"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629709823.8830478/events.out.tfevents.1629709823.74272264b15c.932.93 +3 -0
- model-bin/finetune/base/log/1629710466.9426787/events.out.tfevents.1629710466.74272264b15c.932.95 +3 -0
- model-bin/finetune/base/log/1629711119.662889/events.out.tfevents.1629711119.74272264b15c.932.97 +3 -0
- model-bin/finetune/base/log/1629711757.3439753/events.out.tfevents.1629711757.74272264b15c.932.99 +3 -0
- model-bin/finetune/base/log/1629712398.8804219/events.out.tfevents.1629712398.74272264b15c.932.101 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629709823.74272264b15c.932.92 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629710466.74272264b15c.932.94 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629711119.74272264b15c.932.96 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629711757.74272264b15c.932.98 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629712398.74272264b15c.932.100 +3 -0
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/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 sha256:0279d6274d99aff29875a66ade8497c784842aa2119d1baf7b9acc61b04cdc3a
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/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:c08812173259313450b45a26f7f54c70391884179523eaddfcddbe8366b0dcf7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61e9d1561fc2c97754a65ed1a57222faa352ea414b0fcaf1a6e647ffe648c057
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/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 sha256:0093f8ed51d9b7428b561d65ea23ba6345a4854dd7ee67dd3710814f98b18ba6
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/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 sha256:9e27558b55d0b0d4ff86a26e05fd1ec09d88eb36db5be0ded8883e0a02efd2e5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -155253,11 +155253,806 @@
|
|
| 155253 |
"eval_steps_per_second": 0.674,
|
| 155254 |
"eval_wer": 0.2129204326566899,
|
| 155255 |
"step": 43928
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 155256 |
}
|
| 155257 |
],
|
| 155258 |
-
"max_steps":
|
| 155259 |
"num_train_epochs": 5000,
|
| 155260 |
-
"total_flos": 1.
|
| 155261 |
"trial_name": null,
|
| 155262 |
"trial_params": null
|
| 155263 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
+
"epoch": 358.99598393574297,
|
| 5 |
+
"global_step": 44550,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 155253 |
"eval_steps_per_second": 0.674,
|
| 155254 |
"eval_wer": 0.2129204326566899,
|
| 155255 |
"step": 43928
|
| 155256 |
+
},
|
| 155257 |
+
{
|
| 155258 |
+
"epoch": 351.02,
|
| 155259 |
+
"learning_rate": 9.312291666666667e-06,
|
| 155260 |
+
"loss": 0.4421,
|
| 155261 |
+
"step": 43930
|
| 155262 |
+
},
|
| 155263 |
+
{
|
| 155264 |
+
"epoch": 351.06,
|
| 155265 |
+
"learning_rate": 9.31221153846154e-06,
|
| 155266 |
+
"loss": 0.3264,
|
| 155267 |
+
"step": 43935
|
| 155268 |
+
},
|
| 155269 |
+
{
|
| 155270 |
+
"epoch": 351.1,
|
| 155271 |
+
"learning_rate": 9.312131410256412e-06,
|
| 155272 |
+
"loss": 0.4353,
|
| 155273 |
+
"step": 43940
|
| 155274 |
+
},
|
| 155275 |
+
{
|
| 155276 |
+
"epoch": 351.14,
|
| 155277 |
+
"learning_rate": 9.312051282051283e-06,
|
| 155278 |
+
"loss": 0.3769,
|
| 155279 |
+
"step": 43945
|
| 155280 |
+
},
|
| 155281 |
+
{
|
| 155282 |
+
"epoch": 351.18,
|
| 155283 |
+
"learning_rate": 9.311971153846154e-06,
|
| 155284 |
+
"loss": 0.6903,
|
| 155285 |
+
"step": 43950
|
| 155286 |
+
},
|
| 155287 |
+
{
|
| 155288 |
+
"epoch": 351.22,
|
| 155289 |
+
"learning_rate": 9.311891025641027e-06,
|
| 155290 |
+
"loss": 1.1204,
|
| 155291 |
+
"step": 43955
|
| 155292 |
+
},
|
| 155293 |
+
{
|
| 155294 |
