"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629808165.1768463/events.out.tfevents.1629808165.c435e1c5ee04.920.131 +3 -0
- model-bin/finetune/base/log/1629808795.0569663/events.out.tfevents.1629808795.c435e1c5ee04.920.133 +3 -0
- model-bin/finetune/base/log/1629809420.608819/events.out.tfevents.1629809420.c435e1c5ee04.920.135 +3 -0
- model-bin/finetune/base/log/1629810059.173823/events.out.tfevents.1629810059.c435e1c5ee04.920.137 +3 -0
- model-bin/finetune/base/log/1629810681.6404505/events.out.tfevents.1629810681.c435e1c5ee04.920.139 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629808165.c435e1c5ee04.920.130 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629808795.c435e1c5ee04.920.132 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629809420.c435e1c5ee04.920.134 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629810059.c435e1c5ee04.920.136 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629810681.c435e1c5ee04.920.138 +3 -0
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/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:a290280a3474f2b8bc9387a119cd80b9f8c96dd9fb331d7444409f1cfadc3f18
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/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 sha256:17d820521db824b7c8ca736a8227578c15a02c200974eb4c05bbf63c89e2d79f
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/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:c54ea73fbbfb418a761cbd445a49e0abf8149fad4e85739cc15fd2a0a60951f7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/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 sha256:509952d91db008ba8f08d16f0ae61258f09556f1cad52e8eac185788fc6cba38
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/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:c704047ee5a23a50c622659244f035b7a2ec188e86d6da0e2e74f9128b5a56d9
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -178212,11 +178212,800 @@
|
|
| 178212 |
"eval_steps_per_second": 0.667,
|
| 178213 |
"eval_wer": 0.1950281359071159,
|
| 178214 |
"step": 61973
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 178215 |
}
|
| 178216 |
],
|
| 178217 |
"max_steps": 620000,
|
| 178218 |
"num_train_epochs": 5000,
|
| 178219 |
-
"total_flos": 1.
|
| 178220 |
"trial_name": null,
|
| 178221 |
"trial_params": null
|
| 178222 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 504.0,
|
| 5 |
+
"global_step": 62593,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 178212 |
"eval_steps_per_second": 0.667,
|
| 178213 |
"eval_wer": 0.1950281359071159,
|
| 178214 |
"step": 61973
|
| 178215 |
+
},
|
| 178216 |
+
{
|
| 178217 |
+
"epoch": 499.02,
|
| 178218 |
+
"learning_rate": 9.023205128205129e-06,
|
| 178219 |
+
"loss": 0.3369,
|
| 178220 |
+
"step": 61975
|
| 178221 |
+
},
|
| 178222 |
+
{
|
| 178223 |
+
"epoch": 499.06,
|
| 178224 |
+
"learning_rate": 9.023125e-06,
|
| 178225 |
+
"loss": 0.3677,
|
| 178226 |
+
"step": 61980
|
| 178227 |
+
},
|
| 178228 |
+
{
|
| 178229 |
+
"epoch": 499.1,
|
| 178230 |
+
"learning_rate": 9.023044871794873e-06,
|
| 178231 |
+
"loss": 0.3488,
|
| 178232 |
+
"step": 61985
|
| 178233 |
+
},
|
| 178234 |
+
{
|
| 178235 |
+
"epoch": 499.14,
|
| 178236 |
+
"learning_rate": 9.022964743589745e-06,
|
| 178237 |
+
"loss": 0.3578,
|
| 178238 |
+
"step": 61990
|
| 178239 |
+
},
|
| 178240 |
+
{
|
| 178241 |
+
"epoch": 499.18,
|
| 178242 |
+
"learning_rate": 9.022884615384616e-06,
|
| 178243 |
+
"loss": 0.6207,
|
| 178244 |
+
"step": 61995
|
| 178245 |
+
},
|
| 178246 |
+
{
|
| 178247 |
+
"epoch": 499.22,
|
| 178248 |
+
"learning_rate": 9.022804487179489e-06,
|
| 178249 |
+
"loss": 1.0681,
|
| 178250 |
+
"step": 62000
