"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629918685.2041333/events.out.tfevents.1629918685.7e498afd5545.7645.125 +3 -0
- model-bin/finetune/base/log/1629919158.470076/events.out.tfevents.1629919158.7e498afd5545.7645.127 +3 -0
- model-bin/finetune/base/log/1629919633.5511148/events.out.tfevents.1629919633.7e498afd5545.7645.129 +3 -0
- model-bin/finetune/base/log/1629920107.9077082/events.out.tfevents.1629920107.7e498afd5545.7645.131 +3 -0
- model-bin/finetune/base/log/1629920581.73902/events.out.tfevents.1629920581.7e498afd5545.7645.133 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629918685.7e498afd5545.7645.124 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629919158.7e498afd5545.7645.126 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629919633.7e498afd5545.7645.128 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629920107.7e498afd5545.7645.130 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629920581.7e498afd5545.7645.132 +3 -0
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/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 sha256:35cbe7ab3362f3d9d416def618359265c26b9a47243d5ebfdbce021a9d514ce2
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/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:bb88936fbbed35267cc18d24cb4addb0e1c5f77ccc0bac83f3e42191852ec657
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/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:ef9efe43d872c0b7c4c3f9cf869d4e3159ef3fe70063cc12e48d87f2ee4f2a8b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/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 sha256:42c7556e64554470eb5a108fd2e304f85439dd31cac51b64a3aeafaa85ae7bc8
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/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 sha256:6efdd0e0c70faa9ea3b1397d573a4c203c65597ac8842e40857d44ef64c08d75
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -206241,11 +206241,800 @@
|
|
| 206241 |
"eval_steps_per_second": 0.638,
|
| 206242 |
"eval_wer": 0.19464897572355472,
|
| 206243 |
"step": 84001
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 206244 |
}
|
| 206245 |
],
|
| 206246 |
-
"max_steps":
|
| 206247 |
"num_train_epochs": 5000,
|
| 206248 |
-
"total_flos": 2.
|
| 206249 |
"trial_name": null,
|
| 206250 |
"trial_params": null
|
| 206251 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 681.995983935743,
|
| 5 |
+
"global_step": 84623,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 206241 |
"eval_steps_per_second": 0.638,
|
| 206242 |
"eval_wer": 0.19464897572355472,
|
| 206243 |
"step": 84001
|
| 206244 |
+
},
|
| 206245 |
+
{
|
| 206246 |
+
"epoch": 677.03,
|
| 206247 |
+
"learning_rate": 8.659579967689823e-06,
|
| 206248 |
+
"loss": 0.4234,
|
| 206249 |
+
"step": 84005
|
| 206250 |
+
},
|
| 206251 |
+
{
|
| 206252 |
+
"epoch": 677.07,
|
| 206253 |
+
"learning_rate": 8.659499192245559e-06,
|
| 206254 |
+
"loss": 0.3168,
|
| 206255 |
+
"step": 84010
|
| 206256 |
+
},
|
| 206257 |
+
{
|
| 206258 |
+
"epoch": 677.11,
|
| 206259 |
+
"learning_rate": 8.659418416801293e-06,
|
| 206260 |
+
"loss": 0.3753,
|
| 206261 |
+
"step": 84015
|
| 206262 |
+
},
|
| 206263 |
+
{
|
| 206264 |
+
"epoch": 677.15,
|
| 206265 |
+
"learning_rate": 8.659337641357029e-06,
|
| 206266 |
+
"loss": 0.4001,
|
| 206267 |
+
"step": 84020
|
| 206268 |
+
},
|
| 206269 |
+
{
|
| 206270 |
+
"epoch": 677.19,
|
| 206271 |
+
"learning_rate": 8.659256865912763e-06,
|
| 206272 |
+
"loss": 1.0413,
|
| 206273 |
+
"step": 84025
|
| 206274 |
+
},
|
| 206275 |
+
{
|
| 206276 |
+
"epoch": 677.23,
|
| 206277 |
+
"learning_rate": 8.659176090468498e-06,
|
| 206278 |
+
"loss": 0.7222,
|
