"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630150760.2881565/events.out.tfevents.1630150760.86bb0ddabf9b.4092.141 +3 -0
- model-bin/finetune/base/log/1630151150.1171627/events.out.tfevents.1630151150.86bb0ddabf9b.4092.143 +3 -0
- model-bin/finetune/base/log/1630151541.9840891/events.out.tfevents.1630151541.86bb0ddabf9b.4092.145 +3 -0
- model-bin/finetune/base/log/1630151929.4076536/events.out.tfevents.1630151929.86bb0ddabf9b.4092.147 +3 -0
- model-bin/finetune/base/log/1630152323.041462/events.out.tfevents.1630152323.86bb0ddabf9b.4092.149 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630150760.86bb0ddabf9b.4092.140 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630151150.86bb0ddabf9b.4092.142 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630151541.86bb0ddabf9b.4092.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630151929.86bb0ddabf9b.4092.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630152323.86bb0ddabf9b.4092.148 +3 -0
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/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 sha256:9b3156a8f4e201766c4df6ffa74d8abc97daafe531c5db333d2a9a6e1bbb367c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/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 sha256:380e4ce8c820adecac0ac117754c22dbcbee91001dd4e41eebe002872dbe631c
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/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 sha256:5dee93488770f745feda0ea9fd7c0a22d4ef4429c854d4af13ea78ee1e76527d
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/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 sha256:703b2c951df4efdfabf698ea4ebf8efc6d144661a2d4c8aee86aa621f308f4ea
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/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:d9bc4ee8cc526980e6e93a1f57d2bdc9f5f11496f758041ced6b9b5c7d86ec3e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -260550,11 +260550,806 @@
|
|
| 260550 |
"eval_steps_per_second": 0.742,
|
| 260551 |
"eval_wer": 0.17306996599377758,
|
| 260552 |
"step": 126689
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 260553 |
}
|
| 260554 |
],
|
| 260555 |
-
"max_steps":
|
| 260556 |
"num_train_epochs": 5000,
|
| 260557 |
-
"total_flos": 3.
|
| 260558 |
"trial_name": null,
|
| 260559 |
"trial_params": null
|
| 260560 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1018.0,
|
| 5 |
+
"global_step": 127312,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 260550 |
"eval_steps_per_second": 0.742,
|
| 260551 |
"eval_wer": 0.17306996599377758,
|
| 260552 |
"step": 126689
|
| 260553 |
+
},
|
| 260554 |
+
{
|
| 260555 |
+
"epoch": 1013.01,
|
| 260556 |
+
"learning_rate": 7.986522435897436e-06,
|
| 260557 |
+
"loss": 0.2831,
|
| 260558 |
+
"step": 126690
|
| 260559 |
+
},
|
| 260560 |
+
{
|
| 260561 |
+
"epoch": 1013.05,
|
| 260562 |
+
"learning_rate": 7.986442307692308e-06,
|
| 260563 |
+
"loss": 0.2879,
|
| 260564 |
+
"step": 126695
|
| 260565 |
+
},
|
| 260566 |
+
{
|
| 260567 |
+
"epoch": 1013.09,
|
| 260568 |
+
"learning_rate": 7.98636217948718e-06,
|
| 260569 |
+
"loss": 0.2605,
|
| 260570 |
+
"step": 126700
|
| 260571 |
+
},
|
| 260572 |
+
{
|
| 260573 |
+
"epoch": 1013.13,
|
| 260574 |
+
"learning_rate": 7.986282051282052e-06,
|
| 260575 |
+
"loss": 0.346,
|
| 260576 |
+
"step": 126705
|
| 260577 |
+
},
|
| 260578 |
+
{
|
| 260579 |
+
"epoch": 1013.17,
|
| 260580 |
+
"learning_rate": 7.986201923076923e-06,
|
| 260581 |
+
"loss": 0.5423,
|
| 260582 |
+
"step": 126710
|
| 260583 |
+
},
|
| 260584 |
+
{
|
| 260585 |
+
"epoch": 1013.21,
|
| 260586 |
+
"learning_rate": 7.986121794871795e-06,
|
| 260587 |
+
"loss": 1.0971,
|
| 260588 |
+
"step": 126715
|
| 260589 |
+
},