|
| 178251 |
+
},
|
| 178252 |
+
{
|
| 178253 |
+
"epoch": 499.26,
|
| 178254 |
+
"learning_rate": 9.02272435897436e-06,
|
| 178255 |
+
"loss": 0.3851,
|
| 178256 |
+
"step": 62005
|
| 178257 |
+
},
|
| 178258 |
+
{
|
| 178259 |
+
"epoch": 499.3,
|
| 178260 |
+
"learning_rate": 9.022644230769232e-06,
|
| 178261 |
+
"loss": 0.5174,
|
| 178262 |
+
"step": 62010
|
| 178263 |
+
},
|
| 178264 |
+
{
|
| 178265 |
+
"epoch": 499.34,
|
| 178266 |
+
"learning_rate": 9.022564102564103e-06,
|
| 178267 |
+
"loss": 0.4413,
|
| 178268 |
+
"step": 62015
|
| 178269 |
+
},
|
| 178270 |
+
{
|
| 178271 |
+
"epoch": 499.38,
|
| 178272 |
+
"learning_rate": 9.022483974358976e-06,
|
| 178273 |
+
"loss": 0.8265,
|
| 178274 |
+
"step": 62020
|
| 178275 |
+
},
|
| 178276 |
+
{
|
| 178277 |
+
"epoch": 499.42,
|
| 178278 |
+
"learning_rate": 9.022403846153846e-06,
|
| 178279 |
+
"loss": 1.0571,
|
| 178280 |
+
"step": 62025
|
| 178281 |
+
},
|
| 178282 |
+
{
|
| 178283 |
+
"epoch": 499.46,
|
| 178284 |
+
"learning_rate": 9.022323717948719e-06,
|
| 178285 |
+
"loss": 0.3545,
|
| 178286 |
+
"step": 62030
|
| 178287 |
+
},
|
| 178288 |
+
{
|
| 178289 |
+
"epoch": 499.5,
|
| 178290 |
+
"learning_rate": 9.02224358974359e-06,
|
| 178291 |
+
"loss": 0.3214,
|
| 178292 |
+
"step": 62035
|
| 178293 |
+
},
|
| 178294 |
+
{
|
| 178295 |
+
"epoch": 499.54,
|
| 178296 |
+
"learning_rate": 9.022163461538461e-06,
|
| 178297 |
+
"loss": 0.316,
|
| 178298 |
+
"step": 62040
|
| 178299 |
+
},
|
| 178300 |
+
{
|
| 178301 |
+
"epoch": 499.58,
|
| 178302 |
+
"learning_rate": 9.022083333333333e-06,
|
| 178303 |
+
"loss": 0.718,
|
| 178304 |
+
"step": 62045
|
| 178305 |
+
},
|
| 178306 |
+
{
|
| 178307 |
+
"epoch": 499.62,
|
| 178308 |
+
"learning_rate": 9.022003205128206e-06,
|
| 178309 |
+
"loss": 1.0395,
|
| 178310 |
+
"step": 62050
|
| 178311 |
+
},
|
| 178312 |
+
{
|
| 178313 |
+
"epoch": 499.66,
|
| 178314 |
+
"learning_rate": 9.021923076923079e-06,
|
| 178315 |
+
"loss": 0.3217,
|
| 178316 |
+
"step": 62055
|
| 178317 |
+
},
|
| 178318 |
+
{
|
| 178319 |
+
"epoch": 499.7,
|
| 178320 |
+
"learning_rate": 9.021842948717949e-06,
|
| 178321 |
+
"loss": 0.3342,
|
| 178322 |
+
"step": 62060
|
| 178323 |
+
},
|
| 178324 |
+
{
|
| 178325 |
+
"epoch": 499.74,
|
| 178326 |
+
"learning_rate": 9.021762820512822e-06,
|
| 178327 |
+
"loss": 0.412,
|
| 178328 |
+
"step": 62065
|
| 178329 |
+
},
|
| 178330 |
+
{
|
| 178331 |
+
"epoch": 499.78,
|
| 178332 |
+
"learning_rate": 9.021682692307693e-06,
|
| 178333 |
+
"loss": 0.7316,
|
| 178334 |
+
"step": 62070
|
| 178335 |
+
},
|
| 178336 |
+
{
|
| 178337 |
+
"epoch": 499.82,
|
| 178338 |
+
"learning_rate": 9.021602564102564e-06,
|
| 178339 |
+
"loss": 0.9206,
|
| 178340 |
+
"step": 62075
|
| 178341 |
+
},
|
| 178342 |
+
{
|
| 178343 |
+
"epoch": 499.86,
|
| 178344 |
+
"learning_rate": 9.021522435897436e-06,
|
| 178345 |
+
"loss": 0.3452,
|
| 178346 |
+
"step": 62080
|
| 178347 |
+
},
|
| 178348 |
+
{
|
| 178349 |
+
"epoch": 499.9,
|
| 178350 |
+
"learning_rate": 9.021442307692309e-06,
|
| 178351 |
+
"loss": 0.3346,
|
| 178352 |
+
"step": 62085
|
| 178353 |
+
},
|
| 178354 |
+
{
|
| 178355 |
+
"epoch": 499.94,
|
| 178356 |
+
"learning_rate": 9.02136217948718e-06,
|
| 178357 |
+
"loss": 0.4821,
|
| 178358 |
+
"step": 62090
|
| 178359 |
+
},
|
| 178360 |
+
{
|
| 178361 |
+
"epoch": 499.98,
|
| 178362 |
+
"learning_rate": 9.021282051282052e-06,
|
| 178363 |
+
"loss": 0.8865,
|
| 178364 |
+
"step": 62095
|
| 178365 |
+
},
|
| 178366 |
+
{
|
| 178367 |
+
"epoch": 500.0,
|
| 178368 |
+
"eval_loss": 0.38966497778892517,
|
| 178369 |
+
"eval_runtime": 38.9602,
|
| 178370 |
+
"eval_samples_per_second": 21.689,
|