| 206279 |
+
"step": 84030
|
| 206280 |
+
},
|
| 206281 |
+
{
|
| 206282 |
+
"epoch": 677.27,
|
| 206283 |
+
"learning_rate": 8.659095315024233e-06,
|
| 206284 |
+
"loss": 0.2808,
|
| 206285 |
+
"step": 84035
|
| 206286 |
+
},
|
| 206287 |
+
{
|
| 206288 |
+
"epoch": 677.31,
|
| 206289 |
+
"learning_rate": 8.659014539579968e-06,
|
| 206290 |
+
"loss": 0.3472,
|
| 206291 |
+
"step": 84040
|
| 206292 |
+
},
|
| 206293 |
+
{
|
| 206294 |
+
"epoch": 677.35,
|
| 206295 |
+
"learning_rate": 8.658933764135703e-06,
|
| 206296 |
+
"loss": 0.4189,
|
| 206297 |
+
"step": 84045
|
| 206298 |
+
},
|
| 206299 |
+
{
|
| 206300 |
+
"epoch": 677.39,
|
| 206301 |
+
"learning_rate": 8.658852988691438e-06,
|
| 206302 |
+
"loss": 0.8356,
|
| 206303 |
+
"step": 84050
|
| 206304 |
+
},
|
| 206305 |
+
{
|
| 206306 |
+
"epoch": 677.43,
|
| 206307 |
+
"learning_rate": 8.658772213247174e-06,
|
| 206308 |
+
"loss": 0.6287,
|
| 206309 |
+
"step": 84055
|
| 206310 |
+
},
|
| 206311 |
+
{
|
| 206312 |
+
"epoch": 677.47,
|
| 206313 |
+
"learning_rate": 8.658691437802908e-06,
|
| 206314 |
+
"loss": 0.2809,
|
| 206315 |
+
"step": 84060
|
| 206316 |
+
},
|
| 206317 |
+
{
|
| 206318 |
+
"epoch": 677.51,
|
| 206319 |
+
"learning_rate": 8.658610662358644e-06,
|
| 206320 |
+
"loss": 0.3183,
|
| 206321 |
+
"step": 84065
|
| 206322 |
+
},
|
| 206323 |
+
{
|
| 206324 |
+
"epoch": 677.55,
|
| 206325 |
+
"learning_rate": 8.658529886914378e-06,
|
| 206326 |
+
"loss": 0.3778,
|
| 206327 |
+
"step": 84070
|
| 206328 |
+
},
|
| 206329 |
+
{
|
| 206330 |
+
"epoch": 677.59,
|
| 206331 |
+
"learning_rate": 8.658449111470114e-06,
|
| 206332 |
+
"loss": 0.9175,
|
| 206333 |
+
"step": 84075
|
| 206334 |
+
},
|
| 206335 |
+
{
|
| 206336 |
+
"epoch": 677.63,
|
| 206337 |
+
"learning_rate": 8.658368336025848e-06,
|
| 206338 |
+
"loss": 0.6633,
|
| 206339 |
+
"step": 84080
|
| 206340 |
+
},
|
| 206341 |
+
{
|
| 206342 |
+
"epoch": 677.67,
|
| 206343 |
+
"learning_rate": 8.658287560581584e-06,
|
| 206344 |
+
"loss": 0.2622,
|
| 206345 |
+
"step": 84085
|
| 206346 |
+
},
|
| 206347 |
+
{
|
| 206348 |
+
"epoch": 677.71,
|
| 206349 |
+
"learning_rate": 8.658206785137318e-06,
|
| 206350 |
+
"loss": 0.3968,
|
| 206351 |
+
"step": 84090
|
| 206352 |
+
},
|
| 206353 |
+
{
|
| 206354 |
+
"epoch": 677.76,
|
| 206355 |
+
"learning_rate": 8.658126009693054e-06,
|
| 206356 |
+
"loss": 0.4656,
|
| 206357 |
+
"step": 84095
|
| 206358 |
+
},
|
| 206359 |
+
{
|
| 206360 |
+
"epoch": 677.8,
|
| 206361 |
+
"learning_rate": 8.658045234248788e-06,
|
| 206362 |
+
"loss": 0.9295,
|
| 206363 |
+
"step": 84100
|
| 206364 |
+
},
|
| 206365 |
+
{
|
| 206366 |
+
"epoch": 677.84,
|
| 206367 |
+
"learning_rate": 8.657964458804524e-06,
|
| 206368 |
+
"loss": 0.7699,
|
| 206369 |
+
"step": 84105
|
| 206370 |
+
},
|
| 206371 |
+
{
|
| 206372 |
+
"epoch": 677.88,
|
| 206373 |
+
"learning_rate": 8.657883683360258e-06,
|
| 206374 |
+
"loss": 0.2955,
|
| 206375 |
+
"step": 84110
|
| 206376 |
+
},
|
| 206377 |
+
{
|
| 206378 |
+
"epoch": 677.92,
|
| 206379 |
+
"learning_rate": 8.657802907915994e-06,
|
| 206380 |
+
"loss": 0.3351,
|
| 206381 |
+
"step": 84115
|
| 206382 |
+
},
|
| 206383 |
+
{
|
| 206384 |
+
"epoch": 677.96,
|
| 206385 |
+
"learning_rate": 8.65772213247173e-06,
|
| 206386 |
+
"loss": 0.5043,
|
| 206387 |
+
"step": 84120
|
| 206388 |
+
},
|
| 206389 |
+
{
|
| 206390 |
+
"epoch": 678.0,
|
| 206391 |
+
"learning_rate": 8.657641357027464e-06,
|
| 206392 |
+
"loss": 1.13,
|
| 206393 |
+
"step": 84125
|
| 206394 |
+
},
|
| 206395 |
+
{
|
| 206396 |
+
"epoch": 678.0,
|
| 206397 |
+
"eval_loss": 0.43177497386932373,
|
| 206398 |
+
"eval_runtime": 41.6294,
|
| 206399 |
+
