"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629906505.5518122/events.out.tfevents.1629906505.7e498afd5545.7645.75 +3 -0
- model-bin/finetune/base/log/1629906984.4172366/events.out.tfevents.1629906984.7e498afd5545.7645.77 +3 -0
- model-bin/finetune/base/log/1629907453.1797047/events.out.tfevents.1629907453.7e498afd5545.7645.79 +3 -0
- model-bin/finetune/base/log/1629907922.5338476/events.out.tfevents.1629907922.7e498afd5545.7645.81 +3 -0
- model-bin/finetune/base/log/1629908391.3240607/events.out.tfevents.1629908391.7e498afd5545.7645.83 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629906505.7e498afd5545.7645.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629906983.7e498afd5545.7645.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629907453.7e498afd5545.7645.78 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629907922.7e498afd5545.7645.80 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629908391.7e498afd5545.7645.82 +3 -0
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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:cee9520ecfcf573b7470abc40c8d1a678f4ce8139f7bc8f0a0df02438bcd739b
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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:cb3faac825cb095efe79913ef557cfc08dda78f8c31112d90e14c7511260d3ee
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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:f36484815dd77b257aac189d6d35455009b7c2f31308b82f2185abfeba225193
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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 sha256:87e2d1745c681c91159da5acdbfd8bc474c3ecf40e467f65b9d5603d68c91173
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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:b0e58b6f31048670ccaa6bae7bbd667060549520d587a77664846428f983e350
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/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,
|
|
@@ -202284,11 +202284,800 @@
|
|
| 202284 |
"eval_steps_per_second": 0.631,
|
| 202285 |
"eval_wer": 0.19266525144572139,
|
| 202286 |
"step": 80891
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 202287 |
}
|
| 202288 |
],
|
| 202289 |
"max_steps": 620000,
|
| 202290 |
"num_train_epochs": 5000,
|
| 202291 |
-
"total_flos": 2.
|
| 202292 |
"trial_name": null,
|
| 202293 |
"trial_params": null
|
| 202294 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 656.995983935743,
|
| 5 |
+
"global_step": 81513,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 202284 |
"eval_steps_per_second": 0.631,
|
| 202285 |
"eval_wer": 0.19266525144572139,
|
| 202286 |
"step": 80891
|
| 202287 |
+
},
|
| 202288 |
+
{
|
| 202289 |
+
"epoch": 647.03,
|
| 202290 |
+
"learning_rate": 8.709789983844912e-06,
|
| 202291 |
+
"loss": 0.377,
|
| 202292 |
+
"step": 80895
|
| 202293 |
+
},
|
| 202294 |
+
{
|
| 202295 |
+
"epoch": 647.07,
|
| 202296 |
+