| 178371 |
+
"eval_steps_per_second": 0.693,
|
| 178372 |
+
"eval_wer": 0.1995421313049258,
|
| 178373 |
+
"step": 62097
|
| 178374 |
+
},
|
| 178375 |
+
{
|
| 178376 |
+
"epoch": 500.02,
|
| 178377 |
+
"learning_rate": 9.021201923076925e-06,
|
| 178378 |
+
"loss": 0.3577,
|
| 178379 |
+
"step": 62100
|
| 178380 |
+
},
|
| 178381 |
+
{
|
| 178382 |
+
"epoch": 500.06,
|
| 178383 |
+
"learning_rate": 9.021121794871796e-06,
|
| 178384 |
+
"loss": 0.2985,
|
| 178385 |
+
"step": 62105
|
| 178386 |
+
},
|
| 178387 |
+
{
|
| 178388 |
+
"epoch": 500.1,
|
| 178389 |
+
"learning_rate": 9.021041666666667e-06,
|
| 178390 |
+
"loss": 0.3115,
|
| 178391 |
+
"step": 62110
|
| 178392 |
+
},
|
| 178393 |
+
{
|
| 178394 |
+
"epoch": 500.15,
|
| 178395 |
+
"learning_rate": 9.020961538461539e-06,
|
| 178396 |
+
"loss": 0.4337,
|
| 178397 |
+
"step": 62115
|
| 178398 |
+
},
|
| 178399 |
+
{
|
| 178400 |
+
"epoch": 500.19,
|
| 178401 |
+
"learning_rate": 9.020881410256412e-06,
|
| 178402 |
+
"loss": 0.8495,
|
| 178403 |
+
"step": 62120
|
| 178404 |
+
},
|
| 178405 |
+
{
|
| 178406 |
+
"epoch": 500.23,
|
| 178407 |
+
"learning_rate": 9.020801282051283e-06,
|
| 178408 |
+
"loss": 0.9316,
|
| 178409 |
+
"step": 62125
|
| 178410 |
+
},
|
| 178411 |
+
{
|
| 178412 |
+
"epoch": 500.27,
|
| 178413 |
+
"learning_rate": 9.020721153846154e-06,
|
| 178414 |
+
"loss": 0.3287,
|
| 178415 |
+
"step": 62130
|
| 178416 |
+
},
|
| 178417 |
+
{
|
| 178418 |
+
"epoch": 500.31,
|
| 178419 |
+
"learning_rate": 9.020641025641026e-06,
|
| 178420 |
+
"loss": 0.308,
|
| 178421 |
+
"step": 62135
|
| 178422 |
+
},
|
| 178423 |
+
{
|
| 178424 |
+
"epoch": 500.35,
|
| 178425 |
+
"learning_rate": 9.020560897435899e-06,
|
| 178426 |
+
"loss": 0.3617,
|
| 178427 |
+
"step": 62140
|
| 178428 |
+
},
|
| 178429 |
+
{
|
| 178430 |
+
"epoch": 500.39,
|
| 178431 |
+
"learning_rate": 9.02048076923077e-06,
|
| 178432 |
+
"loss": 0.8829,
|
| 178433 |
+
"step": 62145
|
| 178434 |
+
},
|
| 178435 |
+
{
|
| 178436 |
+
"epoch": 500.43,
|
| 178437 |
+
"learning_rate": 9.020400641025642e-06,
|
| 178438 |
+
"loss": 0.8967,
|
| 178439 |
+
"step": 62150
|
| 178440 |
+
},
|
| 178441 |
+
{
|
| 178442 |
+
"epoch": 500.47,
|
| 178443 |
+
"learning_rate": 9.020320512820515e-06,
|
| 178444 |
+
"loss": 0.3013,
|
| 178445 |
+
"step": 62155
|
| 178446 |
+
},
|
| 178447 |
+
{
|
| 178448 |
+
"epoch": 500.51,
|
| 178449 |
+
"learning_rate": 9.020240384615386e-06,
|
| 178450 |
+
"loss": 0.3501,
|
| 178451 |
+
"step": 62160
|
| 178452 |
+
},
|
| 178453 |
+
{
|
| 178454 |
+
"epoch": 500.55,
|
| 178455 |
+
"learning_rate": 9.020160256410257e-06,
|
| 178456 |
+
"loss": 0.4097,
|
| 178457 |
+
"step": 62165
|
| 178458 |
+
},
|
| 178459 |
+
{
|
| 178460 |
+
"epoch": 500.59,
|
| 178461 |
+
"learning_rate": 9.020080128205129e-06,
|
| 178462 |
+
"loss": 0.9575,
|
| 178463 |
+
"step": 62170
|
| 178464 |
+
},
|
| 178465 |
+
{
|
| 178466 |
+
"epoch": 500.63,
|
| 178467 |
+
"learning_rate": 9.020000000000002e-06,
|
| 178468 |
+
"loss": 0.9411,
|
| 178469 |
+
"step": 62175
|
| 178470 |
+
},
|
| 178471 |
+
{
|
| 178472 |
+
"epoch": 500.67,
|
| 178473 |
+
"learning_rate": 9.019919871794871e-06,
|
| 178474 |
+
"loss": 0.3027,
|
| 178475 |
+
"step": 62180
|
| 178476 |
+
},
|
| 178477 |
+
{
|
| 178478 |
+
"epoch": 500.71,
|
| 178479 |
+
"learning_rate": 9.019839743589744e-06,
|
| 178480 |
+
"loss": 0.3253,
|
| 178481 |
+
"step": 62185
|
| 178482 |
+
},
|
| 178483 |
+
{
|
| 178484 |
+
"epoch": 500.75,
|
| 178485 |
+
"learning_rate": 9.019759615384616e-06,
|
| 178486 |
+
"loss": 0.4113,
|
| 178487 |
+
"step": 62190
|
| 178488 |
+
},
|
| 178489 |
+
{
|
| 178490 |
+
"epoch": 500.79,
|
| 178491 |