"eval_samples_per_second": 20.202,
|
| 206400 |
+
"eval_steps_per_second": 0.649,
|
| 206401 |
+
"eval_wer": 0.19316737684084623,
|
| 206402 |
+
"step": 84125
|
| 206403 |
+
},
|
| 206404 |
+
{
|
| 206405 |
+
"epoch": 678.04,
|
| 206406 |
+
"learning_rate": 8.6575605815832e-06,
|
| 206407 |
+
"loss": 0.387,
|
| 206408 |
+
"step": 84130
|
| 206409 |
+
},
|
| 206410 |
+
{
|
| 206411 |
+
"epoch": 678.08,
|
| 206412 |
+
"learning_rate": 8.657479806138934e-06,
|
| 206413 |
+
"loss": 0.2739,
|
| 206414 |
+
"step": 84135
|
| 206415 |
+
},
|
| 206416 |
+
{
|
| 206417 |
+
"epoch": 678.12,
|
| 206418 |
+
"learning_rate": 8.65739903069467e-06,
|
| 206419 |
+
"loss": 0.3239,
|
| 206420 |
+
"step": 84140
|
| 206421 |
+
},
|
| 206422 |
+
{
|
| 206423 |
+
"epoch": 678.16,
|
| 206424 |
+
"learning_rate": 8.657318255250404e-06,
|
| 206425 |
+
"loss": 0.4463,
|
| 206426 |
+
"step": 84145
|
| 206427 |
+
},
|
| 206428 |
+
{
|
| 206429 |
+
"epoch": 678.2,
|
| 206430 |
+
"learning_rate": 8.65723747980614e-06,
|
| 206431 |
+
"loss": 1.1074,
|
| 206432 |
+
"step": 84150
|
| 206433 |
+
},
|
| 206434 |
+
{
|
| 206435 |
+
"epoch": 678.24,
|
| 206436 |
+
"learning_rate": 8.657156704361874e-06,
|
| 206437 |
+
"loss": 0.3395,
|
| 206438 |
+
"step": 84155
|
| 206439 |
+
},
|
| 206440 |
+
{
|
| 206441 |
+
"epoch": 678.28,
|
| 206442 |
+
"learning_rate": 8.65707592891761e-06,
|
| 206443 |
+
"loss": 0.372,
|
| 206444 |
+
"step": 84160
|
| 206445 |
+
},
|
| 206446 |
+
{
|
| 206447 |
+
"epoch": 678.32,
|
| 206448 |
+
"learning_rate": 8.656995153473344e-06,
|
| 206449 |
+
"loss": 0.3044,
|
| 206450 |
+
"step": 84165
|
| 206451 |
+
},
|
| 206452 |
+
{
|
| 206453 |
+
"epoch": 678.36,
|
| 206454 |
+
"learning_rate": 8.65691437802908e-06,
|
| 206455 |
+
"loss": 0.4819,
|
| 206456 |
+
"step": 84170
|
| 206457 |
+
},
|
| 206458 |
+
{
|
| 206459 |
+
"epoch": 678.4,
|
| 206460 |
+
"learning_rate": 8.656833602584816e-06,
|
| 206461 |
+
"loss": 1.3375,
|
| 206462 |
+
"step": 84175
|
| 206463 |
+
},
|
| 206464 |
+
{
|
| 206465 |
+
"epoch": 678.44,
|
| 206466 |
+
"learning_rate": 8.65675282714055e-06,
|
| 206467 |
+
"loss": 0.4084,
|
| 206468 |
+
"step": 84180
|
| 206469 |
+
},
|
| 206470 |
+
{
|
| 206471 |
+
"epoch": 678.48,
|
| 206472 |
+
"learning_rate": 8.656672051696286e-06,
|
| 206473 |
+
"loss": 0.3376,
|
| 206474 |
+
"step": 84185
|
| 206475 |
+
},
|
| 206476 |
+
{
|
| 206477 |
+
"epoch": 678.52,
|
| 206478 |
+
"learning_rate": 8.65659127625202e-06,
|
| 206479 |
+
"loss": 0.3932,
|
| 206480 |
+
"step": 84190
|
| 206481 |
+
},
|
| 206482 |
+
{
|
| 206483 |
+
"epoch": 678.56,
|
| 206484 |
+
"learning_rate": 8.656510500807756e-06,
|
| 206485 |
+
"loss": 0.4701,
|
| 206486 |
+
"step": 84195
|
| 206487 |
+
},
|
| 206488 |
+
{
|
| 206489 |
+
"epoch": 678.6,
|
| 206490 |
+
"learning_rate": 8.65642972536349e-06,
|
| 206491 |
+
"loss": 1.2044,
|
| 206492 |
+
"step": 84200
|
| 206493 |
+
},
|
| 206494 |
+
{
|
| 206495 |
+
"epoch": 678.64,
|
| 206496 |
+
"learning_rate": 8.656348949919226e-06,
|
| 206497 |
+
"loss": 0.9007,
|
| 206498 |
+
"step": 84205
|
| 206499 |
+
},
|
| 206500 |
+
{
|
| 206501 |
+
"epoch": 678.68,
|
| 206502 |
+
"learning_rate": 8.65626817447496e-06,
|
| 206503 |
+
"loss": 0.3381,
|
| 206504 |
+
"step": 84210
|
| 206505 |
+
},
|
| 206506 |
+
{
|
| 206507 |
+
"epoch": 678.72,
|
| 206508 |
+
"learning_rate": 8.656187399030696e-06,
|
| 206509 |
+
"loss": 0.3584,
|
| 206510 |
+
"step": 84215
|
| 206511 |
+
},
|
| 206512 |
+
{
|
| 206513 |
+
"epoch": 678.76,
|
| 206514 |
+
"learning_rate": 8.65610662358643e-06,
|
| 206515 |
+
"loss": 0.5029,
|
| 206516 |
+
"step": 84220
|
| 206517 |
+
},
|
| 206518 |
+
{
|
| 206519 |
+
"epoch": 678.8,
|
| 206520 |