"learning_rate": 8.709709208400648e-06,
|
| 202297 |
+
"loss": 0.3128,
|
| 202298 |
+
"step": 80900
|
| 202299 |
+
},
|
| 202300 |
+
{
|
| 202301 |
+
"epoch": 647.11,
|
| 202302 |
+
"learning_rate": 8.709628432956382e-06,
|
| 202303 |
+
"loss": 0.3556,
|
| 202304 |
+
"step": 80905
|
| 202305 |
+
},
|
| 202306 |
+
{
|
| 202307 |
+
"epoch": 647.15,
|
| 202308 |
+
"learning_rate": 8.709547657512118e-06,
|
| 202309 |
+
"loss": 0.4434,
|
| 202310 |
+
"step": 80910
|
| 202311 |
+
},
|
| 202312 |
+
{
|
| 202313 |
+
"epoch": 647.19,
|
| 202314 |
+
"learning_rate": 8.709466882067852e-06,
|
| 202315 |
+
"loss": 1.0346,
|
| 202316 |
+
"step": 80915
|
| 202317 |
+
},
|
| 202318 |
+
{
|
| 202319 |
+
"epoch": 647.23,
|
| 202320 |
+
"learning_rate": 8.709386106623588e-06,
|
| 202321 |
+
"loss": 0.6095,
|
| 202322 |
+
"step": 80920
|
| 202323 |
+
},
|
| 202324 |
+
{
|
| 202325 |
+
"epoch": 647.27,
|
| 202326 |
+
"learning_rate": 8.709305331179322e-06,
|
| 202327 |
+
"loss": 0.3047,
|
| 202328 |
+
"step": 80925
|
| 202329 |
+
},
|
| 202330 |
+
{
|
| 202331 |
+
"epoch": 647.31,
|
| 202332 |
+
"learning_rate": 8.709224555735058e-06,
|
| 202333 |
+
"loss": 0.3396,
|
| 202334 |
+
"step": 80930
|
| 202335 |
+
},
|
| 202336 |
+
{
|
| 202337 |
+
"epoch": 647.35,
|
| 202338 |
+
"learning_rate": 8.709143780290792e-06,
|
| 202339 |
+
"loss": 0.3937,
|
| 202340 |
+
"step": 80935
|
| 202341 |
+
},
|
| 202342 |
+
{
|
| 202343 |
+
"epoch": 647.39,
|
| 202344 |
+
"learning_rate": 8.709063004846528e-06,
|
| 202345 |
+
"loss": 0.9028,
|
| 202346 |
+
"step": 80940
|
| 202347 |
+
},
|
| 202348 |
+
{
|
| 202349 |
+
"epoch": 647.43,
|
| 202350 |
+
"learning_rate": 8.708982229402262e-06,
|
| 202351 |
+
"loss": 0.5815,
|
| 202352 |
+
"step": 80945
|
| 202353 |
+
},
|
| 202354 |
+
{
|
| 202355 |
+
"epoch": 647.47,
|
| 202356 |
+
"learning_rate": 8.708901453957997e-06,
|
| 202357 |
+
"loss": 0.276,
|
| 202358 |
+
"step": 80950
|
| 202359 |
+
},
|
| 202360 |
+
{
|
| 202361 |
+
"epoch": 647.51,
|
| 202362 |
+
"learning_rate": 8.708820678513733e-06,
|
| 202363 |
+
"loss": 0.3206,
|
| 202364 |
+
"step": 80955
|
| 202365 |
+
},
|
| 202366 |
+
{
|
| 202367 |
+
"epoch": 647.55,
|
| 202368 |
+
"learning_rate": 8.708739903069467e-06,
|
| 202369 |
+
"loss": 0.408,
|
| 202370 |
+
"step": 80960
|
| 202371 |
+
},
|
| 202372 |
+
{
|
| 202373 |
+
"epoch": 647.59,
|
| 202374 |
+
"learning_rate": 8.708659127625203e-06,
|
| 202375 |
+
"loss": 0.8844,
|
| 202376 |
+
"step": 80965
|
| 202377 |
+
},
|
| 202378 |
+
{
|
| 202379 |
+
"epoch": 647.63,
|
| 202380 |
+
"learning_rate": 8.708578352180937e-06,
|
| 202381 |
+
"loss": 0.6862,
|
| 202382 |
+
"step": 80970
|
| 202383 |
+
},
|
| 202384 |
+
{
|
| 202385 |
+
"epoch": 647.67,
|
| 202386 |
+
"learning_rate": 8.708497576736673e-06,
|
| 202387 |
+
"loss": 0.2837,
|
| 202388 |
+
"step": 80975
|
| 202389 |
+
},
|
| 202390 |
+
{
|
| 202391 |
+
"epoch": 647.71,
|
| 202392 |