|
| 260590 |
+
{
|
| 260591 |
+
"epoch": 1013.25,
|
| 260592 |
+
"learning_rate": 7.986041666666668e-06,
|
| 260593 |
+
"loss": 0.2661,
|
| 260594 |
+
"step": 126720
|
| 260595 |
+
},
|
| 260596 |
+
{
|
| 260597 |
+
"epoch": 1013.29,
|
| 260598 |
+
"learning_rate": 7.98596153846154e-06,
|
| 260599 |
+
"loss": 0.2675,
|
| 260600 |
+
"step": 126725
|
| 260601 |
+
},
|
| 260602 |
+
{
|
| 260603 |
+
"epoch": 1013.33,
|
| 260604 |
+
"learning_rate": 7.98588141025641e-06,
|
| 260605 |
+
"loss": 0.3517,
|
| 260606 |
+
"step": 126730
|
| 260607 |
+
},
|
| 260608 |
+
{
|
| 260609 |
+
"epoch": 1013.37,
|
| 260610 |
+
"learning_rate": 7.985801282051284e-06,
|
| 260611 |
+
"loss": 0.5183,
|
| 260612 |
+
"step": 126735
|
| 260613 |
+
},
|
| 260614 |
+
{
|
| 260615 |
+
"epoch": 1013.41,
|
| 260616 |
+
"learning_rate": 7.985721153846155e-06,
|
| 260617 |
+
"loss": 1.2005,
|
| 260618 |
+
"step": 126740
|
| 260619 |
+
},
|
| 260620 |
+
{
|
| 260621 |
+
"epoch": 1013.45,
|
| 260622 |
+
"learning_rate": 7.985641025641026e-06,
|
| 260623 |
+
"loss": 0.2968,
|
| 260624 |
+
"step": 126745
|
| 260625 |
+
},
|
| 260626 |
+
{
|
| 260627 |
+
"epoch": 1013.49,
|
| 260628 |
+
"learning_rate": 7.985560897435898e-06,
|
| 260629 |
+
"loss": 0.2646,
|
| 260630 |
+
"step": 126750
|
| 260631 |
+
},
|
| 260632 |
+
{
|
| 260633 |
+
"epoch": 1013.53,
|
| 260634 |
+
"learning_rate": 7.98548076923077e-06,
|
| 260635 |
+
"loss": 0.3582,
|
| 260636 |
+
"step": 126755
|
| 260637 |
+
},
|
| 260638 |
+
{
|
| 260639 |
+
"epoch": 1013.57,
|
| 260640 |
+
"learning_rate": 7.98540064102564e-06,
|
| 260641 |
+
"loss": 0.4968,
|
| 260642 |
+
"step": 126760
|
| 260643 |
+
},
|
| 260644 |
+
{
|
| 260645 |
+
"epoch": 1013.61,
|
| 260646 |
+
"learning_rate": 7.985320512820514e-06,
|
| 260647 |
+
"loss": 1.1176,
|
| 260648 |
+
"step": 126765
|
| 260649 |
+
},
|
| 260650 |
+
{
|
| 260651 |
+
"epoch": 1013.65,
|
| 260652 |
+
"learning_rate": 7.985240384615385e-06,
|
| 260653 |
+
"loss": 0.2675,
|
| 260654 |
+
"step": 126770
|
| 260655 |
+
},
|
| 260656 |
+
{
|
| 260657 |
+
"epoch": 1013.69,
|
| 260658 |
+
"learning_rate": 7.985160256410256e-06,
|
| 260659 |
+
"loss": 0.2741,
|
| 260660 |
+
"step": 126775
|
| 260661 |
+
},
|
| 260662 |
+
{
|
| 260663 |
+
"epoch": 1013.73,
|
| 260664 |
+
"learning_rate": 7.98508012820513e-06,
|
| 260665 |
+
"loss": 0.2681,
|
| 260666 |
+
"step": 126780
|
| 260667 |
+
},
|
| 260668 |
+
{
|
| 260669 |
+
"epoch": 1013.77,
|
| 260670 |
+
"learning_rate": 7.985e-06,
|
| 260671 |
+
"loss": 0.4765,
|
| 260672 |
+
"step": 126785
|
| 260673 |
+
},
|
| 260674 |
+
{
|
| 260675 |
+
"epoch": 1013.81,
|
| 260676 |
+
"learning_rate": 7.984919871794872e-06,
|
| 260677 |
+
"loss": 1.0048,
|
| 260678 |
+
"step": 126790
|
| 260679 |
+
},
|
| 260680 |
+
{
|
| 260681 |
+
"epoch": 1013.85,
|
| 260682 |
+
"learning_rate": 7.984839743589743e-06,
|
| 260683 |
+
"loss": 0.3135,
|
| 260684 |
+
"step": 126795
|
| 260685 |
+
},
|
| 260686 |
+
{
|
| 260687 |
+
"epoch": 1013.89,
|
| 260688 |
+
"learning_rate": 7.984759615384616e-06,
|
| 260689 |
+
"loss": 0.2559,
|
| 260690 |
+
"step": 126800
|
| 260691 |
+
},
|
| 260692 |
+
{
|
| 260693 |
+
"epoch": 1013.93,
|
| 260694 |
+
"learning_rate": 7.984679487179488e-06,
|
| 260695 |
+
"loss": 0.285,
|
| 260696 |
+
"step": 126805
|
| 260697 |
+
},
|
| 260698 |
+
{
|
| 260699 |
+
"epoch": 1013.97,
|
| 260700 |
+
"learning_rate": 7.984599358974359e-06,
|
| 260701 |
+
"loss": 0.5271,
|
| 260702 |
+
"step": 126810
|
| 260703 |
+
},
|
| 260704 |
+
{
|
| 260705 |
+
"epoch": 1014.0,
|
| 260706 |
+
"eval_loss": 0.39917340874671936,
|
| 260707 |
+
"eval_runtime": 35.7013,
|
| 260708 |
+
"eval_samples_per_second": 23.557,
|
| 260709 |
+
"eval_steps_per_second": 0.756,
