"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629736186.820362/events.out.tfevents.1629736186.74272264b15c.932.173 +3 -0
- model-bin/finetune/base/log/1629736861.6649554/events.out.tfevents.1629736861.74272264b15c.932.175 +3 -0
- model-bin/finetune/base/log/1629737496.0693905/events.out.tfevents.1629737496.74272264b15c.932.177 +3 -0
- model-bin/finetune/base/log/1629738136.82416/events.out.tfevents.1629738136.74272264b15c.932.179 +3 -0
- model-bin/finetune/base/log/1629738769.9003348/events.out.tfevents.1629738769.74272264b15c.932.181 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629736186.74272264b15c.932.172 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629736861.74272264b15c.932.174 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629737496.74272264b15c.932.176 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629738136.74272264b15c.932.178 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629738769.74272264b15c.932.180 +3 -0
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/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:ba72d974503ea478c5b1729780f9276d10be80ca6fa1c6cd08db6a45a220a068
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/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 sha256:92cb28cec2c0eee45942a06d4108f95570fa53db5e6fd7d5a34fde871d8bdfec
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/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 sha256:7cf89f648b1dff6e8fb245e4c3a61d5629be972463315d60dd829635437bcb14
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/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 sha256:9d3cadb12dc304d2d13a4672789668d30e7ae619e0398705d5c7682eaaebbdf8
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/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:f7969b68e4aa6b630303de92298d5193fab81b96e65d08c15e07bcc1987850e1
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -161589,11 +161589,806 @@
|
|
| 161589 |
"eval_steps_per_second": 0.697,
|
| 161590 |
"eval_wer": 0.2061362152752593,
|
| 161591 |
"step": 48909
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 161592 |
}
|
| 161593 |
],
|
| 161594 |
"max_steps": 620000,
|
| 161595 |
"num_train_epochs": 5000,
|
| 161596 |
-
"total_flos": 1.
|
| 161597 |
"trial_name": null,
|
| 161598 |
"trial_params": null
|
| 161599 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 398.99598393574297,
|
| 5 |
+
"global_step": 49530,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 161589 |
"eval_steps_per_second": 0.697,
|
| 161590 |
"eval_wer": 0.2061362152752593,
|
| 161591 |
"step": 48909
|
| 161592 |
+
},
|
| 161593 |
+
{
|
| 161594 |
+
"epoch": 394.01,
|
| 161595 |
+
"learning_rate": 9.2325e-06,
|
| 161596 |
+
"loss": 0.3377,
|
| 161597 |
+
"step": 48910
|
| 161598 |
+
},
|
| 161599 |
+
{
|
| 161600 |
+
"epoch": 394.05,
|
| 161601 |
+
"learning_rate": 9.232419871794871e-06,
|
| 161602 |
+
"loss": 0.3536,
|
| 161603 |
+
"step": 48915
|
| 161604 |
+
},
|
| 161605 |
+
{
|
| 161606 |
+
"epoch": 394.09,
|
| 161607 |
+
"learning_rate": 9.232339743589745e-06,
|
| 161608 |
+
"loss": 0.3795,
|
| 161609 |
+
"step": 48920
|
| 161610 |
+
},
|
| 161611 |
+
{
|
| 161612 |
+
"epoch": 394.13,
|
| 161613 |
+
"learning_rate": 9.232259615384616e-06,
|
| 161614 |
+
"loss": 0.3885,
|
| 161615 |
+
"step": 48925
|
| 161616 |
+
},
|
| 161617 |
+
{
|
| 161618 |
+
"epoch": 394.17,
|
| 161619 |
+
"learning_rate": 9.232179487179487e-06,
|
| 161620 |
+
"loss": 0.5964,
|
| 161621 |
+
"step": 48930
|
| 161622 |
+
},
|
| 161623 |
+
{
|
| 161624 |
+
"epoch": 394.21,
|
| 161625 |
+
"learning_rate": 9.23209935897436e-06,
|
| 161626 |
+
"loss": 1.2616,
|
| 161627 |
+
"step": 48935
|