+
"learning_rate": 9.019679487179487e-06,
|
| 178492 |
+
"loss": 0.8241,
|
| 178493 |
+
"step": 62195
|
| 178494 |
+
},
|
| 178495 |
+
{
|
| 178496 |
+
"epoch": 500.83,
|
| 178497 |
+
"learning_rate": 9.01959935897436e-06,
|
| 178498 |
+
"loss": 0.8903,
|
| 178499 |
+
"step": 62200
|
| 178500 |
+
},
|
| 178501 |
+
{
|
| 178502 |
+
"epoch": 500.87,
|
| 178503 |
+
"learning_rate": 9.019519230769232e-06,
|
| 178504 |
+
"loss": 0.265,
|
| 178505 |
+
"step": 62205
|
| 178506 |
+
},
|
| 178507 |
+
{
|
| 178508 |
+
"epoch": 500.91,
|
| 178509 |
+
"learning_rate": 9.019439102564103e-06,
|
| 178510 |
+
"loss": 0.5923,
|
| 178511 |
+
"step": 62210
|
| 178512 |
+
},
|
| 178513 |
+
{
|
| 178514 |
+
"epoch": 500.95,
|
| 178515 |
+
"learning_rate": 9.019358974358974e-06,
|
| 178516 |
+
"loss": 0.3643,
|
| 178517 |
+
"step": 62215
|
| 178518 |
+
},
|
| 178519 |
+
{
|
| 178520 |
+
"epoch": 500.99,
|
| 178521 |
+
"learning_rate": 9.019278846153847e-06,
|
| 178522 |
+
"loss": 1.0937,
|
| 178523 |
+
"step": 62220
|
| 178524 |
+
},
|
| 178525 |
+
{
|
| 178526 |
+
"epoch": 501.0,
|
| 178527 |
+
"eval_loss": 0.41493460536003113,
|
| 178528 |
+
"eval_runtime": 39.3216,
|
| 178529 |
+
"eval_samples_per_second": 21.489,
|
| 178530 |
+
"eval_steps_per_second": 0.687,
|
| 178531 |
+
"eval_wer": 0.1906767354051699,
|
| 178532 |
+
"step": 62221
|
| 178533 |
+
},
|
| 178534 |
+
{
|
| 178535 |
+
"epoch": 501.03,
|
| 178536 |
+
"learning_rate": 9.019198717948719e-06,
|
| 178537 |
+
"loss": 0.3448,
|
| 178538 |
+
"step": 62225
|
| 178539 |
+
},
|
| 178540 |
+
{
|
| 178541 |
+
"epoch": 501.07,
|
| 178542 |
+
"learning_rate": 9.01911858974359e-06,
|
| 178543 |
+
"loss": 0.3584,
|
| 178544 |
+
"step": 62230
|
| 178545 |
+
},
|
| 178546 |
+
{
|
| 178547 |
+
"epoch": 501.11,
|
| 178548 |
+
"learning_rate": 9.019038461538461e-06,
|
| 178549 |
+
"loss": 0.3391,
|
| 178550 |
+
"step": 62235
|
| 178551 |
+
},
|
| 178552 |
+
{
|
| 178553 |
+
"epoch": 501.15,
|
| 178554 |
+
"learning_rate": 9.018958333333334e-06,
|
| 178555 |
+
"loss": 0.4869,
|
| 178556 |
+
"step": 62240
|
| 178557 |
+
},
|
| 178558 |
+
{
|
| 178559 |
+
"epoch": 501.19,
|
| 178560 |
+
"learning_rate": 9.018878205128206e-06,
|
| 178561 |
+
"loss": 0.9897,
|
| 178562 |
+
"step": 62245
|
| 178563 |
+
},
|
| 178564 |
+
{
|
| 178565 |
+
"epoch": 501.23,
|
| 178566 |
+
"learning_rate": 9.018798076923077e-06,
|
| 178567 |
+
"loss": 0.6692,
|
| 178568 |
+
"step": 62250
|
| 178569 |
+
},
|
| 178570 |
+
{
|
| 178571 |
+
"epoch": 501.27,
|
| 178572 |
+
"learning_rate": 9.01871794871795e-06,
|
| 178573 |
+
"loss": 0.332,
|
| 178574 |
+
"step": 62255
|
| 178575 |
+
},
|
| 178576 |
+
{
|
| 178577 |
+
"epoch": 501.31,
|
| 178578 |
+
"learning_rate": 9.018637820512822e-06,
|
| 178579 |
+
"loss": 0.3738,
|
| 178580 |
+
"step": 62260
|
| 178581 |
+
},
|
| 178582 |
+
{
|
| 178583 |
+
"epoch": 501.35,
|
| 178584 |
+
"learning_rate": 9.018557692307693e-06,
|
| 178585 |
+
"loss": 0.4152,
|
| 178586 |
+
"step": 62265
|
| 178587 |
+
},
|
| 178588 |
+
{
|
| 178589 |
+
"epoch": 501.4,
|
| 178590 |
+
"learning_rate": 9.018477564102564e-06,
|
| 178591 |
+
"loss": 1.0347,
|
| 178592 |
+
"step": 62270
|
| 178593 |
+
},
|
| 178594 |
+
{
|
| 178595 |
+
"epoch": 501.44,
|
| 178596 |
+
"learning_rate": 9.018397435897437e-06,
|
| 178597 |
+
"loss": 0.6941,
|
| 178598 |
+
"step": 62275
|
| 178599 |
+
},
|
| 178600 |
+
{
|
| 178601 |
+
"epoch": 501.48,
|
| 178602 |
+
"learning_rate": 9.018317307692309e-06,
|
| 178603 |
+
"loss": 0.3831,
|
| 178604 |
+
"step": 62280
|
| 178605 |
+
},
|
| 178606 |
+
{
|
| 178607 |
+
"epoch": 501.52,
|
| 178608 |
+
"learning_rate": 9.01823717948718e-06,
|
| 178609 |
+