|
| 260710 |
+
"eval_wer": 0.17787259084890392,
|
| 260711 |
+
"step": 126814
|
| 260712 |
+
},
|
| 260713 |
+
{
|
| 260714 |
+
"epoch": 1014.01,
|
| 260715 |
+
"learning_rate": 7.984535256410258e-06,
|
| 260716 |
+
"loss": 0.3605,
|
| 260717 |
+
"step": 126815
|
| 260718 |
+
},
|
| 260719 |
+
{
|
| 260720 |
+
"epoch": 1014.05,
|
| 260721 |
+
"learning_rate": 7.984455128205128e-06,
|
| 260722 |
+
"loss": 0.2761,
|
| 260723 |
+
"step": 126820
|
| 260724 |
+
},
|
| 260725 |
+
{
|
| 260726 |
+
"epoch": 1014.09,
|
| 260727 |
+
"learning_rate": 7.984375000000001e-06,
|
| 260728 |
+
"loss": 0.2825,
|
| 260729 |
+
"step": 126825
|
| 260730 |
+
},
|
| 260731 |
+
{
|
| 260732 |
+
"epoch": 1014.13,
|
| 260733 |
+
"learning_rate": 7.984294871794872e-06,
|
| 260734 |
+
"loss": 0.2817,
|
| 260735 |
+
"step": 126830
|
| 260736 |
+
},
|
| 260737 |
+
{
|
| 260738 |
+
"epoch": 1014.17,
|
| 260739 |
+
"learning_rate": 7.984214743589744e-06,
|
| 260740 |
+
"loss": 0.5611,
|
| 260741 |
+
"step": 126835
|
| 260742 |
+
},
|
| 260743 |
+
{
|
| 260744 |
+
"epoch": 1014.21,
|
| 260745 |
+
"learning_rate": 7.984134615384615e-06,
|
| 260746 |
+
"loss": 1.0804,
|
| 260747 |
+
"step": 126840
|
| 260748 |
+
},
|
| 260749 |
+
{
|
| 260750 |
+
"epoch": 1014.25,
|
| 260751 |
+
"learning_rate": 7.984054487179488e-06,
|
| 260752 |
+
"loss": 0.2839,
|
| 260753 |
+
"step": 126845
|
| 260754 |
+
},
|
| 260755 |
+
{
|
| 260756 |
+
"epoch": 1014.29,
|
| 260757 |
+
"learning_rate": 7.98397435897436e-06,
|
| 260758 |
+
"loss": 0.305,
|
| 260759 |
+
"step": 126850
|
| 260760 |
+
},
|
| 260761 |
+
{
|
| 260762 |
+
"epoch": 1014.33,
|
| 260763 |
+
"learning_rate": 7.98389423076923e-06,
|
| 260764 |
+
"loss": 0.3367,
|
| 260765 |
+
"step": 126855
|
| 260766 |
+
},
|
| 260767 |
+
{
|
| 260768 |
+
"epoch": 1014.37,
|
| 260769 |
+
"learning_rate": 7.983814102564104e-06,
|
| 260770 |
+
"loss": 0.5284,
|
| 260771 |
+
"step": 126860
|
| 260772 |
+
},
|
| 260773 |
+
{
|
| 260774 |
+
"epoch": 1014.41,
|
| 260775 |
+
"learning_rate": 7.983733974358975e-06,
|
| 260776 |
+
"loss": 1.2354,
|
| 260777 |
+
"step": 126865
|
| 260778 |
+
},
|
| 260779 |
+
{
|
| 260780 |
+
"epoch": 1014.45,
|
| 260781 |
+
"learning_rate": 7.983653846153847e-06,
|
| 260782 |
+
"loss": 0.2866,
|
| 260783 |
+
"step": 126870
|
| 260784 |
+
},
|
| 260785 |
+
{
|
| 260786 |
+
"epoch": 1014.49,
|
| 260787 |
+
"learning_rate": 7.983573717948718e-06,
|
| 260788 |
+
"loss": 0.2693,
|
| 260789 |
+
"step": 126875
|
| 260790 |
+
},
|
| 260791 |
+
{
|
| 260792 |
+
"epoch": 1014.53,
|
| 260793 |
+
"learning_rate": 7.983493589743591e-06,
|
| 260794 |
+
"loss": 0.2434,
|
| 260795 |
+
"step": 126880
|
| 260796 |
+
},
|
| 260797 |
+
{
|
| 260798 |
+
"epoch": 1014.57,
|
| 260799 |
+
"learning_rate": 7.983413461538462e-06,
|
| 260800 |
+
"loss": 0.6202,
|
| 260801 |
+
"step": 126885
|
| 260802 |
+
},
|
| 260803 |
+
{
|
| 260804 |
+
"epoch": 1014.61,
|
| 260805 |
+
"learning_rate": 7.983333333333334e-06,
|
| 260806 |
+
"loss": 1.2398,
|
| 260807 |
+
"step": 126890
|
| 260808 |
+
},
|
| 260809 |
+
{
|
| 260810 |
+
"epoch": 1014.65,
|
| 260811 |
+
"learning_rate": 7.983253205128205e-06,
|
| 260812 |
+
"loss": 0.3453,
|
| 260813 |
+
"step": 126895
|
| 260814 |
+
},
|
| 260815 |
+
{
|
| 260816 |
+
"epoch": 1014.69,
|
| 260817 |
+
"learning_rate": 7.983173076923078e-06,
|
| 260818 |
+
"loss": 0.297,
|
| 260819 |
+
"step": 126900
|
| 260820 |
+
},
|
| 260821 |
+
{
|
| 260822 |
+
"epoch": 1014.73,
|
| 260823 |
+
"learning_rate": 7.98309294871795e-06,
|
| 260824 |
+
"loss": 0.3543,
|
| 260825 |
+
"step": 126905
|
| 260826 |
+
},
|
| 260827 |
+
{
|
| 260828 |
+
"epoch": 1014.77,
|
| 260829 |
+
"learning_rate": 7.98301282051282e-06,
|
| 260830 |
+
