"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629930714.9815521/events.out.tfevents.1629930714.7e498afd5545.7645.175 +3 -0
- model-bin/finetune/base/log/1629931194.0387235/events.out.tfevents.1629931194.7e498afd5545.7645.177 +3 -0
- model-bin/finetune/base/log/1629931664.7699187/events.out.tfevents.1629931665.7e498afd5545.7645.179 +3 -0
- model-bin/finetune/base/log/1629932126.928111/events.out.tfevents.1629932126.7e498afd5545.7645.181 +3 -0
- model-bin/finetune/base/log/1629932592.8902197/events.out.tfevents.1629932592.7e498afd5545.7645.183 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629930714.7e498afd5545.7645.174 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629931192.7e498afd5545.7645.176 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629931664.7e498afd5545.7645.178 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629932126.7e498afd5545.7645.180 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629932592.7e498afd5545.7645.182 +3 -0
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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 sha256:2d09bfeb91151a041e1a2586f4f0c88f953bb4961eb791e48dff997958709da3
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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:d0f668373a5aabfcb5d6228b13adab83f210eb9604d96cf85ec06a0a9f51e74a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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 sha256:0b42b9b28f11aa022d01becad0204fbef71ff74ea4ca43eda3c0405300ea7ecc
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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:d93d43bbeecc2e2dd56d18ef4fdffea967e794dca54efdfd4a801ca408b71336
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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:f41e1b270a3f2a331f7e7c3da5883fc44c8bf2833663e98a5dc1e4ab7134eaf2
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/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,
|
|
@@ -210198,11 +210198,806 @@
|
|
| 210198 |
"eval_steps_per_second": 0.652,
|
| 210199 |
"eval_wer": 0.18761075014766687,
|
| 210200 |
"step": 87113
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 210201 |
}
|
| 210202 |
],
|
| 210203 |
-
"max_steps":
|
| 210204 |
"num_train_epochs": 5000,
|
| 210205 |
-
"total_flos": 2.
|
| 210206 |
"trial_name": null,
|
| 210207 |
"trial_params": null
|
| 210208 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 706.995983935743,
|
| 5 |
+
"global_step": 87737,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 210198 |
"eval_steps_per_second": 0.652,
|
| 210199 |
"eval_wer": 0.18761075014766687,
|
| 210200 |
"step": 87113
|
| 210201 |
+
},
|
| 210202 |
+
{
|
| 210203 |
+
"epoch": 696.02,
|
| 210204 |
+
"learning_rate": 8.60935379644588e-06,
|
| 210205 |
+
"loss": 0.4733,
|
| 210206 |
+
"step": 87115
|
| 210207 |
+
},
|
| 210208 |
+
{
|
| 210209 |
+
"epoch": 696.06,
|
| 210210 |
+
"learning_rate": 8.609273021001616e-06,
|
| 210211 |
+
"loss": 0.366,
|
| 210212 |
+
"step": 87120
|
| 210213 |
+
},
|
| 210214 |
+
{
|
| 210215 |
+
"epoch": 696.1,
|
| 210216 |
+
"learning_rate": 8.609192245557352e-06,
|
| 210217 |
+
"loss": 0.3383,
|
| 210218 |
+
"step": 87125
|
| 210219 |
+
},
|
| 210220 |
+
{
|
| 210221 |
+
"epoch": 696.14,
|
| 210222 |
+
"learning_rate": 8.609111470113086e-06,
|
| 210223 |
+
"loss": 0.3779,
|
| 210224 |
+
"step": 87130
|
| 210225 |
+
},
|
| 210226 |
+
{
|
| 210227 |
+
"epoch": 696.18,
|
| 210228 |
+
"learning_rate": 8.609030694668822e-06,
|
| 210229 |
+
"loss": 0.6634,
|
| 210230 |
+
"step": 87135
|
| 210231 |
+
},
|
| 210232 |
+
{
|
| 210233 |
+
"epoch": 696.22,
|
| 210234 |
+
"learning_rate": 8.608949919224556e-06,
|
| 210235 |