+
"epoch": 351.26,
|
| 155295 |
+
"learning_rate": 9.311810897435897e-06,
|
| 155296 |
+
"loss": 0.3226,
|
| 155297 |
+
"step": 43960
|
| 155298 |
+
},
|
| 155299 |
+
{
|
| 155300 |
+
"epoch": 351.3,
|
| 155301 |
+
"learning_rate": 9.31173076923077e-06,
|
| 155302 |
+
"loss": 0.3742,
|
| 155303 |
+
"step": 43965
|
| 155304 |
+
},
|
| 155305 |
+
{
|
| 155306 |
+
"epoch": 351.34,
|
| 155307 |
+
"learning_rate": 9.311650641025642e-06,
|
| 155308 |
+
"loss": 0.4141,
|
| 155309 |
+
"step": 43970
|
| 155310 |
+
},
|
| 155311 |
+
{
|
| 155312 |
+
"epoch": 351.38,
|
| 155313 |
+
"learning_rate": 9.311570512820513e-06,
|
| 155314 |
+
"loss": 0.6841,
|
| 155315 |
+
"step": 43975
|
| 155316 |
+
},
|
| 155317 |
+
{
|
| 155318 |
+
"epoch": 351.42,
|
| 155319 |
+
"learning_rate": 9.311490384615386e-06,
|
| 155320 |
+
"loss": 1.1666,
|
| 155321 |
+
"step": 43980
|
| 155322 |
+
},
|
| 155323 |
+
{
|
| 155324 |
+
"epoch": 351.46,
|
| 155325 |
+
"learning_rate": 9.311410256410257e-06,
|
| 155326 |
+
"loss": 0.3489,
|
| 155327 |
+
"step": 43985
|
| 155328 |
+
},
|
| 155329 |
+
{
|
| 155330 |
+
"epoch": 351.5,
|
| 155331 |
+
"learning_rate": 9.311330128205129e-06,
|
| 155332 |
+
"loss": 0.4329,
|
| 155333 |
+
"step": 43990
|
| 155334 |
+
},
|
| 155335 |
+
{
|
| 155336 |
+
"epoch": 351.54,
|
| 155337 |
+
"learning_rate": 9.31125e-06,
|
| 155338 |
+
"loss": 0.4582,
|
| 155339 |
+
"step": 43995
|
| 155340 |
+
},
|
| 155341 |
+
{
|
| 155342 |
+
"epoch": 351.58,
|
| 155343 |
+
"learning_rate": 9.311169871794873e-06,
|
| 155344 |
+
"loss": 0.8117,
|
| 155345 |
+
"step": 44000
|
| 155346 |
+
},
|
| 155347 |
+
{
|
| 155348 |
+
"epoch": 351.62,
|
| 155349 |
+
"learning_rate": 9.311089743589744e-06,
|
| 155350 |
+
"loss": 1.0366,
|
| 155351 |
+
"step": 44005
|
| 155352 |
+
},
|
| 155353 |
+
{
|
| 155354 |
+
"epoch": 351.66,
|
| 155355 |
+
"learning_rate": 9.311009615384616e-06,
|
| 155356 |
+
"loss": 0.3585,
|
| 155357 |
+
"step": 44010
|
| 155358 |
+
},
|
| 155359 |
+
{
|
| 155360 |
+
"epoch": 351.7,
|
| 155361 |
+
"learning_rate": 9.310929487179487e-06,
|
| 155362 |
+
"loss": 0.356,
|
| 155363 |
+
"step": 44015
|
| 155364 |
+
},
|
| 155365 |
+
{
|
| 155366 |
+
"epoch": 351.74,
|
| 155367 |
+
"learning_rate": 9.31084935897436e-06,
|
| 155368 |
+
"loss": 0.3427,
|
| 155369 |
+
"step": 44020
|
| 155370 |
+
},
|
| 155371 |
+
{
|
| 155372 |
+
"epoch": 351.78,
|
| 155373 |
+
"learning_rate": 9.310769230769232e-06,
|
| 155374 |
+
"loss": 1.0205,
|
| 155375 |
+
"step": 44025
|
| 155376 |
+
},
|
| 155377 |
+
{
|
| 155378 |
+
"epoch": 351.82,
|
| 155379 |
+
"learning_rate": 9.310689102564103e-06,
|
| 155380 |
+
"loss": 1.1595,
|
| 155381 |
+
"step": 44030
|
| 155382 |
+
},
|
| 155383 |
+
{
|
| 155384 |
+
"epoch": 351.86,
|
| 155385 |
+
"learning_rate": 9.310608974358976e-06,
|
| 155386 |
+
"loss": 0.3654,
|
| 155387 |
+
"step": 44035
|
| 155388 |
+
},
|
| 155389 |
+
{
|
| 155390 |
+
"epoch": 351.9,
|
| 155391 |
+
"learning_rate": 9.310528846153847e-06,
|
| 155392 |
+
"loss": 0.3289,
|
| 155393 |
+
"step": 44040
|
| 155394 |
+
},
|
| 155395 |
+
{
|
| 155396 |
+
"epoch": 351.94,
|
| 155397 |
+
"learning_rate": 9.310448717948719e-06,
|
| 155398 |
+
"loss": 0.3902,
|
| 155399 |
+
"step": 44045
|
| 155400 |
+
},
|
| 155401 |
+
{
|
| 155402 |
+
"epoch": 351.98,
|
| 155403 |
+
"learning_rate": 9.31036858974359e-06,
|
| 155404 |
+
"loss": 0.8672,
|
| 155405 |
+
"step": 44050
|
| 155406 |
+
},
|
| 155407 |
+
{
|
| 155408 |
+
"epoch": 352.0,
|
| 155409 |
+
"eval_loss": 0.41955453157424927,
|
| 155410 |
+
"eval_runtime": 40.7711,
|
| 155411 |
+
"eval_samples_per_second": 20.676,
|
| 155412 |
+
"eval_steps_per_second": 0.662,
|
| 155413 |
+
