
| 213460 |
+
{
|
| 213461 |
+
"epoch": 722.62,
|
| 213462 |
+
"learning_rate": 8.579423076923078e-06,
|
| 213463 |
+
"loss": 1.0589,
|
| 213464 |
+
"step": 89680
|
| 213465 |
+
},
|
| 213466 |
+
{
|
| 213467 |
+
"epoch": 722.66,
|
| 213468 |
+
"learning_rate": 8.579342948717949e-06,
|
| 213469 |
+
"loss": 0.2908,
|
| 213470 |
+
"step": 89685
|
| 213471 |
+
},
|
| 213472 |
+
{
|
| 213473 |
+
"epoch": 722.7,
|
| 213474 |
+
"learning_rate": 8.579262820512822e-06,
|
| 213475 |
+
"loss": 0.3075,
|
| 213476 |
+
"step": 89690
|
| 213477 |
+
},
|
| 213478 |
+
{
|
| 213479 |
+
"epoch": 722.74,
|
| 213480 |
+
"learning_rate": 8.579182692307693e-06,
|
| 213481 |
+
"loss": 0.4415,
|
| 213482 |
+
"step": 89695
|
| 213483 |
+
},
|
| 213484 |
+
{
|
| 213485 |
+
"epoch": 722.78,
|
| 213486 |
+
"learning_rate": 8.579102564102565e-06,
|
| 213487 |
+
"loss": 0.8988,
|
| 213488 |
+
"step": 89700
|
| 213489 |
+
},
|
| 213490 |
+
{
|
| 213491 |
+
"epoch": 722.82,
|
| 213492 |
+
"learning_rate": 8.579022435897436e-06,
|
| 213493 |
+
"loss": 1.005,
|
| 213494 |
+
"step": 89705
|
| 213495 |
+
},
|
| 213496 |
+
{
|
| 213497 |
+
"epoch": 722.86,
|
| 213498 |
+
"learning_rate": 8.57894230769231e-06,
|
| 213499 |
+
"loss": 0.2915,
|
| 213500 |
+
"step": 89710
|
| 213501 |
+
},
|
| 213502 |
+
{
|
| 213503 |
+
"epoch": 722.9,
|
| 213504 |
+
"learning_rate": 8.57886217948718e-06,
|
| 213505 |
+
"loss": 0.2801,
|
| 213506 |
+
"step": 89715
|
| 213507 |
+
},
|
| 213508 |
+
{
|
| 213509 |
+
"epoch": 722.94,
|
| 213510 |
+
"learning_rate": 8.578782051282052e-06,
|
| 213511 |
+
"loss": 0.3683,
|
| 213512 |
+
"step": 89720
|
| 213513 |
+
},
|
| 213514 |
+
{
|
| 213515 |
+
"epoch": 722.98,
|
| 213516 |
+
"learning_rate": 8.578701923076925e-06,
|
| 213517 |
+
"loss": 0.8019,
|
| 213518 |
+
"step": 89725
|
| 213519 |
+
},
|
| 213520 |
+
{
|
| 213521 |
+
"epoch": 723.0,
|
| 213522 |
+
"eval_loss": 0.3949987590312958,
|
| 213523 |
+
"eval_runtime": 40.4989,
|
| 213524 |
+
"eval_samples_per_second": 20.791,
|
| 213525 |
+
"eval_steps_per_second": 0.667,
|
| 213526 |
+
"eval_wer": 0.18734617055161432,
|
| 213527 |
+
"step": 89727
|
| 213528 |
+
},
|
| 213529 |
+
{
|
| 213530 |
+
"epoch": 723.02,
|
| 213531 |
+
"learning_rate": 8.578621794871795e-06,
|
| 213532 |
+
"loss": 0.3285,
|
| 213533 |
+
"step": 89730
|
| 213534 |
+
},
|
| 213535 |
+
{
|
| 213536 |
+
"epoch": 723.06,
|
| 213537 |
+
"learning_rate": 8.578541666666668e-06,
|
| 213538 |
+
"loss": 0.3242,
|
| 213539 |
+
"step": 89735
|
| 213540 |
+
},
|
| 213541 |
+
{
|
| 213542 |
+
"epoch": 723.1,
|
| 213543 |
+
"learning_rate": 8.578461538461539e-06,
|
| 213544 |
+
"loss": 0.3054,
|
| 213545 |
+
"step": 89740
|
| 213546 |
+
},
|
| 213547 |
+
{
|
| 213548 |
+
"epoch": 723.15,
|
| 213549 |
+
"learning_rate": 8.57838141025641e-06,
|
| 213550 |
+
"loss": 0.4542,
|
| 213551 |
+
"step": 89745
|
| 213552 |
+
},
|
| 213553 |
+
{
|
| 213554 |
+
"epoch": 723.19,
|
| 213555 |
+
"learning_rate": 8.578301282051282e-06,
|
| 213556 |
