"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630031322.4475768/events.out.tfevents.1630031322.52f5c7e305a3.886.11 +3 -0
- model-bin/finetune/base/log/1630031785.6304755/events.out.tfevents.1630031785.52f5c7e305a3.886.13 +3 -0
- model-bin/finetune/base/log/1630032238.9460363/events.out.tfevents.1630032238.52f5c7e305a3.886.15 +3 -0
- model-bin/finetune/base/log/1630032701.7654727/events.out.tfevents.1630032701.52f5c7e305a3.886.17 +3 -0
- model-bin/finetune/base/log/1630033167.5690465/events.out.tfevents.1630033167.52f5c7e305a3.886.19 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630031322.52f5c7e305a3.886.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630031785.52f5c7e305a3.886.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630032238.52f5c7e305a3.886.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630032701.52f5c7e305a3.886.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630033167.52f5c7e305a3.886.18 +3 -0
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/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 sha256:0511e945eee2cce74e2bfc8fd97bf0523d45f090f5c07f33ee8bd60c48a5f958
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/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 sha256:93d22d7241d82d3fd12ad078edeec5ea81399ca51fda453dfa711412cdcff622
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14567
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0b0e0099cd1648151ebfcab823d77bb4b8564946ffa0efed906f9b73a025ce93
|
| 3 |
size 14567
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/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:d91b756cefbdc186ed1eb5f7ac935ad1f9a9f4aef4c7e23ad9339c02ad4894ab
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/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:d4ac6939ce2595ba56cbd3511ff5c20e93fbea3ebe253dffbcf31939e03f80ee
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -235536,11 +235536,800 @@
|
|
| 235536 |
"eval_steps_per_second": 0.635,
|
| 235537 |
"eval_wer": 0.18572358429160338,
|
| 235538 |
"step": 107027
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 235539 |
}
|
| 235540 |
],
|
| 235541 |
-
"max_steps":
|
| 235542 |
"num_train_epochs": 5000,
|
| 235543 |
-
"total_flos": 3.
|
| 235544 |
"trial_name": null,
|
| 235545 |
"trial_params": null
|
| 235546 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 867.995983935743,
|
| 5 |
+
"global_step": 107649,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 235536 |
"eval_steps_per_second": 0.635,
|
| 235537 |
"eval_wer": 0.18572358429160338,
|
| 235538 |
"step": 107027
|
| 235539 |
+
},
|
| 235540 |
+
{
|
| 235541 |
+
"epoch": 863.02,
|
| 235542 |
+
"learning_rate": 8.287770597738288e-06,
|
| 235543 |
+
"loss": 0.3168,
|
| 235544 |
+
"step": 107030
|
| 235545 |
+
},
|
| 235546 |
+
{
|
| 235547 |
+
"epoch": 863.06,
|
| 235548 |
+
"learning_rate": 8.287689822294022e-06,
|
| 235549 |
+
"loss": 0.3031,
|
| 235550 |
+
"step": 107035
|
| 235551 |
+
},
|
| 235552 |
+
{
|
| 235553 |
+
"epoch": 863.1,
|
| 235554 |
+
"learning_rate": 8.287609046849758e-06,
|
| 235555 |
+
"loss": 0.2755,
|
| 235556 |
+
"step": 107040
|
| 235557 |
+
},
|
| 235558 |
+
{
|
| 235559 |
+
"epoch": 863.14,
|
| 235560 |
+
"learning_rate": 8.287528271405494e-06,
|
| 235561 |
+
"loss": 0.3647,
|
| 235562 |
+
"step": 107045
|
| 235563 |
+
},
|
| 235564 |
+
{
|
| 235565 |
+
"epoch": 863.18,
|
| 235566 |
+
"learning_rate": 8.287447495961228e-06,
|
| 235567 |
+
"loss": 0.6837,
|
| 235568 |
+
"step": 107050
|
| 235569 |
+
},
|
| 235570 |
+
{
|
| 235571 |
+
"epoch": 863.22,
|
| 235572 |
+
"learning_rate": 8.287366720516964e-06,
|
| 235573 |
+
"loss": 0.8166,
|
| 235574 |
+