+
"learning_rate": 8.656025848142166e-06,
|
| 206521 |
+
"loss": 1.2432,
|
| 206522 |
+
"step": 84225
|
| 206523 |
+
},
|
| 206524 |
+
{
|
| 206525 |
+
"epoch": 678.84,
|
| 206526 |
+
"learning_rate": 8.655945072697901e-06,
|
| 206527 |
+
"loss": 0.3662,
|
| 206528 |
+
"step": 84230
|
| 206529 |
+
},
|
| 206530 |
+
{
|
| 206531 |
+
"epoch": 678.88,
|
| 206532 |
+
"learning_rate": 8.655864297253636e-06,
|
| 206533 |
+
"loss": 0.2631,
|
| 206534 |
+
"step": 84235
|
| 206535 |
+
},
|
| 206536 |
+
{
|
| 206537 |
+
"epoch": 678.92,
|
| 206538 |
+
"learning_rate": 8.655783521809371e-06,
|
| 206539 |
+
"loss": 0.3069,
|
| 206540 |
+
"step": 84240
|
| 206541 |
+
},
|
| 206542 |
+
{
|
| 206543 |
+
"epoch": 678.96,
|
| 206544 |
+
"learning_rate": 8.655702746365105e-06,
|
| 206545 |
+
"loss": 0.4808,
|
| 206546 |
+
"step": 84245
|
| 206547 |
+
},
|
| 206548 |
+
{
|
| 206549 |
+
"epoch": 679.0,
|
| 206550 |
+
"eval_loss": 0.345355361700058,
|
| 206551 |
+
"eval_runtime": 41.7372,
|
| 206552 |
+
"eval_samples_per_second": 20.15,
|
| 206553 |
+
"eval_steps_per_second": 0.647,
|
| 206554 |
+
"eval_wer": 0.1902878488244342,
|
| 206555 |
+
"step": 84249
|
| 206556 |
+
},
|
| 206557 |
+
{
|
| 206558 |
+
"epoch": 673.01,
|
| 206559 |
+
"learning_rate": 8.655621970920841e-06,
|
| 206560 |
+
"loss": 0.4596,
|
| 206561 |
+
"step": 84250
|
| 206562 |
+
},
|
| 206563 |
+
{
|
| 206564 |
+
"epoch": 673.05,
|
| 206565 |
+
"learning_rate": 8.655541195476575e-06,
|
| 206566 |
+
"loss": 0.3423,
|
| 206567 |
+
"step": 84255
|
| 206568 |
+
},
|
| 206569 |
+
{
|
| 206570 |
+
"epoch": 673.09,
|
| 206571 |
+
"learning_rate": 8.655460420032311e-06,
|
| 206572 |
+
"loss": 0.3191,
|
| 206573 |
+
"step": 84260
|
| 206574 |
+
},
|
| 206575 |
+
{
|
| 206576 |
+
"epoch": 673.13,
|
| 206577 |
+
"learning_rate": 8.655379644588045e-06,
|
| 206578 |
+
"loss": 0.2949,
|
| 206579 |
+
"step": 84265
|
| 206580 |
+
},
|
| 206581 |
+
{
|
| 206582 |
+
"epoch": 673.17,
|
| 206583 |
+
"learning_rate": 8.655298869143781e-06,
|
| 206584 |
+
"loss": 0.6105,
|
| 206585 |
+
"step": 84270
|
| 206586 |
+
},
|
| 206587 |
+
{
|
| 206588 |
+
"epoch": 673.21,
|
| 206589 |
+
"learning_rate": 8.655218093699515e-06,
|
| 206590 |
+
"loss": 1.1691,
|
| 206591 |
+
"step": 84275
|
| 206592 |
+
},
|
| 206593 |
+
{
|
| 206594 |
+
"epoch": 673.25,
|
| 206595 |
+
"learning_rate": 8.655137318255251e-06,
|
| 206596 |
+
"loss": 0.3023,
|
| 206597 |
+
"step": 84280
|
| 206598 |
+
},
|
| 206599 |
+
{
|
| 206600 |
+
"epoch": 673.29,
|
| 206601 |
+
"learning_rate": 8.655056542810985e-06,
|
| 206602 |
+
"loss": 0.2756,
|
| 206603 |
+
"step": 84285
|
| 206604 |
+
},
|
| 206605 |
+
{
|
| 206606 |
+
"epoch": 673.33,
|
| 206607 |
+
"learning_rate": 8.654975767366721e-06,
|
| 206608 |
+
"loss": 0.3465,
|
| 206609 |
+
"step": 84290
|
| 206610 |
+
},
|
| 206611 |
+
{
|
| 206612 |
+
"epoch": 673.37,
|
| 206613 |
+
"learning_rate": 8.654894991922457e-06,
|
| 206614 |
+
"loss": 0.6593,
|
| 206615 |
+
"step": 84295
|
| 206616 |
+
},
|
| 206617 |
+
{
|
| 206618 |
+
"epoch": 673.41,
|
| 206619 |
+
"learning_rate": 8.654814216478191e-06,
|
| 206620 |
+
"loss": 1.3793,
|
| 206621 |
+
"step": 84300
|
| 206622 |
+
},
|
| 206623 |
+
{
|
| 206624 |
+
"epoch": 673.45,
|
| 206625 |
+
"learning_rate": 8.654733441033927e-06,
|
| 206626 |
+
"loss": 0.3088,
|
| 206627 |
+
"step": 84305
|
| 206628 |
+
},
|
| 206629 |
+
{
|
| 206630 |
+
"epoch": 673.49,
|
| 206631 |
+
"learning_rate": 8.654652665589661e-06,
|
| 206632 |
+
"loss": 0.3144,
|
| 206633 |
+
"step": 84310
|
| 206634 |
+
},
|
| 206635 |
+
{
|
| 206636 |
+
"epoch": 673.53,
|
| 206637 |
+
"learning_rate": 8.654571890145397e-06,
|
| 206638 |
+