+
"loss": 1.1731,
|
| 210236 |
+
"step": 87140
|
| 210237 |
+
},
|
| 210238 |
+
{
|
| 210239 |
+
"epoch": 696.26,
|
| 210240 |
+
"learning_rate": 8.608869143780292e-06,
|
| 210241 |
+
"loss": 0.2799,
|
| 210242 |
+
"step": 87145
|
| 210243 |
+
},
|
| 210244 |
+
{
|
| 210245 |
+
"epoch": 696.3,
|
| 210246 |
+
"learning_rate": 8.608788368336026e-06,
|
| 210247 |
+
"loss": 0.3293,
|
| 210248 |
+
"step": 87150
|
| 210249 |
+
},
|
| 210250 |
+
{
|
| 210251 |
+
"epoch": 696.34,
|
| 210252 |
+
"learning_rate": 8.608707592891762e-06,
|
| 210253 |
+
"loss": 0.3596,
|
| 210254 |
+
"step": 87155
|
| 210255 |
+
},
|
| 210256 |
+
{
|
| 210257 |
+
"epoch": 696.38,
|
| 210258 |
+
"learning_rate": 8.608626817447496e-06,
|
| 210259 |
+
"loss": 0.706,
|
| 210260 |
+
"step": 87160
|
| 210261 |
+
},
|
| 210262 |
+
{
|
| 210263 |
+
"epoch": 696.42,
|
| 210264 |
+
"learning_rate": 8.608546042003232e-06,
|
| 210265 |
+
"loss": 0.957,
|
| 210266 |
+
"step": 87165
|
| 210267 |
+
},
|
| 210268 |
+
{
|
| 210269 |
+
"epoch": 696.46,
|
| 210270 |
+
"learning_rate": 8.608465266558966e-06,
|
| 210271 |
+
"loss": 0.293,
|
| 210272 |
+
"step": 87170
|
| 210273 |
+
},
|
| 210274 |
+
{
|
| 210275 |
+
"epoch": 696.5,
|
| 210276 |
+
"learning_rate": 8.608384491114702e-06,
|
| 210277 |
+
"loss": 0.3336,
|
| 210278 |
+
"step": 87175
|
| 210279 |
+
},
|
| 210280 |
+
{
|
| 210281 |
+
"epoch": 696.54,
|
| 210282 |
+
"learning_rate": 8.608303715670438e-06,
|
| 210283 |
+
"loss": 0.3706,
|
| 210284 |
+
"step": 87180
|
| 210285 |
+
},
|
| 210286 |
+
{
|
| 210287 |
+
"epoch": 696.58,
|
| 210288 |
+
"learning_rate": 8.608222940226172e-06,
|
| 210289 |
+
"loss": 0.6474,
|
| 210290 |
+
"step": 87185
|
| 210291 |
+
},
|
| 210292 |
+
{
|
| 210293 |
+
"epoch": 696.62,
|
| 210294 |
+
"learning_rate": 8.608142164781908e-06,
|
| 210295 |
+
"loss": 1.1947,
|
| 210296 |
+
"step": 87190
|
| 210297 |
+
},
|
| 210298 |
+
{
|
| 210299 |
+
"epoch": 696.66,
|
| 210300 |
+
"learning_rate": 8.608061389337642e-06,
|
| 210301 |
+
"loss": 0.2743,
|
| 210302 |
+
"step": 87195
|
| 210303 |
+
},
|
| 210304 |
+
{
|
| 210305 |
+
"epoch": 696.7,
|
| 210306 |
+
"learning_rate": 8.607980613893378e-06,
|
| 210307 |
+
"loss": 0.2946,
|
| 210308 |
+
"step": 87200
|
| 210309 |
+
},
|
| 210310 |
+
{
|
| 210311 |
+
"epoch": 696.74,
|
| 210312 |
+
"learning_rate": 8.607899838449112e-06,
|
| 210313 |
+
"loss": 0.3946,
|
| 210314 |
+
"step": 87205
|
| 210315 |
+
},
|
| 210316 |
+
{
|
| 210317 |
+
"epoch": 696.78,
|
| 210318 |
+
"learning_rate": 8.607819063004848e-06,
|
| 210319 |
+
"loss": 0.6193,
|
| 210320 |
+
"step": 87210
|
| 210321 |
+
},
|
| 210322 |
+
{
|
| 210323 |
+
"epoch": 696.82,
|
| 210324 |
+
"learning_rate": 8.607738287560582e-06,
|
| 210325 |
+
"loss": 0.9692,
|
| 210326 |
+
"step": 87215
|
| 210327 |
+
},
|
| 210328 |
+
{
|
| 210329 |
+
"epoch": 696.86,
|
| 210330 |
+
"learning_rate": 8.607657512116318e-06,
|
| 210331 |
+
"loss": 0.3383,
|
| 210332 |
+
"step": 87220
|
| 210333 |
+
},
|
| 210334 |
+
{
|
| 210335 |
+
"epoch": 696.9,
|
| 210336 |
+
"learning_rate": 8.607576736672052e-06,
|
| 210337 |
+
"loss": 0.2939,
|
| 210338 |
+
"step": 87225
|
| 210339 |
+
},
|
| 210340 |
+
{
|
| 210341 |
+
"epoch": 696.94,
|
| 210342 |
+
"learning_rate": 8.607495961227788e-06,
|
| 210343 |
+
"loss": 0.517,
|
| 210344 |
+
"step": 87230
|
| 210345 |
+
},
|
| 210346 |
+
{
|
| 210347 |
+
"epoch": 696.98,
|
| 210348 |
+
"learning_rate": 8.607415185783522e-06,
|
| 210349 |
+
"loss": 0.7363,
|
| 210350 |
+
"step": 87235
|
| 210351 |
+
},
|
| 210352 |
+
{
|
| 210353 |
+
"epoch": 697.0,
|
| 210354 |
+
"eval_loss": 0.3449787199497223,
|
| 210355 |
+
"eval_runtime": 43.777,
|
| 210356 |
+