"step": 107055
|
| 235575 |
+
},
|
| 235576 |
+
{
|
| 235577 |
+
"epoch": 863.27,
|
| 235578 |
+
"learning_rate": 8.287285945072698e-06,
|
| 235579 |
+
"loss": 0.2991,
|
| 235580 |
+
"step": 107060
|
| 235581 |
+
},
|
| 235582 |
+
{
|
| 235583 |
+
"epoch": 863.31,
|
| 235584 |
+
"learning_rate": 8.287205169628434e-06,
|
| 235585 |
+
"loss": 0.2825,
|
| 235586 |
+
"step": 107065
|
| 235587 |
+
},
|
| 235588 |
+
{
|
| 235589 |
+
"epoch": 863.35,
|
| 235590 |
+
"learning_rate": 8.287124394184168e-06,
|
| 235591 |
+
"loss": 0.3772,
|
| 235592 |
+
"step": 107070
|
| 235593 |
+
},
|
| 235594 |
+
{
|
| 235595 |
+
"epoch": 863.39,
|
| 235596 |
+
"learning_rate": 8.287043618739904e-06,
|
| 235597 |
+
"loss": 0.6845,
|
| 235598 |
+
"step": 107075
|
| 235599 |
+
},
|
| 235600 |
+
{
|
| 235601 |
+
"epoch": 863.43,
|
| 235602 |
+
"learning_rate": 8.286962843295638e-06,
|
| 235603 |
+
"loss": 0.7852,
|
| 235604 |
+
"step": 107080
|
| 235605 |
+
},
|
| 235606 |
+
{
|
| 235607 |
+
"epoch": 863.47,
|
| 235608 |
+
"learning_rate": 8.286882067851374e-06,
|
| 235609 |
+
"loss": 0.2787,
|
| 235610 |
+
"step": 107085
|
| 235611 |
+
},
|
| 235612 |
+
{
|
| 235613 |
+
"epoch": 863.51,
|
| 235614 |
+
"learning_rate": 8.286801292407108e-06,
|
| 235615 |
+
"loss": 0.2554,
|
| 235616 |
+
"step": 107090
|
| 235617 |
+
},
|
| 235618 |
+
{
|
| 235619 |
+
"epoch": 863.55,
|
| 235620 |
+
"learning_rate": 8.286720516962844e-06,
|
| 235621 |
+
"loss": 0.3733,
|
| 235622 |
+
"step": 107095
|
| 235623 |
+
},
|
| 235624 |
+
{
|
| 235625 |
+
"epoch": 863.59,
|
| 235626 |
+
"learning_rate": 8.28663974151858e-06,
|
| 235627 |
+
"loss": 0.7647,
|
| 235628 |
+
"step": 107100
|
| 235629 |
+
},
|
| 235630 |
+
{
|
| 235631 |
+
"epoch": 863.63,
|
| 235632 |
+
"learning_rate": 8.286558966074314e-06,
|
| 235633 |
+
"loss": 0.874,
|
| 235634 |
+
"step": 107105
|
| 235635 |
+
},
|
| 235636 |
+
{
|
| 235637 |
+
"epoch": 863.67,
|
| 235638 |
+
"learning_rate": 8.28647819063005e-06,
|
| 235639 |
+
"loss": 0.2574,
|
| 235640 |
+
"step": 107110
|
| 235641 |
+
},
|
| 235642 |
+
{
|
| 235643 |
+
"epoch": 863.71,
|
| 235644 |
+
"learning_rate": 8.286397415185784e-06,
|
| 235645 |
+
"loss": 0.2416,
|
| 235646 |
+
"step": 107115
|
| 235647 |
+
},
|
| 235648 |
+
{
|
| 235649 |
+
"epoch": 863.75,
|
| 235650 |
+
"learning_rate": 8.28631663974152e-06,
|
| 235651 |
+
"loss": 0.4198,
|
| 235652 |
+
"step": 107120
|
| 235653 |
+
},
|
| 235654 |
+
{
|
| 235655 |
+
"epoch": 863.79,
|
| 235656 |
+
"learning_rate": 8.286235864297254e-06,
|
| 235657 |
+
"loss": 0.7976,
|
| 235658 |
+
"step": 107125
|
| 235659 |
+
},
|
| 235660 |
+
{
|
| 235661 |
+
"epoch": 863.83,
|
| 235662 |
+
"learning_rate": 8.28615508885299e-06,
|
| 235663 |
+
"loss": 0.8667,
|
| 235664 |
+
"step": 107130
|
| 235665 |
+
},
|
| 235666 |
+
{
|
| 235667 |
+
"epoch": 863.87,
|
| 235668 |
+
"learning_rate": 8.286074313408724e-06,
|
| 235669 |
+
"loss": 0.2495,
|
| 235670 |
+
"step": 107135
|
| 235671 |
+
},
|
| 235672 |
+
{
|
| 235673 |
+
"epoch": 863.91,
|
| 235674 |
+
"learning_rate": 8.28599353796446e-06,
|
| 235675 |
+
"loss": 0.3132,
|
| 235676 |
+
"step": 107140
|
| 235677 |
+
},
|
| 235678 |
+
{
|
| 235679 |
+
"epoch": 863.95,
|
| 235680 |
+
"learning_rate": 8.285912762520193e-06,
|
| 235681 |
+
"loss": 0.4003,
|
| 235682 |
+
"step": 107145
|
| 235683 |
+
},
|
| 235684 |
+
{
|
| 235685 |
+
"epoch": 863.99,
|
| 235686 |
+
"learning_rate": 8.28583198707593e-06,
|
| 235687 |
+
"loss": 0.8067,
|
| 235688 |
+
"step": 107150
|
| 235689 |
+
},
|
| 235690 |
+
{
|
| 235691 |
+
"epoch": 864.0,
|
| 235692 |
+
"eval_loss": 0.37224456667900085,
|
| 235693 |
+
"eval_runtime": 42.1498,
|
| 235694 |
+