"loss": 0.4673,
|
| 206639 |
+
"step": 84315
|
| 206640 |
+
},
|
| 206641 |
+
{
|
| 206642 |
+
"epoch": 673.57,
|
| 206643 |
+
"learning_rate": 8.654491114701131e-06,
|
| 206644 |
+
"loss": 0.5631,
|
| 206645 |
+
"step": 84320
|
| 206646 |
+
},
|
| 206647 |
+
{
|
| 206648 |
+
"epoch": 673.61,
|
| 206649 |
+
"learning_rate": 8.654410339256867e-06,
|
| 206650 |
+
"loss": 1.1314,
|
| 206651 |
+
"step": 84325
|
| 206652 |
+
},
|
| 206653 |
+
{
|
| 206654 |
+
"epoch": 673.65,
|
| 206655 |
+
"learning_rate": 8.654329563812601e-06,
|
| 206656 |
+
"loss": 0.3437,
|
| 206657 |
+
"step": 84330
|
| 206658 |
+
},
|
| 206659 |
+
{
|
| 206660 |
+
"epoch": 673.69,
|
| 206661 |
+
"learning_rate": 8.654248788368337e-06,
|
| 206662 |
+
"loss": 0.3081,
|
| 206663 |
+
"step": 84335
|
| 206664 |
+
},
|
| 206665 |
+
{
|
| 206666 |
+
"epoch": 673.73,
|
| 206667 |
+
"learning_rate": 8.654168012924071e-06,
|
| 206668 |
+
"loss": 0.3422,
|
| 206669 |
+
"step": 84340
|
| 206670 |
+
},
|
| 206671 |
+
{
|
| 206672 |
+
"epoch": 673.77,
|
| 206673 |
+
"learning_rate": 8.654087237479807e-06,
|
| 206674 |
+
"loss": 0.5191,
|
| 206675 |
+
"step": 84345
|
| 206676 |
+
},
|
| 206677 |
+
{
|
| 206678 |
+
"epoch": 673.81,
|
| 206679 |
+
"learning_rate": 8.654006462035543e-06,
|
| 206680 |
+
"loss": 1.1358,
|
| 206681 |
+
"step": 84350
|
| 206682 |
+
},
|
| 206683 |
+
{
|
| 206684 |
+
"epoch": 673.85,
|
| 206685 |
+
"learning_rate": 8.653925686591277e-06,
|
| 206686 |
+
"loss": 0.3092,
|
| 206687 |
+
"step": 84355
|
| 206688 |
+
},
|
| 206689 |
+
{
|
| 206690 |
+
"epoch": 673.89,
|
| 206691 |
+
"learning_rate": 8.653844911147013e-06,
|
| 206692 |
+
"loss": 0.2932,
|
| 206693 |
+
"step": 84360
|
| 206694 |
+
},
|
| 206695 |
+
{
|
| 206696 |
+
"epoch": 673.93,
|
| 206697 |
+
"learning_rate": 8.653764135702747e-06,
|
| 206698 |
+
"loss": 0.336,
|
| 206699 |
+
"step": 84365
|
| 206700 |
+
},
|
| 206701 |
+
{
|
| 206702 |
+
"epoch": 673.97,
|
| 206703 |
+
"learning_rate": 8.653683360258483e-06,
|
| 206704 |
+
"loss": 0.5495,
|
| 206705 |
+
"step": 84370
|
| 206706 |
+
},
|
| 206707 |
+
{
|
| 206708 |
+
"epoch": 674.0,
|
| 206709 |
+
"eval_loss": 0.3955157697200775,
|
| 206710 |
+
"eval_runtime": 43.0467,
|
| 206711 |
+
"eval_samples_per_second": 19.514,
|
| 206712 |
+
"eval_steps_per_second": 0.627,
|
| 206713 |
+
"eval_wer": 0.18563897991166833,
|
| 206714 |
+
"step": 84374
|
| 206715 |
+
},
|
| 206716 |
+
{
|
| 206717 |
+
"epoch": 674.01,
|
| 206718 |
+
"learning_rate": 8.653602584814217e-06,
|
| 206719 |
+
"loss": 0.3837,
|
| 206720 |
+
"step": 84375
|
| 206721 |
+
},
|
| 206722 |
+
{
|
| 206723 |
+
"epoch": 674.05,
|
| 206724 |
+
"learning_rate": 8.653521809369953e-06,
|
| 206725 |
+
"loss": 0.318,
|
| 206726 |
+
"step": 84380
|
| 206727 |
+
},
|
| 206728 |
+
{
|
| 206729 |
+
"epoch": 674.09,
|
| 206730 |
+
"learning_rate": 8.653441033925687e-06,
|
| 206731 |
+
"loss": 0.2944,
|
| 206732 |
+
"step": 84385
|
| 206733 |
+
},
|
| 206734 |
+
{
|
| 206735 |
+
"epoch": 674.13,
|
| 206736 |
+
"learning_rate": 8.653360258481423e-06,
|
| 206737 |
+
"loss": 0.3259,
|
| 206738 |
+
"step": 84390
|
| 206739 |
+
},
|
| 206740 |
+
{
|
| 206741 |
+
"epoch": 674.17,
|
| 206742 |
+
"learning_rate": 8.653279483037157e-06,
|
| 206743 |
+
"loss": 0.5183,
|
| 206744 |
+
"step": 84395
|
| 206745 |
+
},
|
| 206746 |
+
{
|
| 206747 |
+
"epoch": 674.21,
|
| 206748 |
+
"learning_rate": 8.653198707592893e-06,
|
| 206749 |
+
"loss": 1.1968,
|
| 206750 |
+
"step": 84400
|
| 206751 |
+
},
|
| 206752 |
+
{
|
| 206753 |
+
"epoch": 674.25,
|
| 206754 |
+
"learning_rate": 8.653117932148628e-06,
|
| 206755 |
+
"loss": 0.4829,
|
| 206756 |
+
"step": 84405
|
| 206757 |
+
},