"eval_samples_per_second": 19.165,
|
| 210357 |
+
"eval_steps_per_second": 0.617,
|
| 210358 |
+
"eval_wer": 0.18407169184071692,
|
| 210359 |
+
"step": 87238
|
| 210360 |
+
},
|
| 210361 |
+
{
|
| 210362 |
+
"epoch": 697.02,
|
| 210363 |
+
"learning_rate": 8.607334410339258e-06,
|
| 210364 |
+
"loss": 0.3579,
|
| 210365 |
+
"step": 87240
|
| 210366 |
+
},
|
| 210367 |
+
{
|
| 210368 |
+
"epoch": 697.06,
|
| 210369 |
+
"learning_rate": 8.607253634894994e-06,
|
| 210370 |
+
"loss": 0.3034,
|
| 210371 |
+
"step": 87245
|
| 210372 |
+
},
|
| 210373 |
+
{
|
| 210374 |
+
"epoch": 697.1,
|
| 210375 |
+
"learning_rate": 8.607172859450728e-06,
|
| 210376 |
+
"loss": 0.3203,
|
| 210377 |
+
"step": 87250
|
| 210378 |
+
},
|
| 210379 |
+
{
|
| 210380 |
+
"epoch": 697.14,
|
| 210381 |
+
"learning_rate": 8.607092084006464e-06,
|
| 210382 |
+
"loss": 0.3518,
|
| 210383 |
+
"step": 87255
|
| 210384 |
+
},
|
| 210385 |
+
{
|
| 210386 |
+
"epoch": 697.18,
|
| 210387 |
+
"learning_rate": 8.607011308562198e-06,
|
| 210388 |
+
"loss": 0.7261,
|
| 210389 |
+
"step": 87260
|
| 210390 |
+
},
|
| 210391 |
+
{
|
| 210392 |
+
"epoch": 697.22,
|
| 210393 |
+
"learning_rate": 8.606930533117933e-06,
|
| 210394 |
+
"loss": 0.9333,
|
| 210395 |
+
"step": 87265
|
| 210396 |
+
},
|
| 210397 |
+
{
|
| 210398 |
+
"epoch": 697.26,
|
| 210399 |
+
"learning_rate": 8.606849757673668e-06,
|
| 210400 |
+
"loss": 0.3645,
|
| 210401 |
+
"step": 87270
|
| 210402 |
+
},
|
| 210403 |
+
{
|
| 210404 |
+
"epoch": 697.3,
|
| 210405 |
+
"learning_rate": 8.606768982229403e-06,
|
| 210406 |
+
"loss": 0.3068,
|
| 210407 |
+
"step": 87275
|
| 210408 |
+
},
|
| 210409 |
+
{
|
| 210410 |
+
"epoch": 697.34,
|
| 210411 |
+
"learning_rate": 8.606688206785138e-06,
|
| 210412 |
+
"loss": 0.3019,
|
| 210413 |
+
"step": 87280
|
| 210414 |
+
},
|
| 210415 |
+
{
|
| 210416 |
+
"epoch": 697.38,
|
| 210417 |
+
"learning_rate": 8.606607431340873e-06,
|
| 210418 |
+
"loss": 0.6634,
|
| 210419 |
+
"step": 87285
|
| 210420 |
+
},
|
| 210421 |
+
{
|
| 210422 |
+
"epoch": 697.42,
|
| 210423 |
+
"learning_rate": 8.606526655896608e-06,
|
| 210424 |
+
"loss": 1.0637,
|
| 210425 |
+
"step": 87290
|
| 210426 |
+
},
|
| 210427 |
+
{
|
| 210428 |
+
"epoch": 697.46,
|
| 210429 |
+
"learning_rate": 8.606445880452343e-06,
|
| 210430 |
+
"loss": 0.3017,
|
| 210431 |
+
"step": 87295
|
| 210432 |
+
},
|
| 210433 |
+
{
|
| 210434 |
+
"epoch": 697.5,
|
| 210435 |
+
"learning_rate": 8.60636510500808e-06,
|
| 210436 |
+
"loss": 0.3878,
|
| 210437 |
+
"step": 87300
|
| 210438 |
+
},
|
| 210439 |
+
{
|
| 210440 |
+
"epoch": 697.54,
|
| 210441 |
+
"learning_rate": 8.606284329563813e-06,
|
| 210442 |
+
"loss": 0.3294,
|
| 210443 |
+
"step": 87305
|
| 210444 |
+
},
|
| 210445 |
+
{
|
| 210446 |
+
"epoch": 697.58,
|
| 210447 |
+
"learning_rate": 8.60620355411955e-06,
|
| 210448 |
+
"loss": 0.5655,
|
| 210449 |
+
"step": 87310
|
| 210450 |
+
},
|
| 210451 |
+
{
|
| 210452 |
+
"epoch": 697.62,
|
| 210453 |
+
"learning_rate": 8.606122778675283e-06,
|
| 210454 |
+
"loss": 0.978,
|
| 210455 |
+
"step": 87315
|
| 210456 |
+
},
|
| 210457 |
+
{
|
| 210458 |
+
"epoch": 697.66,
|
| 210459 |
+
"learning_rate": 8.60604200323102e-06,
|
| 210460 |
+
"loss": 0.3328,
|
| 210461 |
+
"step": 87320
|
| 210462 |
+
},
|
| 210463 |
+
{
|
| 210464 |
+
"epoch": 697.7,
|
| 210465 |
+
"learning_rate": 8.605961227786753e-06,
|
| 210466 |
+
"loss": 0.4727,
|
| 210467 |
+
"step": 87325
|
| 210468 |
+
},
|
| 210469 |
+
{
|
| 210470 |
+
"epoch": 697.74,
|
| 210471 |
+
"learning_rate": 8.60588045234249e-06,
|
| 210472 |
+
"loss": 0.3351,
|
| 210473 |
+
"step": 87330
|
| 210474 |
+
},
|
| 210475 |
+
{
|
| 210476 |
+
"epoch": 697.78,
|
| 210477 |
+