"eval_samples_per_second": 19.881,
|
| 235695 |
+
"eval_steps_per_second": 0.641,
|
| 235696 |
+
"eval_wer": 0.18059855521155832,
|
| 235697 |
+
"step": 107151
|
| 235698 |
+
},
|
| 235699 |
+
{
|
| 235700 |
+
"epoch": 857.03,
|
| 235701 |
+
"learning_rate": 8.285751211631665e-06,
|
| 235702 |
+
"loss": 0.3574,
|
| 235703 |
+
"step": 107155
|
| 235704 |
+
},
|
| 235705 |
+
{
|
| 235706 |
+
"epoch": 857.07,
|
| 235707 |
+
"learning_rate": 8.2856704361874e-06,
|
| 235708 |
+
"loss": 0.2889,
|
| 235709 |
+
"step": 107160
|
| 235710 |
+
},
|
| 235711 |
+
{
|
| 235712 |
+
"epoch": 857.11,
|
| 235713 |
+
"learning_rate": 8.285589660743135e-06,
|
| 235714 |
+
"loss": 0.3086,
|
| 235715 |
+
"step": 107165
|
| 235716 |
+
},
|
| 235717 |
+
{
|
| 235718 |
+
"epoch": 857.15,
|
| 235719 |
+
"learning_rate": 8.28550888529887e-06,
|
| 235720 |
+
"loss": 0.3949,
|
| 235721 |
+
"step": 107170
|
| 235722 |
+
},
|
| 235723 |
+
{
|
| 235724 |
+
"epoch": 857.19,
|
| 235725 |
+
"learning_rate": 8.285428109854605e-06,
|
| 235726 |
+
"loss": 0.8241,
|
| 235727 |
+
"step": 107175
|
| 235728 |
+
},
|
| 235729 |
+
{
|
| 235730 |
+
"epoch": 857.23,
|
| 235731 |
+
"learning_rate": 8.28534733441034e-06,
|
| 235732 |
+
"loss": 0.7566,
|
| 235733 |
+
"step": 107180
|
| 235734 |
+
},
|
| 235735 |
+
{
|
| 235736 |
+
"epoch": 857.27,
|
| 235737 |
+
"learning_rate": 8.285266558966075e-06,
|
| 235738 |
+
"loss": 0.2608,
|
| 235739 |
+
"step": 107185
|
| 235740 |
+
},
|
| 235741 |
+
{
|
| 235742 |
+
"epoch": 857.31,
|
| 235743 |
+
"learning_rate": 8.28518578352181e-06,
|
| 235744 |
+
"loss": 0.2851,
|
| 235745 |
+
"step": 107190
|
| 235746 |
+
},
|
| 235747 |
+
{
|
| 235748 |
+
"epoch": 857.35,
|
| 235749 |
+
"learning_rate": 8.285105008077545e-06,
|
| 235750 |
+
"loss": 0.3554,
|
| 235751 |
+
"step": 107195
|
| 235752 |
+
},
|
| 235753 |
+
{
|
| 235754 |
+
"epoch": 857.39,
|
| 235755 |
+
"learning_rate": 8.28502423263328e-06,
|
| 235756 |
+
"loss": 0.8735,
|
| 235757 |
+
"step": 107200
|
| 235758 |
+
},
|
| 235759 |
+
{
|
| 235760 |
+
"epoch": 857.43,
|
| 235761 |
+
"learning_rate": 8.284943457189015e-06,
|
| 235762 |
+
"loss": 0.7487,
|
| 235763 |
+
"step": 107205
|
| 235764 |
+
},
|
| 235765 |
+
{
|
| 235766 |
+
"epoch": 857.47,
|
| 235767 |
+
"learning_rate": 8.28486268174475e-06,
|
| 235768 |
+
"loss": 0.3156,
|
| 235769 |
+
"step": 107210
|
| 235770 |
+
},
|
| 235771 |
+
{
|
| 235772 |
+
"epoch": 857.51,
|
| 235773 |
+
"learning_rate": 8.284781906300485e-06,
|
| 235774 |
+
"loss": 0.3211,
|
| 235775 |
+
"step": 107215
|
| 235776 |
+
},
|
| 235777 |
+
{
|
| 235778 |
+
"epoch": 857.55,
|
| 235779 |
+
"learning_rate": 8.284701130856221e-06,
|
| 235780 |
+
"loss": 0.3652,
|
| 235781 |
+
"step": 107220
|
| 235782 |
+
},
|
| 235783 |
+
{
|
| 235784 |
+
"epoch": 857.59,
|
| 235785 |
+
"learning_rate": 8.284620355411955e-06,
|
| 235786 |
+
"loss": 0.9481,
|
| 235787 |
+
"step": 107225
|
| 235788 |
+
},
|
| 235789 |
+
{
|
| 235790 |
+
"epoch": 857.63,
|
| 235791 |
+
"learning_rate": 8.28453957996769e-06,
|
| 235792 |
+
"loss": 0.6768,
|
| 235793 |
+
"step": 107230
|
| 235794 |
+
},
|
| 235795 |
+
{
|
| 235796 |
+
"epoch": 857.67,
|
| 235797 |
+
"learning_rate": 8.284458804523425e-06,
|
| 235798 |
+
"loss": 0.3285,
|
| 235799 |
+
"step": 107235
|
| 235800 |
+
},
|
| 235801 |
+
{
|
| 235802 |
+
"epoch": 857.71,
|
| 235803 |
+
"learning_rate": 8.28437802907916e-06,
|
| 235804 |
+
"loss": 0.2585,
|
| 235805 |
+
"step": 107240
|
| 235806 |
+
},
|
| 235807 |
+
{
|
| 235808 |
+
"epoch": 857.75,
|
| 235809 |
+
"learning_rate": 8.284297253634895e-06,
|
| 235810 |
+
"loss": 0.446,
|
| 235811 |
+
"step": 107245
|
| 235812 |
+
},
|
| 235813 |
+
{
|
| 235814 |
+