"loss": 0.4713,
|
| 178610 |
+
"step": 62285
|
| 178611 |
+
},
|
| 178612 |
+
{
|
| 178613 |
+
"epoch": 501.56,
|
| 178614 |
+
"learning_rate": 9.018157051282051e-06,
|
| 178615 |
+
"loss": 0.4299,
|
| 178616 |
+
"step": 62290
|
| 178617 |
+
},
|
| 178618 |
+
{
|
| 178619 |
+
"epoch": 501.6,
|
| 178620 |
+
"learning_rate": 9.018076923076925e-06,
|
| 178621 |
+
"loss": 0.9001,
|
| 178622 |
+
"step": 62295
|
| 178623 |
+
},
|
| 178624 |
+
{
|
| 178625 |
+
"epoch": 501.64,
|
| 178626 |
+
"learning_rate": 9.017996794871796e-06,
|
| 178627 |
+
"loss": 0.6992,
|
| 178628 |
+
"step": 62300
|
| 178629 |
+
},
|
| 178630 |
+
{
|
| 178631 |
+
"epoch": 501.68,
|
| 178632 |
+
"learning_rate": 9.017916666666667e-06,
|
| 178633 |
+
"loss": 0.3267,
|
| 178634 |
+
"step": 62305
|
| 178635 |
+
},
|
| 178636 |
+
{
|
| 178637 |
+
"epoch": 501.72,
|
| 178638 |
+
"learning_rate": 9.01783653846154e-06,
|
| 178639 |
+
"loss": 0.283,
|
| 178640 |
+
"step": 62310
|
| 178641 |
+
},
|
| 178642 |
+
{
|
| 178643 |
+
"epoch": 501.76,
|
| 178644 |
+
"learning_rate": 9.017756410256412e-06,
|
| 178645 |
+
"loss": 0.436,
|
| 178646 |
+
"step": 62315
|
| 178647 |
+
},
|
| 178648 |
+
{
|
| 178649 |
+
"epoch": 501.8,
|
| 178650 |
+
"learning_rate": 9.017676282051283e-06,
|
| 178651 |
+
"loss": 0.9756,
|
| 178652 |
+
"step": 62320
|
| 178653 |
+
},
|
| 178654 |
+
{
|
| 178655 |
+
"epoch": 501.84,
|
| 178656 |
+
"learning_rate": 9.017596153846154e-06,
|
| 178657 |
+
"loss": 0.8162,
|
| 178658 |
+
"step": 62325
|
| 178659 |
+
},
|
| 178660 |
+
{
|
| 178661 |
+
"epoch": 501.88,
|
| 178662 |
+
"learning_rate": 9.017516025641027e-06,
|
| 178663 |
+
"loss": 0.2937,
|
| 178664 |
+
"step": 62330
|
| 178665 |
+
},
|
| 178666 |
+
{
|
| 178667 |
+
"epoch": 501.92,
|
| 178668 |
+
"learning_rate": 9.017435897435897e-06,
|
| 178669 |
+
"loss": 0.3419,
|
| 178670 |
+
"step": 62335
|
| 178671 |
+
},
|
| 178672 |
+
{
|
| 178673 |
+
"epoch": 501.96,
|
| 178674 |
+
"learning_rate": 9.01735576923077e-06,
|
| 178675 |
+
"loss": 0.4361,
|
| 178676 |
+
"step": 62340
|
| 178677 |
+
},
|
| 178678 |
+
{
|
| 178679 |
+
"epoch": 502.0,
|
| 178680 |
+
"learning_rate": 9.017275641025641e-06,
|
| 178681 |
+
"loss": 1.5338,
|
| 178682 |
+
"step": 62345
|
| 178683 |
+
},
|
| 178684 |
+
{
|
| 178685 |
+
"epoch": 502.0,
|
| 178686 |
+
"eval_loss": 0.4359527826309204,
|
| 178687 |
+
"eval_runtime": 39.306,
|
| 178688 |
+
"eval_samples_per_second": 21.473,
|
| 178689 |
+
"eval_steps_per_second": 0.687,
|
| 178690 |
+
"eval_wer": 0.1983987072131629,
|
| 178691 |
+
"step": 62345
|
| 178692 |
+
},
|
| 178693 |
+
{
|
| 178694 |
+
"epoch": 502.04,
|
| 178695 |
+
"learning_rate": 9.017195512820513e-06,
|
| 178696 |
+
"loss": 0.3277,
|
| 178697 |
+
"step": 62350
|
| 178698 |
+
},
|
| 178699 |
+
{
|
| 178700 |
+
"epoch": 502.08,
|
| 178701 |
+
"learning_rate": 9.017115384615386e-06,
|
| 178702 |
+
"loss": 0.2857,
|
| 178703 |
+
"step": 62355
|
| 178704 |
+
},
|
| 178705 |
+
{
|
| 178706 |
+
"epoch": 502.12,
|
| 178707 |
+
"learning_rate": 9.017035256410257e-06,
|
| 178708 |
+
"loss": 0.3327,
|
| 178709 |
+
"step": 62360
|
| 178710 |
+
},
|
| 178711 |
+
{
|
| 178712 |
+
"epoch": 502.16,
|
| 178713 |
+
"learning_rate": 9.016955128205129e-06,
|
| 178714 |
+
"loss": 0.4918,
|
| 178715 |
+
"step": 62365
|
| 178716 |
+
},
|
| 178717 |
+
{
|
| 178718 |
+
"epoch": 502.2,
|
| 178719 |
+
"learning_rate": 9.016875e-06,
|
| 178720 |
+
"loss": 1.2648,
|
| 178721 |
+
"step": 62370
|
| 178722 |
+
},
|
| 178723 |
+
{
|
| 178724 |
+
"epoch": 502.24,
|
| 178725 |
+
"learning_rate": 9.016794871794873e-06,
|
| 178726 |
+
"loss": 0.3506,
|
| 178727 |
+
"step": 62375
|
| 178728 |
+
},
|
| 178729 |