"learning_rate": 8.605799676898223e-06,
|
| 210478 |
+
"loss": 0.6038,
|
| 210479 |
+
"step": 87335
|
| 210480 |
+
},
|
| 210481 |
+
{
|
| 210482 |
+
"epoch": 697.82,
|
| 210483 |
+
"learning_rate": 8.605718901453959e-06,
|
| 210484 |
+
"loss": 0.9593,
|
| 210485 |
+
"step": 87340
|
| 210486 |
+
},
|
| 210487 |
+
{
|
| 210488 |
+
"epoch": 697.86,
|
| 210489 |
+
"learning_rate": 8.605638126009693e-06,
|
| 210490 |
+
"loss": 0.321,
|
| 210491 |
+
"step": 87345
|
| 210492 |
+
},
|
| 210493 |
+
{
|
| 210494 |
+
"epoch": 697.9,
|
| 210495 |
+
"learning_rate": 8.605557350565429e-06,
|
| 210496 |
+
"loss": 0.3431,
|
| 210497 |
+
"step": 87350
|
| 210498 |
+
},
|
| 210499 |
+
{
|
| 210500 |
+
"epoch": 697.94,
|
| 210501 |
+
"learning_rate": 8.605476575121165e-06,
|
| 210502 |
+
"loss": 0.3429,
|
| 210503 |
+
"step": 87355
|
| 210504 |
+
},
|
| 210505 |
+
{
|
| 210506 |
+
"epoch": 697.98,
|
| 210507 |
+
"learning_rate": 8.605395799676899e-06,
|
| 210508 |
+
"loss": 0.7337,
|
| 210509 |
+
"step": 87360
|
| 210510 |
+
},
|
| 210511 |
+
{
|
| 210512 |
+
"epoch": 698.0,
|
| 210513 |
+
"eval_loss": 0.42268607020378113,
|
| 210514 |
+
"eval_runtime": 41.7967,
|
| 210515 |
+
"eval_samples_per_second": 20.073,
|
| 210516 |
+
"eval_steps_per_second": 0.646,
|
| 210517 |
+
"eval_wer": 0.1897736313010776,
|
| 210518 |
+
"step": 87363
|
| 210519 |
+
},
|
| 210520 |
+
{
|
| 210521 |
+
"epoch": 698.02,
|
| 210522 |
+
"learning_rate": 8.605315024232635e-06,
|
| 210523 |
+
"loss": 0.3079,
|
| 210524 |
+
"step": 87365
|
| 210525 |
+
},
|
| 210526 |
+
{
|
| 210527 |
+
"epoch": 698.06,
|
| 210528 |
+
"learning_rate": 8.605234248788369e-06,
|
| 210529 |
+
"loss": 0.317,
|
| 210530 |
+
"step": 87370
|
| 210531 |
+
},
|
| 210532 |
+
{
|
| 210533 |
+
"epoch": 698.1,
|
| 210534 |
+
"learning_rate": 8.605153473344105e-06,
|
| 210535 |
+
"loss": 0.3125,
|
| 210536 |
+
"step": 87375
|
| 210537 |
+
},
|
| 210538 |
+
{
|
| 210539 |
+
"epoch": 698.14,
|
| 210540 |
+
"learning_rate": 8.605072697899839e-06,
|
| 210541 |
+
"loss": 0.3947,
|
| 210542 |
+
"step": 87380
|
| 210543 |
+
},
|
| 210544 |
+
{
|
| 210545 |
+
"epoch": 698.18,
|
| 210546 |
+
"learning_rate": 8.604991922455575e-06,
|
| 210547 |
+
"loss": 0.6674,
|
| 210548 |
+
"step": 87385
|
| 210549 |
+
},
|
| 210550 |
+
{
|
| 210551 |
+
"epoch": 698.22,
|
| 210552 |
+
"learning_rate": 8.604911147011309e-06,
|
| 210553 |
+
"loss": 1.0877,
|
| 210554 |
+
"step": 87390
|
| 210555 |
+
},
|
| 210556 |
+
{
|
| 210557 |
+
"epoch": 698.26,
|
| 210558 |
+
"learning_rate": 8.604830371567045e-06,
|
| 210559 |
+
"loss": 0.354,
|
| 210560 |
+
"step": 87395
|
| 210561 |
+
},
|
| 210562 |
+
{
|
| 210563 |
+
"epoch": 698.3,
|
| 210564 |
+
"learning_rate": 8.604749596122779e-06,
|
| 210565 |
+
"loss": 0.3048,
|
| 210566 |
+
"step": 87400
|
| 210567 |
+
},
|
| 210568 |
+
{
|
| 210569 |
+
"epoch": 698.34,
|
| 210570 |
+
"learning_rate": 8.604668820678515e-06,
|
| 210571 |
+
"loss": 0.3772,
|
| 210572 |
+
"step": 87405
|
| 210573 |
+
},
|
| 210574 |
+
{
|
| 210575 |
+
"epoch": 698.38,
|
| 210576 |
+
"learning_rate": 8.604588045234249e-06,
|
| 210577 |
+
"loss": 0.6627,
|
| 210578 |
+
"step": 87410
|
| 210579 |
+
},
|
| 210580 |
+
{
|
| 210581 |
+
"epoch": 698.42,
|
| 210582 |
+
"learning_rate": 8.604507269789985e-06,
|
| 210583 |
+
"loss": 0.9726,
|
| 210584 |
+
"step": 87415
|
| 210585 |
+
},
|
| 210586 |
+
{
|
| 210587 |
+
"epoch": 698.46,
|
| 210588 |
+
"learning_rate": 8.60442649434572e-06,
|
| 210589 |
+
"loss": 0.2566,
|
| 210590 |
+
"step": 87420
|
| 210591 |
+
},
|
| 210592 |
+
{
|
| 210593 |
+
"epoch": 698.5,
|
| 210594 |
+
"learning_rate": 8.604345718901455e-06,
|
| 210595 |
+
"loss": 0.3177,
|
| 210596 |
+
"step": 87425
|