"loss": 0.5697,
|
| 260831 |
+
"step": 126910
|
| 260832 |
+
},
|
| 260833 |
+
{
|
| 260834 |
+
"epoch": 1014.81,
|
| 260835 |
+
"learning_rate": 7.982932692307694e-06,
|
| 260836 |
+
"loss": 1.1563,
|
| 260837 |
+
"step": 126915
|
| 260838 |
+
},
|
| 260839 |
+
{
|
| 260840 |
+
"epoch": 1014.85,
|
| 260841 |
+
"learning_rate": 7.982852564102565e-06,
|
| 260842 |
+
"loss": 0.2929,
|
| 260843 |
+
"step": 126920
|
| 260844 |
+
},
|
| 260845 |
+
{
|
| 260846 |
+
"epoch": 1014.89,
|
| 260847 |
+
"learning_rate": 7.982772435897437e-06,
|
| 260848 |
+
"loss": 0.2657,
|
| 260849 |
+
"step": 126925
|
| 260850 |
+
},
|
| 260851 |
+
{
|
| 260852 |
+
"epoch": 1014.93,
|
| 260853 |
+
"learning_rate": 7.982692307692308e-06,
|
| 260854 |
+
"loss": 0.3184,
|
| 260855 |
+
"step": 126930
|
| 260856 |
+
},
|
| 260857 |
+
{
|
| 260858 |
+
"epoch": 1014.97,
|
| 260859 |
+
"learning_rate": 7.982612179487181e-06,
|
| 260860 |
+
"loss": 0.5176,
|
| 260861 |
+
"step": 126935
|
| 260862 |
+
},
|
| 260863 |
+
{
|
| 260864 |
+
"epoch": 1015.0,
|
| 260865 |
+
"eval_loss": 0.37343114614486694,
|
| 260866 |
+
"eval_runtime": 36.0496,
|
| 260867 |
+
"eval_samples_per_second": 23.329,
|
| 260868 |
+
"eval_steps_per_second": 0.749,
|
| 260869 |
+
"eval_wer": 0.18139708629062384,
|
| 260870 |
+
"step": 126939
|
| 260871 |
+
},
|
| 260872 |
+
{
|
| 260873 |
+
"epoch": 1023.01,
|
| 260874 |
+
"learning_rate": 7.98253205128205e-06,
|
| 260875 |
+
"loss": 0.4567,
|
| 260876 |
+
"step": 126940
|
| 260877 |
+
},
|
| 260878 |
+
{
|
| 260879 |
+
"epoch": 1023.05,
|
| 260880 |
+
"learning_rate": 7.982451923076924e-06,
|
| 260881 |
+
"loss": 0.2387,
|
| 260882 |
+
"step": 126945
|
| 260883 |
+
},
|
| 260884 |
+
{
|
| 260885 |
+
"epoch": 1023.09,
|
| 260886 |
+
"learning_rate": 7.982371794871797e-06,
|
| 260887 |
+
"loss": 0.3367,
|
| 260888 |
+
"step": 126950
|
| 260889 |
+
},
|
| 260890 |
+
{
|
| 260891 |
+
"epoch": 1023.13,
|
| 260892 |
+
"learning_rate": 7.982291666666666e-06,
|
| 260893 |
+
"loss": 0.2928,
|
| 260894 |
+
"step": 126955
|
| 260895 |
+
},
|
| 260896 |
+
{
|
| 260897 |
+
"epoch": 1023.17,
|
| 260898 |
+
"learning_rate": 7.98221153846154e-06,
|
| 260899 |
+
"loss": 0.5915,
|
| 260900 |
+
"step": 126960
|
| 260901 |
+
},
|
| 260902 |
+
{
|
| 260903 |
+
"epoch": 1023.21,
|
| 260904 |
+
"learning_rate": 7.982131410256411e-06,
|
| 260905 |
+
"loss": 1.155,
|
| 260906 |
+
"step": 126965
|
| 260907 |
+
},
|
| 260908 |
+
{
|
| 260909 |
+
"epoch": 1023.25,
|
| 260910 |
+
"learning_rate": 7.982051282051282e-06,
|
| 260911 |
+
"loss": 0.3101,
|
| 260912 |
+
"step": 126970
|
| 260913 |
+
},
|
| 260914 |
+
{
|
| 260915 |
+
"epoch": 1023.29,
|
| 260916 |
+
"learning_rate": 7.981971153846154e-06,
|
| 260917 |
+
"loss": 0.3658,
|
| 260918 |
+
"step": 126975
|
| 260919 |
+
},
|
| 260920 |
+
{
|
| 260921 |
+
"epoch": 1023.33,
|
| 260922 |
+
"learning_rate": 7.981891025641027e-06,
|
| 260923 |
+
"loss": 0.3383,
|
| 260924 |
+
"step": 126980
|
| 260925 |
+
},
|
| 260926 |
+
{
|
| 260927 |
+
"epoch": 1023.37,
|
| 260928 |
+
"learning_rate": 7.981810897435898e-06,
|
| 260929 |
+
"loss": 0.536,
|
| 260930 |
+
"step": 126985
|
| 260931 |
+
},
|
| 260932 |
+
{
|
| 260933 |
+
"epoch": 1023.41,
|
| 260934 |
+
"learning_rate": 7.98173076923077e-06,
|
| 260935 |
+
"loss": 1.1083,
|
| 260936 |
+
"step": 126990
|
| 260937 |
+
},
|
| 260938 |
+
{
|
| 260939 |
+
"epoch": 1023.45,
|
| 260940 |
+
"learning_rate": 7.98165064102564e-06,
|
| 260941 |
+
"loss": 0.2801,
|
| 260942 |
+
"step": 126995
|
| 260943 |
+
},
|
| 260944 |
+
{
|
| 260945 |
+
"epoch": 1023.49,
|
| 260946 |
+
"learning_rate": 7.981570512820514e-06,
|
| 260947 |
+
"loss": 0.3007,
|
| 260948 |
+
"step": 127000
|
| 260949 |
+
},
|
| 260950 |