+
"loss": 0.712,
|
| 213557 |
+
"step": 89750
|
| 213558 |
+
},
|
| 213559 |
+
{
|
| 213560 |
+
"epoch": 723.23,
|
| 213561 |
+
"learning_rate": 8.578221153846155e-06,
|
| 213562 |
+
"loss": 0.8091,
|
| 213563 |
+
"step": 89755
|
| 213564 |
+
},
|
| 213565 |
+
{
|
| 213566 |
+
"epoch": 723.27,
|
| 213567 |
+
"learning_rate": 8.578141025641026e-06,
|
| 213568 |
+
"loss": 0.3459,
|
| 213569 |
+
"step": 89760
|
| 213570 |
+
},
|
| 213571 |
+
{
|
| 213572 |
+
"epoch": 723.31,
|
| 213573 |
+
"learning_rate": 8.578060897435898e-06,
|
| 213574 |
+
"loss": 0.2816,
|
| 213575 |
+
"step": 89765
|
| 213576 |
+
},
|
| 213577 |
+
{
|
| 213578 |
+
"epoch": 723.35,
|
| 213579 |
+
"learning_rate": 8.57798076923077e-06,
|
| 213580 |
+
"loss": 0.3782,
|
| 213581 |
+
"step": 89770
|
| 213582 |
+
},
|
| 213583 |
+
{
|
| 213584 |
+
"epoch": 723.39,
|
| 213585 |
+
"learning_rate": 8.577900641025642e-06,
|
| 213586 |
+
"loss": 0.73,
|
| 213587 |
+
"step": 89775
|
| 213588 |
+
},
|
| 213589 |
+
{
|
| 213590 |
+
"epoch": 723.43,
|
| 213591 |
+
"learning_rate": 8.577820512820513e-06,
|
| 213592 |
+
"loss": 0.9373,
|
| 213593 |
+
"step": 89780
|
| 213594 |
+
},
|
| 213595 |
+
{
|
| 213596 |
+
"epoch": 723.47,
|
| 213597 |
+
"learning_rate": 8.577740384615385e-06,
|
| 213598 |
+
"loss": 0.3174,
|
| 213599 |
+
"step": 89785
|
| 213600 |
+
},
|
| 213601 |
+
{
|
| 213602 |
+
"epoch": 723.51,
|
| 213603 |
+
"learning_rate": 8.577660256410258e-06,
|
| 213604 |
+
"loss": 0.333,
|
| 213605 |
+
"step": 89790
|
| 213606 |
+
},
|
| 213607 |
+
{
|
| 213608 |
+
"epoch": 723.55,
|
| 213609 |
+
"learning_rate": 8.577580128205129e-06,
|
| 213610 |
+
"loss": 0.4021,
|
| 213611 |
+
"step": 89795
|
| 213612 |
+
},
|
| 213613 |
+
{
|
| 213614 |
+
"epoch": 723.59,
|
| 213615 |
+
"learning_rate": 8.5775e-06,
|
| 213616 |
+
"loss": 0.7118,
|
| 213617 |
+
"step": 89800
|
| 213618 |
+
},
|
| 213619 |
+
{
|
| 213620 |
+
"epoch": 723.63,
|
| 213621 |
+
"learning_rate": 8.577419871794872e-06,
|
| 213622 |
+
"loss": 0.8595,
|
| 213623 |
+
"step": 89805
|
| 213624 |
+
},
|
| 213625 |
+
{
|
| 213626 |
+
"epoch": 723.67,
|
| 213627 |
+
"learning_rate": 8.577339743589745e-06,
|
| 213628 |
+
"loss": 0.4852,
|
| 213629 |
+
"step": 89810
|
| 213630 |
+
},
|
| 213631 |
+
{
|
| 213632 |
+
"epoch": 723.71,
|
| 213633 |
+
"learning_rate": 8.577259615384616e-06,
|
| 213634 |
+
"loss": 0.3192,
|
| 213635 |
+
"step": 89815
|
| 213636 |
+
},
|
| 213637 |
+
{
|
| 213638 |
+
"epoch": 723.75,
|
| 213639 |
+
"learning_rate": 8.577179487179488e-06,
|
| 213640 |
+
"loss": 0.3907,
|
| 213641 |
+
"step": 89820
|
| 213642 |
+
},
|
| 213643 |
+
{
|
| 213644 |
+
"epoch": 723.79,
|
| 213645 |
+
"learning_rate": 8.57709935897436e-06,
|
| 213646 |
+
"loss": 0.7862,
|
| 213647 |
+
"step": 89825
|
| 213648 |
+
},
|
| 213649 |
+
{
|
| 213650 |
+
"epoch": 723.83,
|
| 213651 |
+
"learning_rate": 8.577019230769232e-06,
|
| 213652 |
+
"loss": 0.9558,
|
| 213653 |
+
"step": 89830
|
| 213654 |
+
},
|
| 213655 |
+
{
|
| 213656 |