+
{
|
| 178730 |
+
"epoch": 502.28,
|
| 178731 |
+
"learning_rate": 9.016714743589744e-06,
|
| 178732 |
+
"loss": 0.2926,
|
| 178733 |
+
"step": 62380
|
| 178734 |
+
},
|
| 178735 |
+
{
|
| 178736 |
+
"epoch": 502.32,
|
| 178737 |
+
"learning_rate": 9.016634615384616e-06,
|
| 178738 |
+
"loss": 0.3372,
|
| 178739 |
+
"step": 62385
|
| 178740 |
+
},
|
| 178741 |
+
{
|
| 178742 |
+
"epoch": 502.36,
|
| 178743 |
+
"learning_rate": 9.016554487179487e-06,
|
| 178744 |
+
"loss": 0.5222,
|
| 178745 |
+
"step": 62390
|
| 178746 |
+
},
|
| 178747 |
+
{
|
| 178748 |
+
"epoch": 502.4,
|
| 178749 |
+
"learning_rate": 9.01647435897436e-06,
|
| 178750 |
+
"loss": 1.2323,
|
| 178751 |
+
"step": 62395
|
| 178752 |
+
},
|
| 178753 |
+
{
|
| 178754 |
+
"epoch": 502.44,
|
| 178755 |
+
"learning_rate": 9.016394230769231e-06,
|
| 178756 |
+
"loss": 0.3076,
|
| 178757 |
+
"step": 62400
|
| 178758 |
+
},
|
| 178759 |
+
{
|
| 178760 |
+
"epoch": 502.48,
|
| 178761 |
+
"learning_rate": 9.016314102564103e-06,
|
| 178762 |
+
"loss": 0.3123,
|
| 178763 |
+
"step": 62405
|
| 178764 |
+
},
|
| 178765 |
+
{
|
| 178766 |
+
"epoch": 502.52,
|
| 178767 |
+
"learning_rate": 9.016233974358976e-06,
|
| 178768 |
+
"loss": 0.367,
|
| 178769 |
+
"step": 62410
|
| 178770 |
+
},
|
| 178771 |
+
{
|
| 178772 |
+
"epoch": 502.56,
|
| 178773 |
+
"learning_rate": 9.016153846153847e-06,
|
| 178774 |
+
"loss": 0.5039,
|
| 178775 |
+
"step": 62415
|
| 178776 |
+
},
|
| 178777 |
+
{
|
| 178778 |
+
"epoch": 502.6,
|
| 178779 |
+
"learning_rate": 9.016073717948719e-06,
|
| 178780 |
+
"loss": 1.2038,
|
| 178781 |
+
"step": 62420
|
| 178782 |
+
},
|
| 178783 |
+
{
|
| 178784 |
+
"epoch": 502.64,
|
| 178785 |
+
"learning_rate": 9.01599358974359e-06,
|
| 178786 |
+
"loss": 0.3573,
|
| 178787 |
+
"step": 62425
|
| 178788 |
+
},
|
| 178789 |
+
{
|
| 178790 |
+
"epoch": 502.68,
|
| 178791 |
+
"learning_rate": 9.015913461538463e-06,
|
| 178792 |
+
"loss": 0.361,
|
| 178793 |
+
"step": 62430
|
| 178794 |
+
},
|
| 178795 |
+
{
|
| 178796 |
+
"epoch": 502.72,
|
| 178797 |
+
"learning_rate": 9.015833333333334e-06,
|
| 178798 |
+
"loss": 0.3163,
|
| 178799 |
+
"step": 62435
|
| 178800 |
+
},
|
| 178801 |
+
{
|
| 178802 |
+
"epoch": 502.76,
|
| 178803 |
+
"learning_rate": 9.015753205128206e-06,
|
| 178804 |
+
"loss": 0.4155,
|
| 178805 |
+
"step": 62440
|
| 178806 |
+
},
|
| 178807 |
+
{
|
| 178808 |
+
"epoch": 502.8,
|
| 178809 |
+
"learning_rate": 9.015673076923077e-06,
|
| 178810 |
+
"loss": 1.3244,
|
| 178811 |
+
"step": 62445
|
| 178812 |
+
},
|
| 178813 |
+
{
|
| 178814 |
+
"epoch": 502.84,
|
| 178815 |
+
"learning_rate": 9.01559294871795e-06,
|
| 178816 |
+
"loss": 0.2981,
|
| 178817 |
+
"step": 62450
|
| 178818 |
+
},
|
| 178819 |
+
{
|
| 178820 |
+
"epoch": 502.88,
|
| 178821 |
+
"learning_rate": 9.015512820512822e-06,
|
| 178822 |
+
"loss": 0.2625,
|
| 178823 |
+
"step": 62455
|
| 178824 |
+
},
|
| 178825 |
+
{
|
| 178826 |
+
"epoch": 502.92,
|
| 178827 |
+
"learning_rate": 9.015432692307693e-06,
|
| 178828 |
+
"loss": 0.3824,
|
| 178829 |
+
"step": 62460
|
| 178830 |
+
},
|
| 178831 |
+
{
|
| 178832 |
+
"epoch": 502.96,
|
| 178833 |
+
"learning_rate": 9.015352564102566e-06,
|
| 178834 |
+
"loss": 0.6499,
|
| 178835 |
+
"step": 62465
|
| 178836 |
+
},
|
| 178837 |
+
{
|
| 178838 |
+
"epoch": 503.0,
|
| 178839 |
+
"eval_loss": 0.4520353376865387,
|
| 178840 |
+
"eval_runtime": 39.4174,
|
| 178841 |
+
"eval_samples_per_second": 21.412,
|
| 178842 |
+
"eval_steps_per_second": 0.685,
|
| 178843 |
+
"eval_wer": 0.20146465401285307,
|
| 178844 |
+
"step": 62469
|
| 178845 |
+
},
|
| 178846 |
+
{
|
| 178847 |
+
"epoch": 503.01,
|
| 178848 |
+
