"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629896881.4320064/events.out.tfevents.1629896881.7e498afd5545.7645.35 +3 -0
- model-bin/finetune/base/log/1629897354.142085/events.out.tfevents.1629897354.7e498afd5545.7645.37 +3 -0
- model-bin/finetune/base/log/1629897817.6067963/events.out.tfevents.1629897817.7e498afd5545.7645.39 +3 -0
- model-bin/finetune/base/log/1629898283.2853954/events.out.tfevents.1629898283.7e498afd5545.7645.41 +3 -0
- model-bin/finetune/base/log/1629898756.3973496/events.out.tfevents.1629898756.7e498afd5545.7645.43 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629896881.7e498afd5545.7645.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629897354.7e498afd5545.7645.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629897817.7e498afd5545.7645.38 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629898283.7e498afd5545.7645.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629898756.7e498afd5545.7645.42 +3 -0
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/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 sha256:83f64e4cff009a5eded3493ab62103df495f530c979f6cf26437d1f53ff63216
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/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 sha256:4c7efd6688eb76b3c85ac4778a8f34067f39e8f7792b58fe6f5237972aecfccd
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e2855477cf4f44f6936d902653afb56e3aec4404923ab7e41a06c920deac1b3
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/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 sha256:22dd04c6be986b933dbf36885b7a308ab00202782f956c433993fd7d5ef14eee
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/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 sha256:5953ed201d78a6b8ec30853512b2de7c9867c7338b9888cde9a2224b7fb7f365
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -199116,11 +199116,800 @@
|
|
| 199116 |
"eval_steps_per_second": 0.65,
|
| 199117 |
"eval_wer": 0.19250363901018921,
|
| 199118 |
"step": 78401
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 199119 |
}
|
| 199120 |
],
|
| 199121 |
-
"max_steps":
|
| 199122 |
"num_train_epochs": 5000,
|
| 199123 |
-
"total_flos": 2.
|
| 199124 |
"trial_name": null,
|
| 199125 |
"trial_params": null
|
| 199126 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 636.995983935743,
|
| 5 |
+
"global_step": 79024,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 199116 |
"eval_steps_per_second": 0.65,
|
| 199117 |
"eval_wer": 0.19250363901018921,
|
| 199118 |
"step": 78401
|
| 199119 |
+
},
|
| 199120 |
+
{
|
| 199121 |
+
"epoch": 627.03,
|
| 199122 |
+
"learning_rate": 8.750016155088854e-06,
|
| 199123 |
+
"loss": 0.3852,
|
| 199124 |
+
"step": 78405
|
| 199125 |
+
},
|
| 199126 |
+
{
|
| 199127 |
+
"epoch": 627.07,
|
| 199128 |
+
"learning_rate": 8.749935379644588e-06,
|
| 199129 |
+
"loss": 0.334,
|
| 199130 |
+
"step": 78410
|
| 199131 |
+
},
|
| 199132 |
+
{
|
| 199133 |
+
"epoch": 627.11,
|
| 199134 |
+
"learning_rate": 8.749854604200324e-06,
|
| 199135 |
+
"loss": 0.299,
|
| 199136 |
+
"step": 78415
|
| 199137 |
+
},
|
| 199138 |
+
{
|
| 199139 |
+
"epoch": 627.15,
|
| 199140 |
+
"learning_rate": 8.749773828756058e-06,
|
| 199141 |
+
"loss": 0.4201,
|
| 199142 |
+
"step": 78420
|
| 199143 |
+
},
|
| 199144 |
+
{
|
| 199145 |
+
"epoch": 627.19,
|
| 199146 |
+
"learning_rate": 8.749693053311794e-06,
|
| 199147 |
+
"loss": 0.9708,
|
| 199148 |
+
"step": 78425
|
| 199149 |
+
},
|
| 199150 |
+
{
|
| 199151 |
+
"epoch": 627.23,
|
| 199152 |
+
"learning_rate": 8.749612277867528e-06,
|
| 199153 |
+
"loss": 0.6433,
|
| 199154 |
+
"step": 78430
|
| 199155 |
+
},
|
| 199156 |