| 210597 |
+
},
|
| 210598 |
+
{
|
| 210599 |
+
"epoch": 698.54,
|
| 210600 |
+
"learning_rate": 8.60426494345719e-06,
|
| 210601 |
+
"loss": 0.3497,
|
| 210602 |
+
"step": 87430
|
| 210603 |
+
},
|
| 210604 |
+
{
|
| 210605 |
+
"epoch": 698.58,
|
| 210606 |
+
"learning_rate": 8.604184168012925e-06,
|
| 210607 |
+
"loss": 0.5742,
|
| 210608 |
+
"step": 87435
|
| 210609 |
+
},
|
| 210610 |
+
{
|
| 210611 |
+
"epoch": 698.62,
|
| 210612 |
+
"learning_rate": 8.60410339256866e-06,
|
| 210613 |
+
"loss": 0.9627,
|
| 210614 |
+
"step": 87440
|
| 210615 |
+
},
|
| 210616 |
+
{
|
| 210617 |
+
"epoch": 698.66,
|
| 210618 |
+
"learning_rate": 8.604022617124395e-06,
|
| 210619 |
+
"loss": 0.2875,
|
| 210620 |
+
"step": 87445
|
| 210621 |
+
},
|
| 210622 |
+
{
|
| 210623 |
+
"epoch": 698.7,
|
| 210624 |
+
"learning_rate": 8.60394184168013e-06,
|
| 210625 |
+
"loss": 0.3195,
|
| 210626 |
+
"step": 87450
|
| 210627 |
+
},
|
| 210628 |
+
{
|
| 210629 |
+
"epoch": 698.74,
|
| 210630 |
+
"learning_rate": 8.603861066235865e-06,
|
| 210631 |
+
"loss": 0.3011,
|
| 210632 |
+
"step": 87455
|
| 210633 |
+
},
|
| 210634 |
+
{
|
| 210635 |
+
"epoch": 698.78,
|
| 210636 |
+
"learning_rate": 8.6037802907916e-06,
|
| 210637 |
+
"loss": 0.6043,
|
| 210638 |
+
"step": 87460
|
| 210639 |
+
},
|
| 210640 |
+
{
|
| 210641 |
+
"epoch": 698.82,
|
| 210642 |
+
"learning_rate": 8.603699515347335e-06,
|
| 210643 |
+
"loss": 0.9924,
|
| 210644 |
+
"step": 87465
|
| 210645 |
+
},
|
| 210646 |
+
{
|
| 210647 |
+
"epoch": 698.86,
|
| 210648 |
+
"learning_rate": 8.60361873990307e-06,
|
| 210649 |
+
"loss": 0.3596,
|
| 210650 |
+
"step": 87470
|
| 210651 |
+
},
|
| 210652 |
+
{
|
| 210653 |
+
"epoch": 698.9,
|
| 210654 |
+
"learning_rate": 8.603537964458806e-06,
|
| 210655 |
+
"loss": 0.3265,
|
| 210656 |
+
"step": 87475
|
| 210657 |
+
},
|
| 210658 |
+
{
|
| 210659 |
+
"epoch": 698.94,
|
| 210660 |
+
"learning_rate": 8.60345718901454e-06,
|
| 210661 |
+
"loss": 0.3486,
|
| 210662 |
+
"step": 87480
|
| 210663 |
+
},
|
| 210664 |
+
{
|
| 210665 |
+
"epoch": 698.98,
|
| 210666 |
+
"learning_rate": 8.603376413570276e-06,
|
| 210667 |
+
"loss": 0.7913,
|
| 210668 |
+
"step": 87485
|
| 210669 |
+
},
|
| 210670 |
+
{
|
| 210671 |
+
"epoch": 699.0,
|
| 210672 |
+
"eval_loss": 0.4103078842163086,
|
| 210673 |
+
"eval_runtime": 40.1454,
|
| 210674 |
+
"eval_samples_per_second": 20.899,
|
| 210675 |
+
"eval_steps_per_second": 0.673,
|
| 210676 |
+
"eval_wer": 0.18871318294236678,
|
| 210677 |
+
"step": 87488
|
| 210678 |
+
},
|
| 210679 |
+
{
|
| 210680 |
+
"epoch": 699.02,
|
| 210681 |
+
"learning_rate": 8.603311793214864e-06,
|
| 210682 |
+
"loss": 0.3567,
|
| 210683 |
+
"step": 87490
|
| 210684 |
+
},
|
| 210685 |
+
{
|
| 210686 |
+
"epoch": 699.06,
|
| 210687 |
+
"learning_rate": 8.603231017770598e-06,
|
| 210688 |
+
"loss": 0.2734,
|
| 210689 |
+
"step": 87495
|
| 210690 |
+
},
|
| 210691 |
+
{
|
| 210692 |
+
"epoch": 699.1,
|
| 210693 |
+
"learning_rate": 8.603150242326334e-06,
|
| 210694 |
+
"loss": 0.3132,
|
| 210695 |
+
"step": 87500
|
| 210696 |
+
},
|
| 210697 |
+
{
|
| 210698 |
+
"epoch": 699.14,
|
| 210699 |
+
"learning_rate": 8.603069466882068e-06,
|
| 210700 |
+
"loss": 0.3374,
|
| 210701 |
+
"step": 87505
|
| 210702 |
+
},
|
| 210703 |
+
{
|
| 210704 |
+
"epoch": 699.18,
|
| 210705 |
+
"learning_rate": 8.602988691437804e-06,
|
| 210706 |
+
"loss": 0.5836,
|
| 210707 |
+
"step": 87510
|
| 210708 |
+
},
|
| 210709 |
+
{
|
| 210710 |
+
"epoch": 699.22,
|
| 210711 |
+
"learning_rate": 8.60290791599354e-06,
|
| 210712 |
+
"loss": 0.9619,
|
| 210713 |
+
"step": 87515
|
| 210714 |
+
},
|
| 210715 |
+
{
|
| 210716 |
+
"epoch": 699.26,
|
| 210717 |
+