"epoch": 857.79,
|
| 235815 |
+
"learning_rate": 8.28421647819063e-06,
|
| 235816 |
+
"loss": 0.9258,
|
| 235817 |
+
"step": 107250
|
| 235818 |
+
},
|
| 235819 |
+
{
|
| 235820 |
+
"epoch": 857.83,
|
| 235821 |
+
"learning_rate": 8.284135702746365e-06,
|
| 235822 |
+
"loss": 0.6147,
|
| 235823 |
+
"step": 107255
|
| 235824 |
+
},
|
| 235825 |
+
{
|
| 235826 |
+
"epoch": 857.87,
|
| 235827 |
+
"learning_rate": 8.2840549273021e-06,
|
| 235828 |
+
"loss": 0.3745,
|
| 235829 |
+
"step": 107260
|
| 235830 |
+
},
|
| 235831 |
+
{
|
| 235832 |
+
"epoch": 857.91,
|
| 235833 |
+
"learning_rate": 8.283974151857835e-06,
|
| 235834 |
+
"loss": 0.3359,
|
| 235835 |
+
"step": 107265
|
| 235836 |
+
},
|
| 235837 |
+
{
|
| 235838 |
+
"epoch": 857.95,
|
| 235839 |
+
"learning_rate": 8.28389337641357e-06,
|
| 235840 |
+
"loss": 0.4183,
|
| 235841 |
+
"step": 107270
|
| 235842 |
+
},
|
| 235843 |
+
{
|
| 235844 |
+
"epoch": 857.99,
|
| 235845 |
+
"learning_rate": 8.283812600969307e-06,
|
| 235846 |
+
"loss": 0.9921,
|
| 235847 |
+
"step": 107275
|
| 235848 |
+
},
|
| 235849 |
+
{
|
| 235850 |
+
"epoch": 858.0,
|
| 235851 |
+
"eval_loss": 0.41086187958717346,
|
| 235852 |
+
"eval_runtime": 42.2944,
|
| 235853 |
+
"eval_samples_per_second": 19.814,
|
| 235854 |
+
"eval_steps_per_second": 0.638,
|
| 235855 |
+
"eval_wer": 0.18910348803689825,
|
| 235856 |
+
"step": 107276
|
| 235857 |
+
},
|
| 235858 |
+
{
|
| 235859 |
+
"epoch": 865.03,
|
| 235860 |
+
"learning_rate": 8.28373182552504e-06,
|
| 235861 |
+
"loss": 0.3192,
|
| 235862 |
+
"step": 107280
|
| 235863 |
+
},
|
| 235864 |
+
{
|
| 235865 |
+
"epoch": 865.07,
|
| 235866 |
+
"learning_rate": 8.283651050080777e-06,
|
| 235867 |
+
"loss": 0.3178,
|
| 235868 |
+
"step": 107285
|
| 235869 |
+
},
|
| 235870 |
+
{
|
| 235871 |
+
"epoch": 865.11,
|
| 235872 |
+
"learning_rate": 8.28357027463651e-06,
|
| 235873 |
+
"loss": 0.2855,
|
| 235874 |
+
"step": 107290
|
| 235875 |
+
},
|
| 235876 |
+
{
|
| 235877 |
+
"epoch": 865.15,
|
| 235878 |
+
"learning_rate": 8.283489499192247e-06,
|
| 235879 |
+
"loss": 0.4488,
|
| 235880 |
+
"step": 107295
|
| 235881 |
+
},
|
| 235882 |
+
{
|
| 235883 |
+
"epoch": 865.19,
|
| 235884 |
+
"learning_rate": 8.28340872374798e-06,
|
| 235885 |
+
"loss": 0.9815,
|
| 235886 |
+
"step": 107300
|
| 235887 |
+
},
|
| 235888 |
+
{
|
| 235889 |
+
"epoch": 865.23,
|
| 235890 |
+
"learning_rate": 8.283327948303716e-06,
|
| 235891 |
+
"loss": 0.6211,
|
| 235892 |
+
"step": 107305
|
| 235893 |
+
},
|
| 235894 |
+
{
|
| 235895 |
+
"epoch": 865.27,
|
| 235896 |
+
"learning_rate": 8.28324717285945e-06,
|
| 235897 |
+
"loss": 0.2939,
|
| 235898 |
+
"step": 107310
|
| 235899 |
+
},
|
| 235900 |
+
{
|
| 235901 |
+
"epoch": 865.31,
|
| 235902 |
+
"learning_rate": 8.283166397415186e-06,
|
| 235903 |
+
"loss": 0.3034,
|
| 235904 |
+
"step": 107315
|
| 235905 |
+
},
|
| 235906 |
+
{
|
| 235907 |
+
"epoch": 865.35,
|
| 235908 |
+
"learning_rate": 8.28308562197092e-06,
|
| 235909 |
+
"loss": 0.3967,
|
| 235910 |
+
"step": 107320
|
| 235911 |
+
},
|
| 235912 |
+
{
|
| 235913 |
+
"epoch": 865.39,
|
| 235914 |
+
"learning_rate": 8.283004846526656e-06,
|
| 235915 |
+
"loss": 0.9682,
|
| 235916 |
+
"step": 107325
|
| 235917 |
+
},
|
| 235918 |
+
{
|
| 235919 |
+
"epoch": 865.43,
|
| 235920 |
+
"learning_rate": 8.282924071082392e-06,
|
| 235921 |
+
"loss": 0.684,
|
| 235922 |
+
"step": 107330
|
| 235923 |
+
},
|
| 235924 |
+
{
|
| 235925 |
+
"epoch": 865.47,
|
| 235926 |
+
"learning_rate": 8.282843295638126e-06,
|
| 235927 |
+
"loss": 0.2839,
|
| 235928 |
+
"step": 107335
|
| 235929 |
+
},
|
| 235930 |
+
{
|
| 235931 |
+
"epoch": 865.51,
|
| 235932 |
+
"learning_rate": 8.282762520193862e-06,