+
{
|
| 199157 |
+
"epoch": 627.27,
|
| 199158 |
+
"learning_rate": 8.749531502423264e-06,
|
| 199159 |
+
"loss": 0.3363,
|
| 199160 |
+
"step": 78435
|
| 199161 |
+
},
|
| 199162 |
+
{
|
| 199163 |
+
"epoch": 627.31,
|
| 199164 |
+
"learning_rate": 8.749450726978998e-06,
|
| 199165 |
+
"loss": 0.3599,
|
| 199166 |
+
"step": 78440
|
| 199167 |
+
},
|
| 199168 |
+
{
|
| 199169 |
+
"epoch": 627.35,
|
| 199170 |
+
"learning_rate": 8.749369951534734e-06,
|
| 199171 |
+
"loss": 0.4116,
|
| 199172 |
+
"step": 78445
|
| 199173 |
+
},
|
| 199174 |
+
{
|
| 199175 |
+
"epoch": 627.39,
|
| 199176 |
+
"learning_rate": 8.74928917609047e-06,
|
| 199177 |
+
"loss": 1.0044,
|
| 199178 |
+
"step": 78450
|
| 199179 |
+
},
|
| 199180 |
+
{
|
| 199181 |
+
"epoch": 627.43,
|
| 199182 |
+
"learning_rate": 8.749208400646204e-06,
|
| 199183 |
+
"loss": 0.8095,
|
| 199184 |
+
"step": 78455
|
| 199185 |
+
},
|
| 199186 |
+
{
|
| 199187 |
+
"epoch": 627.47,
|
| 199188 |
+
"learning_rate": 8.74912762520194e-06,
|
| 199189 |
+
"loss": 0.3486,
|
| 199190 |
+
"step": 78460
|
| 199191 |
+
},
|
| 199192 |
+
{
|
| 199193 |
+
"epoch": 627.51,
|
| 199194 |
+
"learning_rate": 8.749046849757674e-06,
|
| 199195 |
+
"loss": 0.2948,
|
| 199196 |
+
"step": 78465
|
| 199197 |
+
},
|
| 199198 |
+
{
|
| 199199 |
+
"epoch": 627.55,
|
| 199200 |
+
"learning_rate": 8.74896607431341e-06,
|
| 199201 |
+
"loss": 0.4802,
|
| 199202 |
+
"step": 78470
|
| 199203 |
+
},
|
| 199204 |
+
{
|
| 199205 |
+
"epoch": 627.59,
|
| 199206 |
+
"learning_rate": 8.748885298869144e-06,
|
| 199207 |
+
"loss": 0.9969,
|
| 199208 |
+
"step": 78475
|
| 199209 |
+
},
|
| 199210 |
+
{
|
| 199211 |
+
"epoch": 627.63,
|
| 199212 |
+
"learning_rate": 8.74880452342488e-06,
|
| 199213 |
+
"loss": 0.662,
|
| 199214 |
+
"step": 78480
|
| 199215 |
+
},
|
| 199216 |
+
{
|
| 199217 |
+
"epoch": 627.67,
|
| 199218 |
+
"learning_rate": 8.748723747980614e-06,
|
| 199219 |
+
"loss": 0.3773,
|
| 199220 |
+
"step": 78485
|
| 199221 |
+
},
|
| 199222 |
+
{
|
| 199223 |
+
"epoch": 627.71,
|
| 199224 |
+
"learning_rate": 8.74864297253635e-06,
|
| 199225 |
+
"loss": 0.3748,
|
| 199226 |
+
"step": 78490
|
| 199227 |
+
},
|
| 199228 |
+
{
|
| 199229 |
+
"epoch": 627.75,
|
| 199230 |
+
"learning_rate": 8.748562197092084e-06,
|
| 199231 |
+
"loss": 0.4032,
|
| 199232 |
+
"step": 78495
|
| 199233 |
+
},
|
| 199234 |
+
{
|
| 199235 |
+
"epoch": 627.79,
|
| 199236 |
+
"learning_rate": 8.74848142164782e-06,
|
| 199237 |
+
"loss": 0.9498,
|
| 199238 |
+
"step": 78500
|
| 199239 |
+
},
|
| 199240 |
+
{
|
| 199241 |
+
"epoch": 627.83,
|
| 199242 |
+
"learning_rate": 8.748400646203556e-06,
|
| 199243 |
+
"loss": 0.6955,
|
| 199244 |
+
"step": 78505
|
| 199245 |
+
},
|
| 199246 |
+
{
|
| 199247 |
+
"epoch": 627.87,
|
| 199248 |
+
"learning_rate": 8.74831987075929e-06,
|
| 199249 |
+
"loss": 0.3037,
|
| 199250 |
+
"step": 78510
|
| 199251 |
+
},
|
| 199252 |
+
{
|
| 199253 |
+
"epoch": 627.91,
|
| 199254 |
+
"learning_rate": 8.748239095315026e-06,
|
| 199255 |
+
"loss": 0.3241,
|
| 199256 |
+
"step": 78515
|
| 199257 |
+
},
|
| 199258 |
+
{
|
| 199259 |
+
"epoch": 627.95,
|
| 199260 |
+
"learning_rate": 8.74815831987076e-06,
|
| 199261 |
+
"loss": 0.4145,
|
| 199262 |
+
"step": 78520
|
| 199263 |
+
},
|
| 199264 |
+
{
|
| 199265 |
+
"epoch": 627.99,
|
| 199266 |
+
"learning_rate": 8.748077544426495e-06,
|
| 199267 |
+
"loss": 0.907,
|
| 199268 |
+
"step": 78525
|
| 199269 |
+
},
|
| 199270 |
+
{
|
| 199271 |
+
"epoch": 628.0,
|
| 199272 |
+
"eval_loss": 0.4018295109272003,
|
| 199273 |
+
"eval_runtime": 44.0091,
|
| 199274 |
+
"eval_samples_per_second": 18.996,
|
| 199275 |
+