| 161628 |
+
},
|
| 161629 |
+
{
|
| 161630 |
+
"epoch": 394.25,
|
| 161631 |
+
"learning_rate": 9.232019230769232e-06,
|
| 161632 |
+
"loss": 0.3901,
|
| 161633 |
+
"step": 48940
|
| 161634 |
+
},
|
| 161635 |
+
{
|
| 161636 |
+
"epoch": 394.29,
|
| 161637 |
+
"learning_rate": 9.231939102564103e-06,
|
| 161638 |
+
"loss": 0.3503,
|
| 161639 |
+
"step": 48945
|
| 161640 |
+
},
|
| 161641 |
+
{
|
| 161642 |
+
"epoch": 394.33,
|
| 161643 |
+
"learning_rate": 9.231858974358974e-06,
|
| 161644 |
+
"loss": 0.4277,
|
| 161645 |
+
"step": 48950
|
| 161646 |
+
},
|
| 161647 |
+
{
|
| 161648 |
+
"epoch": 394.37,
|
| 161649 |
+
"learning_rate": 9.231778846153847e-06,
|
| 161650 |
+
"loss": 0.5298,
|
| 161651 |
+
"step": 48955
|
| 161652 |
+
},
|
| 161653 |
+
{
|
| 161654 |
+
"epoch": 394.41,
|
| 161655 |
+
"learning_rate": 9.231698717948719e-06,
|
| 161656 |
+
"loss": 1.1753,
|
| 161657 |
+
"step": 48960
|
| 161658 |
+
},
|
| 161659 |
+
{
|
| 161660 |
+
"epoch": 394.45,
|
| 161661 |
+
"learning_rate": 9.23161858974359e-06,
|
| 161662 |
+
"loss": 0.4082,
|
| 161663 |
+
"step": 48965
|
| 161664 |
+
},
|
| 161665 |
+
{
|
| 161666 |
+
"epoch": 394.49,
|
| 161667 |
+
"learning_rate": 9.231538461538462e-06,
|
| 161668 |
+
"loss": 0.2778,
|
| 161669 |
+
"step": 48970
|
| 161670 |
+
},
|
| 161671 |
+
{
|
| 161672 |
+
"epoch": 394.53,
|
| 161673 |
+
"learning_rate": 9.231458333333335e-06,
|
| 161674 |
+
"loss": 0.399,
|
| 161675 |
+
"step": 48975
|
| 161676 |
+
},
|
| 161677 |
+
{
|
| 161678 |
+
"epoch": 394.57,
|
| 161679 |
+
"learning_rate": 9.231378205128206e-06,
|
| 161680 |
+
"loss": 0.8602,
|
| 161681 |
+
"step": 48980
|
| 161682 |
+
},
|
| 161683 |
+
{
|
| 161684 |
+
"epoch": 394.61,
|
| 161685 |
+
"learning_rate": 9.231298076923077e-06,
|
| 161686 |
+
"loss": 1.3946,
|
| 161687 |
+
"step": 48985
|
| 161688 |
+
},
|
| 161689 |
+
{
|
| 161690 |
+
"epoch": 394.65,
|
| 161691 |
+
"learning_rate": 9.23121794871795e-06,
|
| 161692 |
+
"loss": 0.3442,
|
| 161693 |
+
"step": 48990
|
| 161694 |
+
},
|
| 161695 |
+
{
|
| 161696 |
+
"epoch": 394.69,
|
| 161697 |
+
"learning_rate": 9.231137820512822e-06,
|
| 161698 |
+
"loss": 0.3421,
|
| 161699 |
+
"step": 48995
|
| 161700 |
+
},
|
| 161701 |
+
{
|
| 161702 |
+
"epoch": 394.73,
|
| 161703 |
+
"learning_rate": 9.231057692307693e-06,
|
| 161704 |
+
"loss": 0.4279,
|
| 161705 |
+
"step": 49000
|
| 161706 |
+
},
|
| 161707 |
+
{
|
| 161708 |
+
"epoch": 394.77,
|
| 161709 |
+
"learning_rate": 9.230977564102564e-06,
|
| 161710 |
+
"loss": 0.5507,
|
| 161711 |
+
"step": 49005
|
| 161712 |
+
},
|
| 161713 |
+
{
|
| 161714 |
+
"epoch": 394.81,
|
| 161715 |
+
"learning_rate": 9.230897435897437e-06,
|
| 161716 |
+
"loss": 1.263,
|
| 161717 |
+
"step": 49010
|
| 161718 |
+
},
|
| 161719 |
+
{
|
| 161720 |
+
"epoch": 394.85,
|
| 161721 |
+
"learning_rate": 9.230817307692309e-06,
|
| 161722 |
+
"loss": 0.3819,
|
| 161723 |
+
"step": 49015
|
| 161724 |
+
},
|
| 161725 |
+
{
|
| 161726 |
+
"epoch": 394.89,
|
| 161727 |
+
"learning_rate": 9.23073717948718e-06,
|
| 161728 |
+
"loss": 0.2942,
|
| 161729 |
+
"step": 49020
|
| 161730 |
+
},
|
| 161731 |
+
{
|
| 161732 |
+
"epoch": 394.93,
|
| 161733 |
+
"learning_rate": 9.230657051282053e-06,
|
| 161734 |
+
"loss": 0.4264,
|
| 161735 |
+
"step": 49025
|
| 161736 |
+
},
|
| 161737 |
+
{
|
| 161738 |
+
"epoch": 394.97,
|
| 161739 |
+
"learning_rate": 9.230576923076925e-06,
|
| 161740 |
+
"loss": 0.7415,
|
| 161741 |
+
"step": 49030
|
| 161742 |
+
},
|
| 161743 |
+
{
|
| 161744 |
+
"epoch": 395.0,
|
| 161745 |
+
"eval_loss": 0.3846549093723297,
|
| 161746 |
+
"eval_runtime": 41.7511,
|
| 161747 |
+
"eval_samples_per_second": 20.095,
|
| 161748 |
+