"learning_rate": 8.602827140549274e-06,
|
| 210718 |
+
"loss": 0.2813,
|
| 210719 |
+
"step": 87520
|
| 210720 |
+
},
|
| 210721 |
+
{
|
| 210722 |
+
"epoch": 699.3,
|
| 210723 |
+
"learning_rate": 8.60274636510501e-06,
|
| 210724 |
+
"loss": 0.3142,
|
| 210725 |
+
"step": 87525
|
| 210726 |
+
},
|
| 210727 |
+
{
|
| 210728 |
+
"epoch": 699.34,
|
| 210729 |
+
"learning_rate": 8.602665589660744e-06,
|
| 210730 |
+
"loss": 0.3315,
|
| 210731 |
+
"step": 87530
|
| 210732 |
+
},
|
| 210733 |
+
{
|
| 210734 |
+
"epoch": 699.38,
|
| 210735 |
+
"learning_rate": 8.60258481421648e-06,
|
| 210736 |
+
"loss": 0.5975,
|
| 210737 |
+
"step": 87535
|
| 210738 |
+
},
|
| 210739 |
+
{
|
| 210740 |
+
"epoch": 699.42,
|
| 210741 |
+
"learning_rate": 8.602520193861067e-06,
|
| 210742 |
+
"loss": 0.9943,
|
| 210743 |
+
"step": 87540
|
| 210744 |
+
},
|
| 210745 |
+
{
|
| 210746 |
+
"epoch": 699.46,
|
| 210747 |
+
"learning_rate": 8.602439418416803e-06,
|
| 210748 |
+
"loss": 0.272,
|
| 210749 |
+
"step": 87545
|
| 210750 |
+
},
|
| 210751 |
+
{
|
| 210752 |
+
"epoch": 699.5,
|
| 210753 |
+
"learning_rate": 8.602358642972537e-06,
|
| 210754 |
+
"loss": 0.3074,
|
| 210755 |
+
"step": 87550
|
| 210756 |
+
},
|
| 210757 |
+
{
|
| 210758 |
+
"epoch": 699.54,
|
| 210759 |
+
"learning_rate": 8.602277867528272e-06,
|
| 210760 |
+
"loss": 0.3044,
|
| 210761 |
+
"step": 87555
|
| 210762 |
+
},
|
| 210763 |
+
{
|
| 210764 |
+
"epoch": 699.58,
|
| 210765 |
+
"learning_rate": 8.602197092084007e-06,
|
| 210766 |
+
"loss": 0.6669,
|
| 210767 |
+
"step": 87560
|
| 210768 |
+
},
|
| 210769 |
+
{
|
| 210770 |
+
"epoch": 699.62,
|
| 210771 |
+
"learning_rate": 8.602116316639742e-06,
|
| 210772 |
+
"loss": 1.0493,
|
| 210773 |
+
"step": 87565
|
| 210774 |
+
},
|
| 210775 |
+
{
|
| 210776 |
+
"epoch": 699.66,
|
| 210777 |
+
"learning_rate": 8.602035541195477e-06,
|
| 210778 |
+
"loss": 0.3005,
|
| 210779 |
+
"step": 87570
|
| 210780 |
+
},
|
| 210781 |
+
{
|
| 210782 |
+
"epoch": 699.7,
|
| 210783 |
+
"learning_rate": 8.601954765751212e-06,
|
| 210784 |
+
"loss": 0.3076,
|
| 210785 |
+
"step": 87575
|
| 210786 |
+
},
|
| 210787 |
+
{
|
| 210788 |
+
"epoch": 699.74,
|
| 210789 |
+
"learning_rate": 8.601873990306947e-06,
|
| 210790 |
+
"loss": 0.3347,
|
| 210791 |
+
"step": 87580
|
| 210792 |
+
},
|
| 210793 |
+
{
|
| 210794 |
+
"epoch": 699.78,
|
| 210795 |
+
"learning_rate": 8.601793214862682e-06,
|
| 210796 |
+
"loss": 0.6441,
|
| 210797 |
+
"step": 87585
|
| 210798 |
+
},
|
| 210799 |
+
{
|
| 210800 |
+
"epoch": 699.82,
|
| 210801 |
+
"learning_rate": 8.601712439418417e-06,
|
| 210802 |
+
"loss": 1.043,
|
| 210803 |
+
"step": 87590
|
| 210804 |
+
},
|
| 210805 |
+
{
|
| 210806 |
+
"epoch": 699.86,
|
| 210807 |
+
"learning_rate": 8.601631663974152e-06,
|
| 210808 |
+
"loss": 0.5242,
|
| 210809 |
+
"step": 87595
|
| 210810 |
+
},
|
| 210811 |
+
{
|
| 210812 |
+
"epoch": 699.9,
|
| 210813 |
+
"learning_rate": 8.601550888529887e-06,
|
| 210814 |
+
"loss": 0.2898,
|
| 210815 |
+
"step": 87600
|
| 210816 |
+
},
|
| 210817 |
+
{
|
| 210818 |
+
"epoch": 699.94,
|
| 210819 |
+
"learning_rate": 8.601470113085622e-06,
|
| 210820 |
+
"loss": 0.4386,
|
| 210821 |
+
"step": 87605
|
| 210822 |
+
},
|
| 210823 |
+
{
|
| 210824 |
+
"epoch": 699.98,
|
| 210825 |
+
"learning_rate": 8.601389337641358e-06,
|
| 210826 |
+
"loss": 0.7206,
|
| 210827 |
+
"step": 87610
|
| 210828 |
+
},
|
| 210829 |
+
{
|
| 210830 |
+
"epoch": 700.0,
|
| 210831 |
+
"eval_loss": 0.3917600214481354,
|
| 210832 |
+
"eval_runtime": 40.6477,
|
| 210833 |
+
"eval_samples_per_second": 20.665,
|
| 210834 |
+
"eval_steps_per_second": 0.664,
|
| 210835 |
+
"eval_wer": 0.18650414861590425,
|
| 210836 |
+
"step": 87613
|
| 210837 |