+
"epoch": 723.87,
|
| 213657 |
+
"learning_rate": 8.576939102564103e-06,
|
| 213658 |
+
"loss": 0.2887,
|
| 213659 |
+
"step": 89835
|
| 213660 |
+
},
|
| 213661 |
+
{
|
| 213662 |
+
"epoch": 723.91,
|
| 213663 |
+
"learning_rate": 8.576858974358975e-06,
|
| 213664 |
+
"loss": 0.3314,
|
| 213665 |
+
"step": 89840
|
| 213666 |
+
},
|
| 213667 |
+
{
|
| 213668 |
+
"epoch": 723.95,
|
| 213669 |
+
"learning_rate": 8.576778846153848e-06,
|
| 213670 |
+
"loss": 0.4156,
|
| 213671 |
+
"step": 89845
|
| 213672 |
+
},
|
| 213673 |
+
{
|
| 213674 |
+
"epoch": 723.99,
|
| 213675 |
+
"learning_rate": 8.576698717948717e-06,
|
| 213676 |
+
"loss": 0.879,
|
| 213677 |
+
"step": 89850
|
| 213678 |
+
},
|
| 213679 |
+
{
|
| 213680 |
+
"epoch": 724.0,
|
| 213681 |
+
"eval_loss": 0.3989127278327942,
|
| 213682 |
+
"eval_runtime": 40.1239,
|
| 213683 |
+
"eval_samples_per_second": 20.985,
|
| 213684 |
+
"eval_steps_per_second": 0.673,
|
| 213685 |
+
"eval_wer": 0.1779624588049864,
|
| 213686 |
+
"step": 89851
|
| 213687 |
+
},
|
| 213688 |
+
{
|
| 213689 |
+
"epoch": 724.03,
|
| 213690 |
+
"learning_rate": 8.57661858974359e-06,
|
| 213691 |
+
"loss": 0.3534,
|
| 213692 |
+
"step": 89855
|
| 213693 |
+
},
|
| 213694 |
+
{
|
| 213695 |
+
"epoch": 724.07,
|
| 213696 |
+
"learning_rate": 8.576538461538462e-06,
|
| 213697 |
+
"loss": 0.2546,
|
| 213698 |
+
"step": 89860
|
| 213699 |
+
},
|
| 213700 |
+
{
|
| 213701 |
+
"epoch": 724.11,
|
| 213702 |
+
"learning_rate": 8.576458333333333e-06,
|
| 213703 |
+
"loss": 0.3423,
|
| 213704 |
+
"step": 89865
|
| 213705 |
+
},
|
| 213706 |
+
{
|
| 213707 |
+
"epoch": 724.15,
|
| 213708 |
+
"learning_rate": 8.576378205128206e-06,
|
| 213709 |
+
"loss": 0.4294,
|
| 213710 |
+
"step": 89870
|
| 213711 |
+
},
|
| 213712 |
+
{
|
| 213713 |
+
"epoch": 724.19,
|
| 213714 |
+
"learning_rate": 8.576298076923078e-06,
|
| 213715 |
+
"loss": 0.9324,
|
| 213716 |
+
"step": 89875
|
| 213717 |
+
},
|
| 213718 |
+
{
|
| 213719 |
+
"epoch": 724.23,
|
| 213720 |
+
"learning_rate": 8.576217948717949e-06,
|
| 213721 |
+
"loss": 0.6242,
|
| 213722 |
+
"step": 89880
|
| 213723 |
+
},
|
| 213724 |
+
{
|
| 213725 |
+
"epoch": 724.27,
|
| 213726 |
+
"learning_rate": 8.57613782051282e-06,
|
| 213727 |
+
"loss": 0.259,
|
| 213728 |
+
"step": 89885
|
| 213729 |
+
},
|
| 213730 |
+
{
|
| 213731 |
+
"epoch": 724.31,
|
| 213732 |
+
"learning_rate": 8.576057692307693e-06,
|
| 213733 |
+
"loss": 0.3533,
|
| 213734 |
+
"step": 89890
|
| 213735 |
+
},
|
| 213736 |
+
{
|
| 213737 |
+
"epoch": 724.35,
|
| 213738 |
+
"learning_rate": 8.575977564102565e-06,
|
| 213739 |
+
"loss": 0.3916,
|
| 213740 |
+
"step": 89895
|
| 213741 |
+
},
|
| 213742 |
+
{
|
| 213743 |
+
"epoch": 724.4,
|
| 213744 |
+
"learning_rate": 8.575897435897436e-06,
|
| 213745 |
+
"loss": 0.9333,
|
| 213746 |
+
"step": 89900
|
| 213747 |
+
},
|
| 213748 |
+
{
|
| 213749 |
+
"epoch": 724.44,
|
| 213750 |
+
"learning_rate": 8.575817307692307e-06,
|
| 213751 |
+
"loss": 0.6326,
|