"eval_steps_per_second": 0.647,
|
| 161749 |
+
"eval_wer": 0.20638499018253217,
|
| 161750 |
+
"step": 49033
|
| 161751 |
+
},
|
| 161752 |
+
{
|
| 161753 |
+
"epoch": 395.02,
|
| 161754 |
+
"learning_rate": 9.230496794871796e-06,
|
| 161755 |
+
"loss": 0.3296,
|
| 161756 |
+
"step": 49035
|
| 161757 |
+
},
|
| 161758 |
+
{
|
| 161759 |
+
"epoch": 395.06,
|
| 161760 |
+
"learning_rate": 9.230416666666667e-06,
|
| 161761 |
+
"loss": 0.5311,
|
| 161762 |
+
"step": 49040
|
| 161763 |
+
},
|
| 161764 |
+
{
|
| 161765 |
+
"epoch": 395.1,
|
| 161766 |
+
"learning_rate": 9.23033653846154e-06,
|
| 161767 |
+
"loss": 0.3304,
|
| 161768 |
+
"step": 49045
|
| 161769 |
+
},
|
| 161770 |
+
{
|
| 161771 |
+
"epoch": 395.14,
|
| 161772 |
+
"learning_rate": 9.23025641025641e-06,
|
| 161773 |
+
"loss": 0.3987,
|
| 161774 |
+
"step": 49050
|
| 161775 |
+
},
|
| 161776 |
+
{
|
| 161777 |
+
"epoch": 395.18,
|
| 161778 |
+
"learning_rate": 9.230176282051283e-06,
|
| 161779 |
+
"loss": 0.6405,
|
| 161780 |
+
"step": 49055
|
| 161781 |
+
},
|
| 161782 |
+
{
|
| 161783 |
+
"epoch": 395.22,
|
| 161784 |
+
"learning_rate": 9.230096153846154e-06,
|
| 161785 |
+
"loss": 1.3818,
|
| 161786 |
+
"step": 49060
|
| 161787 |
+
},
|
| 161788 |
+
{
|
| 161789 |
+
"epoch": 395.26,
|
| 161790 |
+
"learning_rate": 9.230016025641026e-06,
|
| 161791 |
+
"loss": 0.3342,
|
| 161792 |
+
"step": 49065
|
| 161793 |
+
},
|
| 161794 |
+
{
|
| 161795 |
+
"epoch": 395.3,
|
| 161796 |
+
"learning_rate": 9.229935897435897e-06,
|
| 161797 |
+
"loss": 0.3353,
|
| 161798 |
+
"step": 49070
|
| 161799 |
+
},
|
| 161800 |
+
{
|
| 161801 |
+
"epoch": 395.34,
|
| 161802 |
+
"learning_rate": 9.22985576923077e-06,
|
| 161803 |
+
"loss": 0.3485,
|
| 161804 |
+
"step": 49075
|
| 161805 |
+
},
|
| 161806 |
+
{
|
| 161807 |
+
"epoch": 395.38,
|
| 161808 |
+
"learning_rate": 9.229775641025642e-06,
|
| 161809 |
+
"loss": 0.6765,
|
| 161810 |
+
"step": 49080
|
| 161811 |
+
},
|
| 161812 |
+
{
|
| 161813 |
+
"epoch": 395.42,
|
| 161814 |
+
"learning_rate": 9.229695512820513e-06,
|
| 161815 |
+
"loss": 1.1464,
|
| 161816 |
+
"step": 49085
|
| 161817 |
+
},
|
| 161818 |
+
{
|
| 161819 |
+
"epoch": 395.46,
|
| 161820 |
+
"learning_rate": 9.229615384615386e-06,
|
| 161821 |
+
"loss": 0.3754,
|
| 161822 |
+
"step": 49090
|
| 161823 |
+
},
|
| 161824 |
+
{
|
| 161825 |
+
"epoch": 395.5,
|
| 161826 |
+
"learning_rate": 9.229535256410257e-06,
|
| 161827 |
+
"loss": 0.3995,
|
| 161828 |
+
"step": 49095
|
| 161829 |
+
},
|
| 161830 |
+
{
|
| 161831 |
+
"epoch": 395.54,
|
| 161832 |
+
"learning_rate": 9.229455128205129e-06,
|
| 161833 |
+
"loss": 0.4875,
|
| 161834 |
+
"step": 49100
|
| 161835 |
+
},
|
| 161836 |
+
{
|
| 161837 |
+
"epoch": 395.58,
|
| 161838 |
+
"learning_rate": 9.229375e-06,
|
| 161839 |
+
"loss": 0.7236,
|
| 161840 |
+
"step": 49105
|
| 161841 |
+
},
|
| 161842 |
+
{
|
| 161843 |
+
"epoch": 395.62,
|
| 161844 |
+
"learning_rate": 9.229294871794873e-06,
|
| 161845 |
+
"loss": 1.1748,
|
| 161846 |
+
"step": 49110
|
| 161847 |
+
},
|
| 161848 |
+
{
|
| 161849 |
+
"epoch": 395.66,
|
| 161850 |
+
"learning_rate": 9.229214743589744e-06,
|
| 161851 |
+
"loss": 0.3442,
|
| 161852 |
+
"step": 49115
|
| 161853 |
+
},
|
| 161854 |
+
{
|
| 161855 |
+
"epoch": 395.7,
|
| 161856 |
+
"learning_rate": 9.229134615384616e-06,
|
| 161857 |
+
"loss": 0.3461,
|
| 161858 |
+
"step": 49120
|
| 161859 |
+
},
|
| 161860 |
+
{
|
| 161861 |
+
"epoch": 395.74,
|
| 161862 |
+
"learning_rate": 9.229054487179489e-06,
|
| 161863 |
+
"loss": 0.4033,
|
| 161864 |
+
"step": 49125
|
| 161865 |
+
},
|
| 161866 |
+
{
|
| 161867 |
+
"epoch": 395.78,
|
| 161868 |
+
"learning_rate": 9.22897435897436e-06,
|
| 161869 |
+