+
{
|
| 260951 |
+
"epoch": 1023.53,
|
| 260952 |
+
"learning_rate": 7.981490384615385e-06,
|
| 260953 |
+
"loss": 0.3513,
|
| 260954 |
+
"step": 127005
|
| 260955 |
+
},
|
| 260956 |
+
{
|
| 260957 |
+
"epoch": 1023.57,
|
| 260958 |
+
"learning_rate": 7.981410256410256e-06,
|
| 260959 |
+
"loss": 0.6116,
|
| 260960 |
+
"step": 127010
|
| 260961 |
+
},
|
| 260962 |
+
{
|
| 260963 |
+
"epoch": 1023.61,
|
| 260964 |
+
"learning_rate": 7.98133012820513e-06,
|
| 260965 |
+
"loss": 1.1935,
|
| 260966 |
+
"step": 127015
|
| 260967 |
+
},
|
| 260968 |
+
{
|
| 260969 |
+
"epoch": 1023.65,
|
| 260970 |
+
"learning_rate": 7.981250000000001e-06,
|
| 260971 |
+
"loss": 0.3328,
|
| 260972 |
+
"step": 127020
|
| 260973 |
+
},
|
| 260974 |
+
{
|
| 260975 |
+
"epoch": 1023.69,
|
| 260976 |
+
"learning_rate": 7.981169871794872e-06,
|
| 260977 |
+
"loss": 0.2515,
|
| 260978 |
+
"step": 127025
|
| 260979 |
+
},
|
| 260980 |
+
{
|
| 260981 |
+
"epoch": 1023.73,
|
| 260982 |
+
"learning_rate": 7.981089743589744e-06,
|
| 260983 |
+
"loss": 0.2931,
|
| 260984 |
+
"step": 127030
|
| 260985 |
+
},
|
| 260986 |
+
{
|
| 260987 |
+
"epoch": 1023.77,
|
| 260988 |
+
"learning_rate": 7.981009615384617e-06,
|
| 260989 |
+
"loss": 0.499,
|
| 260990 |
+
"step": 127035
|
| 260991 |
+
},
|
| 260992 |
+
{
|
| 260993 |
+
"epoch": 1023.81,
|
| 260994 |
+
"learning_rate": 7.980929487179488e-06,
|
| 260995 |
+
"loss": 1.0857,
|
| 260996 |
+
"step": 127040
|
| 260997 |
+
},
|
| 260998 |
+
{
|
| 260999 |
+
"epoch": 1023.85,
|
| 261000 |
+
"learning_rate": 7.98084935897436e-06,
|
| 261001 |
+
"loss": 0.3193,
|
| 261002 |
+
"step": 127045
|
| 261003 |
+
},
|
| 261004 |
+
{
|
| 261005 |
+
"epoch": 1023.9,
|
| 261006 |
+
"learning_rate": 7.980769230769232e-06,
|
| 261007 |
+
"loss": 0.4392,
|
| 261008 |
+
"step": 127050
|
| 261009 |
+
},
|
| 261010 |
+
{
|
| 261011 |
+
"epoch": 1023.94,
|
| 261012 |
+
"learning_rate": 7.980689102564104e-06,
|
| 261013 |
+
"loss": 0.3994,
|
| 261014 |
+
"step": 127055
|
| 261015 |
+
},
|
| 261016 |
+
{
|
| 261017 |
+
"epoch": 1023.98,
|
| 261018 |
+
"learning_rate": 7.980608974358975e-06,
|
| 261019 |
+
"loss": 0.6098,
|
| 261020 |
+
"step": 127060
|
| 261021 |
+
},
|
| 261022 |
+
{
|
| 261023 |
+
"epoch": 1024.0,
|
| 261024 |
+
"eval_loss": 0.4028739035129547,
|
| 261025 |
+
"eval_runtime": 36.5089,
|
| 261026 |
+
"eval_samples_per_second": 23.036,
|
| 261027 |
+
"eval_steps_per_second": 0.74,
|
| 261028 |
+
"eval_wer": 0.17880398173789386,
|
| 261029 |
+
"step": 127063
|
| 261030 |
+
},
|
| 261031 |
+
{
|
| 261032 |
+
"epoch": 1024.02,
|
| 261033 |
+
"learning_rate": 7.980528846153846e-06,
|
| 261034 |
+
"loss": 0.3767,
|
| 261035 |
+
"step": 127065
|
| 261036 |
+
},
|
| 261037 |
+
{
|
| 261038 |
+
"epoch": 1024.06,
|
| 261039 |
+
"learning_rate": 7.98044871794872e-06,
|
| 261040 |
+
"loss": 0.2873,
|
| 261041 |
+
"step": 127070
|
| 261042 |
+
},
|
| 261043 |
+
{
|
| 261044 |
+
"epoch": 1024.1,
|
| 261045 |
+
"learning_rate": 7.980368589743591e-06,
|
| 261046 |
+
"loss": 0.2949,
|
| 261047 |
+
"step": 127075
|
| 261048 |
+
},
|
| 261049 |
+
{
|
| 261050 |
+
"epoch": 1024.14,
|
| 261051 |
+
"learning_rate": 7.980288461538462e-06,
|
| 261052 |
+
"loss": 0.3629,
|
| 261053 |
+
"step": 127080
|
| 261054 |
+
},
|
| 261055 |
+
{
|
| 261056 |
+
"epoch": 1024.18,
|
| 261057 |
+
"learning_rate": 7.980208333333334e-06,
|
| 261058 |
+
"loss": 0.5691,
|
| 261059 |
+
"step": 127085
|
| 261060 |
+
},
|
| 261061 |
+
{
|
| 261062 |
+
"epoch": 1024.22,
|
| 261063 |
+
"learning_rate": 7.980128205128207e-06,
|
| 261064 |
+
"loss": 1.0522,
|
| 261065 |
+
"step": 127090
|
| 261066 |
+
},
|
| 261067 |
+
{
|
| 261068 |
+
"epoch": 1024.26,
|
| 261069 |
+
"learning_rate": 7.980048076923076e-06,