"eval_wer": 0.20655926352128884,
|
| 155414 |
+
"step": 44053
|
| 155415 |
+
},
|
| 155416 |
+
{
|
| 155417 |
+
"epoch": 355.02,
|
| 155418 |
+
"learning_rate": 9.310288461538463e-06,
|
| 155419 |
+
"loss": 0.4501,
|
| 155420 |
+
"step": 44055
|
| 155421 |
+
},
|
| 155422 |
+
{
|
| 155423 |
+
"epoch": 355.06,
|
| 155424 |
+
"learning_rate": 9.310208333333334e-06,
|
| 155425 |
+
"loss": 0.3245,
|
| 155426 |
+
"step": 44060
|
| 155427 |
+
},
|
| 155428 |
+
{
|
| 155429 |
+
"epoch": 355.1,
|
| 155430 |
+
"learning_rate": 9.310128205128206e-06,
|
| 155431 |
+
"loss": 0.3245,
|
| 155432 |
+
"step": 44065
|
| 155433 |
+
},
|
| 155434 |
+
{
|
| 155435 |
+
"epoch": 355.14,
|
| 155436 |
+
"learning_rate": 9.310048076923077e-06,
|
| 155437 |
+
"loss": 0.4605,
|
| 155438 |
+
"step": 44070
|
| 155439 |
+
},
|
| 155440 |
+
{
|
| 155441 |
+
"epoch": 355.18,
|
| 155442 |
+
"learning_rate": 9.30996794871795e-06,
|
| 155443 |
+
"loss": 0.6796,
|
| 155444 |
+
"step": 44075
|
| 155445 |
+
},
|
| 155446 |
+
{
|
| 155447 |
+
"epoch": 355.22,
|
| 155448 |
+
"learning_rate": 9.309887820512822e-06,
|
| 155449 |
+
"loss": 1.2864,
|
| 155450 |
+
"step": 44080
|
| 155451 |
+
},
|
| 155452 |
+
{
|
| 155453 |
+
"epoch": 355.26,
|
| 155454 |
+
"learning_rate": 9.309807692307693e-06,
|
| 155455 |
+
"loss": 0.3521,
|
| 155456 |
+
"step": 44085
|
| 155457 |
+
},
|
| 155458 |
+
{
|
| 155459 |
+
"epoch": 355.3,
|
| 155460 |
+
"learning_rate": 9.309727564102566e-06,
|
| 155461 |
+
"loss": 0.474,
|
| 155462 |
+
"step": 44090
|
| 155463 |
+
},
|
| 155464 |
+
{
|
| 155465 |
+
"epoch": 355.34,
|
| 155466 |
+
"learning_rate": 9.309647435897436e-06,
|
| 155467 |
+
"loss": 0.4906,
|
| 155468 |
+
"step": 44095
|
| 155469 |
+
},
|
| 155470 |
+
{
|
| 155471 |
+
"epoch": 355.38,
|
| 155472 |
+
"learning_rate": 9.309567307692309e-06,
|
| 155473 |
+
"loss": 0.8119,
|
| 155474 |
+
"step": 44100
|
| 155475 |
+
},
|
| 155476 |
+
{
|
| 155477 |
+
"epoch": 355.42,
|
| 155478 |
+
"learning_rate": 9.30948717948718e-06,
|
| 155479 |
+
"loss": 1.1036,
|
| 155480 |
+
"step": 44105
|
| 155481 |
+
},
|
| 155482 |
+
{
|
| 155483 |
+
"epoch": 355.46,
|
| 155484 |
+
"learning_rate": 9.309407051282051e-06,
|
| 155485 |
+
"loss": 0.3622,
|
| 155486 |
+
"step": 44110
|
| 155487 |
+
},
|
| 155488 |
+
{
|
| 155489 |
+
"epoch": 355.5,
|
| 155490 |
+
"learning_rate": 9.309326923076923e-06,
|
| 155491 |
+
"loss": 0.3163,
|
| 155492 |
+
"step": 44115
|
| 155493 |
+
},
|
| 155494 |
+
{
|
| 155495 |
+
"epoch": 355.54,
|
| 155496 |
+
"learning_rate": 9.309246794871796e-06,
|
| 155497 |
+
"loss": 0.4837,
|
| 155498 |
+
"step": 44120
|
| 155499 |
+
},
|
| 155500 |
+
{
|
| 155501 |
+
"epoch": 355.58,
|
| 155502 |
+
"learning_rate": 9.309166666666667e-06,
|
| 155503 |
+
"loss": 0.7545,
|
| 155504 |
+
"step": 44125
|
| 155505 |
+
},
|
| 155506 |
+
{
|
| 155507 |
+
"epoch": 355.62,
|
| 155508 |
+
"learning_rate": 9.309086538461539e-06,
|
| 155509 |
+
"loss": 1.0636,
|
| 155510 |
+
"step": 44130
|
| 155511 |
+
},
|
| 155512 |
+
{
|
| 155513 |
+
"epoch": 355.66,
|
| 155514 |
+
"learning_rate": 9.309006410256412e-06,
|
| 155515 |
+
"loss": 0.3603,
|
| 155516 |
+
"step": 44135
|
| 155517 |
+
},
|
| 155518 |
+
{
|
| 155519 |
+
"epoch": 355.7,
|
| 155520 |
+
"learning_rate": 9.308926282051283e-06,
|
| 155521 |
+
"loss": 0.3325,
|
| 155522 |
+
"step": 44140
|
| 155523 |
+
},
|
| 155524 |
+
{
|
| 155525 |
+
"epoch": 355.74,
|
| 155526 |
+
"learning_rate": 9.308846153846154e-06,
|
| 155527 |
+
"loss": 0.4486,
|
| 155528 |
+
"step": 44145
|
| 155529 |
+
},
|
| 155530 |
+
{
|
| 155531 |
+
"epoch": 355.78,
|
| 155532 |
+
"learning_rate": 9.308766025641026e-06,
|
| 155533 |
+
"loss": 0.7067,
|
| 155534 |
+
"step": 44150