"learning_rate": 9.015272435897436e-06,
|
| 178849 |
+
"loss": 0.5768,
|
| 178850 |
+
"step": 62470
|
| 178851 |
+
},
|
| 178852 |
+
{
|
| 178853 |
+
"epoch": 503.05,
|
| 178854 |
+
"learning_rate": 9.015192307692309e-06,
|
| 178855 |
+
"loss": 0.3239,
|
| 178856 |
+
"step": 62475
|
| 178857 |
+
},
|
| 178858 |
+
{
|
| 178859 |
+
"epoch": 503.09,
|
| 178860 |
+
"learning_rate": 9.01511217948718e-06,
|
| 178861 |
+
"loss": 0.3187,
|
| 178862 |
+
"step": 62480
|
| 178863 |
+
},
|
| 178864 |
+
{
|
| 178865 |
+
"epoch": 503.13,
|
| 178866 |
+
"learning_rate": 9.015032051282051e-06,
|
| 178867 |
+
"loss": 0.3518,
|
| 178868 |
+
"step": 62485
|
| 178869 |
+
},
|
| 178870 |
+
{
|
| 178871 |
+
"epoch": 503.17,
|
| 178872 |
+
"learning_rate": 9.014951923076923e-06,
|
| 178873 |
+
"loss": 0.5604,
|
| 178874 |
+
"step": 62490
|
| 178875 |
+
},
|
| 178876 |
+
{
|
| 178877 |
+
"epoch": 503.21,
|
| 178878 |
+
"learning_rate": 9.014871794871796e-06,
|
| 178879 |
+
"loss": 1.1861,
|
| 178880 |
+
"step": 62495
|
| 178881 |
+
},
|
| 178882 |
+
{
|
| 178883 |
+
"epoch": 503.25,
|
| 178884 |
+
"learning_rate": 9.014791666666667e-06,
|
| 178885 |
+
"loss": 0.3453,
|
| 178886 |
+
"step": 62500
|
| 178887 |
+
},
|
| 178888 |
+
{
|
| 178889 |
+
"epoch": 503.29,
|
| 178890 |
+
"learning_rate": 9.014711538461538e-06,
|
| 178891 |
+
"loss": 0.3375,
|
| 178892 |
+
"step": 62505
|
| 178893 |
+
},
|
| 178894 |
+
{
|
| 178895 |
+
"epoch": 503.33,
|
| 178896 |
+
"learning_rate": 9.014631410256412e-06,
|
| 178897 |
+
"loss": 0.2996,
|
| 178898 |
+
"step": 62510
|
| 178899 |
+
},
|
| 178900 |
+
{
|
| 178901 |
+
"epoch": 503.37,
|
| 178902 |
+
"learning_rate": 9.014551282051283e-06,
|
| 178903 |
+
"loss": 0.6461,
|
| 178904 |
+
"step": 62515
|
| 178905 |
+
},
|
| 178906 |
+
{
|
| 178907 |
+
"epoch": 503.41,
|
| 178908 |
+
"learning_rate": 9.014471153846154e-06,
|
| 178909 |
+
"loss": 1.2417,
|
| 178910 |
+
"step": 62520
|
| 178911 |
+
},
|
| 178912 |
+
{
|
| 178913 |
+
"epoch": 503.45,
|
| 178914 |
+
"learning_rate": 9.014391025641026e-06,
|
| 178915 |
+
"loss": 0.3491,
|
| 178916 |
+
"step": 62525
|
| 178917 |
+
},
|
| 178918 |
+
{
|
| 178919 |
+
"epoch": 503.49,
|
| 178920 |
+
"learning_rate": 9.014310897435899e-06,
|
| 178921 |
+
"loss": 0.351,
|
| 178922 |
+
"step": 62530
|
| 178923 |
+
},
|
| 178924 |
+
{
|
| 178925 |
+
"epoch": 503.53,
|
| 178926 |
+
"learning_rate": 9.01423076923077e-06,
|
| 178927 |
+
"loss": 0.3876,
|
| 178928 |
+
"step": 62535
|
| 178929 |
+
},
|
| 178930 |
+
{
|
| 178931 |
+
"epoch": 503.57,
|
| 178932 |
+
"learning_rate": 9.014150641025641e-06,
|
| 178933 |
+
"loss": 0.5868,
|
| 178934 |
+
"step": 62540
|
| 178935 |
+
},
|
| 178936 |
+
{
|
| 178937 |
+
"epoch": 503.61,
|
| 178938 |
+
"learning_rate": 9.014070512820513e-06,
|
| 178939 |
+
"loss": 1.1291,
|
| 178940 |
+
"step": 62545
|
| 178941 |
+
},
|
| 178942 |
+
{
|
| 178943 |
+
"epoch": 503.65,
|
| 178944 |
+
"learning_rate": 9.013990384615386e-06,
|
| 178945 |
+
"loss": 0.3,
|
| 178946 |
+
"step": 62550
|
| 178947 |
+
},
|
| 178948 |
+
{
|
| 178949 |
+
"epoch": 503.69,
|
| 178950 |
+
"learning_rate": 9.013910256410257e-06,
|
| 178951 |
+
"loss": 0.3338,
|
| 178952 |
+
"step": 62555
|
| 178953 |
+
},
|
| 178954 |
+
{
|
| 178955 |
+
"epoch": 503.73,
|
| 178956 |
+
"learning_rate": 9.013830128205129e-06,
|
| 178957 |
+
"loss": 0.4333,
|
| 178958 |
+
"step": 62560
|
| 178959 |
+
},
|
| 178960 |
+
{
|
| 178961 |
+
"epoch": 503.77,
|
| 178962 |
+
"learning_rate": 9.013750000000002e-06,
|
| 178963 |
+
"loss": 0.6022,
|
| 178964 |
+
"step": 62565
|
| 178965 |
+
},
|
| 178966 |
+
{
|
| 178967 |
+
"epoch": 503.81,
|
| 178968 |
+
"learning_rate": 9.013669871794873e-06,