"loss": 0.7508,
|
| 161870 |
+
"step": 49130
|
| 161871 |
+
},
|
| 161872 |
+
{
|
| 161873 |
+
"epoch": 395.82,
|
| 161874 |
+
"learning_rate": 9.228894230769232e-06,
|
| 161875 |
+
"loss": 1.134,
|
| 161876 |
+
"step": 49135
|
| 161877 |
+
},
|
| 161878 |
+
{
|
| 161879 |
+
"epoch": 395.86,
|
| 161880 |
+
"learning_rate": 9.228814102564103e-06,
|
| 161881 |
+
"loss": 0.3769,
|
| 161882 |
+
"step": 49140
|
| 161883 |
+
},
|
| 161884 |
+
{
|
| 161885 |
+
"epoch": 395.9,
|
| 161886 |
+
"learning_rate": 9.228733974358976e-06,
|
| 161887 |
+
"loss": 0.6254,
|
| 161888 |
+
"step": 49145
|
| 161889 |
+
},
|
| 161890 |
+
{
|
| 161891 |
+
"epoch": 395.94,
|
| 161892 |
+
"learning_rate": 9.228653846153847e-06,
|
| 161893 |
+
"loss": 0.4059,
|
| 161894 |
+
"step": 49150
|
| 161895 |
+
},
|
| 161896 |
+
{
|
| 161897 |
+
"epoch": 395.98,
|
| 161898 |
+
"learning_rate": 9.228573717948719e-06,
|
| 161899 |
+
"loss": 1.0096,
|
| 161900 |
+
"step": 49155
|
| 161901 |
+
},
|
| 161902 |
+
{
|
| 161903 |
+
"epoch": 396.0,
|
| 161904 |
+
"eval_loss": 0.443951278924942,
|
| 161905 |
+
"eval_runtime": 39.9548,
|
| 161906 |
+
"eval_samples_per_second": 21.024,
|
| 161907 |
+
"eval_steps_per_second": 0.676,
|
| 161908 |
+
"eval_wer": 0.19998531679025036,
|
| 161909 |
+
"step": 49157
|
| 161910 |
+
},
|
| 161911 |
+
{
|
| 161912 |
+
"epoch": 393.02,
|
| 161913 |
+
"learning_rate": 9.22849358974359e-06,
|
| 161914 |
+
"loss": 0.3615,
|
| 161915 |
+
"step": 49160
|
| 161916 |
+
},
|
| 161917 |
+
{
|
| 161918 |
+
"epoch": 393.06,
|
| 161919 |
+
"learning_rate": 9.228413461538463e-06,
|
| 161920 |
+
"loss": 0.3503,
|
| 161921 |
+
"step": 49165
|
| 161922 |
+
},
|
| 161923 |
+
{
|
| 161924 |
+
"epoch": 393.1,
|
| 161925 |
+
"learning_rate": 9.228333333333333e-06,
|
| 161926 |
+
"loss": 0.385,
|
| 161927 |
+
"step": 49170
|
| 161928 |
+
},
|
| 161929 |
+
{
|
| 161930 |
+
"epoch": 393.14,
|
| 161931 |
+
"learning_rate": 9.228253205128206e-06,
|
| 161932 |
+
"loss": 0.44,
|
| 161933 |
+
"step": 49175
|
| 161934 |
+
},
|
| 161935 |
+
{
|
| 161936 |
+
"epoch": 393.18,
|
| 161937 |
+
"learning_rate": 9.228173076923079e-06,
|
| 161938 |
+
"loss": 0.8611,
|
| 161939 |
+
"step": 49180
|
| 161940 |
+
},
|
| 161941 |
+
{
|
| 161942 |
+
"epoch": 393.22,
|
| 161943 |
+
"learning_rate": 9.228092948717949e-06,
|
| 161944 |
+
"loss": 1.0425,
|
| 161945 |
+
"step": 49185
|
| 161946 |
+
},
|
| 161947 |
+
{
|
| 161948 |
+
"epoch": 393.26,
|
| 161949 |
+
"learning_rate": 9.228012820512822e-06,
|
| 161950 |
+
"loss": 0.3258,
|
| 161951 |
+
"step": 49190
|
| 161952 |
+
},
|
| 161953 |
+
{
|
| 161954 |
+
"epoch": 393.3,
|
| 161955 |
+
"learning_rate": 9.227932692307693e-06,
|
| 161956 |
+
"loss": 0.3474,
|
| 161957 |
+
"step": 49195
|
| 161958 |
+
},
|
| 161959 |
+
{
|
| 161960 |
+
"epoch": 393.34,
|
| 161961 |
+
"learning_rate": 9.227852564102564e-06,
|
| 161962 |
+
"loss": 0.3853,
|
| 161963 |
+
"step": 49200
|
| 161964 |
+
},
|
| 161965 |
+
{
|
| 161966 |
+
"epoch": 393.38,
|
| 161967 |
+
"learning_rate": 9.227772435897436e-06,
|
| 161968 |
+
"loss": 0.7701,
|
| 161969 |
+
"step": 49205
|
| 161970 |
+
},
|
| 161971 |
+
{
|
| 161972 |
+
"epoch": 393.42,
|
| 161973 |
+
"learning_rate": 9.227692307692309e-06,
|
| 161974 |
+
"loss": 1.0871,
|
| 161975 |
+
"step": 49210
|
| 161976 |
+
},
|
| 161977 |
+
{
|
| 161978 |
+
"epoch": 393.46,
|
| 161979 |
+
"learning_rate": 9.22761217948718e-06,
|
| 161980 |
+
"loss": 0.346,
|
| 161981 |
+
"step": 49215
|
| 161982 |
+
},
|
| 161983 |
+
{
|
| 161984 |
+
"epoch": 393.5,
|
| 161985 |
+
"learning_rate": 9.227532051282051e-06,
|
| 161986 |
+
"loss": 0.3688,
|
| 161987 |
+
"step": 49220
|
| 161988 |
+
},
|
| 161989 |
+
{
|
| 161990 |