|
| 155535 |
+
},
|
| 155536 |
+
{
|
| 155537 |
+
"epoch": 355.82,
|
| 155538 |
+
"learning_rate": 9.308685897435899e-06,
|
| 155539 |
+
"loss": 1.1456,
|
| 155540 |
+
"step": 44155
|
| 155541 |
+
},
|
| 155542 |
+
{
|
| 155543 |
+
"epoch": 355.86,
|
| 155544 |
+
"learning_rate": 9.30860576923077e-06,
|
| 155545 |
+
"loss": 0.3518,
|
| 155546 |
+
"step": 44160
|
| 155547 |
+
},
|
| 155548 |
+
{
|
| 155549 |
+
"epoch": 355.9,
|
| 155550 |
+
"learning_rate": 9.308525641025641e-06,
|
| 155551 |
+
"loss": 0.3302,
|
| 155552 |
+
"step": 44165
|
| 155553 |
+
},
|
| 155554 |
+
{
|
| 155555 |
+
"epoch": 355.94,
|
| 155556 |
+
"learning_rate": 9.308445512820513e-06,
|
| 155557 |
+
"loss": 0.507,
|
| 155558 |
+
"step": 44170
|
| 155559 |
+
},
|
| 155560 |
+
{
|
| 155561 |
+
"epoch": 355.98,
|
| 155562 |
+
"learning_rate": 9.308365384615386e-06,
|
| 155563 |
+
"loss": 0.7606,
|
| 155564 |
+
"step": 44175
|
| 155565 |
+
},
|
| 155566 |
+
{
|
| 155567 |
+
"epoch": 356.0,
|
| 155568 |
+
"eval_loss": 0.39258870482444763,
|
| 155569 |
+
"eval_runtime": 40.2688,
|
| 155570 |
+
"eval_samples_per_second": 20.959,
|
| 155571 |
+
"eval_steps_per_second": 0.67,
|
| 155572 |
+
"eval_wer": 0.20093727469358327,
|
| 155573 |
+
"step": 44177
|
| 155574 |
+
},
|
| 155575 |
+
{
|
| 155576 |
+
"epoch": 356.02,
|
| 155577 |
+
"learning_rate": 9.308285256410257e-06,
|
| 155578 |
+
"loss": 0.3451,
|
| 155579 |
+
"step": 44180
|
| 155580 |
+
},
|
| 155581 |
+
{
|
| 155582 |
+
"epoch": 356.06,
|
| 155583 |
+
"learning_rate": 9.308205128205129e-06,
|
| 155584 |
+
"loss": 0.3565,
|
| 155585 |
+
"step": 44185
|
| 155586 |
+
},
|
| 155587 |
+
{
|
| 155588 |
+
"epoch": 356.1,
|
| 155589 |
+
"learning_rate": 9.308125000000002e-06,
|
| 155590 |
+
"loss": 0.3717,
|
| 155591 |
+
"step": 44190
|
| 155592 |
+
},
|
| 155593 |
+
{
|
| 155594 |
+
"epoch": 356.15,
|
| 155595 |
+
"learning_rate": 9.308044871794873e-06,
|
| 155596 |
+
"loss": 0.4751,
|
| 155597 |
+
"step": 44195
|
| 155598 |
+
},
|
| 155599 |
+
{
|
| 155600 |
+
"epoch": 356.19,
|
| 155601 |
+
"learning_rate": 9.307964743589744e-06,
|
| 155602 |
+
"loss": 0.8632,
|
| 155603 |
+
"step": 44200
|
| 155604 |
+
},
|
| 155605 |
+
{
|
| 155606 |
+
"epoch": 356.23,
|
| 155607 |
+
"learning_rate": 9.307884615384616e-06,
|
| 155608 |
+
"loss": 0.9419,
|
| 155609 |
+
"step": 44205
|
| 155610 |
+
},
|
| 155611 |
+
{
|
| 155612 |
+
"epoch": 356.27,
|
| 155613 |
+
"learning_rate": 9.307804487179489e-06,
|
| 155614 |
+
"loss": 0.3371,
|
| 155615 |
+
"step": 44210
|
| 155616 |
+
},
|
| 155617 |
+
{
|
| 155618 |
+
"epoch": 356.31,
|
| 155619 |
+
"learning_rate": 9.30772435897436e-06,
|
| 155620 |
+
"loss": 0.393,
|
| 155621 |
+
"step": 44215
|
| 155622 |
+
},
|
| 155623 |
+
{
|
| 155624 |
+
"epoch": 356.35,
|
| 155625 |
+
"learning_rate": 9.307644230769231e-06,
|
| 155626 |
+
"loss": 0.4187,
|
| 155627 |
+
"step": 44220
|
| 155628 |
+
},
|
| 155629 |
+
{
|
| 155630 |
+
"epoch": 356.39,
|
| 155631 |
+
"learning_rate": 9.307564102564103e-06,
|
| 155632 |
+
"loss": 0.904,
|
| 155633 |
+
"step": 44225
|
| 155634 |
+
},
|
| 155635 |
+
{
|
| 155636 |
+
"epoch": 356.43,
|
| 155637 |
+
"learning_rate": 9.307483974358976e-06,
|
| 155638 |
+
"loss": 1.0664,
|
| 155639 |
+
"step": 44230
|
| 155640 |
+
},
|
| 155641 |
+
{
|
| 155642 |
+
"epoch": 356.47,
|
| 155643 |
+
"learning_rate": 9.307403846153847e-06,
|
| 155644 |
+
"loss": 0.3177,
|
| 155645 |
+
"step": 44235
|
| 155646 |
+
},
|
| 155647 |
+
{
|
| 155648 |
+
"epoch": 356.51,
|
| 155649 |
+
"learning_rate": 9.307323717948719e-06,
|
| 155650 |
+
"loss": 0.3885,
|
| 155651 |
+
"step": 44240
|
| 155652 |
+
},
|
| 155653 |
+
{
|
| 155654 |
+
"epoch": 356.55,
|