| 213752 |
+
"step": 89905
|
| 213753 |
+
},
|
| 213754 |
+
{
|
| 213755 |
+
"epoch": 724.48,
|
| 213756 |
+
"learning_rate": 8.57573717948718e-06,
|
| 213757 |
+
"loss": 0.321,
|
| 213758 |
+
"step": 89910
|
| 213759 |
+
},
|
| 213760 |
+
{
|
| 213761 |
+
"epoch": 724.52,
|
| 213762 |
+
"learning_rate": 8.575657051282052e-06,
|
| 213763 |
+
"loss": 0.3128,
|
| 213764 |
+
"step": 89915
|
| 213765 |
+
},
|
| 213766 |
+
{
|
| 213767 |
+
"epoch": 724.56,
|
| 213768 |
+
"learning_rate": 8.575576923076923e-06,
|
| 213769 |
+
"loss": 0.4245,
|
| 213770 |
+
"step": 89920
|
| 213771 |
+
},
|
| 213772 |
+
{
|
| 213773 |
+
"epoch": 724.6,
|
| 213774 |
+
"learning_rate": 8.575496794871796e-06,
|
| 213775 |
+
"loss": 0.7346,
|
| 213776 |
+
"step": 89925
|
| 213777 |
+
},
|
| 213778 |
+
{
|
| 213779 |
+
"epoch": 724.64,
|
| 213780 |
+
"learning_rate": 8.575416666666668e-06,
|
| 213781 |
+
"loss": 0.6214,
|
| 213782 |
+
"step": 89930
|
| 213783 |
+
},
|
| 213784 |
+
{
|
| 213785 |
+
"epoch": 724.68,
|
| 213786 |
+
"learning_rate": 8.575336538461539e-06,
|
| 213787 |
+
"loss": 0.309,
|
| 213788 |
+
"step": 89935
|
| 213789 |
+
},
|
| 213790 |
+
{
|
| 213791 |
+
"epoch": 724.72,
|
| 213792 |
+
"learning_rate": 8.57525641025641e-06,
|
| 213793 |
+
"loss": 0.3766,
|
| 213794 |
+
"step": 89940
|
| 213795 |
+
},
|
| 213796 |
+
{
|
| 213797 |
+
"epoch": 724.76,
|
| 213798 |
+
"learning_rate": 8.575176282051283e-06,
|
| 213799 |
+
"loss": 0.3731,
|
| 213800 |
+
"step": 89945
|
| 213801 |
+
},
|
| 213802 |
+
{
|
| 213803 |
+
"epoch": 724.8,
|
| 213804 |
+
"learning_rate": 8.575096153846155e-06,
|
| 213805 |
+
"loss": 0.9065,
|
| 213806 |
+
"step": 89950
|
| 213807 |
+
},
|
| 213808 |
+
{
|
| 213809 |
+
"epoch": 724.84,
|
| 213810 |
+
"learning_rate": 8.575016025641026e-06,
|
| 213811 |
+
"loss": 0.7299,
|
| 213812 |
+
"step": 89955
|
| 213813 |
+
},
|
| 213814 |
+
{
|
| 213815 |
+
"epoch": 724.88,
|
| 213816 |
+
"learning_rate": 8.574935897435897e-06,
|
| 213817 |
+
"loss": 0.2867,
|
| 213818 |
+
"step": 89960
|
| 213819 |
+
},
|
| 213820 |
+
{
|
| 213821 |
+
"epoch": 724.92,
|
| 213822 |
+
"learning_rate": 8.57485576923077e-06,
|
| 213823 |
+
"loss": 0.3506,
|
| 213824 |
+
"step": 89965
|
| 213825 |
+
},
|
| 213826 |
+
{
|
| 213827 |
+
"epoch": 724.96,
|
| 213828 |
+
"learning_rate": 8.574775641025642e-06,
|
| 213829 |
+
"loss": 0.4504,
|
| 213830 |
+
"step": 89970
|
| 213831 |
+
},
|
| 213832 |
+
{
|
| 213833 |
+
"epoch": 725.0,
|
| 213834 |
+
"learning_rate": 8.574695512820513e-06,
|
| 213835 |
+
"loss": 1.1994,
|
| 213836 |
+
"step": 89975
|
| 213837 |
+
},
|
| 213838 |
+
{
|
| 213839 |
+
"epoch": 725.0,
|
| 213840 |
+
"eval_loss": 0.3205711245536804,
|
| 213841 |
+
"eval_runtime": 39.7661,
|
| 213842 |
+
"eval_samples_per_second": 21.149,
|
| 213843 |
+
"eval_steps_per_second": 0.679,
|
| 213844 |
+
"eval_wer": 0.18327764056370768,
|
| 213845 |
+
"step": 89975
|
| 213846 |
+
},
|
| 213847 |
+
{
|
| 213848 |
+
"epoch": 725.04,
|
| 213849 |
+