"eval_steps_per_second": 0.614,
|
| 199276 |
+
"eval_wer": 0.1865575183110728,
|
| 199277 |
+
"step": 78526
|
| 199278 |
+
},
|
| 199279 |
+
{
|
| 199280 |
+
"epoch": 628.03,
|
| 199281 |
+
"learning_rate": 8.74799676898223e-06,
|
| 199282 |
+
"loss": 0.3616,
|
| 199283 |
+
"step": 78530
|
| 199284 |
+
},
|
| 199285 |
+
{
|
| 199286 |
+
"epoch": 628.07,
|
| 199287 |
+
"learning_rate": 8.747915993537965e-06,
|
| 199288 |
+
"loss": 0.3102,
|
| 199289 |
+
"step": 78535
|
| 199290 |
+
},
|
| 199291 |
+
{
|
| 199292 |
+
"epoch": 628.11,
|
| 199293 |
+
"learning_rate": 8.7478352180937e-06,
|
| 199294 |
+
"loss": 0.327,
|
| 199295 |
+
"step": 78540
|
| 199296 |
+
},
|
| 199297 |
+
{
|
| 199298 |
+
"epoch": 628.15,
|
| 199299 |
+
"learning_rate": 8.747754442649435e-06,
|
| 199300 |
+
"loss": 0.4266,
|
| 199301 |
+
"step": 78545
|
| 199302 |
+
},
|
| 199303 |
+
{
|
| 199304 |
+
"epoch": 628.19,
|
| 199305 |
+
"learning_rate": 8.74767366720517e-06,
|
| 199306 |
+
"loss": 1.018,
|
| 199307 |
+
"step": 78550
|
| 199308 |
+
},
|
| 199309 |
+
{
|
| 199310 |
+
"epoch": 628.23,
|
| 199311 |
+
"learning_rate": 8.747592891760905e-06,
|
| 199312 |
+
"loss": 0.6763,
|
| 199313 |
+
"step": 78555
|
| 199314 |
+
},
|
| 199315 |
+
{
|
| 199316 |
+
"epoch": 628.27,
|
| 199317 |
+
"learning_rate": 8.747512116316641e-06,
|
| 199318 |
+
"loss": 0.2833,
|
| 199319 |
+
"step": 78560
|
| 199320 |
+
},
|
| 199321 |
+
{
|
| 199322 |
+
"epoch": 628.31,
|
| 199323 |
+
"learning_rate": 8.747431340872375e-06,
|
| 199324 |
+
"loss": 0.3231,
|
| 199325 |
+
"step": 78565
|
| 199326 |
+
},
|
| 199327 |
+
{
|
| 199328 |
+
"epoch": 628.35,
|
| 199329 |
+
"learning_rate": 8.747350565428111e-06,
|
| 199330 |
+
"loss": 0.362,
|
| 199331 |
+
"step": 78570
|
| 199332 |
+
},
|
| 199333 |
+
{
|
| 199334 |
+
"epoch": 628.39,
|
| 199335 |
+
"learning_rate": 8.747269789983845e-06,
|
| 199336 |
+
"loss": 0.8538,
|
| 199337 |
+
"step": 78575
|
| 199338 |
+
},
|
| 199339 |
+
{
|
| 199340 |
+
"epoch": 628.43,
|
| 199341 |
+
"learning_rate": 8.747189014539581e-06,
|
| 199342 |
+
"loss": 0.6172,
|
| 199343 |
+
"step": 78580
|
| 199344 |
+
},
|
| 199345 |
+
{
|
| 199346 |
+
"epoch": 628.47,
|
| 199347 |
+
"learning_rate": 8.747108239095315e-06,
|
| 199348 |
+
"loss": 0.2573,
|
| 199349 |
+
"step": 78585
|
| 199350 |
+
},
|
| 199351 |
+
{
|
| 199352 |
+
"epoch": 628.51,
|
| 199353 |
+
"learning_rate": 8.747027463651051e-06,
|
| 199354 |
+
"loss": 0.2711,
|
| 199355 |
+
"step": 78590
|
| 199356 |
+
},
|
| 199357 |
+
{
|
| 199358 |
+
"epoch": 628.55,
|
| 199359 |
+
"learning_rate": 8.746946688206785e-06,
|
| 199360 |
+
"loss": 0.5042,
|
| 199361 |
+
"step": 78595
|
| 199362 |
+
},
|
| 199363 |
+
{
|
| 199364 |
+
"epoch": 628.59,
|
| 199365 |
+
"learning_rate": 8.746865912762521e-06,
|
| 199366 |
+
"loss": 1.0298,
|
| 199367 |
+
"step": 78600
|
| 199368 |
+
},
|
| 199369 |
+
{
|
| 199370 |
+
"epoch": 628.63,
|
| 199371 |
+
"learning_rate": 8.746785137318255e-06,
|
| 199372 |
+
"loss": 0.7017,
|
| 199373 |
+
"step": 78605
|
| 199374 |
+
},
|
| 199375 |
+
{
|
| 199376 |
+
"epoch": 628.67,
|
| 199377 |
+
"learning_rate": 8.746704361873991e-06,
|
| 199378 |
+
"loss": 0.3638,
|
| 199379 |
+
"step": 78610
|
| 199380 |
+
},
|
| 199381 |
+
{
|
| 199382 |
+
"epoch": 628.71,
|
| 199383 |
+
"learning_rate": 8.746623586429725e-06,
|
| 199384 |
+
"loss": 0.3454,
|
| 199385 |
+
"step": 78615
|
| 199386 |
+
},
|
| 199387 |
+
{
|
| 199388 |
+
"epoch": 628.75,
|
| 199389 |
+
"learning_rate": 8.746542810985461e-06,
|
| 199390 |
+
"loss": 0.4958,
|
| 199391 |
+
"step": 78620
|
| 199392 |
+
},
|
| 199393 |
+
{
|
| 199394 |
+
"epoch": 628.79,
|
| 199395 |
+