+
"learning_rate": 8.708416801292407e-06,
|
| 202393 |
+
"loss": 0.3988,
|
| 202394 |
+
"step": 80980
|
| 202395 |
+
},
|
| 202396 |
+
{
|
| 202397 |
+
"epoch": 647.75,
|
| 202398 |
+
"learning_rate": 8.708336025848143e-06,
|
| 202399 |
+
"loss": 0.4274,
|
| 202400 |
+
"step": 80985
|
| 202401 |
+
},
|
| 202402 |
+
{
|
| 202403 |
+
"epoch": 647.79,
|
| 202404 |
+
"learning_rate": 8.708255250403877e-06,
|
| 202405 |
+
"loss": 0.916,
|
| 202406 |
+
"step": 80990
|
| 202407 |
+
},
|
| 202408 |
+
{
|
| 202409 |
+
"epoch": 647.83,
|
| 202410 |
+
"learning_rate": 8.708174474959613e-06,
|
| 202411 |
+
"loss": 0.8558,
|
| 202412 |
+
"step": 80995
|
| 202413 |
+
},
|
| 202414 |
+
{
|
| 202415 |
+
"epoch": 647.87,
|
| 202416 |
+
"learning_rate": 8.708093699515347e-06,
|
| 202417 |
+
"loss": 0.3033,
|
| 202418 |
+
"step": 81000
|
| 202419 |
+
},
|
| 202420 |
+
{
|
| 202421 |
+
"epoch": 647.91,
|
| 202422 |
+
"learning_rate": 8.708012924071083e-06,
|
| 202423 |
+
"loss": 0.2949,
|
| 202424 |
+
"step": 81005
|
| 202425 |
+
},
|
| 202426 |
+
{
|
| 202427 |
+
"epoch": 647.95,
|
| 202428 |
+
"learning_rate": 8.707932148626819e-06,
|
| 202429 |
+
"loss": 0.4069,
|
| 202430 |
+
"step": 81010
|
| 202431 |
+
},
|
| 202432 |
+
{
|
| 202433 |
+
"epoch": 647.99,
|
| 202434 |
+
"learning_rate": 8.707851373182553e-06,
|
| 202435 |
+
"loss": 0.9068,
|
| 202436 |
+
"step": 81015
|
| 202437 |
+
},
|
| 202438 |
+
{
|
| 202439 |
+
"epoch": 648.0,
|
| 202440 |
+
"eval_loss": 0.37509772181510925,
|
| 202441 |
+
"eval_runtime": 43.9854,
|
| 202442 |
+
"eval_samples_per_second": 19.12,
|
| 202443 |
+
"eval_steps_per_second": 0.614,
|
| 202444 |
+
"eval_wer": 0.19617328519855595,
|
| 202445 |
+
"step": 81016
|
| 202446 |
+
},
|
| 202447 |
+
{
|
| 202448 |
+
"epoch": 648.03,
|
| 202449 |
+
"learning_rate": 8.707770597738289e-06,
|
| 202450 |
+
"loss": 0.3808,
|
| 202451 |
+
"step": 81020
|
| 202452 |
+
},
|
| 202453 |
+
{
|
| 202454 |
+
"epoch": 648.07,
|
| 202455 |
+
"learning_rate": 8.707689822294023e-06,
|
| 202456 |
+
"loss": 0.3607,
|
| 202457 |
+
"step": 81025
|
| 202458 |
+
},
|
| 202459 |
+
{
|
| 202460 |
+
"epoch": 648.11,
|
| 202461 |
+
"learning_rate": 8.707609046849759e-06,
|
| 202462 |
+
"loss": 0.3475,
|
| 202463 |
+
"step": 81030
|
| 202464 |
+
},
|
| 202465 |
+
{
|
| 202466 |
+
"epoch": 648.15,
|
| 202467 |
+
"learning_rate": 8.707528271405493e-06,
|
| 202468 |
+
"loss": 0.4446,
|
| 202469 |
+
"step": 81035
|
| 202470 |
+
},
|
| 202471 |
+
{
|
| 202472 |
+
"epoch": 648.19,
|
| 202473 |
+
"learning_rate": 8.707447495961229e-06,
|
| 202474 |
+
"loss": 0.9625,
|
| 202475 |
+
"step": 81040
|
| 202476 |
+
},
|
| 202477 |
+
{
|
| 202478 |
+
"epoch": 648.23,
|
| 202479 |
+
"learning_rate": 8.707366720516963e-06,
|
| 202480 |
+
"loss": 0.5669,
|
| 202481 |
+
"step": 81045
|
| 202482 |
+
},
|
| 202483 |
+
{
|
| 202484 |
+
"epoch": 648.27,
|
| 202485 |
+
"learning_rate": 8.707285945072699e-06,