+
"epoch": 393.54,
|
| 161991 |
+
"learning_rate": 9.227451923076923e-06,
|
| 161992 |
+
"loss": 0.4968,
|
| 161993 |
+
"step": 49225
|
| 161994 |
+
},
|
| 161995 |
+
{
|
| 161996 |
+
"epoch": 393.58,
|
| 161997 |
+
"learning_rate": 9.227371794871796e-06,
|
| 161998 |
+
"loss": 0.8089,
|
| 161999 |
+
"step": 49230
|
| 162000 |
+
},
|
| 162001 |
+
{
|
| 162002 |
+
"epoch": 393.62,
|
| 162003 |
+
"learning_rate": 9.227291666666667e-06,
|
| 162004 |
+
"loss": 0.8807,
|
| 162005 |
+
"step": 49235
|
| 162006 |
+
},
|
| 162007 |
+
{
|
| 162008 |
+
"epoch": 393.66,
|
| 162009 |
+
"learning_rate": 9.227211538461539e-06,
|
| 162010 |
+
"loss": 0.3152,
|
| 162011 |
+
"step": 49240
|
| 162012 |
+
},
|
| 162013 |
+
{
|
| 162014 |
+
"epoch": 393.7,
|
| 162015 |
+
"learning_rate": 9.227131410256412e-06,
|
| 162016 |
+
"loss": 0.3571,
|
| 162017 |
+
"step": 49245
|
| 162018 |
+
},
|
| 162019 |
+
{
|
| 162020 |
+
"epoch": 393.74,
|
| 162021 |
+
"learning_rate": 9.227051282051283e-06,
|
| 162022 |
+
"loss": 0.6453,
|
| 162023 |
+
"step": 49250
|
| 162024 |
+
},
|
| 162025 |
+
{
|
| 162026 |
+
"epoch": 393.78,
|
| 162027 |
+
"learning_rate": 9.226971153846154e-06,
|
| 162028 |
+
"loss": 0.8375,
|
| 162029 |
+
"step": 49255
|
| 162030 |
+
},
|
| 162031 |
+
{
|
| 162032 |
+
"epoch": 393.82,
|
| 162033 |
+
"learning_rate": 9.226891025641026e-06,
|
| 162034 |
+
"loss": 0.9072,
|
| 162035 |
+
"step": 49260
|
| 162036 |
+
},
|
| 162037 |
+
{
|
| 162038 |
+
"epoch": 393.86,
|
| 162039 |
+
"learning_rate": 9.226810897435899e-06,
|
| 162040 |
+
"loss": 0.3948,
|
| 162041 |
+
"step": 49265
|
| 162042 |
+
},
|
| 162043 |
+
{
|
| 162044 |
+
"epoch": 393.9,
|
| 162045 |
+
"learning_rate": 9.22673076923077e-06,
|
| 162046 |
+
"loss": 0.4225,
|
| 162047 |
+
"step": 49270
|
| 162048 |
+
},
|
| 162049 |
+
{
|
| 162050 |
+
"epoch": 393.94,
|
| 162051 |
+
"learning_rate": 9.226650641025641e-06,
|
| 162052 |
+
"loss": 0.4801,
|
| 162053 |
+
"step": 49275
|
| 162054 |
+
},
|
| 162055 |
+
{
|
| 162056 |
+
"epoch": 393.98,
|
| 162057 |
+
"learning_rate": 9.226570512820515e-06,
|
| 162058 |
+
"loss": 0.7899,
|
| 162059 |
+
"step": 49280
|
| 162060 |
+
},
|
| 162061 |
+
{
|
| 162062 |
+
"epoch": 394.0,
|
| 162063 |
+
"eval_loss": 0.4228745996952057,
|
| 162064 |
+
"eval_runtime": 39.0975,
|
| 162065 |
+
"eval_samples_per_second": 21.485,
|
| 162066 |
+
"eval_steps_per_second": 0.691,
|
| 162067 |
+
"eval_wer": 0.20292917516758963,
|
| 162068 |
+
"step": 49282
|
| 162069 |
+
},
|
| 162070 |
+
{
|
| 162071 |
+
"epoch": 397.02,
|
| 162072 |
+
"learning_rate": 9.226490384615386e-06,
|
| 162073 |
+
"loss": 0.4219,
|
| 162074 |
+
"step": 49285
|
| 162075 |
+
},
|
| 162076 |
+
{
|
| 162077 |
+
"epoch": 397.06,
|
| 162078 |
+
"learning_rate": 9.226410256410257e-06,
|
| 162079 |
+
"loss": 0.3077,
|
| 162080 |
+
"step": 49290
|
| 162081 |
+
},
|
| 162082 |
+
{
|
| 162083 |
+
"epoch": 397.1,
|
| 162084 |
+
"learning_rate": 9.226330128205129e-06,
|
| 162085 |
+
"loss": 0.2918,
|
| 162086 |
+
"step": 49295
|
| 162087 |
+
},
|
| 162088 |
+
{
|
| 162089 |
+
"epoch": 397.14,
|
| 162090 |
+
"learning_rate": 9.226250000000002e-06,
|
| 162091 |
+
"loss": 0.4312,
|
| 162092 |
+
"step": 49300
|
| 162093 |
+
},
|
| 162094 |
+
{
|
| 162095 |
+
"epoch": 397.18,
|
| 162096 |
+
"learning_rate": 9.226169871794873e-06,
|
| 162097 |
+
"loss": 0.7957,
|
| 162098 |
+
"step": 49305
|
| 162099 |
+
},
|
| 162100 |
+
{
|
| 162101 |
+
"epoch": 397.22,
|
| 162102 |
+
"learning_rate": 9.226089743589744e-06,
|
| 162103 |
+
"loss": 0.9193,
|
| 162104 |
+
"step": 49310
|
| 162105 |
+
},
|
| 162106 |
+
{
|
| 162107 |
+
"epoch": 397.27,
|
| 162108 |
+
"learning_rate": 9.226009615384616e-06,
|