| 155655 |
+
"learning_rate": 9.307243589743592e-06,
|
| 155656 |
+
"loss": 0.4047,
|
| 155657 |
+
"step": 44245
|
| 155658 |
+
},
|
| 155659 |
+
{
|
| 155660 |
+
"epoch": 356.59,
|
| 155661 |
+
"learning_rate": 9.307163461538461e-06,
|
| 155662 |
+
"loss": 0.8789,
|
| 155663 |
+
"step": 44250
|
| 155664 |
+
},
|
| 155665 |
+
{
|
| 155666 |
+
"epoch": 356.63,
|
| 155667 |
+
"learning_rate": 9.307083333333334e-06,
|
| 155668 |
+
"loss": 0.9838,
|
| 155669 |
+
"step": 44255
|
| 155670 |
+
},
|
| 155671 |
+
{
|
| 155672 |
+
"epoch": 356.67,
|
| 155673 |
+
"learning_rate": 9.307003205128206e-06,
|
| 155674 |
+
"loss": 0.406,
|
| 155675 |
+
"step": 44260
|
| 155676 |
+
},
|
| 155677 |
+
{
|
| 155678 |
+
"epoch": 356.71,
|
| 155679 |
+
"learning_rate": 9.306923076923077e-06,
|
| 155680 |
+
"loss": 0.3532,
|
| 155681 |
+
"step": 44265
|
| 155682 |
+
},
|
| 155683 |
+
{
|
| 155684 |
+
"epoch": 356.75,
|
| 155685 |
+
"learning_rate": 9.306842948717948e-06,
|
| 155686 |
+
"loss": 0.4139,
|
| 155687 |
+
"step": 44270
|
| 155688 |
+
},
|
| 155689 |
+
{
|
| 155690 |
+
"epoch": 356.79,
|
| 155691 |
+
"learning_rate": 9.306762820512821e-06,
|
| 155692 |
+
"loss": 0.8771,
|
| 155693 |
+
"step": 44275
|
| 155694 |
+
},
|
| 155695 |
+
{
|
| 155696 |
+
"epoch": 356.83,
|
| 155697 |
+
"learning_rate": 9.306682692307693e-06,
|
| 155698 |
+
"loss": 1.0188,
|
| 155699 |
+
"step": 44280
|
| 155700 |
+
},
|
| 155701 |
+
{
|
| 155702 |
+
"epoch": 356.87,
|
| 155703 |
+
"learning_rate": 9.306602564102564e-06,
|
| 155704 |
+
"loss": 0.3093,
|
| 155705 |
+
"step": 44285
|
| 155706 |
+
},
|
| 155707 |
+
{
|
| 155708 |
+
"epoch": 356.91,
|
| 155709 |
+
"learning_rate": 9.306522435897437e-06,
|
| 155710 |
+
"loss": 0.3349,
|
| 155711 |
+
"step": 44290
|
| 155712 |
+
},
|
| 155713 |
+
{
|
| 155714 |
+
"epoch": 356.95,
|
| 155715 |
+
"learning_rate": 9.306442307692309e-06,
|
| 155716 |
+
"loss": 0.5235,
|
| 155717 |
+
"step": 44295
|
| 155718 |
+
},
|
| 155719 |
+
{
|
| 155720 |
+
"epoch": 356.99,
|
| 155721 |
+
"learning_rate": 9.30636217948718e-06,
|
| 155722 |
+
"loss": 1.0774,
|
| 155723 |
+
"step": 44300
|
| 155724 |
+
},
|
| 155725 |
+
{
|
| 155726 |
+
"epoch": 357.0,
|
| 155727 |
+
"eval_loss": 0.39617443084716797,
|
| 155728 |
+
"eval_runtime": 39.1797,
|
| 155729 |
+
"eval_samples_per_second": 21.542,
|
| 155730 |
+
"eval_steps_per_second": 0.689,
|
| 155731 |
+
"eval_wer": 0.20717247399432603,
|
| 155732 |
+
"step": 44301
|
| 155733 |
+
},
|
| 155734 |
+
{
|
| 155735 |
+
"epoch": 354.03,
|
| 155736 |
+
"learning_rate": 9.306282051282051e-06,
|
| 155737 |
+
"loss": 0.3491,
|
| 155738 |
+
"step": 44305
|
| 155739 |
+
},
|
| 155740 |
+
{
|
| 155741 |
+
"epoch": 354.07,
|
| 155742 |
+
"learning_rate": 9.306201923076924e-06,
|
| 155743 |
+
"loss": 0.4052,
|
| 155744 |
+
"step": 44310
|
| 155745 |
+
},
|
| 155746 |
+
{
|
| 155747 |
+
"epoch": 354.11,
|
| 155748 |
+
"learning_rate": 9.306121794871796e-06,
|
| 155749 |
+
"loss": 0.4445,
|
| 155750 |
+
"step": 44315
|
| 155751 |
+
},
|
| 155752 |
+
{
|
| 155753 |
+
"epoch": 354.15,
|
| 155754 |
+
"learning_rate": 9.306041666666667e-06,
|
| 155755 |
+
"loss": 0.473,
|
| 155756 |
+
"step": 44320
|
| 155757 |
+
},
|
| 155758 |
+
{
|
| 155759 |
+
"epoch": 354.19,
|
| 155760 |
+
"learning_rate": 9.305961538461538e-06,
|
| 155761 |
+
"loss": 0.9977,
|
| 155762 |
+
"step": 44325
|
| 155763 |
+
},
|
| 155764 |
+
{
|
| 155765 |
+
"epoch": 354.23,
|
| 155766 |
+
"learning_rate": 9.305881410256412e-06,
|
| 155767 |
+
"loss": 0.7714,
|
| 155768 |
+
"step": 44330
|
| 155769 |
+
},
|
| 155770 |
+
{
|
| 155771 |
+
"epoch": 354.27,
|
| 155772 |
+
"learning_rate": 9.305801282051283e-06,
|
| 155773 |
+