|
| 178969 |
+
"loss": 1.3034,
|
| 178970 |
+
"step": 62570
|
| 178971 |
+
},
|
| 178972 |
+
{
|
| 178973 |
+
"epoch": 503.85,
|
| 178974 |
+
"learning_rate": 9.013589743589744e-06,
|
| 178975 |
+
"loss": 0.3255,
|
| 178976 |
+
"step": 62575
|
| 178977 |
+
},
|
| 178978 |
+
{
|
| 178979 |
+
"epoch": 503.9,
|
| 178980 |
+
"learning_rate": 9.013509615384616e-06,
|
| 178981 |
+
"loss": 0.3653,
|
| 178982 |
+
"step": 62580
|
| 178983 |
+
},
|
| 178984 |
+
{
|
| 178985 |
+
"epoch": 503.94,
|
| 178986 |
+
"learning_rate": 9.013429487179489e-06,
|
| 178987 |
+
"loss": 0.3396,
|
| 178988 |
+
"step": 62585
|
| 178989 |
+
},
|
| 178990 |
+
{
|
| 178991 |
+
"epoch": 503.98,
|
| 178992 |
+
"learning_rate": 9.01334935897436e-06,
|
| 178993 |
+
"loss": 0.5734,
|
| 178994 |
+
"step": 62590
|
| 178995 |
+
},
|
| 178996 |
+
{
|
| 178997 |
+
"epoch": 504.0,
|
| 178998 |
+
"eval_loss": 0.4007973372936249,
|
| 178999 |
+
"eval_runtime": 39.7118,
|
| 179000 |
+
"eval_samples_per_second": 21.253,
|
| 179001 |
+
"eval_steps_per_second": 0.68,
|
| 179002 |
+
"eval_wer": 0.2003999111308598,
|
| 179003 |
+
"step": 62593
|
| 179004 |
}
|
| 179005 |
],
|
| 179006 |
"max_steps": 620000,
|
| 179007 |
"num_train_epochs": 5000,
|
| 179008 |
+
"total_flos": 1.7614464185546205e+20,
|
| 179009 |
"trial_name": null,
|
| 179010 |
"trial_params": null
|
| 179011 |
}
|
model-bin/finetune/base/{checkpoint-61973 → checkpoint-62593}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629808165.1768463/events.out.tfevents.1629808165.c435e1c5ee04.920.131
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:608b3428755ac70cfe641f1e408fd50a23496bf36865a51e00f9d27ebb167568
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629808795.0569663/events.out.tfevents.1629808795.c435e1c5ee04.920.133
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72f0712b413e9cf4134d88ecd3b55297ef19abec20351883f9dae5e057e9b311
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629809420.608819/events.out.tfevents.1629809420.c435e1c5ee04.920.135
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71b56095ad13b9493cae15fe078e2950452c84a11166860e3411610dfeea11f7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629810059.173823/events.out.tfevents.1629810059.c435e1c5ee04.920.137
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:616b455d4a26cd1025e83571ab28150abc4179e0d6a84f4a840f1c5853937ac3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629810681.6404505/events.out.tfevents.1629810681.c435e1c5ee04.920.139
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:82eb21e014236382cb41a82e05450d9477161dc406586feff882436b12400915
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629808165.c435e1c5ee04.920.130
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1757822e761fdd2d02160700bb57a166ede9c17fd5b91ef027270cb1ef0a94b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629808795.c435e1c5ee04.920.132
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eff5915003ba3a0048c15cbdc29a68a3606f147285a121565767f87d190d0b34
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629809420.c435e1c5ee04.920.134
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30946592ea203a82f3955ec67de5bcea6630c29260c47b1cf6849fa4064f022a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629810059.c435e1c5ee04.920.136
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6a963f808dfcc9c4a56a6f750a811fb81cc053770e308c26c7f99cb95eacac6
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629810681.c435e1c5ee04.920.138
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72dc5f1026a2a5b43007bf14df4bc86bc5181b39c9f1720ffd877deaaf42f53f
|
| 3 |
+
size 8622
|