| 162109 |
+
"loss": 0.3935,
|
| 162110 |
+
"step": 49315
|
| 162111 |
+
},
|
| 162112 |
+
{
|
| 162113 |
+
"epoch": 397.31,
|
| 162114 |
+
"learning_rate": 9.225929487179489e-06,
|
| 162115 |
+
"loss": 0.2814,
|
| 162116 |
+
"step": 49320
|
| 162117 |
+
},
|
| 162118 |
+
{
|
| 162119 |
+
"epoch": 397.35,
|
| 162120 |
+
"learning_rate": 9.225849358974358e-06,
|
| 162121 |
+
"loss": 0.4674,
|
| 162122 |
+
"step": 49325
|
| 162123 |
+
},
|
| 162124 |
+
{
|
| 162125 |
+
"epoch": 397.39,
|
| 162126 |
+
"learning_rate": 9.225769230769232e-06,
|
| 162127 |
+
"loss": 0.8111,
|
| 162128 |
+
"step": 49330
|
| 162129 |
+
},
|
| 162130 |
+
{
|
| 162131 |
+
"epoch": 397.43,
|
| 162132 |
+
"learning_rate": 9.225689102564105e-06,
|
| 162133 |
+
"loss": 0.8121,
|
| 162134 |
+
"step": 49335
|
| 162135 |
+
},
|
| 162136 |
+
{
|
| 162137 |
+
"epoch": 397.47,
|
| 162138 |
+
"learning_rate": 9.225608974358974e-06,
|
| 162139 |
+
"loss": 0.3561,
|
| 162140 |
+
"step": 49340
|
| 162141 |
+
},
|
| 162142 |
+
{
|
| 162143 |
+
"epoch": 397.51,
|
| 162144 |
+
"learning_rate": 9.225528846153847e-06,
|
| 162145 |
+
"loss": 0.3344,
|
| 162146 |
+
"step": 49345
|
| 162147 |
+
},
|
| 162148 |
+
{
|
| 162149 |
+
"epoch": 397.55,
|
| 162150 |
+
"learning_rate": 9.225448717948719e-06,
|
| 162151 |
+
"loss": 0.4519,
|
| 162152 |
+
"step": 49350
|
| 162153 |
+
},
|
| 162154 |
+
{
|
| 162155 |
+
"epoch": 397.59,
|
| 162156 |
+
"learning_rate": 9.22536858974359e-06,
|
| 162157 |
+
"loss": 0.7212,
|
| 162158 |
+
"step": 49355
|
| 162159 |
+
},
|
| 162160 |
+
{
|
| 162161 |
+
"epoch": 397.63,
|
| 162162 |
+
"learning_rate": 9.225288461538461e-06,
|
| 162163 |
+
"loss": 0.9421,
|
| 162164 |
+
"step": 49360
|
| 162165 |
+
},
|
| 162166 |
+
{
|
| 162167 |
+
"epoch": 397.67,
|
| 162168 |
+
"learning_rate": 9.225208333333334e-06,
|
| 162169 |
+
"loss": 0.3276,
|
| 162170 |
+
"step": 49365
|
| 162171 |
+
},
|
| 162172 |
+
{
|
| 162173 |
+
"epoch": 397.71,
|
| 162174 |
+
"learning_rate": 9.225128205128206e-06,
|
| 162175 |
+
"loss": 0.3171,
|
| 162176 |
+
"step": 49370
|
| 162177 |
+
},
|
| 162178 |
+
{
|
| 162179 |
+
"epoch": 397.75,
|
| 162180 |
+
"learning_rate": 9.225048076923077e-06,
|
| 162181 |
+
"loss": 0.4455,
|
| 162182 |
+
"step": 49375
|
| 162183 |
+
},
|
| 162184 |
+
{
|
| 162185 |
+
"epoch": 397.79,
|
| 162186 |
+
"learning_rate": 9.22496794871795e-06,
|
| 162187 |
+
"loss": 0.7965,
|
| 162188 |
+
"step": 49380
|
| 162189 |
+
},
|
| 162190 |
+
{
|
| 162191 |
+
"epoch": 397.83,
|
| 162192 |
+
"learning_rate": 9.224887820512822e-06,
|
| 162193 |
+
"loss": 1.0549,
|
| 162194 |
+
"step": 49385
|
| 162195 |
+
},
|
| 162196 |
+
{
|
| 162197 |
+
"epoch": 397.87,
|
| 162198 |
+
"learning_rate": 9.224807692307693e-06,
|
| 162199 |
+
"loss": 0.4175,
|
| 162200 |
+
"step": 49390
|
| 162201 |
+
},
|
| 162202 |
+
{
|
| 162203 |
+
"epoch": 397.91,
|
| 162204 |
+
"learning_rate": 9.224727564102564e-06,
|
| 162205 |
+
"loss": 0.3971,
|
| 162206 |
+
"step": 49395
|
| 162207 |
+
},
|
| 162208 |
+
{
|
| 162209 |
+
"epoch": 397.95,
|
| 162210 |
+
"learning_rate": 9.224647435897437e-06,
|
| 162211 |
+
"loss": 0.424,
|
| 162212 |
+
"step": 49400
|
| 162213 |
+
},
|
| 162214 |
+
{
|
| 162215 |
+
"epoch": 397.99,
|
| 162216 |
+
"learning_rate": 9.224567307692309e-06,
|
| 162217 |
+
"loss": 0.9309,
|
| 162218 |
+
"step": 49405
|
| 162219 |
+
},
|
| 162220 |
+
{
|
| 162221 |
+
"epoch": 398.0,
|
| 162222 |
+
"eval_loss": 0.3627434968948364,
|
| 162223 |
+
"eval_runtime": 39.3755,
|
| 162224 |
+
"eval_samples_per_second": 21.333,
|
| 162225 |
+
"eval_steps_per_second": 0.686,
|
| 162226 |
+
"eval_wer": 0.2020069808027923,
|
| 162227 |
+
"step": 49406
|
| 162228 |
+
},
|
| 162229 |