+
},
|
| 210838 |
+
{
|
| 210839 |
+
"epoch": 706.02,
|
| 210840 |
+
"learning_rate": 8.601308562197092e-06,
|
| 210841 |
+
"loss": 0.4084,
|
| 210842 |
+
"step": 87615
|
| 210843 |
+
},
|
| 210844 |
+
{
|
| 210845 |
+
"epoch": 706.06,
|
| 210846 |
+
"learning_rate": 8.601227786752828e-06,
|
| 210847 |
+
"loss": 0.3104,
|
| 210848 |
+
"step": 87620
|
| 210849 |
+
},
|
| 210850 |
+
{
|
| 210851 |
+
"epoch": 706.1,
|
| 210852 |
+
"learning_rate": 8.601147011308562e-06,
|
| 210853 |
+
"loss": 0.3198,
|
| 210854 |
+
"step": 87625
|
| 210855 |
+
},
|
| 210856 |
+
{
|
| 210857 |
+
"epoch": 706.14,
|
| 210858 |
+
"learning_rate": 8.601066235864298e-06,
|
| 210859 |
+
"loss": 0.3715,
|
| 210860 |
+
"step": 87630
|
| 210861 |
+
},
|
| 210862 |
+
{
|
| 210863 |
+
"epoch": 706.18,
|
| 210864 |
+
"learning_rate": 8.600985460420032e-06,
|
| 210865 |
+
"loss": 0.6981,
|
| 210866 |
+
"step": 87635
|
| 210867 |
+
},
|
| 210868 |
+
{
|
| 210869 |
+
"epoch": 706.22,
|
| 210870 |
+
"learning_rate": 8.600904684975768e-06,
|
| 210871 |
+
"loss": 1.048,
|
| 210872 |
+
"step": 87640
|
| 210873 |
+
},
|
| 210874 |
+
{
|
| 210875 |
+
"epoch": 706.26,
|
| 210876 |
+
"learning_rate": 8.600823909531502e-06,
|
| 210877 |
+
"loss": 0.2971,
|
| 210878 |
+
"step": 87645
|
| 210879 |
+
},
|
| 210880 |
+
{
|
| 210881 |
+
"epoch": 706.3,
|
| 210882 |
+
"learning_rate": 8.600743134087238e-06,
|
| 210883 |
+
"loss": 0.2827,
|
| 210884 |
+
"step": 87650
|
| 210885 |
+
},
|
| 210886 |
+
{
|
| 210887 |
+
"epoch": 706.34,
|
| 210888 |
+
"learning_rate": 8.600662358642972e-06,
|
| 210889 |
+
"loss": 0.3783,
|
| 210890 |
+
"step": 87655
|
| 210891 |
+
},
|
| 210892 |
+
{
|
| 210893 |
+
"epoch": 706.38,
|
| 210894 |
+
"learning_rate": 8.600581583198708e-06,
|
| 210895 |
+
"loss": 0.7016,
|
| 210896 |
+
"step": 87660
|
| 210897 |
+
},
|
| 210898 |
+
{
|
| 210899 |
+
"epoch": 706.42,
|
| 210900 |
+
"learning_rate": 8.600500807754444e-06,
|
| 210901 |
+
"loss": 1.0775,
|
| 210902 |
+
"step": 87665
|
| 210903 |
+
},
|
| 210904 |
+
{
|
| 210905 |
+
"epoch": 706.46,
|
| 210906 |
+
"learning_rate": 8.600420032310178e-06,
|
| 210907 |
+
"loss": 0.3293,
|
| 210908 |
+
"step": 87670
|
| 210909 |
+
},
|
| 210910 |
+
{
|
| 210911 |
+
"epoch": 706.5,
|
| 210912 |
+
"learning_rate": 8.600339256865914e-06,
|
| 210913 |
+
"loss": 0.2872,
|
| 210914 |
+
"step": 87675
|
| 210915 |
+
},
|
| 210916 |
+
{
|
| 210917 |
+
"epoch": 706.54,
|
| 210918 |
+
"learning_rate": 8.600258481421648e-06,
|
| 210919 |
+
"loss": 0.39,
|
| 210920 |
+
"step": 87680
|
| 210921 |
+
},
|
| 210922 |
+
{
|
| 210923 |
+
"epoch": 706.58,
|
| 210924 |
+
"learning_rate": 8.600177705977384e-06,
|
| 210925 |
+
"loss": 0.7223,
|
| 210926 |
+
"step": 87685
|
| 210927 |
+
},
|
| 210928 |
+
{
|
| 210929 |
+
"epoch": 706.62,
|
| 210930 |
+
"learning_rate": 8.600096930533118e-06,
|
| 210931 |
+
"loss": 1.2287,
|
| 210932 |
+
"step": 87690
|
| 210933 |
+
},
|
| 210934 |
+
{
|
| 210935 |
+
"epoch": 706.66,
|
| 210936 |
+
"learning_rate": 8.600016155088854e-06,
|
| 210937 |
+
"loss": 0.2745,
|
| 210938 |
+
"step": 87695
|
| 210939 |
+
},
|
| 210940 |
+
{
|
| 210941 |
+
"epoch": 706.7,
|
| 210942 |
+
"learning_rate": 8.599935379644588e-06,
|
| 210943 |
+
"loss": 0.2764,
|
| 210944 |
+
"step": 87700
|
| 210945 |
+
},
|
| 210946 |
+
{
|
| 210947 |
+
"epoch": 706.74,
|
| 210948 |
+
"learning_rate": 8.599854604200324e-06,
|
| 210949 |
+
"loss": 0.3831,
|
| 210950 |
+
"step": 87705
|
| 210951 |
+
},
|
| 210952 |
+
{
|
| 210953 |
+
"epoch": 706.78,
|
| 210954 |
+
"learning_rate": 8.599773828756058e-06,
|
| 210955 |
+
"loss": 0.6149,
|
| 210956 |
+
"step": 87710
|
| 210957 |
+
},
|
| 210958 |
+
{
|
| 210959 |
+
"epoch": 706.82,