"loss": 0.3936,
|
| 155774 |
+
"step": 44335
|
| 155775 |
+
},
|
| 155776 |
+
{
|
| 155777 |
+
"epoch": 354.31,
|
| 155778 |
+
"learning_rate": 9.305721153846154e-06,
|
| 155779 |
+
"loss": 0.3338,
|
| 155780 |
+
"step": 44340
|
| 155781 |
+
},
|
| 155782 |
+
{
|
| 155783 |
+
"epoch": 354.35,
|
| 155784 |
+
"learning_rate": 9.305641025641027e-06,
|
| 155785 |
+
"loss": 0.5747,
|
| 155786 |
+
"step": 44345
|
| 155787 |
+
},
|
| 155788 |
+
{
|
| 155789 |
+
"epoch": 354.39,
|
| 155790 |
+
"learning_rate": 9.305560897435899e-06,
|
| 155791 |
+
"loss": 1.1114,
|
| 155792 |
+
"step": 44350
|
| 155793 |
+
},
|
| 155794 |
+
{
|
| 155795 |
+
"epoch": 354.43,
|
| 155796 |
+
"learning_rate": 9.30548076923077e-06,
|
| 155797 |
+
"loss": 0.7386,
|
| 155798 |
+
"step": 44355
|
| 155799 |
+
},
|
| 155800 |
+
{
|
| 155801 |
+
"epoch": 354.47,
|
| 155802 |
+
"learning_rate": 9.305400641025641e-06,
|
| 155803 |
+
"loss": 0.3316,
|
| 155804 |
+
"step": 44360
|
| 155805 |
+
},
|
| 155806 |
+
{
|
| 155807 |
+
"epoch": 354.51,
|
| 155808 |
+
"learning_rate": 9.305320512820514e-06,
|
| 155809 |
+
"loss": 0.4235,
|
| 155810 |
+
"step": 44365
|
| 155811 |
+
},
|
| 155812 |
+
{
|
| 155813 |
+
"epoch": 354.55,
|
| 155814 |
+
"learning_rate": 9.305240384615384e-06,
|
| 155815 |
+
"loss": 0.478,
|
| 155816 |
+
"step": 44370
|
| 155817 |
+
},
|
| 155818 |
+
{
|
| 155819 |
+
"epoch": 354.59,
|
| 155820 |
+
"learning_rate": 9.305160256410257e-06,
|
| 155821 |
+
"loss": 1.0239,
|
| 155822 |
+
"step": 44375
|
| 155823 |
+
},
|
| 155824 |
+
{
|
| 155825 |
+
"epoch": 354.63,
|
| 155826 |
+
"learning_rate": 9.30508012820513e-06,
|
| 155827 |
+
"loss": 0.8528,
|
| 155828 |
+
"step": 44380
|
| 155829 |
+
},
|
| 155830 |
+
{
|
| 155831 |
+
"epoch": 354.67,
|
| 155832 |
+
"learning_rate": 9.305000000000002e-06,
|
| 155833 |
+
"loss": 0.3236,
|
| 155834 |
+
"step": 44385
|
| 155835 |
+
},
|
| 155836 |
+
{
|
| 155837 |
+
"epoch": 354.71,
|
| 155838 |
+
"learning_rate": 9.304919871794873e-06,
|
| 155839 |
+
"loss": 0.3199,
|
| 155840 |
+
"step": 44390
|
| 155841 |
+
},
|
| 155842 |
+
{
|
| 155843 |
+
"epoch": 354.75,
|
| 155844 |
+
"learning_rate": 9.304839743589744e-06,
|
| 155845 |
+
"loss": 0.5057,
|
| 155846 |
+
"step": 44395
|
| 155847 |
+
},
|
| 155848 |
+
{
|
| 155849 |
+
"epoch": 354.79,
|
| 155850 |
+
"learning_rate": 9.304759615384617e-06,
|
| 155851 |
+
"loss": 0.985,
|
| 155852 |
+
"step": 44400
|
| 155853 |
+
},
|
| 155854 |
+
{
|
| 155855 |
+
"epoch": 354.83,
|
| 155856 |
+
"learning_rate": 9.304679487179487e-06,
|
| 155857 |
+
"loss": 0.7583,
|
| 155858 |
+
"step": 44405
|
| 155859 |
+
},
|
| 155860 |
+
{
|
| 155861 |
+
"epoch": 354.87,
|
| 155862 |
+
"learning_rate": 9.30459935897436e-06,
|
| 155863 |
+
"loss": 0.3995,
|
| 155864 |
+
"step": 44410
|
| 155865 |
+
},
|
| 155866 |
+
{
|
| 155867 |
+
"epoch": 354.91,
|
| 155868 |
+
"learning_rate": 9.304519230769231e-06,
|
| 155869 |
+
"loss": 0.3757,
|
| 155870 |
+
"step": 44415
|
| 155871 |
+
},
|
| 155872 |
+
{
|
| 155873 |
+
"epoch": 354.95,
|
| 155874 |
+
"learning_rate": 9.304439102564103e-06,
|
| 155875 |
+
"loss": 0.4814,
|
| 155876 |
+
"step": 44420
|
| 155877 |
+
},
|
| 155878 |
+
{
|
| 155879 |
+
"epoch": 354.99,
|
| 155880 |
+
"learning_rate": 9.304358974358974e-06,
|
| 155881 |
+
"loss": 1.0594,
|
| 155882 |
+
"step": 44425
|
| 155883 |
+
},
|
| 155884 |
+
{
|
| 155885 |
+
"epoch": 355.0,
|
| 155886 |
+
"eval_loss": 0.3944911062717438,
|
| 155887 |
+
"eval_runtime": 40.0166,
|
| 155888 |
+
"eval_samples_per_second": 21.091,
|
| 155889 |
+
"eval_steps_per_second": 0.675,
|
| 155890 |
+
"eval_wer": 0.21263776634827333,
|
| 155891 |
+
"step": 44426
|
| 155892 |
+
},
|
| 155893 |
+
{