"learning_rate": 8.746462035541197e-06,
|
| 199396 |
+
"loss": 1.0455,
|
| 199397 |
+
"step": 78625
|
| 199398 |
+
},
|
| 199399 |
+
{
|
| 199400 |
+
"epoch": 628.83,
|
| 199401 |
+
"learning_rate": 8.746381260096931e-06,
|
| 199402 |
+
"loss": 0.7148,
|
| 199403 |
+
"step": 78630
|
| 199404 |
+
},
|
| 199405 |
+
{
|
| 199406 |
+
"epoch": 628.87,
|
| 199407 |
+
"learning_rate": 8.746300484652667e-06,
|
| 199408 |
+
"loss": 0.2964,
|
| 199409 |
+
"step": 78635
|
| 199410 |
+
},
|
| 199411 |
+
{
|
| 199412 |
+
"epoch": 628.91,
|
| 199413 |
+
"learning_rate": 8.746219709208401e-06,
|
| 199414 |
+
"loss": 0.3861,
|
| 199415 |
+
"step": 78640
|
| 199416 |
+
},
|
| 199417 |
+
{
|
| 199418 |
+
"epoch": 628.95,
|
| 199419 |
+
"learning_rate": 8.746138933764137e-06,
|
| 199420 |
+
"loss": 0.4008,
|
| 199421 |
+
"step": 78645
|
| 199422 |
+
},
|
| 199423 |
+
{
|
| 199424 |
+
"epoch": 628.99,
|
| 199425 |
+
"learning_rate": 8.746058158319871e-06,
|
| 199426 |
+
"loss": 0.9843,
|
| 199427 |
+
"step": 78650
|
| 199428 |
+
},
|
| 199429 |
+
{
|
| 199430 |
+
"epoch": 629.0,
|
| 199431 |
+
"eval_loss": 0.40389513969421387,
|
| 199432 |
+
"eval_runtime": 41.4895,
|
| 199433 |
+
"eval_samples_per_second": 20.174,
|
| 199434 |
+
"eval_steps_per_second": 0.651,
|
| 199435 |
+
"eval_wer": 0.19676647574856812,
|
| 199436 |
+
"step": 78651
|
| 199437 |
+
},
|
| 199438 |
+
{
|
| 199439 |
+
"epoch": 634.03,
|
| 199440 |
+
"learning_rate": 8.745977382875607e-06,
|
| 199441 |
+
"loss": 0.3614,
|
| 199442 |
+
"step": 78655
|
| 199443 |
+
},
|
| 199444 |
+
{
|
| 199445 |
+
"epoch": 634.07,
|
| 199446 |
+
"learning_rate": 8.745896607431341e-06,
|
| 199447 |
+
"loss": 0.3363,
|
| 199448 |
+
"step": 78660
|
| 199449 |
+
},
|
| 199450 |
+
{
|
| 199451 |
+
"epoch": 634.11,
|
| 199452 |
+
"learning_rate": 8.745815831987077e-06,
|
| 199453 |
+
"loss": 0.291,
|
| 199454 |
+
"step": 78665
|
| 199455 |
+
},
|
| 199456 |
+
{
|
| 199457 |
+
"epoch": 634.15,
|
| 199458 |
+
"learning_rate": 8.745735056542811e-06,
|
| 199459 |
+
"loss": 0.4561,
|
| 199460 |
+
"step": 78670
|
| 199461 |
+
},
|
| 199462 |
+
{
|
| 199463 |
+
"epoch": 634.19,
|
| 199464 |
+
"learning_rate": 8.745654281098547e-06,
|
| 199465 |
+
"loss": 1.0624,
|
| 199466 |
+
"step": 78675
|
| 199467 |
+
},
|
| 199468 |
+
{
|
| 199469 |
+
"epoch": 634.23,
|
| 199470 |
+
"learning_rate": 8.745573505654283e-06,
|
| 199471 |
+
"loss": 0.7522,
|
| 199472 |
+
"step": 78680
|
| 199473 |
+
},
|
| 199474 |
+
{
|
| 199475 |
+
"epoch": 634.27,
|
| 199476 |
+
"learning_rate": 8.745492730210017e-06,
|
| 199477 |
+
"loss": 0.3011,
|
| 199478 |
+
"step": 78685
|
| 199479 |
+
},
|
| 199480 |
+
{
|
| 199481 |
+
"epoch": 634.31,
|
| 199482 |
+
"learning_rate": 8.745411954765753e-06,
|
| 199483 |
+
"loss": 0.2983,
|
| 199484 |
+
"step": 78690
|
| 199485 |
+
},
|
| 199486 |
+
{
|
| 199487 |
+
"epoch": 634.35,
|
| 199488 |
+
"learning_rate": 8.745331179321487e-06,
|
| 199489 |
+
"loss": 0.4159,
|
| 199490 |
+
"step": 78695
|
| 199491 |
+
},
|
| 199492 |
+
{
|
| 199493 |
+
"epoch": 634.39,
|
| 199494 |
+
"learning_rate": 8.745250403877223e-06,
|
| 199495 |
+
"loss": 0.8889,
|
| 199496 |
+
"step": 78700
|
| 199497 |
+
},
|
| 199498 |
+
{
|
| 199499 |
+
"epoch": 634.43,
|
| 199500 |
+
"learning_rate": 8.745169628432957e-06,
|
| 199501 |
+
"loss": 0.6537,
|
| 199502 |
+
"step": 78705
|
| 199503 |
+
},
|
| 199504 |
+
{
|
| 199505 |
+
"epoch": 634.47,
|
| 199506 |
+
"learning_rate": 8.745088852988693e-06,
|
| 199507 |
+
"loss": 0.2519,
|
| 199508 |
+
"step": 78710
|
| 199509 |
+
},
|
| 199510 |
+
{
|
| 199511 |
+
"epoch": 634.51,
|
| 199512 |
+
"learning_rate": 8.745008077544427e-06,
|
| 199513 |
+
"loss": 0.3454,