|
| 235933 |
+
"loss": 0.2923,
|
| 235934 |
+
"step": 107340
|
| 235935 |
+
},
|
| 235936 |
+
{
|
| 235937 |
+
"epoch": 865.55,
|
| 235938 |
+
"learning_rate": 8.282681744749596e-06,
|
| 235939 |
+
"loss": 0.398,
|
| 235940 |
+
"step": 107345
|
| 235941 |
+
},
|
| 235942 |
+
{
|
| 235943 |
+
"epoch": 865.59,
|
| 235944 |
+
"learning_rate": 8.282600969305332e-06,
|
| 235945 |
+
"loss": 1.0223,
|
| 235946 |
+
"step": 107350
|
| 235947 |
+
},
|
| 235948 |
+
{
|
| 235949 |
+
"epoch": 865.63,
|
| 235950 |
+
"learning_rate": 8.282520193861066e-06,
|
| 235951 |
+
"loss": 0.717,
|
| 235952 |
+
"step": 107355
|
| 235953 |
+
},
|
| 235954 |
+
{
|
| 235955 |
+
"epoch": 865.67,
|
| 235956 |
+
"learning_rate": 8.282439418416802e-06,
|
| 235957 |
+
"loss": 0.2891,
|
| 235958 |
+
"step": 107360
|
| 235959 |
+
},
|
| 235960 |
+
{
|
| 235961 |
+
"epoch": 865.71,
|
| 235962 |
+
"learning_rate": 8.282358642972536e-06,
|
| 235963 |
+
"loss": 0.3408,
|
| 235964 |
+
"step": 107365
|
| 235965 |
+
},
|
| 235966 |
+
{
|
| 235967 |
+
"epoch": 865.76,
|
| 235968 |
+
"learning_rate": 8.282277867528272e-06,
|
| 235969 |
+
"loss": 0.3692,
|
| 235970 |
+
"step": 107370
|
| 235971 |
+
},
|
| 235972 |
+
{
|
| 235973 |
+
"epoch": 865.8,
|
| 235974 |
+
"learning_rate": 8.282197092084006e-06,
|
| 235975 |
+
"loss": 0.96,
|
| 235976 |
+
"step": 107375
|
| 235977 |
+
},
|
| 235978 |
+
{
|
| 235979 |
+
"epoch": 865.84,
|
| 235980 |
+
"learning_rate": 8.282116316639742e-06,
|
| 235981 |
+
"loss": 0.6772,
|
| 235982 |
+
"step": 107380
|
| 235983 |
+
},
|
| 235984 |
+
{
|
| 235985 |
+
"epoch": 865.88,
|
| 235986 |
+
"learning_rate": 8.282035541195476e-06,
|
| 235987 |
+
"loss": 0.2813,
|
| 235988 |
+
"step": 107385
|
| 235989 |
+
},
|
| 235990 |
+
{
|
| 235991 |
+
"epoch": 865.92,
|
| 235992 |
+
"learning_rate": 8.281954765751212e-06,
|
| 235993 |
+
"loss": 0.2974,
|
| 235994 |
+
"step": 107390
|
| 235995 |
+
},
|
| 235996 |
+
{
|
| 235997 |
+
"epoch": 865.96,
|
| 235998 |
+
"learning_rate": 8.281873990306948e-06,
|
| 235999 |
+
"loss": 0.4065,
|
| 236000 |
+
"step": 107395
|
| 236001 |
+
},
|
| 236002 |
+
{
|
| 236003 |
+
"epoch": 866.0,
|
| 236004 |
+
"learning_rate": 8.281793214862682e-06,
|
| 236005 |
+
"loss": 1.0347,
|
| 236006 |
+
"step": 107400
|
| 236007 |
+
},
|
| 236008 |
+
{
|
| 236009 |
+
"epoch": 866.0,
|
| 236010 |
+
"eval_loss": 0.3686482906341553,
|
| 236011 |
+
"eval_runtime": 41.8142,
|
| 236012 |
+
"eval_samples_per_second": 20.041,
|
| 236013 |
+
"eval_steps_per_second": 0.646,
|
| 236014 |
+
"eval_wer": 0.17922694882314835,
|
| 236015 |
+
"step": 107400
|
| 236016 |
+
},
|
| 236017 |
+
{
|
| 236018 |
+
"epoch": 859.04,
|
| 236019 |
+
"learning_rate": 8.281712439418418e-06,
|
| 236020 |
+
"loss": 0.3211,
|
| 236021 |
+
"step": 107405
|
| 236022 |
+
},
|
| 236023 |
+
{
|
| 236024 |
+
"epoch": 859.08,
|
| 236025 |
+
"learning_rate": 8.281631663974152e-06,
|
| 236026 |
+
"loss": 0.3117,
|
| 236027 |
+
"step": 107410
|
| 236028 |
+
},
|
| 236029 |
+
{
|
| 236030 |
+
"epoch": 859.12,
|
| 236031 |
+
"learning_rate": 8.281550888529888e-06,
|
| 236032 |
+
"loss": 0.3672,
|
| 236033 |
+
"step": 107415
|
| 236034 |
+
},
|
| 236035 |
+
{
|
| 236036 |
+
"epoch": 859.16,
|
| 236037 |
+
"learning_rate": 8.281470113085622e-06,
|
| 236038 |
+
"loss": 0.4845,
|
| 236039 |
+
"step": 107420
|
| 236040 |
+
},
|
| 236041 |
+
{
|
| 236042 |
+
"epoch": 859.2,
|
| 236043 |
+
"learning_rate": 8.281389337641358e-06,
|
| 236044 |
+
"loss": 1.2302,
|
| 236045 |
+
"step": 107425
|
| 236046 |
+
},
|
| 236047 |
+
{
|
| 236048 |
+
"epoch": 859.24,
|
| 236049 |
+
"learning_rate": 8.281308562197092e-06,
|
| 236050 |
+
"loss": 0.3565,
|
| 236051 |
+