|
| 155894 |
+
"epoch": 358.03,
|
| 155895 |
+
"learning_rate": 9.304278846153847e-06,
|
| 155896 |
+
"loss": 0.3816,
|
| 155897 |
+
"step": 44430
|
| 155898 |
+
},
|
| 155899 |
+
{
|
| 155900 |
+
"epoch": 358.07,
|
| 155901 |
+
"learning_rate": 9.304198717948719e-06,
|
| 155902 |
+
"loss": 0.2979,
|
| 155903 |
+
"step": 44435
|
| 155904 |
+
},
|
| 155905 |
+
{
|
| 155906 |
+
"epoch": 358.11,
|
| 155907 |
+
"learning_rate": 9.30411858974359e-06,
|
| 155908 |
+
"loss": 0.3596,
|
| 155909 |
+
"step": 44440
|
| 155910 |
+
},
|
| 155911 |
+
{
|
| 155912 |
+
"epoch": 358.15,
|
| 155913 |
+
"learning_rate": 9.304038461538463e-06,
|
| 155914 |
+
"loss": 0.4234,
|
| 155915 |
+
"step": 44445
|
| 155916 |
+
},
|
| 155917 |
+
{
|
| 155918 |
+
"epoch": 358.19,
|
| 155919 |
+
"learning_rate": 9.303958333333334e-06,
|
| 155920 |
+
"loss": 1.0441,
|
| 155921 |
+
"step": 44450
|
| 155922 |
+
},
|
| 155923 |
+
{
|
| 155924 |
+
"epoch": 358.23,
|
| 155925 |
+
"learning_rate": 9.303878205128206e-06,
|
| 155926 |
+
"loss": 0.614,
|
| 155927 |
+
"step": 44455
|
| 155928 |
+
},
|
| 155929 |
+
{
|
| 155930 |
+
"epoch": 358.27,
|
| 155931 |
+
"learning_rate": 9.303798076923077e-06,
|
| 155932 |
+
"loss": 0.3835,
|
| 155933 |
+
"step": 44460
|
| 155934 |
+
},
|
| 155935 |
+
{
|
| 155936 |
+
"epoch": 358.31,
|
| 155937 |
+
"learning_rate": 9.30371794871795e-06,
|
| 155938 |
+
"loss": 0.396,
|
| 155939 |
+
"step": 44465
|
| 155940 |
+
},
|
| 155941 |
+
{
|
| 155942 |
+
"epoch": 358.35,
|
| 155943 |
+
"learning_rate": 9.303637820512821e-06,
|
| 155944 |
+
"loss": 0.4357,
|
| 155945 |
+
"step": 44470
|
| 155946 |
+
},
|
| 155947 |
+
{
|
| 155948 |
+
"epoch": 358.39,
|
| 155949 |
+
"learning_rate": 9.303557692307693e-06,
|
| 155950 |
+
"loss": 0.8927,
|
| 155951 |
+
"step": 44475
|
| 155952 |
+
},
|
| 155953 |
+
{
|
| 155954 |
+
"epoch": 358.43,
|
| 155955 |
+
"learning_rate": 9.303477564102566e-06,
|
| 155956 |
+
"loss": 0.6783,
|
| 155957 |
+
"step": 44480
|
| 155958 |
+
},
|
| 155959 |
+
{
|
| 155960 |
+
"epoch": 358.47,
|
| 155961 |
+
"learning_rate": 9.303397435897437e-06,
|
| 155962 |
+
"loss": 0.3421,
|
| 155963 |
+
"step": 44485
|
| 155964 |
+
},
|
| 155965 |
+
{
|
| 155966 |
+
"epoch": 358.51,
|
| 155967 |
+
"learning_rate": 9.303317307692309e-06,
|
| 155968 |
+
"loss": 0.3063,
|
| 155969 |
+
"step": 44490
|
| 155970 |
+
},
|
| 155971 |
+
{
|
| 155972 |
+
"epoch": 358.55,
|
| 155973 |
+
"learning_rate": 9.30323717948718e-06,
|
| 155974 |
+
"loss": 0.5544,
|
| 155975 |
+
"step": 44495
|
| 155976 |
+
},
|
| 155977 |
+
{
|
| 155978 |
+
"epoch": 358.59,
|
| 155979 |
+
"learning_rate": 9.303157051282053e-06,
|
| 155980 |
+
"loss": 1.0348,
|
| 155981 |
+
"step": 44500
|
| 155982 |
+
},
|
| 155983 |
+
{
|
| 155984 |
+
"epoch": 358.63,
|
| 155985 |
+
"learning_rate": 9.303076923076924e-06,
|
| 155986 |
+
"loss": 0.7595,
|
| 155987 |
+
"step": 44505
|
| 155988 |
+
},
|
| 155989 |
+
{
|
| 155990 |
+
"epoch": 358.67,
|
| 155991 |
+
"learning_rate": 9.302996794871796e-06,
|
| 155992 |
+
"loss": 0.3285,
|
| 155993 |
+
"step": 44510
|
| 155994 |
+
},
|
| 155995 |
+
{
|
| 155996 |
+
"epoch": 358.71,
|
| 155997 |
+
"learning_rate": 9.302916666666667e-06,
|
| 155998 |
+
"loss": 0.3316,
|
| 155999 |
+
"step": 44515
|
| 156000 |
+
},
|
| 156001 |
+
{
|
| 156002 |
+
"epoch": 358.76,
|
| 156003 |
+
"learning_rate": 9.30283653846154e-06,
|
| 156004 |
+
"loss": 0.5228,
|
| 156005 |
+
"step": 44520
|
| 156006 |
+
},
|
| 156007 |
+
{
|
| 156008 |
+
"epoch": 358.8,
|
| 156009 |
+
"learning_rate": 9.30275641025641e-06,
|
| 156010 |
+
"loss": 0.968,
|
| 156011 |
+
"step": 44525
|
| 156012 |
+
},
|
| 156013 |
+
{
|
| 156014 |
+
"epoch": 358.84,
|
| 156015 |