|
| 206758 |
+
{
|
| 206759 |
+
"epoch": 674.29,
|
| 206760 |
+
"learning_rate": 8.653037156704363e-06,
|
| 206761 |
+
"loss": 0.3363,
|
| 206762 |
+
"step": 84410
|
| 206763 |
+
},
|
| 206764 |
+
{
|
| 206765 |
+
"epoch": 674.33,
|
| 206766 |
+
"learning_rate": 8.652956381260098e-06,
|
| 206767 |
+
"loss": 0.3947,
|
| 206768 |
+
"step": 84415
|
| 206769 |
+
},
|
| 206770 |
+
{
|
| 206771 |
+
"epoch": 674.37,
|
| 206772 |
+
"learning_rate": 8.652875605815833e-06,
|
| 206773 |
+
"loss": 0.6282,
|
| 206774 |
+
"step": 84420
|
| 206775 |
+
},
|
| 206776 |
+
{
|
| 206777 |
+
"epoch": 674.41,
|
| 206778 |
+
"learning_rate": 8.652794830371568e-06,
|
| 206779 |
+
"loss": 1.218,
|
| 206780 |
+
"step": 84425
|
| 206781 |
+
},
|
| 206782 |
+
{
|
| 206783 |
+
"epoch": 674.45,
|
| 206784 |
+
"learning_rate": 8.652714054927303e-06,
|
| 206785 |
+
"loss": 0.3229,
|
| 206786 |
+
"step": 84430
|
| 206787 |
+
},
|
| 206788 |
+
{
|
| 206789 |
+
"epoch": 674.49,
|
| 206790 |
+
"learning_rate": 8.652633279483038e-06,
|
| 206791 |
+
"loss": 0.3122,
|
| 206792 |
+
"step": 84435
|
| 206793 |
+
},
|
| 206794 |
+
{
|
| 206795 |
+
"epoch": 674.53,
|
| 206796 |
+
"learning_rate": 8.652552504038773e-06,
|
| 206797 |
+
"loss": 0.3114,
|
| 206798 |
+
"step": 84440
|
| 206799 |
+
},
|
| 206800 |
+
{
|
| 206801 |
+
"epoch": 674.57,
|
| 206802 |
+
"learning_rate": 8.652471728594508e-06,
|
| 206803 |
+
"loss": 0.525,
|
| 206804 |
+
"step": 84445
|
| 206805 |
+
},
|
| 206806 |
+
{
|
| 206807 |
+
"epoch": 674.61,
|
| 206808 |
+
"learning_rate": 8.652390953150243e-06,
|
| 206809 |
+
"loss": 1.1697,
|
| 206810 |
+
"step": 84450
|
| 206811 |
+
},
|
| 206812 |
+
{
|
| 206813 |
+
"epoch": 674.65,
|
| 206814 |
+
"learning_rate": 8.652310177705978e-06,
|
| 206815 |
+
"loss": 0.3382,
|
| 206816 |
+
"step": 84455
|
| 206817 |
+
},
|
| 206818 |
+
{
|
| 206819 |
+
"epoch": 674.69,
|
| 206820 |
+
"learning_rate": 8.652229402261713e-06,
|
| 206821 |
+
"loss": 0.2772,
|
| 206822 |
+
"step": 84460
|
| 206823 |
+
},
|
| 206824 |
+
{
|
| 206825 |
+
"epoch": 674.73,
|
| 206826 |
+
"learning_rate": 8.652148626817448e-06,
|
| 206827 |
+
"loss": 0.4419,
|
| 206828 |
+
"step": 84465
|
| 206829 |
+
},
|
| 206830 |
+
{
|
| 206831 |
+
"epoch": 674.77,
|
| 206832 |
+
"learning_rate": 8.652067851373184e-06,
|
| 206833 |
+
"loss": 0.4908,
|
| 206834 |
+
"step": 84470
|
| 206835 |
+
},
|
| 206836 |
+
{
|
| 206837 |
+
"epoch": 674.81,
|
| 206838 |
+
"learning_rate": 8.651987075928918e-06,
|
| 206839 |
+
"loss": 1.2181,
|
| 206840 |
+
"step": 84475
|
| 206841 |
+
},
|
| 206842 |
+
{
|
| 206843 |
+
"epoch": 674.85,
|
| 206844 |
+
"learning_rate": 8.651906300484654e-06,
|
| 206845 |
+
"loss": 0.3422,
|
| 206846 |
+
"step": 84480
|
| 206847 |
+
},
|
| 206848 |
+
{
|
| 206849 |
+
"epoch": 674.89,
|
| 206850 |
+
"learning_rate": 8.651825525040388e-06,
|
| 206851 |
+
"loss": 0.2761,
|
| 206852 |
+
"step": 84485
|
| 206853 |
+
},
|
| 206854 |
+
{
|
| 206855 |
+
"epoch": 674.93,
|
| 206856 |
+
"learning_rate": 8.651744749596124e-06,
|
| 206857 |
+
"loss": 0.3842,
|
| 206858 |
+
"step": 84490
|
| 206859 |
+
},
|
| 206860 |
+
{
|
| 206861 |
+
"epoch": 674.97,
|
| 206862 |
+
"learning_rate": 8.651663974151858e-06,
|
| 206863 |
+
"loss": 0.595,
|
| 206864 |
+
"step": 84495
|
| 206865 |
+
},
|
| 206866 |
+
{
|
| 206867 |
+
"epoch": 675.0,
|
| 206868 |
+
"eval_loss": 0.3872044086456299,
|
| 206869 |
+
"eval_runtime": 42.1888,
|
| 206870 |
+
"eval_samples_per_second": 19.91,
|
| 206871 |
+
"eval_steps_per_second": 0.64,
|
| 206872 |
+
"eval_wer": 0.1955033508683433,
|
| 206873 |
+
"step": 84499
|
| 206874 |
+
},
|
| 206875 |
+
{
|
| 206876 |
+
"epoch": 681.01,
|
| 206877 |
+