"step": 107430
|
| 236052 |
+
},
|
| 236053 |
+
{
|
| 236054 |
+
"epoch": 859.28,
|
| 236055 |
+
"learning_rate": 8.281227786752828e-06,
|
| 236056 |
+
"loss": 0.3029,
|
| 236057 |
+
"step": 107435
|
| 236058 |
+
},
|
| 236059 |
+
{
|
| 236060 |
+
"epoch": 859.32,
|
| 236061 |
+
"learning_rate": 8.281147011308562e-06,
|
| 236062 |
+
"loss": 0.3457,
|
| 236063 |
+
"step": 107440
|
| 236064 |
+
},
|
| 236065 |
+
{
|
| 236066 |
+
"epoch": 859.36,
|
| 236067 |
+
"learning_rate": 8.281066235864298e-06,
|
| 236068 |
+
"loss": 0.4207,
|
| 236069 |
+
"step": 107445
|
| 236070 |
+
},
|
| 236071 |
+
{
|
| 236072 |
+
"epoch": 859.4,
|
| 236073 |
+
"learning_rate": 8.280985460420034e-06,
|
| 236074 |
+
"loss": 1.3461,
|
| 236075 |
+
"step": 107450
|
| 236076 |
+
},
|
| 236077 |
+
{
|
| 236078 |
+
"epoch": 859.44,
|
| 236079 |
+
"learning_rate": 8.280904684975768e-06,
|
| 236080 |
+
"loss": 0.339,
|
| 236081 |
+
"step": 107455
|
| 236082 |
+
},
|
| 236083 |
+
{
|
| 236084 |
+
"epoch": 859.48,
|
| 236085 |
+
"learning_rate": 8.280823909531504e-06,
|
| 236086 |
+
"loss": 0.2571,
|
| 236087 |
+
"step": 107460
|
| 236088 |
+
},
|
| 236089 |
+
{
|
| 236090 |
+
"epoch": 859.52,
|
| 236091 |
+
"learning_rate": 8.280743134087238e-06,
|
| 236092 |
+
"loss": 0.2763,
|
| 236093 |
+
"step": 107465
|
| 236094 |
+
},
|
| 236095 |
+
{
|
| 236096 |
+
"epoch": 859.56,
|
| 236097 |
+
"learning_rate": 8.280662358642974e-06,
|
| 236098 |
+
"loss": 0.441,
|
| 236099 |
+
"step": 107470
|
| 236100 |
+
},
|
| 236101 |
+
{
|
| 236102 |
+
"epoch": 859.6,
|
| 236103 |
+
"learning_rate": 8.280581583198708e-06,
|
| 236104 |
+
"loss": 1.1244,
|
| 236105 |
+
"step": 107475
|
| 236106 |
+
},
|
| 236107 |
+
{
|
| 236108 |
+
"epoch": 859.64,
|
| 236109 |
+
"learning_rate": 8.280500807754444e-06,
|
| 236110 |
+
"loss": 0.3058,
|
| 236111 |
+
"step": 107480
|
| 236112 |
+
},
|
| 236113 |
+
{
|
| 236114 |
+
"epoch": 859.68,
|
| 236115 |
+
"learning_rate": 8.280420032310178e-06,
|
| 236116 |
+
"loss": 0.2942,
|
| 236117 |
+
"step": 107485
|
| 236118 |
+
},
|
| 236119 |
+
{
|
| 236120 |
+
"epoch": 859.72,
|
| 236121 |
+
"learning_rate": 8.280339256865914e-06,
|
| 236122 |
+
"loss": 0.313,
|
| 236123 |
+
"step": 107490
|
| 236124 |
+
},
|
| 236125 |
+
{
|
| 236126 |
+
"epoch": 859.76,
|
| 236127 |
+
"learning_rate": 8.280258481421648e-06,
|
| 236128 |
+
"loss": 0.4273,
|
| 236129 |
+
"step": 107495
|
| 236130 |
+
},
|
| 236131 |
+
{
|
| 236132 |
+
"epoch": 859.8,
|
| 236133 |
+
"learning_rate": 8.280177705977384e-06,
|
| 236134 |
+
"loss": 1.1123,
|
| 236135 |
+
"step": 107500
|
| 236136 |
+
},
|
| 236137 |
+
{
|
| 236138 |
+
"epoch": 859.84,
|
| 236139 |
+
"learning_rate": 8.28009693053312e-06,
|
| 236140 |
+
"loss": 0.3428,
|
| 236141 |
+
"step": 107505
|
| 236142 |
+
},
|
| 236143 |
+
{
|
| 236144 |
+
"epoch": 859.88,
|
| 236145 |
+
"learning_rate": 8.280016155088854e-06,
|
| 236146 |
+
"loss": 0.2697,
|
| 236147 |
+
"step": 107510
|
| 236148 |
+
},
|
| 236149 |
+
{
|
| 236150 |
+
"epoch": 859.92,
|
| 236151 |
+
"learning_rate": 8.27993537964459e-06,
|
| 236152 |
+
"loss": 0.4264,
|
| 236153 |
+
"step": 107515
|
| 236154 |
+
},
|
| 236155 |
+
{
|
| 236156 |
+
"epoch": 859.96,
|
| 236157 |
+
"learning_rate": 8.279854604200323e-06,
|
| 236158 |
+
"loss": 0.4448,
|
| 236159 |
+
"step": 107520
|
| 236160 |
+
},
|
| 236161 |
+
{
|
| 236162 |
+
"epoch": 860.0,
|
| 236163 |
+
"learning_rate": 8.27977382875606e-06,
|
| 236164 |
+
"loss": 1.3759,
|
| 236165 |
+
"step": 107525
|
| 236166 |
+
},
|
| 236167 |
+
{
|
| 236168 |
+
"epoch": 860.0,
|
| 236169 |
+
"eval_loss": 0.3905540108680725,
|
| 236170 |
+
"eval_runtime": 42.1088,
|
| 236171 |
+
"eval_samples_per_second": 19.901,