+
"learning_rate": 9.302676282051283e-06,
|
| 156016 |
+
"loss": 0.7137,
|
| 156017 |
+
"step": 44530
|
| 156018 |
+
},
|
| 156019 |
+
{
|
| 156020 |
+
"epoch": 358.88,
|
| 156021 |
+
"learning_rate": 9.302596153846156e-06,
|
| 156022 |
+
"loss": 0.274,
|
| 156023 |
+
"step": 44535
|
| 156024 |
+
},
|
| 156025 |
+
{
|
| 156026 |
+
"epoch": 358.92,
|
| 156027 |
+
"learning_rate": 9.302516025641026e-06,
|
| 156028 |
+
"loss": 0.3549,
|
| 156029 |
+
"step": 44540
|
| 156030 |
+
},
|
| 156031 |
+
{
|
| 156032 |
+
"epoch": 358.96,
|
| 156033 |
+
"learning_rate": 9.302435897435899e-06,
|
| 156034 |
+
"loss": 0.451,
|
| 156035 |
+
"step": 44545
|
| 156036 |
+
},
|
| 156037 |
+
{
|
| 156038 |
+
"epoch": 359.0,
|
| 156039 |
+
"learning_rate": 9.30235576923077e-06,
|
| 156040 |
+
"loss": 1.1733,
|
| 156041 |
+
"step": 44550
|
| 156042 |
+
},
|
| 156043 |
+
{
|
| 156044 |
+
"epoch": 359.0,
|
| 156045 |
+
"eval_loss": 0.3972232937812805,
|
| 156046 |
+
"eval_runtime": 39.7191,
|
| 156047 |
+
"eval_samples_per_second": 21.249,
|
| 156048 |
+
"eval_steps_per_second": 0.68,
|
| 156049 |
+
"eval_wer": 0.20374774774774776,
|
| 156050 |
+
"step": 44550
|
| 156051 |
}
|
| 156052 |
],
|
| 156053 |
+
"max_steps": 620000,
|
| 156054 |
"num_train_epochs": 5000,
|
| 156055 |
+
"total_flos": 1.2534703100755942e+20,
|
| 156056 |
"trial_name": null,
|
| 156057 |
"trial_params": null
|
| 156058 |
}
|
model-bin/finetune/base/{checkpoint-43928 → checkpoint-44550}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629709823.8830478/events.out.tfevents.1629709823.74272264b15c.932.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:920bee5809b108e75bd3deab7454e8ac53798398b6ff99609c726b6b0ddd1e0c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629710466.9426787/events.out.tfevents.1629710466.74272264b15c.932.95
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2ed0845e3bad626bbb5659068a76f153c7583e641e0e630d5c4525b8f6e7195
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629711119.662889/events.out.tfevents.1629711119.74272264b15c.932.97
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b061ad48ac9d7d2106124a74faaf9acfb11643cf2487affa859fd50c1323439a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629711757.3439753/events.out.tfevents.1629711757.74272264b15c.932.99
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6338f0bdd55b48f5eee847c2f0caf98e68859425217ba66cc647646927e07eec
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629712398.8804219/events.out.tfevents.1629712398.74272264b15c.932.101
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ad8032a70925b756335f64b70df47411d746b201ac430c3d479992609879646
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629709823.74272264b15c.932.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da1b273885f3e37d669b4202635adc7d80b81558216746c8b88e752d1825c55a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629710466.74272264b15c.932.94
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c3e94a9f3864daa5bb3559fb9ca9a7373a64d8b71e4144e68a78b139349c550c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629711119.74272264b15c.932.96
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:487f25e170842ee5c30fb39617ecb1f4bee0a543cf3520cb7404e3c9131bcc2c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629711757.74272264b15c.932.98
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:167b05c44a2c7f1e2152ebb8323fd8e555aa3faa0fe105b0cf1527f6a5073a1a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629712398.74272264b15c.932.100
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a27f0fce2d6bce03fe5c6bdeb154eb126f703ff93d07cb14303e541479c260f6
|
| 3 |
+
size 8622
|