|
| 199514 |
+
"step": 78715
|
| 199515 |
+
},
|
| 199516 |
+
{
|
| 199517 |
+
"epoch": 634.55,
|
| 199518 |
+
"learning_rate": 8.744927302100163e-06,
|
| 199519 |
+
"loss": 0.6675,
|
| 199520 |
+
"step": 78720
|
| 199521 |
+
},
|
| 199522 |
+
{
|
| 199523 |
+
"epoch": 634.59,
|
| 199524 |
+
"learning_rate": 8.744846526655897e-06,
|
| 199525 |
+
"loss": 0.8333,
|
| 199526 |
+
"step": 78725
|
| 199527 |
+
},
|
| 199528 |
+
{
|
| 199529 |
+
"epoch": 634.63,
|
| 199530 |
+
"learning_rate": 8.744765751211633e-06,
|
| 199531 |
+
"loss": 0.6796,
|
| 199532 |
+
"step": 78730
|
| 199533 |
+
},
|
| 199534 |
+
{
|
| 199535 |
+
"epoch": 634.67,
|
| 199536 |
+
"learning_rate": 8.744684975767368e-06,
|
| 199537 |
+
"loss": 0.3397,
|
| 199538 |
+
"step": 78735
|
| 199539 |
+
},
|
| 199540 |
+
{
|
| 199541 |
+
"epoch": 634.71,
|
| 199542 |
+
"learning_rate": 8.744604200323102e-06,
|
| 199543 |
+
"loss": 0.3553,
|
| 199544 |
+
"step": 78740
|
| 199545 |
+
},
|
| 199546 |
+
{
|
| 199547 |
+
"epoch": 634.76,
|
| 199548 |
+
"learning_rate": 8.744523424878838e-06,
|
| 199549 |
+
"loss": 0.4072,
|
| 199550 |
+
"step": 78745
|
| 199551 |
+
},
|
| 199552 |
+
{
|
| 199553 |
+
"epoch": 634.8,
|
| 199554 |
+
"learning_rate": 8.744442649434572e-06,
|
| 199555 |
+
"loss": 0.8611,
|
| 199556 |
+
"step": 78750
|
| 199557 |
+
},
|
| 199558 |
+
{
|
| 199559 |
+
"epoch": 634.84,
|
| 199560 |
+
"learning_rate": 8.744361873990308e-06,
|
| 199561 |
+
"loss": 0.6651,
|
| 199562 |
+
"step": 78755
|
| 199563 |
+
},
|
| 199564 |
+
{
|
| 199565 |
+
"epoch": 634.88,
|
| 199566 |
+
"learning_rate": 8.744281098546042e-06,
|
| 199567 |
+
"loss": 0.3281,
|
| 199568 |
+
"step": 78760
|
| 199569 |
+
},
|
| 199570 |
+
{
|
| 199571 |
+
"epoch": 634.92,
|
| 199572 |
+
"learning_rate": 8.744200323101778e-06,
|
| 199573 |
+
"loss": 0.3102,
|
| 199574 |
+
"step": 78765
|
| 199575 |
+
},
|
| 199576 |
+
{
|
| 199577 |
+
"epoch": 634.96,
|
| 199578 |
+
"learning_rate": 8.744119547657512e-06,
|
| 199579 |
+
"loss": 0.4566,
|
| 199580 |
+
"step": 78770
|
| 199581 |
+
},
|
| 199582 |
+
{
|
| 199583 |
+
"epoch": 635.0,
|
| 199584 |
+
"learning_rate": 8.744038772213248e-06,
|
| 199585 |
+
"loss": 1.1649,
|
| 199586 |
+
"step": 78775
|
| 199587 |
+
},
|
| 199588 |
+
{
|
| 199589 |
+
"epoch": 635.0,
|
| 199590 |
+
"eval_loss": 0.42814478278160095,
|
| 199591 |
+
"eval_runtime": 40.8229,
|
| 199592 |
+
"eval_samples_per_second": 20.503,
|
| 199593 |
+
"eval_steps_per_second": 0.661,
|
| 199594 |
+
"eval_wer": 0.1919994007041726,
|
| 199595 |
+
"step": 78775
|
| 199596 |
+
},
|
| 199597 |
+
{
|
| 199598 |
+
"epoch": 630.04,
|
| 199599 |
+
"learning_rate": 8.743957996768982e-06,
|
| 199600 |
+
"loss": 0.3452,
|
| 199601 |
+
"step": 78780
|
| 199602 |
+
},
|
| 199603 |
+
{
|
| 199604 |
+
"epoch": 630.08,
|
| 199605 |
+
"learning_rate": 8.743877221324718e-06,
|
| 199606 |
+
"loss": 0.3256,
|
| 199607 |
+
"step": 78785
|
| 199608 |
+
},
|
| 199609 |
+
{
|
| 199610 |
+
"epoch": 630.12,
|
| 199611 |
+
"learning_rate": 8.743796445880452e-06,
|
| 199612 |
+
"loss": 0.4387,
|
| 199613 |
+
"step": 78790
|
| 199614 |
+
},
|
| 199615 |
+
{
|
| 199616 |
+
"epoch": 630.16,
|
| 199617 |
+
"learning_rate": 8.743715670436188e-06,
|
| 199618 |
+
"loss": 0.395,
|
| 199619 |
+
"step": 78795
|
| 199620 |
+
},
|
| 199621 |
+
{
|
| 199622 |
+
"epoch": 630.2,
|
| 199623 |
+
"learning_rate": 8.743634894991924e-06,
|
| 199624 |
+
"loss": 1.3145,
|
| 199625 |
+
"step": 78800
|
| 199626 |
+
},
|
| 199627 |
+
{
|
| 199628 |
+
"epoch": 630.24,
|
| 199629 |
+
"learning_rate": 8.743554119547658e-06,
|
| 199630 |
+
"loss": 0.3666,
|
| 199631 |
+
"step": 78805
|
| 199632 |