"learning_rate": 8.651583198707594e-06,
|
| 206878 |
+
"loss": 0.3997,
|
| 206879 |
+
"step": 84500
|
| 206880 |
+
},
|
| 206881 |
+
{
|
| 206882 |
+
"epoch": 681.05,
|
| 206883 |
+
"learning_rate": 8.651502423263328e-06,
|
| 206884 |
+
"loss": 0.3117,
|
| 206885 |
+
"step": 84505
|
| 206886 |
+
},
|
| 206887 |
+
{
|
| 206888 |
+
"epoch": 681.09,
|
| 206889 |
+
"learning_rate": 8.651421647819064e-06,
|
| 206890 |
+
"loss": 0.2902,
|
| 206891 |
+
"step": 84510
|
| 206892 |
+
},
|
| 206893 |
+
{
|
| 206894 |
+
"epoch": 681.13,
|
| 206895 |
+
"learning_rate": 8.651340872374798e-06,
|
| 206896 |
+
"loss": 0.3179,
|
| 206897 |
+
"step": 84515
|
| 206898 |
+
},
|
| 206899 |
+
{
|
| 206900 |
+
"epoch": 681.17,
|
| 206901 |
+
"learning_rate": 8.651260096930534e-06,
|
| 206902 |
+
"loss": 0.6692,
|
| 206903 |
+
"step": 84520
|
| 206904 |
+
},
|
| 206905 |
+
{
|
| 206906 |
+
"epoch": 681.21,
|
| 206907 |
+
"learning_rate": 8.65117932148627e-06,
|
| 206908 |
+
"loss": 1.0544,
|
| 206909 |
+
"step": 84525
|
| 206910 |
+
},
|
| 206911 |
+
{
|
| 206912 |
+
"epoch": 681.25,
|
| 206913 |
+
"learning_rate": 8.651098546042004e-06,
|
| 206914 |
+
"loss": 0.3278,
|
| 206915 |
+
"step": 84530
|
| 206916 |
+
},
|
| 206917 |
+
{
|
| 206918 |
+
"epoch": 681.29,
|
| 206919 |
+
"learning_rate": 8.65101777059774e-06,
|
| 206920 |
+
"loss": 0.3176,
|
| 206921 |
+
"step": 84535
|
| 206922 |
+
},
|
| 206923 |
+
{
|
| 206924 |
+
"epoch": 681.33,
|
| 206925 |
+
"learning_rate": 8.650936995153474e-06,
|
| 206926 |
+
"loss": 0.3031,
|
| 206927 |
+
"step": 84540
|
| 206928 |
+
},
|
| 206929 |
+
{
|
| 206930 |
+
"epoch": 681.37,
|
| 206931 |
+
"learning_rate": 8.65085621970921e-06,
|
| 206932 |
+
"loss": 0.5016,
|
| 206933 |
+
"step": 84545
|
| 206934 |
+
},
|
| 206935 |
+
{
|
| 206936 |
+
"epoch": 681.41,
|
| 206937 |
+
"learning_rate": 8.650775444264944e-06,
|
| 206938 |
+
"loss": 1.1741,
|
| 206939 |
+
"step": 84550
|
| 206940 |
+
},
|
| 206941 |
+
{
|
| 206942 |
+
"epoch": 681.45,
|
| 206943 |
+
"learning_rate": 8.65069466882068e-06,
|
| 206944 |
+
"loss": 0.3562,
|
| 206945 |
+
"step": 84555
|
| 206946 |
+
},
|
| 206947 |
+
{
|
| 206948 |
+
"epoch": 681.49,
|
| 206949 |
+
"learning_rate": 8.650613893376414e-06,
|
| 206950 |
+
"loss": 0.3223,
|
| 206951 |
+
"step": 84560
|
| 206952 |
+
},
|
| 206953 |
+
{
|
| 206954 |
+
"epoch": 681.53,
|
| 206955 |
+
"learning_rate": 8.65053311793215e-06,
|
| 206956 |
+
"loss": 0.4556,
|
| 206957 |
+
"step": 84565
|
| 206958 |
+
},
|
| 206959 |
+
{
|
| 206960 |
+
"epoch": 681.57,
|
| 206961 |
+
"learning_rate": 8.650452342487884e-06,
|
| 206962 |
+
"loss": 0.5309,
|
| 206963 |
+
"step": 84570
|
| 206964 |
+
},
|
| 206965 |
+
{
|
| 206966 |
+
"epoch": 681.61,
|
| 206967 |
+
"learning_rate": 8.65037156704362e-06,
|
| 206968 |
+
"loss": 1.1481,
|
| 206969 |
+
"step": 84575
|
| 206970 |
+
},
|
| 206971 |
+
{
|
| 206972 |
+
"epoch": 681.65,
|
| 206973 |
+
"learning_rate": 8.650290791599356e-06,
|
| 206974 |
+
"loss": 0.2845,
|
| 206975 |
+
"step": 84580
|
| 206976 |
+
},
|
| 206977 |
+
{
|
| 206978 |
+
"epoch": 681.69,
|
| 206979 |
+
"learning_rate": 8.65021001615509e-06,
|
| 206980 |
+
"loss": 0.2835,
|
| 206981 |
+
"step": 84585
|
| 206982 |
+
},
|
| 206983 |
+
{
|
| 206984 |
+
"epoch": 681.73,
|
| 206985 |
+
"learning_rate": 8.650129240710826e-06,
|
| 206986 |
+
"loss": 0.3701,
|
| 206987 |
+
"step": 84590
|
| 206988 |
+
},
|
| 206989 |
+
{
|
| 206990 |
+
"epoch": 681.77,
|
| 206991 |
+
"learning_rate": 8.65004846526656e-06,
|
| 206992 |
+
"loss": 0.6443,
|
| 206993 |
+
"step": 84595
|
| 206994 |
+
},
|
| 206995 |
+
{
|
| 206996 |
+
"epoch": 681.81,
|
| 206997 |
+
"learning_rate": 8.649967689822296e-06,