"learning_rate": 8.574615384615386e-06,
|
| 213850 |
+
"loss": 0.3563,
|
| 213851 |
+
"step": 89980
|
| 213852 |
+
},
|
| 213853 |
+
{
|
| 213854 |
+
"epoch": 725.08,
|
| 213855 |
+
"learning_rate": 8.574535256410258e-06,
|
| 213856 |
+
"loss": 0.3117,
|
| 213857 |
+
"step": 89985
|
| 213858 |
+
},
|
| 213859 |
+
{
|
| 213860 |
+
"epoch": 725.12,
|
| 213861 |
+
"learning_rate": 8.574455128205129e-06,
|
| 213862 |
+
"loss": 0.3239,
|
| 213863 |
+
"step": 89990
|
| 213864 |
+
},
|
| 213865 |
+
{
|
| 213866 |
+
"epoch": 725.16,
|
| 213867 |
+
"learning_rate": 8.574375e-06,
|
| 213868 |
+
"loss": 0.4605,
|
| 213869 |
+
"step": 89995
|
| 213870 |
+
},
|
| 213871 |
+
{
|
| 213872 |
+
"epoch": 725.2,
|
| 213873 |
+
"learning_rate": 8.574294871794873e-06,
|
| 213874 |
+
"loss": 1.2709,
|
| 213875 |
+
"step": 90000
|
| 213876 |
+
},
|
| 213877 |
+
{
|
| 213878 |
+
"epoch": 725.24,
|
| 213879 |
+
"learning_rate": 8.574214743589743e-06,
|
| 213880 |
+
"loss": 0.3652,
|
| 213881 |
+
"step": 90005
|
| 213882 |
+
},
|
| 213883 |
+
{
|
| 213884 |
+
"epoch": 725.28,
|
| 213885 |
+
"learning_rate": 8.574134615384616e-06,
|
| 213886 |
+
"loss": 0.3051,
|
| 213887 |
+
"step": 90010
|
| 213888 |
+
},
|
| 213889 |
+
{
|
| 213890 |
+
"epoch": 725.32,
|
| 213891 |
+
"learning_rate": 8.57405448717949e-06,
|
| 213892 |
+
"loss": 0.3216,
|
| 213893 |
+
"step": 90015
|
| 213894 |
+
},
|
| 213895 |
+
{
|
| 213896 |
+
"epoch": 725.36,
|
| 213897 |
+
"learning_rate": 8.573974358974359e-06,
|
| 213898 |
+
"loss": 0.4698,
|
| 213899 |
+
"step": 90020
|
| 213900 |
+
},
|
| 213901 |
+
{
|
| 213902 |
+
"epoch": 725.4,
|
| 213903 |
+
"learning_rate": 8.573894230769232e-06,
|
| 213904 |
+
"loss": 1.2027,
|
| 213905 |
+
"step": 90025
|
| 213906 |
+
},
|
| 213907 |
+
{
|
| 213908 |
+
"epoch": 725.44,
|
| 213909 |
+
"learning_rate": 8.573814102564103e-06,
|
| 213910 |
+
"loss": 0.3991,
|
| 213911 |
+
"step": 90030
|
| 213912 |
+
},
|
| 213913 |
+
{
|
| 213914 |
+
"epoch": 725.48,
|
| 213915 |
+
"learning_rate": 8.573733974358975e-06,
|
| 213916 |
+
"loss": 0.3068,
|
| 213917 |
+
"step": 90035
|
| 213918 |
+
},
|
| 213919 |
+
{
|
| 213920 |
+
"epoch": 725.52,
|
| 213921 |
+
"learning_rate": 8.573653846153846e-06,
|
| 213922 |
+
"loss": 0.3173,
|
| 213923 |
+
"step": 90040
|
| 213924 |
+
},
|
| 213925 |
+
{
|
| 213926 |
+
"epoch": 725.56,
|
| 213927 |
+
"learning_rate": 8.573573717948719e-06,
|
| 213928 |
+
"loss": 0.4505,
|
| 213929 |
+
"step": 90045
|
| 213930 |
+
},
|
| 213931 |
+
{
|
| 213932 |
+
"epoch": 725.6,
|
| 213933 |
+
"learning_rate": 8.57349358974359e-06,
|
| 213934 |
+
"loss": 1.1834,
|
| 213935 |
+
"step": 90050
|
| 213936 |
+
},
|
| 213937 |
+
{
|
| 213938 |
+
"epoch": 725.64,
|
| 213939 |
+
"learning_rate": 8.573413461538462e-06,
|
| 213940 |
+
"loss": 0.3149,
|
| 213941 |
+
"step": 90055
|
| 213942 |
+
},
|
| 213943 |
+
{
|
| 213944 |
+
"epoch": 725.68,
|
| 213945 |
+
"learning_rate": 8.573333333333333e-06,
|
| 213946 |
+
"loss": 0.286,
|
| 213947 |
+
"step": 90060