+
},
|
| 199633 |
+
{
|
| 199634 |
+
"epoch": 630.28,
|
| 199635 |
+
"learning_rate": 8.743473344103394e-06,
|
| 199636 |
+
"loss": 0.3053,
|
| 199637 |
+
"step": 78810
|
| 199638 |
+
},
|
| 199639 |
+
{
|
| 199640 |
+
"epoch": 630.32,
|
| 199641 |
+
"learning_rate": 8.743392568659128e-06,
|
| 199642 |
+
"loss": 0.3094,
|
| 199643 |
+
"step": 78815
|
| 199644 |
+
},
|
| 199645 |
+
{
|
| 199646 |
+
"epoch": 630.36,
|
| 199647 |
+
"learning_rate": 8.743311793214864e-06,
|
| 199648 |
+
"loss": 0.4993,
|
| 199649 |
+
"step": 78820
|
| 199650 |
+
},
|
| 199651 |
+
{
|
| 199652 |
+
"epoch": 630.4,
|
| 199653 |
+
"learning_rate": 8.743231017770598e-06,
|
| 199654 |
+
"loss": 1.2233,
|
| 199655 |
+
"step": 78825
|
| 199656 |
+
},
|
| 199657 |
+
{
|
| 199658 |
+
"epoch": 630.44,
|
| 199659 |
+
"learning_rate": 8.743150242326334e-06,
|
| 199660 |
+
"loss": 0.3475,
|
| 199661 |
+
"step": 78830
|
| 199662 |
+
},
|
| 199663 |
+
{
|
| 199664 |
+
"epoch": 630.48,
|
| 199665 |
+
"learning_rate": 8.743069466882068e-06,
|
| 199666 |
+
"loss": 0.5334,
|
| 199667 |
+
"step": 78835
|
| 199668 |
+
},
|
| 199669 |
+
{
|
| 199670 |
+
"epoch": 630.52,
|
| 199671 |
+
"learning_rate": 8.742988691437804e-06,
|
| 199672 |
+
"loss": 0.3864,
|
| 199673 |
+
"step": 78840
|
| 199674 |
+
},
|
| 199675 |
+
{
|
| 199676 |
+
"epoch": 630.56,
|
| 199677 |
+
"learning_rate": 8.742907915993538e-06,
|
| 199678 |
+
"loss": 0.483,
|
| 199679 |
+
"step": 78845
|
| 199680 |
+
},
|
| 199681 |
+
{
|
| 199682 |
+
"epoch": 630.6,
|
| 199683 |
+
"learning_rate": 8.742827140549274e-06,
|
| 199684 |
+
"loss": 1.0561,
|
| 199685 |
+
"step": 78850
|
| 199686 |
+
},
|
| 199687 |
+
{
|
| 199688 |
+
"epoch": 630.64,
|
| 199689 |
+
"learning_rate": 8.74274636510501e-06,
|
| 199690 |
+
"loss": 0.3234,
|
| 199691 |
+
"step": 78855
|
| 199692 |
+
},
|
| 199693 |
+
{
|
| 199694 |
+
"epoch": 630.68,
|
| 199695 |
+
"learning_rate": 8.742665589660744e-06,
|
| 199696 |
+
"loss": 0.3517,
|
| 199697 |
+
"step": 78860
|
| 199698 |
+
},
|
| 199699 |
+
{
|
| 199700 |
+
"epoch": 630.72,
|
| 199701 |
+
"learning_rate": 8.74258481421648e-06,
|
| 199702 |
+
"loss": 0.3457,
|
| 199703 |
+
"step": 78865
|
| 199704 |
+
},
|
| 199705 |
+
{
|
| 199706 |
+
"epoch": 630.76,
|
| 199707 |
+
"learning_rate": 8.742504038772214e-06,
|
| 199708 |
+
"loss": 0.4469,
|
| 199709 |
+
"step": 78870
|
| 199710 |
+
},
|
| 199711 |
+
{
|
| 199712 |
+
"epoch": 630.8,
|
| 199713 |
+
"learning_rate": 8.74242326332795e-06,
|
| 199714 |
+
"loss": 1.2126,
|
| 199715 |
+
"step": 78875
|
| 199716 |
+
},
|
| 199717 |
+
{
|
| 199718 |
+
"epoch": 630.84,
|
| 199719 |
+
"learning_rate": 8.742342487883684e-06,
|
| 199720 |
+
"loss": 0.3057,
|
| 199721 |
+
"step": 78880
|
| 199722 |
+
},
|
| 199723 |
+
{
|
| 199724 |
+
"epoch": 630.88,
|
| 199725 |
+
"learning_rate": 8.74226171243942e-06,
|
| 199726 |
+
"loss": 0.3012,
|
| 199727 |
+
"step": 78885
|
| 199728 |
+
},
|
| 199729 |
+
{
|
| 199730 |
+
"epoch": 630.92,
|
| 199731 |
+
"learning_rate": 8.742180936995154e-06,
|
| 199732 |
+
"loss": 0.3956,
|
| 199733 |
+
"step": 78890
|
| 199734 |
+
},
|
| 199735 |
+
{
|
| 199736 |
+
"epoch": 630.96,
|
| 199737 |
+
"learning_rate": 8.74210016155089e-06,
|
| 199738 |
+
"loss": 0.4191,
|
| 199739 |
+
"step": 78895
|
| 199740 |
+
},
|
| 199741 |
+
{
|
| 199742 |
+
"epoch": 631.0,
|
| 199743 |
+
"learning_rate": 8.742019386106624e-06,
|
| 199744 |
+
"loss": 1.1315,
|
| 199745 |
+
"step": 78900
|
| 199746 |
+
},
|
| 199747 |
+
{
|
| 199748 |
+
"epoch": 631.0,
|
| 199749 |
+
"eval_loss": 0.35649973154067993,
|
| 199750 |
+
"eval_runtime": 41.8167,
|
| 199751 |
+
"eval_samples_per_second": 19.992,
|
| 199752 |