|
| 261070 |
+
"loss": 0.2721,
|
| 261071 |
+
"step": 127095
|
| 261072 |
+
},
|
| 261073 |
+
{
|
| 261074 |
+
"epoch": 1024.3,
|
| 261075 |
+
"learning_rate": 7.97996794871795e-06,
|
| 261076 |
+
"loss": 0.2598,
|
| 261077 |
+
"step": 127100
|
| 261078 |
+
},
|
| 261079 |
+
{
|
| 261080 |
+
"epoch": 1024.34,
|
| 261081 |
+
"learning_rate": 7.979887820512822e-06,
|
| 261082 |
+
"loss": 0.3676,
|
| 261083 |
+
"step": 127105
|
| 261084 |
+
},
|
| 261085 |
+
{
|
| 261086 |
+
"epoch": 1024.38,
|
| 261087 |
+
"learning_rate": 7.979807692307692e-06,
|
| 261088 |
+
"loss": 0.5932,
|
| 261089 |
+
"step": 127110
|
| 261090 |
+
},
|
| 261091 |
+
{
|
| 261092 |
+
"epoch": 1024.42,
|
| 261093 |
+
"learning_rate": 7.979727564102565e-06,
|
| 261094 |
+
"loss": 0.9843,
|
| 261095 |
+
"step": 127115
|
| 261096 |
+
},
|
| 261097 |
+
{
|
| 261098 |
+
"epoch": 1024.46,
|
| 261099 |
+
"learning_rate": 7.979647435897437e-06,
|
| 261100 |
+
"loss": 0.2915,
|
| 261101 |
+
"step": 127120
|
| 261102 |
+
},
|
| 261103 |
+
{
|
| 261104 |
+
"epoch": 1024.5,
|
| 261105 |
+
"learning_rate": 7.979567307692308e-06,
|
| 261106 |
+
"loss": 0.2842,
|
| 261107 |
+
"step": 127125
|
| 261108 |
+
},
|
| 261109 |
+
{
|
| 261110 |
+
"epoch": 1024.54,
|
| 261111 |
+
"learning_rate": 7.97948717948718e-06,
|
| 261112 |
+
"loss": 0.3159,
|
| 261113 |
+
"step": 127130
|
| 261114 |
+
},
|
| 261115 |
+
{
|
| 261116 |
+
"epoch": 1024.58,
|
| 261117 |
+
"learning_rate": 7.979407051282052e-06,
|
| 261118 |
+
"loss": 0.5888,
|
| 261119 |
+
"step": 127135
|
| 261120 |
+
},
|
| 261121 |
+
{
|
| 261122 |
+
"epoch": 1024.62,
|
| 261123 |
+
"learning_rate": 7.979326923076924e-06,
|
| 261124 |
+
"loss": 0.9817,
|
| 261125 |
+
"step": 127140
|
| 261126 |
+
},
|
| 261127 |
+
{
|
| 261128 |
+
"epoch": 1024.66,
|
| 261129 |
+
"learning_rate": 7.979246794871795e-06,
|
| 261130 |
+
"loss": 0.2712,
|
| 261131 |
+
"step": 127145
|
| 261132 |
+
},
|
| 261133 |
+
{
|
| 261134 |
+
"epoch": 1024.7,
|
| 261135 |
+
"learning_rate": 7.979166666666668e-06,
|
| 261136 |
+
"loss": 0.3096,
|
| 261137 |
+
"step": 127150
|
| 261138 |
+
},
|
| 261139 |
+
{
|
| 261140 |
+
"epoch": 1024.74,
|
| 261141 |
+
"learning_rate": 7.97908653846154e-06,
|
| 261142 |
+
"loss": 0.3049,
|
| 261143 |
+
"step": 127155
|
| 261144 |
+
},
|
| 261145 |
+
{
|
| 261146 |
+
"epoch": 1024.78,
|
| 261147 |
+
"learning_rate": 7.97900641025641e-06,
|
| 261148 |
+
"loss": 0.6031,
|
| 261149 |
+
"step": 127160
|
| 261150 |
+
},
|
| 261151 |
+
{
|
| 261152 |
+
"epoch": 1024.82,
|
| 261153 |
+
"learning_rate": 7.978926282051282e-06,
|
| 261154 |
+
"loss": 0.9333,
|
| 261155 |
+
"step": 127165
|
| 261156 |
+
},
|
| 261157 |
+
{
|
| 261158 |
+
"epoch": 1024.86,
|
| 261159 |
+
"learning_rate": 7.978846153846155e-06,
|
| 261160 |
+
"loss": 0.2702,
|
| 261161 |
+
"step": 127170
|
| 261162 |
+
},
|
| 261163 |
+
{
|
| 261164 |
+
"epoch": 1024.9,
|
| 261165 |
+
"learning_rate": 7.978766025641027e-06,
|
| 261166 |
+
"loss": 0.2849,
|
| 261167 |
+
"step": 127175
|
| 261168 |
+
},
|
| 261169 |
+
{
|
| 261170 |
+
"epoch": 1024.94,
|
| 261171 |
+
"learning_rate": 7.978685897435898e-06,
|
| 261172 |
+
"loss": 0.3361,
|
| 261173 |
+
"step": 127180
|
| 261174 |
+
},
|
| 261175 |
+
{
|
| 261176 |
+
"epoch": 1024.98,
|
| 261177 |
+
"learning_rate": 7.97860576923077e-06,
|
| 261178 |
+
"loss": 0.5515,
|
| 261179 |
+
"step": 127185
|
| 261180 |
+
},
|
| 261181 |
+
{
|
| 261182 |
+
"epoch": 1025.0,
|
| 261183 |
+
"eval_loss": 0.30642202496528625,
|
| 261184 |
+
"eval_runtime": 36.1653,
|
| 261185 |
+
"eval_samples_per_second": 23.254,
|
| 261186 |
+
"eval_steps_per_second": 0.747,
|
| 261187 |
+
"eval_wer": 0.1736700622524052,
|
| 261188 |
+
"step": 127187