|
| 210960 |
+
"learning_rate": 8.599693053311794e-06,
|
| 210961 |
+
"loss": 1.1015,
|
| 210962 |
+
"step": 87715
|
| 210963 |
+
},
|
| 210964 |
+
{
|
| 210965 |
+
"epoch": 706.86,
|
| 210966 |
+
"learning_rate": 8.59961227786753e-06,
|
| 210967 |
+
"loss": 0.3164,
|
| 210968 |
+
"step": 87720
|
| 210969 |
+
},
|
| 210970 |
+
{
|
| 210971 |
+
"epoch": 706.9,
|
| 210972 |
+
"learning_rate": 8.599531502423264e-06,
|
| 210973 |
+
"loss": 0.288,
|
| 210974 |
+
"step": 87725
|
| 210975 |
+
},
|
| 210976 |
+
{
|
| 210977 |
+
"epoch": 706.94,
|
| 210978 |
+
"learning_rate": 8.599450726979e-06,
|
| 210979 |
+
"loss": 0.3646,
|
| 210980 |
+
"step": 87730
|
| 210981 |
+
},
|
| 210982 |
+
{
|
| 210983 |
+
"epoch": 706.98,
|
| 210984 |
+
"learning_rate": 8.599369951534734e-06,
|
| 210985 |
+
"loss": 0.6329,
|
| 210986 |
+
"step": 87735
|
| 210987 |
+
},
|
| 210988 |
+
{
|
| 210989 |
+
"epoch": 707.0,
|
| 210990 |
+
"eval_loss": 0.3942706882953644,
|
| 210991 |
+
"eval_runtime": 41.7893,
|
| 210992 |
+
"eval_samples_per_second": 20.101,
|
| 210993 |
+
"eval_steps_per_second": 0.646,
|
| 210994 |
+
"eval_wer": 0.19399109792284866,
|
| 210995 |
+
"step": 87737
|
| 210996 |
}
|
| 210997 |
],
|
| 210998 |
+
"max_steps": 620000,
|
| 210999 |
"num_train_epochs": 5000,
|
| 211000 |
+
"total_flos": 2.4691105331523263e+20,
|
| 211001 |
"trial_name": null,
|
| 211002 |
"trial_params": null
|
| 211003 |
}
|
model-bin/finetune/base/{checkpoint-87113 → checkpoint-87737}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629930714.9815521/events.out.tfevents.1629930714.7e498afd5545.7645.175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af06aa7dcd8f9cb25319e38df5736860e6a40747e21497e6b1a4db5b8740b96e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629931194.0387235/events.out.tfevents.1629931194.7e498afd5545.7645.177
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb96cd4b7c6e6c88d839a4119789f086b92b4eafa9861a8df32394f55e25133d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629931664.7699187/events.out.tfevents.1629931665.7e498afd5545.7645.179
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd5ef02b15f6d78a5b2b7c30004177e75fae17386dd27d7c8ba9c6f4756a605b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629932126.928111/events.out.tfevents.1629932126.7e498afd5545.7645.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce46ce4476bf18521ecad504974d9920f78f488452e1ed829a1309c24990c6c4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629932592.8902197/events.out.tfevents.1629932592.7e498afd5545.7645.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1535fe8bd16d2cc0071e07af53f96ada5696ede57b615bd9cdc1f2cef58a7db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629930714.7e498afd5545.7645.174
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0804db8425a9385d9682d41a5f72aa0caec02a504d8fc8f89454464de4fd3c93
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629931192.7e498afd5545.7645.176
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:73a3b6177ee8d1ef151dc3eb8964352c504c80132cec39e881b08bdd34c901cf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629931664.7e498afd5545.7645.178
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0eb82878f4e54378c3893d2e71717384c0d04a604289819fdd06839a37199ae
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629932126.7e498afd5545.7645.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a480caa807c50468386799df11a9eb27b73d85c28a17e035ae633b5a0d87596
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629932592.7e498afd5545.7645.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8296a1e7c005f1ba62eaa6492b4233a6b394362b035f6cc714ca1e979d25d8ef
|
| 3 |
+
size 8622
|