|
| 213948 |
+
},
|
| 213949 |
+
{
|
| 213950 |
+
"epoch": 725.72,
|
| 213951 |
+
"learning_rate": 8.573253205128206e-06,
|
| 213952 |
+
"loss": 0.2983,
|
| 213953 |
+
"step": 90065
|
| 213954 |
+
},
|
| 213955 |
+
{
|
| 213956 |
+
"epoch": 725.76,
|
| 213957 |
+
"learning_rate": 8.573173076923078e-06,
|
| 213958 |
+
"loss": 0.4423,
|
| 213959 |
+
"step": 90070
|
| 213960 |
+
},
|
| 213961 |
+
{
|
| 213962 |
+
"epoch": 725.8,
|
| 213963 |
+
"learning_rate": 8.573092948717949e-06,
|
| 213964 |
+
"loss": 1.1159,
|
| 213965 |
+
"step": 90075
|
| 213966 |
+
},
|
| 213967 |
+
{
|
| 213968 |
+
"epoch": 725.84,
|
| 213969 |
+
"learning_rate": 8.573012820512822e-06,
|
| 213970 |
+
"loss": 0.4292,
|
| 213971 |
+
"step": 90080
|
| 213972 |
+
},
|
| 213973 |
+
{
|
| 213974 |
+
"epoch": 725.88,
|
| 213975 |
+
"learning_rate": 8.572932692307693e-06,
|
| 213976 |
+
"loss": 0.2777,
|
| 213977 |
+
"step": 90085
|
| 213978 |
+
},
|
| 213979 |
+
{
|
| 213980 |
+
"epoch": 725.92,
|
| 213981 |
+
"learning_rate": 8.572852564102565e-06,
|
| 213982 |
+
"loss": 0.3598,
|
| 213983 |
+
"step": 90090
|
| 213984 |
+
},
|
| 213985 |
+
{
|
| 213986 |
+
"epoch": 725.96,
|
| 213987 |
+
"learning_rate": 8.572772435897436e-06,
|
| 213988 |
+
"loss": 0.5791,
|
| 213989 |
+
"step": 90095
|
| 213990 |
+
},
|
| 213991 |
+
{
|
| 213992 |
+
"epoch": 726.0,
|
| 213993 |
+
"eval_loss": 0.39957037568092346,
|
| 213994 |
+
"eval_runtime": 40.0708,
|
| 213995 |
+
"eval_samples_per_second": 20.988,
|
| 213996 |
+
"eval_steps_per_second": 0.674,
|
| 213997 |
+
"eval_wer": 0.19568990454150997,
|
| 213998 |
+
"step": 90099
|
| 213999 |
+
},
|
| 214000 |
+
{
|
| 214001 |
+
"epoch": 720.01,
|
| 214002 |
+
"learning_rate": 8.572692307692309e-06,
|
| 214003 |
+
"loss": 0.3887,
|
| 214004 |
+
"step": 90100
|
| 214005 |
+
},
|
| 214006 |
+
{
|
| 214007 |
+
"epoch": 720.05,
|
| 214008 |
+
"learning_rate": 8.57261217948718e-06,
|
| 214009 |
+
"loss": 0.2915,
|
| 214010 |
+
"step": 90105
|
| 214011 |
+
},
|
| 214012 |
+
{
|
| 214013 |
+
"epoch": 720.09,
|
| 214014 |
+
"learning_rate": 8.572532051282052e-06,
|
| 214015 |
+
"loss": 0.3146,
|
| 214016 |
+
"step": 90110
|
| 214017 |
+
},
|
| 214018 |
+
{
|
| 214019 |
+
"epoch": 720.13,
|
| 214020 |
+
"learning_rate": 8.572451923076925e-06,
|
| 214021 |
+
"loss": 0.3634,
|
| 214022 |
+
"step": 90115
|
| 214023 |
+
},
|
| 214024 |
+
{
|
| 214025 |
+
"epoch": 720.17,
|
| 214026 |
+
"learning_rate": 8.572371794871796e-06,
|
| 214027 |
+
"loss": 0.5349,
|
| 214028 |
+
"step": 90120
|
| 214029 |
+
},
|
| 214030 |
+
{
|
| 214031 |
+
"epoch": 720.21,
|
| 214032 |
+
"learning_rate": 8.572291666666668e-06,
|
| 214033 |
+
"loss": 1.1848,
|
| 214034 |
+
"step": 90125
|
| 214035 |
+
},
|
| 214036 |
+
{
|
| 214037 |
+
"epoch": 720.25,
|
| 214038 |
+
"learning_rate": 8.572211538461539e-06,
|
| 214039 |
+
"loss": 0.3404,
|
| 214040 |
+
"step": 90130
|
| 214041 |
+
},
|
| 214042 |
+
{
|
| 214043 |
+
"epoch": 720.29,
|
| 214044 |
+
"learning_rate": 8.572131410256412e-06,