|
| 261189 |
+
},
|
| 261190 |
+
{
|
| 261191 |
+
"epoch": 1017.02,
|
| 261192 |
+
"learning_rate": 7.978525641025642e-06,
|
| 261193 |
+
"loss": 0.4288,
|
| 261194 |
+
"step": 127190
|
| 261195 |
+
},
|
| 261196 |
+
{
|
| 261197 |
+
"epoch": 1017.06,
|
| 261198 |
+
"learning_rate": 7.978445512820514e-06,
|
| 261199 |
+
"loss": 0.2895,
|
| 261200 |
+
"step": 127195
|
| 261201 |
+
},
|
| 261202 |
+
{
|
| 261203 |
+
"epoch": 1017.1,
|
| 261204 |
+
"learning_rate": 7.978365384615385e-06,
|
| 261205 |
+
"loss": 0.2966,
|
| 261206 |
+
"step": 127200
|
| 261207 |
+
},
|
| 261208 |
+
{
|
| 261209 |
+
"epoch": 1017.14,
|
| 261210 |
+
"learning_rate": 7.978285256410258e-06,
|
| 261211 |
+
"loss": 0.3754,
|
| 261212 |
+
"step": 127205
|
| 261213 |
+
},
|
| 261214 |
+
{
|
| 261215 |
+
"epoch": 1017.18,
|
| 261216 |
+
"learning_rate": 7.97820512820513e-06,
|
| 261217 |
+
"loss": 0.8517,
|
| 261218 |
+
"step": 127210
|
| 261219 |
+
},
|
| 261220 |
+
{
|
| 261221 |
+
"epoch": 1017.22,
|
| 261222 |
+
"learning_rate": 7.978125e-06,
|
| 261223 |
+
"loss": 0.8469,
|
| 261224 |
+
"step": 127215
|
| 261225 |
+
},
|
| 261226 |
+
{
|
| 261227 |
+
"epoch": 1017.26,
|
| 261228 |
+
"learning_rate": 7.978044871794872e-06,
|
| 261229 |
+
"loss": 0.2694,
|
| 261230 |
+
"step": 127220
|
| 261231 |
+
},
|
| 261232 |
+
{
|
| 261233 |
+
"epoch": 1017.3,
|
| 261234 |
+
"learning_rate": 7.977964743589745e-06,
|
| 261235 |
+
"loss": 0.2834,
|
| 261236 |
+
"step": 127225
|
| 261237 |
+
},
|
| 261238 |
+
{
|
| 261239 |
+
"epoch": 1017.34,
|
| 261240 |
+
"learning_rate": 7.977884615384615e-06,
|
| 261241 |
+
"loss": 0.4049,
|
| 261242 |
+
"step": 127230
|
| 261243 |
+
},
|
| 261244 |
+
{
|
| 261245 |
+
"epoch": 1017.38,
|
| 261246 |
+
"learning_rate": 7.977804487179488e-06,
|
| 261247 |
+
"loss": 0.6497,
|
| 261248 |
+
"step": 127235
|
| 261249 |
+
},
|
| 261250 |
+
{
|
| 261251 |
+
"epoch": 1017.42,
|
| 261252 |
+
"learning_rate": 7.97772435897436e-06,
|
| 261253 |
+
"loss": 0.8937,
|
| 261254 |
+
"step": 127240
|
| 261255 |
+
},
|
| 261256 |
+
{
|
| 261257 |
+
"epoch": 1017.46,
|
| 261258 |
+
"learning_rate": 7.97764423076923e-06,
|
| 261259 |
+
"loss": 0.2865,
|
| 261260 |
+
"step": 127245
|
| 261261 |
+
},
|
| 261262 |
+
{
|
| 261263 |
+
"epoch": 1017.5,
|
| 261264 |
+
"learning_rate": 7.977564102564104e-06,
|
| 261265 |
+
"loss": 0.2933,
|
| 261266 |
+
"step": 127250
|
| 261267 |
+
},
|
| 261268 |
+
{
|
| 261269 |
+
"epoch": 1017.54,
|
| 261270 |
+
"learning_rate": 7.977483974358975e-06,
|
| 261271 |
+
"loss": 0.3423,
|
| 261272 |
+
"step": 127255
|
| 261273 |
+
},
|
| 261274 |
+
{
|
| 261275 |
+
"epoch": 1017.58,
|
| 261276 |
+
"learning_rate": 7.977403846153846e-06,
|
| 261277 |
+
"loss": 0.636,
|
| 261278 |
+
"step": 127260
|
| 261279 |
+
},
|
| 261280 |
+
{
|
| 261281 |
+
"epoch": 1017.62,
|
| 261282 |
+
"learning_rate": 7.977323717948718e-06,
|
| 261283 |
+
"loss": 0.8939,
|
| 261284 |
+
"step": 127265
|
| 261285 |
+
},
|
| 261286 |
+
{
|
| 261287 |
+
"epoch": 1017.66,
|
| 261288 |
+
"learning_rate": 7.97724358974359e-06,
|
| 261289 |
+
"loss": 0.2902,
|
| 261290 |
+
"step": 127270
|
| 261291 |
+
},
|
| 261292 |
+
{
|
| 261293 |
+
"epoch": 1017.7,
|
| 261294 |
+
"learning_rate": 7.977163461538462e-06,
|
| 261295 |
+
"loss": 0.2895,
|
| 261296 |
+
"step": 127275
|
| 261297 |
+
},
|
| 261298 |
+
{
|
| 261299 |
+
"epoch": 1017.74,
|
| 261300 |
+
"learning_rate": 7.977083333333334e-06,
|
| 261301 |
+
"loss": 0.3549,
|
| 261302 |
+
"step": 127280
|
| 261303 |
+
},
|
| 261304 |
+
{
|
| 261305 |
+
"epoch": 1017.78,
|
| 261306 |
+
"learning_rate": 7.977003205128205e-06,
|
| 261307 |
+
"loss": 0.7446,
|
| 261308 |
+
"step": 127285
|
| 261309 |
+
},
|
| 261310 |
+
{
|
| 261311 |