+
{
|
| 162230 |
+
"epoch": 398.03,
|
| 162231 |
+
"learning_rate": 9.22448717948718e-06,
|
| 162232 |
+
"loss": 0.3855,
|
| 162233 |
+
"step": 49410
|
| 162234 |
+
},
|
| 162235 |
+
{
|
| 162236 |
+
"epoch": 398.07,
|
| 162237 |
+
"learning_rate": 9.224407051282051e-06,
|
| 162238 |
+
"loss": 0.3263,
|
| 162239 |
+
"step": 49415
|
| 162240 |
+
},
|
| 162241 |
+
{
|
| 162242 |
+
"epoch": 398.11,
|
| 162243 |
+
"learning_rate": 9.224326923076924e-06,
|
| 162244 |
+
"loss": 0.3117,
|
| 162245 |
+
"step": 49420
|
| 162246 |
+
},
|
| 162247 |
+
{
|
| 162248 |
+
"epoch": 398.15,
|
| 162249 |
+
"learning_rate": 9.224246794871796e-06,
|
| 162250 |
+
"loss": 0.4375,
|
| 162251 |
+
"step": 49425
|
| 162252 |
+
},
|
| 162253 |
+
{
|
| 162254 |
+
"epoch": 398.19,
|
| 162255 |
+
"learning_rate": 9.224166666666667e-06,
|
| 162256 |
+
"loss": 0.9174,
|
| 162257 |
+
"step": 49430
|
| 162258 |
+
},
|
| 162259 |
+
{
|
| 162260 |
+
"epoch": 398.23,
|
| 162261 |
+
"learning_rate": 9.22408653846154e-06,
|
| 162262 |
+
"loss": 0.6554,
|
| 162263 |
+
"step": 49435
|
| 162264 |
+
},
|
| 162265 |
+
{
|
| 162266 |
+
"epoch": 398.27,
|
| 162267 |
+
"learning_rate": 9.224006410256412e-06,
|
| 162268 |
+
"loss": 0.3556,
|
| 162269 |
+
"step": 49440
|
| 162270 |
+
},
|
| 162271 |
+
{
|
| 162272 |
+
"epoch": 398.31,
|
| 162273 |
+
"learning_rate": 9.223926282051283e-06,
|
| 162274 |
+
"loss": 0.35,
|
| 162275 |
+
"step": 49445
|
| 162276 |
+
},
|
| 162277 |
+
{
|
| 162278 |
+
"epoch": 398.35,
|
| 162279 |
+
"learning_rate": 9.223846153846154e-06,
|
| 162280 |
+
"loss": 0.4128,
|
| 162281 |
+
"step": 49450
|
| 162282 |
+
},
|
| 162283 |
+
{
|
| 162284 |
+
"epoch": 398.39,
|
| 162285 |
+
"learning_rate": 9.223766025641027e-06,
|
| 162286 |
+
"loss": 0.9116,
|
| 162287 |
+
"step": 49455
|
| 162288 |
+
},
|
| 162289 |
+
{
|
| 162290 |
+
"epoch": 398.43,
|
| 162291 |
+
"learning_rate": 9.223685897435897e-06,
|
| 162292 |
+
"loss": 0.6949,
|
| 162293 |
+
"step": 49460
|
| 162294 |
+
},
|
| 162295 |
+
{
|
| 162296 |
+
"epoch": 398.47,
|
| 162297 |
+
"learning_rate": 9.22360576923077e-06,
|
| 162298 |
+
"loss": 0.3063,
|
| 162299 |
+
"step": 49465
|
| 162300 |
+
},
|
| 162301 |
+
{
|
| 162302 |
+
"epoch": 398.51,
|
| 162303 |
+
"learning_rate": 9.223525641025641e-06,
|
| 162304 |
+
"loss": 0.3731,
|
| 162305 |
+
"step": 49470
|
| 162306 |
+
},
|
| 162307 |
+
{
|
| 162308 |
+
"epoch": 398.55,
|
| 162309 |
+
"learning_rate": 9.223445512820513e-06,
|
| 162310 |
+
"loss": 0.4437,
|
| 162311 |
+
"step": 49475
|
| 162312 |
+
},
|
| 162313 |
+
{
|
| 162314 |
+
"epoch": 398.59,
|
| 162315 |
+
"learning_rate": 9.223365384615386e-06,
|
| 162316 |
+
"loss": 0.9937,
|
| 162317 |
+
"step": 49480
|
| 162318 |
+
},
|
| 162319 |
+
{
|
| 162320 |
+
"epoch": 398.63,
|
| 162321 |
+
"learning_rate": 9.223285256410257e-06,
|
| 162322 |
+
"loss": 0.7464,
|
| 162323 |
+
"step": 49485
|
| 162324 |
+
},
|
| 162325 |
+
{
|
| 162326 |
+
"epoch": 398.67,
|
| 162327 |
+
"learning_rate": 9.223205128205129e-06,
|
| 162328 |
+
"loss": 0.3578,
|
| 162329 |
+
"step": 49490
|
| 162330 |
+
},
|
| 162331 |
+
{
|
| 162332 |
+
"epoch": 398.71,
|
| 162333 |
+
"learning_rate": 9.223125e-06,
|
| 162334 |
+
"loss": 0.3104,
|
| 162335 |
+
"step": 49495
|
| 162336 |
+
},
|
| 162337 |
+
{
|
| 162338 |
+
"epoch": 398.76,
|
| 162339 |
+
"learning_rate": 9.223044871794873e-06,
|
| 162340 |
+
"loss": 0.5128,
|
| 162341 |
+
"step": 49500
|
| 162342 |
+
},
|
| 162343 |
+
{
|
| 162344 |
+
"epoch": 398.8,
|
| 162345 |
+
"learning_rate": 9.222964743589744e-06,
|
| 162346 |
+
"loss": 1.0195,
|
| 162347 |
+
"step": 49505
|
| 162348 |
+
},
|
| 162349 |
+
{
|
| 162350 |
+
"epoch": 398.84,
|
| 162351 |
+