+
"eval_steps_per_second": 0.646,
|
| 199753 |
+
"eval_wer": 0.19188435573818607,
|
| 199754 |
+
"step": 78900
|
| 199755 |
+
},
|
| 199756 |
+
{
|
| 199757 |
+
"epoch": 636.04,
|
| 199758 |
+
"learning_rate": 8.74193861066236e-06,
|
| 199759 |
+
"loss": 0.3542,
|
| 199760 |
+
"step": 78905
|
| 199761 |
+
},
|
| 199762 |
+
{
|
| 199763 |
+
"epoch": 636.08,
|
| 199764 |
+
"learning_rate": 8.741857835218095e-06,
|
| 199765 |
+
"loss": 0.3388,
|
| 199766 |
+
"step": 78910
|
| 199767 |
+
},
|
| 199768 |
+
{
|
| 199769 |
+
"epoch": 636.12,
|
| 199770 |
+
"learning_rate": 8.74177705977383e-06,
|
| 199771 |
+
"loss": 0.3729,
|
| 199772 |
+
"step": 78915
|
| 199773 |
+
},
|
| 199774 |
+
{
|
| 199775 |
+
"epoch": 636.16,
|
| 199776 |
+
"learning_rate": 8.741696284329565e-06,
|
| 199777 |
+
"loss": 0.4893,
|
| 199778 |
+
"step": 78920
|
| 199779 |
+
},
|
| 199780 |
+
{
|
| 199781 |
+
"epoch": 636.2,
|
| 199782 |
+
"learning_rate": 8.7416155088853e-06,
|
| 199783 |
+
"loss": 1.2058,
|
| 199784 |
+
"step": 78925
|
| 199785 |
+
},
|
| 199786 |
+
{
|
| 199787 |
+
"epoch": 636.24,
|
| 199788 |
+
"learning_rate": 8.741534733441035e-06,
|
| 199789 |
+
"loss": 0.3539,
|
| 199790 |
+
"step": 78930
|
| 199791 |
+
},
|
| 199792 |
+
{
|
| 199793 |
+
"epoch": 636.28,
|
| 199794 |
+
"learning_rate": 8.74145395799677e-06,
|
| 199795 |
+
"loss": 0.2974,
|
| 199796 |
+
"step": 78935
|
| 199797 |
+
},
|
| 199798 |
+
{
|
| 199799 |
+
"epoch": 636.32,
|
| 199800 |
+
"learning_rate": 8.741373182552505e-06,
|
| 199801 |
+
"loss": 0.3459,
|
| 199802 |
+
"step": 78940
|
| 199803 |
+
},
|
| 199804 |
+
{
|
| 199805 |
+
"epoch": 636.36,
|
| 199806 |
+
"learning_rate": 8.74129240710824e-06,
|
| 199807 |
+
"loss": 0.4149,
|
| 199808 |
+
"step": 78945
|
| 199809 |
+
},
|
| 199810 |
+
{
|
| 199811 |
+
"epoch": 636.4,
|
| 199812 |
+
"learning_rate": 8.741211631663975e-06,
|
| 199813 |
+
"loss": 1.3084,
|
| 199814 |
+
"step": 78950
|
| 199815 |
+
},
|
| 199816 |
+
{
|
| 199817 |
+
"epoch": 636.44,
|
| 199818 |
+
"learning_rate": 8.74113085621971e-06,
|
| 199819 |
+
"loss": 0.3186,
|
| 199820 |
+
"step": 78955
|
| 199821 |
+
},
|
| 199822 |
+
{
|
| 199823 |
+
"epoch": 636.48,
|
| 199824 |
+
"learning_rate": 8.741050080775445e-06,
|
| 199825 |
+
"loss": 0.3063,
|
| 199826 |
+
"step": 78960
|
| 199827 |
+
},
|
| 199828 |
+
{
|
| 199829 |
+
"epoch": 636.52,
|
| 199830 |
+
"learning_rate": 8.74096930533118e-06,
|
| 199831 |
+
"loss": 0.3157,
|
| 199832 |
+
"step": 78965
|
| 199833 |
+
},
|
| 199834 |
+
{
|
| 199835 |
+
"epoch": 636.56,
|
| 199836 |
+
"learning_rate": 8.740888529886915e-06,
|
| 199837 |
+
"loss": 0.4941,
|
| 199838 |
+
"step": 78970
|
| 199839 |
+
},
|
| 199840 |
+
{
|
| 199841 |
+
"epoch": 636.6,
|
| 199842 |
+
"learning_rate": 8.740807754442651e-06,
|
| 199843 |
+
"loss": 1.0257,
|
| 199844 |
+
"step": 78975
|
| 199845 |
+
},
|
| 199846 |
+
{
|
| 199847 |
+
"epoch": 636.64,
|
| 199848 |
+
"learning_rate": 8.740726978998385e-06,
|
| 199849 |
+
"loss": 0.344,
|
| 199850 |
+
"step": 78980
|
| 199851 |
+
},
|
| 199852 |
+
{
|
| 199853 |
+
"epoch": 636.68,
|
| 199854 |
+
"learning_rate": 8.740646203554121e-06,
|
| 199855 |
+
"loss": 0.2603,
|
| 199856 |
+
"step": 78985
|
| 199857 |
+
},
|
| 199858 |
+
{
|
| 199859 |
+
"epoch": 636.72,
|
| 199860 |
+
"learning_rate": 8.740565428109855e-06,
|
| 199861 |
+
"loss": 0.2939,
|
| 199862 |
+
"step": 78990
|
| 199863 |
+
},
|
| 199864 |
+
{
|
| 199865 |
+
"epoch": 636.76,
|
| 199866 |
+
"learning_rate": 8.740484652665591e-06,
|
| 199867 |
+
"loss": 0.4707,
|
| 199868 |
+
"step": 78995
|
| 199869 |
+
},
|
| 199870 |
+
{
|
| 199871 |
+
"epoch": 636.8,
|
| 199872 |
+