|
| 236172 |
+
"eval_steps_per_second": 0.641,
|
| 236173 |
+
"eval_wer": 0.19303890026323486,
|
| 236174 |
+
"step": 107525
|
| 236175 |
+
},
|
| 236176 |
+
{
|
| 236177 |
+
"epoch": 867.04,
|
| 236178 |
+
"learning_rate": 8.279693053311793e-06,
|
| 236179 |
+
"loss": 0.3856,
|
| 236180 |
+
"step": 107530
|
| 236181 |
+
},
|
| 236182 |
+
{
|
| 236183 |
+
"epoch": 867.08,
|
| 236184 |
+
"learning_rate": 8.27961227786753e-06,
|
| 236185 |
+
"loss": 0.2681,
|
| 236186 |
+
"step": 107535
|
| 236187 |
+
},
|
| 236188 |
+
{
|
| 236189 |
+
"epoch": 867.12,
|
| 236190 |
+
"learning_rate": 8.279531502423263e-06,
|
| 236191 |
+
"loss": 0.3065,
|
| 236192 |
+
"step": 107540
|
| 236193 |
+
},
|
| 236194 |
+
{
|
| 236195 |
+
"epoch": 867.16,
|
| 236196 |
+
"learning_rate": 8.279450726979e-06,
|
| 236197 |
+
"loss": 0.4992,
|
| 236198 |
+
"step": 107545
|
| 236199 |
+
},
|
| 236200 |
+
{
|
| 236201 |
+
"epoch": 867.2,
|
| 236202 |
+
"learning_rate": 8.279369951534733e-06,
|
| 236203 |
+
"loss": 1.0747,
|
| 236204 |
+
"step": 107550
|
| 236205 |
+
},
|
| 236206 |
+
{
|
| 236207 |
+
"epoch": 867.24,
|
| 236208 |
+
"learning_rate": 8.27928917609047e-06,
|
| 236209 |
+
"loss": 0.3252,
|
| 236210 |
+
"step": 107555
|
| 236211 |
+
},
|
| 236212 |
+
{
|
| 236213 |
+
"epoch": 867.28,
|
| 236214 |
+
"learning_rate": 8.279208400646203e-06,
|
| 236215 |
+
"loss": 0.2915,
|
| 236216 |
+
"step": 107560
|
| 236217 |
+
},
|
| 236218 |
+
{
|
| 236219 |
+
"epoch": 867.32,
|
| 236220 |
+
"learning_rate": 8.27912762520194e-06,
|
| 236221 |
+
"loss": 0.3444,
|
| 236222 |
+
"step": 107565
|
| 236223 |
+
},
|
| 236224 |
+
{
|
| 236225 |
+
"epoch": 867.36,
|
| 236226 |
+
"learning_rate": 8.279046849757675e-06,
|
| 236227 |
+
"loss": 0.4483,
|
| 236228 |
+
"step": 107570
|
| 236229 |
+
},
|
| 236230 |
+
{
|
| 236231 |
+
"epoch": 867.4,
|
| 236232 |
+
"learning_rate": 8.27896607431341e-06,
|
| 236233 |
+
"loss": 1.4068,
|
| 236234 |
+
"step": 107575
|
| 236235 |
+
},
|
| 236236 |
+
{
|
| 236237 |
+
"epoch": 867.44,
|
| 236238 |
+
"learning_rate": 8.278885298869145e-06,
|
| 236239 |
+
"loss": 0.3054,
|
| 236240 |
+
"step": 107580
|
| 236241 |
+
},
|
| 236242 |
+
{
|
| 236243 |
+
"epoch": 867.48,
|
| 236244 |
+
"learning_rate": 8.27880452342488e-06,
|
| 236245 |
+
"loss": 0.2739,
|
| 236246 |
+
"step": 107585
|
| 236247 |
+
},
|
| 236248 |
+
{
|
| 236249 |
+
"epoch": 867.52,
|
| 236250 |
+
"learning_rate": 8.278723747980615e-06,
|
| 236251 |
+
"loss": 0.3259,
|
| 236252 |
+
"step": 107590
|
| 236253 |
+
},
|
| 236254 |
+
{
|
| 236255 |
+
"epoch": 867.56,
|
| 236256 |
+
"learning_rate": 8.278642972536349e-06,
|
| 236257 |
+
"loss": 0.4701,
|
| 236258 |
+
"step": 107595
|
| 236259 |
+
},
|
| 236260 |
+
{
|
| 236261 |
+
"epoch": 867.6,
|
| 236262 |
+
"learning_rate": 8.278562197092085e-06,
|
| 236263 |
+
"loss": 1.1212,
|
| 236264 |
+
"step": 107600
|
| 236265 |
+
},
|
| 236266 |
+
{
|
| 236267 |
+
"epoch": 867.64,
|
| 236268 |
+
"learning_rate": 8.278481421647819e-06,
|
| 236269 |
+
"loss": 0.3187,
|
| 236270 |
+
"step": 107605
|
| 236271 |
+
},
|
| 236272 |
+
{
|
| 236273 |
+
"epoch": 867.68,
|
| 236274 |
+
"learning_rate": 8.278400646203555e-06,
|
| 236275 |
+
"loss": 0.2914,
|
| 236276 |
+
"step": 107610
|
| 236277 |
+
},
|
| 236278 |
+
{
|
| 236279 |
+
"epoch": 867.72,
|
| 236280 |
+
"learning_rate": 8.278319870759289e-06,
|
| 236281 |
+
"loss": 0.3098,
|
| 236282 |
+
"step": 107615
|
| 236283 |
+
},
|
| 236284 |
+
{
|
| 236285 |
+
"epoch": 867.76,
|
| 236286 |
+
"learning_rate": 8.278239095315025e-06,
|
| 236287 |
+
"loss": 0.5153,
|
| 236288 |
+
"step": 107620
|
| 236289 |
+
},
|
| 236290 |
+
{
|
| 236291 |
+
"epoch": 867.8,