"learning_rate": 9.222884615384616e-06,
|
| 162352 |
+
"loss": 0.7428,
|
| 162353 |
+
"step": 49510
|
| 162354 |
+
},
|
| 162355 |
+
{
|
| 162356 |
+
"epoch": 398.88,
|
| 162357 |
+
"learning_rate": 9.222804487179487e-06,
|
| 162358 |
+
"loss": 0.3895,
|
| 162359 |
+
"step": 49515
|
| 162360 |
+
},
|
| 162361 |
+
{
|
| 162362 |
+
"epoch": 398.92,
|
| 162363 |
+
"learning_rate": 9.22272435897436e-06,
|
| 162364 |
+
"loss": 0.6573,
|
| 162365 |
+
"step": 49520
|
| 162366 |
+
},
|
| 162367 |
+
{
|
| 162368 |
+
"epoch": 398.96,
|
| 162369 |
+
"learning_rate": 9.222644230769231e-06,
|
| 162370 |
+
"loss": 0.4496,
|
| 162371 |
+
"step": 49525
|
| 162372 |
+
},
|
| 162373 |
+
{
|
| 162374 |
+
"epoch": 399.0,
|
| 162375 |
+
"learning_rate": 9.222564102564103e-06,
|
| 162376 |
+
"loss": 1.2678,
|
| 162377 |
+
"step": 49530
|
| 162378 |
+
},
|
| 162379 |
+
{
|
| 162380 |
+
"epoch": 399.0,
|
| 162381 |
+
"eval_loss": 0.4208246171474457,
|
| 162382 |
+
"eval_runtime": 39.1767,
|
| 162383 |
+
"eval_samples_per_second": 21.441,
|
| 162384 |
+
"eval_steps_per_second": 0.689,
|
| 162385 |
+
"eval_wer": 0.19971912188631827,
|
| 162386 |
+
"step": 49530
|
| 162387 |
}
|
| 162388 |
],
|
| 162389 |
"max_steps": 620000,
|
| 162390 |
"num_train_epochs": 5000,
|
| 162391 |
+
"total_flos": 1.3936683466505619e+20,
|
| 162392 |
"trial_name": null,
|
| 162393 |
"trial_params": null
|
| 162394 |
}
|
model-bin/finetune/base/{checkpoint-48909 → checkpoint-49530}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629736186.820362/events.out.tfevents.1629736186.74272264b15c.932.173
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:69996fa3005ee33833023d6d349b824f6cf0ceca6962d31bd93df2455f14b660
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629736861.6649554/events.out.tfevents.1629736861.74272264b15c.932.175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b9029eaadaa44370b4529f6c493f046ab63f12920afa1024664af506ced80240
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629737496.0693905/events.out.tfevents.1629737496.74272264b15c.932.177
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0b8aff60b098d0973ad6a31665d69ce12800b3f2b591fa7a0c507a14be7e5436
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629738136.82416/events.out.tfevents.1629738136.74272264b15c.932.179
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2f2d0058bb5c1bc981a748415e2b7f208c78e7a3cb132b15c66003a8e527f81
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629738769.9003348/events.out.tfevents.1629738769.74272264b15c.932.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:302ab1c43d0c939aeef0625ca4d81a35e70d902c3eb295e4cb767c68fdd84f9d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629736186.74272264b15c.932.172
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:55d4a62ab62aa3a40365b9f301bbae94b8b87920144e888a95d7d290662ce0fc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629736861.74272264b15c.932.174
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:70125df5d23bcc4cebf6d48bfe3989d09b76c37a03c3b5269b4ab63d86597ca2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629737496.74272264b15c.932.176
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:154a20d955612dc4a38d2f6870dfb27cea8090acf9b5798ff209c44120f01790
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629738136.74272264b15c.932.178
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9eccc9ae3008109f905cfb3333f79ddf22846adeab84929d3ce8a84aec546c1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629738769.74272264b15c.932.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fbb140ad6164e1f6ae3c1bc9d7efe4194a090c992cb6d1d4ebbe0ff26e9e8bdc
|
| 3 |
+
size 8622
|