|
| 214045 |
+
"loss": 0.3644,
|
| 214046 |
+
"step": 90135
|
| 214047 |
+
},
|
| 214048 |
+
{
|
| 214049 |
+
"epoch": 720.33,
|
| 214050 |
+
"learning_rate": 8.572051282051282e-06,
|
| 214051 |
+
"loss": 0.3367,
|
| 214052 |
+
"step": 90140
|
| 214053 |
+
},
|
| 214054 |
+
{
|
| 214055 |
+
"epoch": 720.37,
|
| 214056 |
+
"learning_rate": 8.571971153846155e-06,
|
| 214057 |
+
"loss": 0.5637,
|
| 214058 |
+
"step": 90145
|
| 214059 |
+
},
|
| 214060 |
+
{
|
| 214061 |
+
"epoch": 720.41,
|
| 214062 |
+
"learning_rate": 8.571891025641026e-06,
|
| 214063 |
+
"loss": 1.217,
|
| 214064 |
+
"step": 90150
|
| 214065 |
+
},
|
| 214066 |
+
{
|
| 214067 |
+
"epoch": 720.45,
|
| 214068 |
+
"learning_rate": 8.571810897435897e-06,
|
| 214069 |
+
"loss": 0.3629,
|
| 214070 |
+
"step": 90155
|
| 214071 |
+
},
|
| 214072 |
+
{
|
| 214073 |
+
"epoch": 720.49,
|
| 214074 |
+
"learning_rate": 8.571730769230769e-06,
|
| 214075 |
+
"loss": 0.2639,
|
| 214076 |
+
"step": 90160
|
| 214077 |
+
},
|
| 214078 |
+
{
|
| 214079 |
+
"epoch": 720.53,
|
| 214080 |
+
"learning_rate": 8.571650641025642e-06,
|
| 214081 |
+
"loss": 0.327,
|
| 214082 |
+
"step": 90165
|
| 214083 |
+
},
|
| 214084 |
+
{
|
| 214085 |
+
"epoch": 720.57,
|
| 214086 |
+
"learning_rate": 8.571570512820513e-06,
|
| 214087 |
+
"loss": 0.4916,
|
| 214088 |
+
"step": 90170
|
| 214089 |
+
},
|
| 214090 |
+
{
|
| 214091 |
+
"epoch": 720.61,
|
| 214092 |
+
"learning_rate": 8.571490384615385e-06,
|
| 214093 |
+
"loss": 1.2959,
|
| 214094 |
+
"step": 90175
|
| 214095 |
+
},
|
| 214096 |
+
{
|
| 214097 |
+
"epoch": 720.65,
|
| 214098 |
+
"learning_rate": 8.571410256410258e-06,
|
| 214099 |
+
"loss": 0.3346,
|
| 214100 |
+
"step": 90180
|
| 214101 |
+
},
|
| 214102 |
+
{
|
| 214103 |
+
"epoch": 720.69,
|
| 214104 |
+
"learning_rate": 8.571330128205129e-06,
|
| 214105 |
+
"loss": 0.2677,
|
| 214106 |
+
"step": 90185
|
| 214107 |
+
},
|
| 214108 |
+
{
|
| 214109 |
+
"epoch": 720.73,
|
| 214110 |
+
"learning_rate": 8.57125e-06,
|
| 214111 |
+
"loss": 0.2892,
|
| 214112 |
+
"step": 90190
|
| 214113 |
+
},
|
| 214114 |
+
{
|
| 214115 |
+
"epoch": 720.77,
|
| 214116 |
+
"learning_rate": 8.571169871794872e-06,
|
| 214117 |
+
"loss": 0.5406,
|
| 214118 |
+
"step": 90195
|
| 214119 |
+
},
|
| 214120 |
+
{
|
| 214121 |
+
"epoch": 720.81,
|
| 214122 |
+
"learning_rate": 8.571089743589745e-06,
|
| 214123 |
+
"loss": 1.1985,
|
| 214124 |
+
"step": 90200
|
| 214125 |
+
},
|
| 214126 |
+
{
|
| 214127 |
+
"epoch": 720.85,
|
| 214128 |
+
"learning_rate": 8.571009615384616e-06,
|
| 214129 |
+
"loss": 0.3474,
|
| 214130 |
+
"step": 90205
|
| 214131 |
+
},
|
| 214132 |
+
{
|
| 214133 |
+
"epoch": 720.89,
|
| 214134 |
+
"learning_rate": 8.570929487179487e-06,
|
| 214135 |
+
"loss": 0.2944,
|
| 214136 |
+
"step": 90210
|
| 214137 |
+
},
|
| 214138 |
+
{
|
| 214139 |
+
"epoch": 720.93,
|
| 214140 |
+
"learning_rate": 8.57084935897436e-06,
|
| 214141 |
+
"loss": 0.3832,
|
| 214142 |
+
"step": 90215
|
| 214143 |