|
| 236292 |
+
"learning_rate": 8.27815831987076e-06,
|
| 236293 |
+
"loss": 1.3541,
|
| 236294 |
+
"step": 107625
|
| 236295 |
+
},
|
| 236296 |
+
{
|
| 236297 |
+
"epoch": 867.84,
|
| 236298 |
+
"learning_rate": 8.278077544426495e-06,
|
| 236299 |
+
"loss": 0.3148,
|
| 236300 |
+
"step": 107630
|
| 236301 |
+
},
|
| 236302 |
+
{
|
| 236303 |
+
"epoch": 867.88,
|
| 236304 |
+
"learning_rate": 8.27799676898223e-06,
|
| 236305 |
+
"loss": 0.2729,
|
| 236306 |
+
"step": 107635
|
| 236307 |
+
},
|
| 236308 |
+
{
|
| 236309 |
+
"epoch": 867.92,
|
| 236310 |
+
"learning_rate": 8.277915993537965e-06,
|
| 236311 |
+
"loss": 0.3692,
|
| 236312 |
+
"step": 107640
|
| 236313 |
+
},
|
| 236314 |
+
{
|
| 236315 |
+
"epoch": 867.96,
|
| 236316 |
+
"learning_rate": 8.2778352180937e-06,
|
| 236317 |
+
"loss": 0.4566,
|
| 236318 |
+
"step": 107645
|
| 236319 |
+
},
|
| 236320 |
+
{
|
| 236321 |
+
"epoch": 868.0,
|
| 236322 |
+
"eval_loss": 0.35562440752983093,
|
| 236323 |
+
"eval_runtime": 42.2953,
|
| 236324 |
+
"eval_samples_per_second": 19.813,
|
| 236325 |
+
"eval_steps_per_second": 0.638,
|
| 236326 |
+
"eval_wer": 0.18214181923859343,
|
| 236327 |
+
"step": 107649
|
| 236328 |
}
|
| 236329 |
],
|
| 236330 |
+
"max_steps": 620000,
|
| 236331 |
"num_train_epochs": 5000,
|
| 236332 |
+
"total_flos": 3.029337911178788e+20,
|
| 236333 |
"trial_name": null,
|
| 236334 |
"trial_params": null
|
| 236335 |
}
|
model-bin/finetune/base/{checkpoint-107027 → checkpoint-107649}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630031322.4475768/events.out.tfevents.1630031322.52f5c7e305a3.886.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe3dc442587b1cbd1052387cdc4487244c95aab79859f25b82074067ad773cb1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630031785.6304755/events.out.tfevents.1630031785.52f5c7e305a3.886.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:35acc94b28a179789ee73988e96ebe4ddaec1853a528be7365cd4b83f1131f74
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630032238.9460363/events.out.tfevents.1630032238.52f5c7e305a3.886.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d97ebc265627985ac28b930344e9f939036cf97e9e5c29a9630737712d6dd14b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630032701.7654727/events.out.tfevents.1630032701.52f5c7e305a3.886.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c37e67a0dbc505b0351ae7e3d241fda6719bb39fb4b5016d7328cbb1cf2dae9f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630033167.5690465/events.out.tfevents.1630033167.52f5c7e305a3.886.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bffb2a43060a678169441deaa7dc602ba7616ab9a12cabd3a77185f0b9179f90
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630031322.52f5c7e305a3.886.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5500f44684514638e91391f0f9e73c415a6f932c0f8c0aea7f027e668eebd52
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630031785.52f5c7e305a3.886.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2df83daa48b6f3c332f0fd5d1dc9cd4c9986920c6e2b634ec0a00742d0bb0080
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630032238.52f5c7e305a3.886.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:78f86afe81b3399c8989c6cf6481241c5e2eeea1a31d5773813f486961a2ba38
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630032701.52f5c7e305a3.886.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a115095892153dad48b81fbaf4b0c4375f9ee38e8a47c0c4f0d5c9e65bb8cca1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630033167.52f5c7e305a3.886.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5e0facf6f50203bd22b90f771233bc10852dad57396ccc4952d4eeaacc07dd9
|
| 3 |
+
size 8462
|