"learning_rate": 8.740403877221325e-06,
|
| 199873 |
+
"loss": 1.1715,
|
| 199874 |
+
"step": 79000
|
| 199875 |
+
},
|
| 199876 |
+
{
|
| 199877 |
+
"epoch": 636.84,
|
| 199878 |
+
"learning_rate": 8.740323101777061e-06,
|
| 199879 |
+
"loss": 0.3283,
|
| 199880 |
+
"step": 79005
|
| 199881 |
+
},
|
| 199882 |
+
{
|
| 199883 |
+
"epoch": 636.88,
|
| 199884 |
+
"learning_rate": 8.740242326332795e-06,
|
| 199885 |
+
"loss": 0.2645,
|
| 199886 |
+
"step": 79010
|
| 199887 |
+
},
|
| 199888 |
+
{
|
| 199889 |
+
"epoch": 636.92,
|
| 199890 |
+
"learning_rate": 8.740161550888531e-06,
|
| 199891 |
+
"loss": 0.3081,
|
| 199892 |
+
"step": 79015
|
| 199893 |
+
},
|
| 199894 |
+
{
|
| 199895 |
+
"epoch": 636.96,
|
| 199896 |
+
"learning_rate": 8.740080775444265e-06,
|
| 199897 |
+
"loss": 0.491,
|
| 199898 |
+
"step": 79020
|
| 199899 |
+
},
|
| 199900 |
+
{
|
| 199901 |
+
"epoch": 637.0,
|
| 199902 |
+
"eval_loss": 0.39863112568855286,
|
| 199903 |
+
"eval_runtime": 40.3355,
|
| 199904 |
+
"eval_samples_per_second": 20.726,
|
| 199905 |
+
"eval_steps_per_second": 0.669,
|
| 199906 |
+
"eval_wer": 0.1863764670925568,
|
| 199907 |
+
"step": 79024
|
| 199908 |
}
|
| 199909 |
],
|
| 199910 |
+
"max_steps": 620000,
|
| 199911 |
"num_train_epochs": 5000,
|
| 199912 |
+
"total_flos": 2.2238986676431526e+20,
|
| 199913 |
"trial_name": null,
|
| 199914 |
"trial_params": null
|
| 199915 |
}
|
model-bin/finetune/base/{checkpoint-78401 → checkpoint-79024}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629896881.4320064/events.out.tfevents.1629896881.7e498afd5545.7645.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2654328f8eb2972242ff195060b7e2ab5745785dce1e3aa2dd62f9fba5e6cf37
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629897354.142085/events.out.tfevents.1629897354.7e498afd5545.7645.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14e876a4c6e9579d18bd67460b6d4403656601d59319132b57b07d3703268315
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629897817.6067963/events.out.tfevents.1629897817.7e498afd5545.7645.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a7b6c4b2e3cf368dd19d47f635771416602cd6108e603471825175c4b150f007
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629898283.2853954/events.out.tfevents.1629898283.7e498afd5545.7645.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5cf63c7933e59788a935ec6e51c974ba66c36185074fe974c88aee444f29f02f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629898756.3973496/events.out.tfevents.1629898756.7e498afd5545.7645.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2877a983cea7887a0baf4c92bf33531660d3ac39f0478f013de74a70b9c0320
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629896881.7e498afd5545.7645.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b49bed7e3d1f265a6baefb1fb4f870ec4edb4c0fdfc381df3c3306c993024f9c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629897354.7e498afd5545.7645.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c396bc68f17f59ee41387235b6cfa971788d25684437bd81454582c3f502fa2b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629897817.7e498afd5545.7645.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c696411aa48c0bd7cba5b62f03cb9b3eb3e25287a9369c0711589f0254373e0b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629898283.7e498afd5545.7645.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c76a338c0ab4deba65a322b8810557a14527eb1f36aa5a4ac8fb422c98db0c4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629898756.7e498afd5545.7645.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7e3869591dc06ff391d5a92da9491bd29cc792c57606d90693c46e7c82e17f86
|
| 3 |
+
size 8462
|