|
| 206998 |
+
"loss": 1.162,
|
| 206999 |
+
"step": 84600
|
| 207000 |
+
},
|
| 207001 |
+
{
|
| 207002 |
+
"epoch": 681.85,
|
| 207003 |
+
"learning_rate": 8.64988691437803e-06,
|
| 207004 |
+
"loss": 0.4954,
|
| 207005 |
+
"step": 84605
|
| 207006 |
+
},
|
| 207007 |
+
{
|
| 207008 |
+
"epoch": 681.89,
|
| 207009 |
+
"learning_rate": 8.649806138933766e-06,
|
| 207010 |
+
"loss": 0.3061,
|
| 207011 |
+
"step": 84610
|
| 207012 |
+
},
|
| 207013 |
+
{
|
| 207014 |
+
"epoch": 681.93,
|
| 207015 |
+
"learning_rate": 8.6497253634895e-06,
|
| 207016 |
+
"loss": 0.3989,
|
| 207017 |
+
"step": 84615
|
| 207018 |
+
},
|
| 207019 |
+
{
|
| 207020 |
+
"epoch": 681.97,
|
| 207021 |
+
"learning_rate": 8.649644588045235e-06,
|
| 207022 |
+
"loss": 0.6317,
|
| 207023 |
+
"step": 84620
|
| 207024 |
+
},
|
| 207025 |
+
{
|
| 207026 |
+
"epoch": 682.0,
|
| 207027 |
+
"eval_loss": 0.39930081367492676,
|
| 207028 |
+
"eval_runtime": 41.9885,
|
| 207029 |
+
"eval_samples_per_second": 20.029,
|
| 207030 |
+
"eval_steps_per_second": 0.643,
|
| 207031 |
+
"eval_wer": 0.1948430329958482,
|
| 207032 |
+
"step": 84623
|
| 207033 |
}
|
| 207034 |
],
|
| 207035 |
+
"max_steps": 620000,
|
| 207036 |
"num_train_epochs": 5000,
|
| 207037 |
+
"total_flos": 2.381427276537856e+20,
|
| 207038 |
"trial_name": null,
|
| 207039 |
"trial_params": null
|
| 207040 |
}
|
model-bin/finetune/base/{checkpoint-84001 → checkpoint-84623}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629918685.2041333/events.out.tfevents.1629918685.7e498afd5545.7645.125
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af115ebcaa932720ca5f61eac12c1708c971478f9031fd7c71f6bbdcd38d3114
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629919158.470076/events.out.tfevents.1629919158.7e498afd5545.7645.127
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:152bd11c3fb723992d5901fdfde6d8724f4bcc1a1bee816c0b0a880eca20636d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629919633.5511148/events.out.tfevents.1629919633.7e498afd5545.7645.129
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:def08badfc86d444a8006546440dbddcd86e44438e527d09dd537f3455e44d1e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629920107.9077082/events.out.tfevents.1629920107.7e498afd5545.7645.131
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97bbc2bebdf36e1a4f62a81001727c236d3d0196edcb6a6654510774d3b1c8b1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629920581.73902/events.out.tfevents.1629920581.7e498afd5545.7645.133
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14b8209ac7cf25e41ee5f9e3c64d3a277beace45ad9b524942a7171ddc5fd0db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629918685.7e498afd5545.7645.124
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:807ea28568a6a9358dd92912de8f1dc795c701fc2c28be7c9c9ead23f8edbbae
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629919158.7e498afd5545.7645.126
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:332cb27651394d717d50ee48bb0807af978196c81a9a44e788d41c0261d8f4ef
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629919633.7e498afd5545.7645.128
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:34af7bbdde9fae14e9fab7148897e7fed6c4bfce53241968c4a782cce479a2e0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629920107.7e498afd5545.7645.130
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95abb7889cbebd87b7ee44cf4e87723817c32595a60340dfdea0f5d6477bf475
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629920581.7e498afd5545.7645.132
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bd9bc5b503e356d1ca2146c19f89352640811ad781f5377fb5908a72a9f979c
|
| 3 |
+
size 8622
|