+
"epoch": 1017.82,
|
| 261312 |
+
"learning_rate": 7.976923076923078e-06,
|
| 261313 |
+
"loss": 0.8723,
|
| 261314 |
+
"step": 127290
|
| 261315 |
+
},
|
| 261316 |
+
{
|
| 261317 |
+
"epoch": 1017.86,
|
| 261318 |
+
"learning_rate": 7.97684294871795e-06,
|
| 261319 |
+
"loss": 0.366,
|
| 261320 |
+
"step": 127295
|
| 261321 |
+
},
|
| 261322 |
+
{
|
| 261323 |
+
"epoch": 1017.9,
|
| 261324 |
+
"learning_rate": 7.97676282051282e-06,
|
| 261325 |
+
"loss": 0.2834,
|
| 261326 |
+
"step": 127300
|
| 261327 |
+
},
|
| 261328 |
+
{
|
| 261329 |
+
"epoch": 1017.94,
|
| 261330 |
+
"learning_rate": 7.976682692307694e-06,
|
| 261331 |
+
"loss": 0.5171,
|
| 261332 |
+
"step": 127305
|
| 261333 |
+
},
|
| 261334 |
+
{
|
| 261335 |
+
"epoch": 1017.98,
|
| 261336 |
+
"learning_rate": 7.976602564102565e-06,
|
| 261337 |
+
"loss": 0.754,
|
| 261338 |
+
"step": 127310
|
| 261339 |
+
},
|
| 261340 |
+
{
|
| 261341 |
+
"epoch": 1018.0,
|
| 261342 |
+
"eval_loss": 0.45526769757270813,
|
| 261343 |
+
"eval_runtime": 35.6393,
|
| 261344 |
+
"eval_samples_per_second": 23.598,
|
| 261345 |
+
"eval_steps_per_second": 0.758,
|
| 261346 |
+
"eval_wer": 0.18508140855736463,
|
| 261347 |
+
"step": 127312
|
| 261348 |
}
|
| 261349 |
],
|
| 261350 |
+
"max_steps": 625000,
|
| 261351 |
"num_train_epochs": 5000,
|
| 261352 |
+
"total_flos": 3.582563971308685e+20,
|
| 261353 |
"trial_name": null,
|
| 261354 |
"trial_params": null
|
| 261355 |
}
|
model-bin/finetune/base/{checkpoint-126689 → checkpoint-127312}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630150760.2881565/events.out.tfevents.1630150760.86bb0ddabf9b.4092.141
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:095dcf1ab0afdc359a9db9f65eb9d40ab61ba0174dddb16d9eec8f49bbf84f03
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630151150.1171627/events.out.tfevents.1630151150.86bb0ddabf9b.4092.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08ab4fbcb92cb5a41866a164ef988d957e600f50c11af4f07e19989b759eebbd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630151541.9840891/events.out.tfevents.1630151541.86bb0ddabf9b.4092.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:417ec6e2a2d39968cae6f4aa173a2a8967a5eb715e4a85c1897ae1798e649597
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630151929.4076536/events.out.tfevents.1630151929.86bb0ddabf9b.4092.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5254878b62ded00fa804510340156dc900ed49ee89d328f177190c04e8b77caa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630152323.041462/events.out.tfevents.1630152323.86bb0ddabf9b.4092.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a496ef544914a1fdb502b6e350fdc6c69eeed216c95e80fc33623f3b72ce3e9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630150760.86bb0ddabf9b.4092.140
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3481ddc6fe8733f703c18a8ff359eced165cb5682bf077a705b29e57b482d529
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630151150.86bb0ddabf9b.4092.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c68c4278c0f61ff4965befa70beccd3ba4bd56725eb8c2916e7a9f29df34bef2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630151541.86bb0ddabf9b.4092.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5aee1d6d4ab9a2cd5fb99ffa3b67bd3bcbad46cc2f6663dde04626e23f94786c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630151929.86bb0ddabf9b.4092.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cd372f6de219ef5c36d546f06292c34ad87d8af7573b5e260b8261b332c99c4a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630152323.86bb0ddabf9b.4092.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c43ca82bc199c61deab2c9f2e1f615d8243eb0b1bacefa233818c677f1a88aa
|
| 3 |
+
size 8622
|