+
},
|
| 214144 |
+
{
|
| 214145 |
+
"epoch": 720.97,
|
| 214146 |
+
"learning_rate": 8.570769230769232e-06,
|
| 214147 |
+
"loss": 0.5242,
|
| 214148 |
+
"step": 90220
|
| 214149 |
+
},
|
| 214150 |
+
{
|
| 214151 |
+
"epoch": 721.0,
|
| 214152 |
+
"eval_loss": 0.46499449014663696,
|
| 214153 |
+
"eval_runtime": 39.44,
|
| 214154 |
+
"eval_samples_per_second": 21.324,
|
| 214155 |
+
"eval_steps_per_second": 0.685,
|
| 214156 |
+
"eval_wer": 0.19095368633674795,
|
| 214157 |
+
"step": 90224
|
| 214158 |
}
|
| 214159 |
],
|
| 214160 |
"max_steps": 625000,
|
| 214161 |
"num_train_epochs": 5000,
|
| 214162 |
+
"total_flos": 2.5390227896927135e+20,
|
| 214163 |
"trial_name": null,
|
| 214164 |
"trial_params": null
|
| 214165 |
}
|
model-bin/finetune/base/{checkpoint-89603 → checkpoint-90224}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629944194.7486594/events.out.tfevents.1629944194.8e89bd551565.924.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7051708776cb9183b6074fc0845310dc9e17de3ede72a0ed70e99cc10790c270
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629944627.279278/events.out.tfevents.1629944628.8e89bd551565.924.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:418c35933551b00ffaa98f3d90e99ef406c876a62ffc8b2f0618f667cadbaa7b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629945066.2757561/events.out.tfevents.1629945066.8e89bd551565.924.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf31ddd57e62fe24c1be7d8f27d771cdebd68cce67ff1fc13738c26ca95d5a0d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629945511.106437/events.out.tfevents.1629945511.8e89bd551565.924.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:059d2bbb31f00a98e12860a757702fa53e4af464c769de56e58495508fcd7d98
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629945940.0520022/events.out.tfevents.1629945941.8e89bd551565.924.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7714cbb651a242f7a490bb3b1a5e2b95427c792d56683e65906757f5ac780e1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629944194.8e89bd551565.924.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4d9bb55e42c8a1f45c434705057ecdbecdebe1793c92c0e7e29de0c3c8ec430
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629944627.8e89bd551565.924.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5c333e46b236b83878bd258cde3d51afdbda4c5d1b54571a5573ec27452e46a5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629945065.8e89bd551565.924.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7a529e6a6d882965e846a26feaa6cdebe0b27150f5f019af1ae105de7b126c5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629945511.8e89bd551565.924.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:206483f5f97009cee5dbb6d61325dcf43835ded6ee9b2ebb4e3b615ca59b4bda
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629945940.8e89bd551565.924.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05c126ba66ed7f53dd76691759b1230125c78e10cc216dac675c176b6392096c
|
| 3 |
+
size 8622
|