|
| 202486 |
+
"loss": 0.3542,
|
| 202487 |
+
"step": 81050
|
| 202488 |
+
},
|
| 202489 |
+
{
|
| 202490 |
+
"epoch": 648.31,
|
| 202491 |
+
"learning_rate": 8.707205169628433e-06,
|
| 202492 |
+
"loss": 0.3174,
|
| 202493 |
+
"step": 81055
|
| 202494 |
+
},
|
| 202495 |
+
{
|
| 202496 |
+
"epoch": 648.35,
|
| 202497 |
+
"learning_rate": 8.707124394184169e-06,
|
| 202498 |
+
"loss": 0.3975,
|
| 202499 |
+
"step": 81060
|
| 202500 |
+
},
|
| 202501 |
+
{
|
| 202502 |
+
"epoch": 648.39,
|
| 202503 |
+
"learning_rate": 8.707043618739905e-06,
|
| 202504 |
+
"loss": 0.8826,
|
| 202505 |
+
"step": 81065
|
| 202506 |
+
},
|
| 202507 |
+
{
|
| 202508 |
+
"epoch": 648.43,
|
| 202509 |
+
"learning_rate": 8.706962843295639e-06,
|
| 202510 |
+
"loss": 0.6087,
|
| 202511 |
+
"step": 81070
|
| 202512 |
+
},
|
| 202513 |
+
{
|
| 202514 |
+
"epoch": 648.47,
|
| 202515 |
+
"learning_rate": 8.706882067851375e-06,
|
| 202516 |
+
"loss": 0.3412,
|
| 202517 |
+
"step": 81075
|
| 202518 |
+
},
|
| 202519 |
+
{
|
| 202520 |
+
"epoch": 648.51,
|
| 202521 |
+
"learning_rate": 8.706801292407109e-06,
|
| 202522 |
+
"loss": 0.3823,
|
| 202523 |
+
"step": 81080
|
| 202524 |
+
},
|
| 202525 |
+
{
|
| 202526 |
+
"epoch": 648.55,
|
| 202527 |
+
"learning_rate": 8.706720516962845e-06,
|
| 202528 |
+
"loss": 0.4852,
|
| 202529 |
+
"step": 81085
|
| 202530 |
+
},
|
| 202531 |
+
{
|
| 202532 |
+
"epoch": 648.59,
|
| 202533 |
+
"learning_rate": 8.706639741518579e-06,
|
| 202534 |
+
"loss": 0.9508,
|
| 202535 |
+
"step": 81090
|
| 202536 |
+
},
|
| 202537 |
+
{
|
| 202538 |
+
"epoch": 648.63,
|
| 202539 |
+
"learning_rate": 8.706558966074315e-06,
|
| 202540 |
+
"loss": 0.6968,
|
| 202541 |
+
"step": 81095
|
| 202542 |
+
},
|
| 202543 |
+
{
|
| 202544 |
+
"epoch": 648.67,
|
| 202545 |
+
"learning_rate": 8.706478190630049e-06,
|
| 202546 |
+
"loss": 0.3213,
|
| 202547 |
+
"step": 81100
|
| 202548 |
+
},
|
| 202549 |
+
{
|
| 202550 |
+
"epoch": 648.71,
|
| 202551 |
+
"learning_rate": 8.706397415185785e-06,
|
| 202552 |
+
"loss": 0.3351,
|
| 202553 |
+
"step": 81105
|
| 202554 |
+
},
|
| 202555 |
+
{
|
| 202556 |
+
"epoch": 648.75,
|
| 202557 |
+
"learning_rate": 8.706316639741519e-06,
|
| 202558 |
+
"loss": 0.4281,
|
| 202559 |
+
"step": 81110
|
| 202560 |
+
},
|
| 202561 |
+
{
|
| 202562 |
+
"epoch": 648.79,
|
| 202563 |
+
"learning_rate": 8.706235864297255e-06,
|
| 202564 |
+
"loss": 0.9528,
|
| 202565 |
+
"step": 81115
|
| 202566 |
+
},
|
| 202567 |
+
{
|
| 202568 |
+
"epoch": 648.83,
|
| 202569 |
+
"learning_rate": 8.706155088852989e-06,
|
| 202570 |
+
"loss": 0.6072,
|
| 202571 |
+
"step": 81120
|
| 202572 |
+
},
|
| 202573 |
+
{
|
| 202574 |
+
"epoch": 648.87,
|
| 202575 |
+
"learning_rate": 8.706074313408725e-06,
|
| 202576 |
+
"loss": 0.2916,
|
| 202577 |
+
"step": 81125
|
| 202578 |
+
},
|
| 202579 |
+
{
|
| 202580 |
+
"epoch": 648.91,
|
| 202581 |
+
"learning_rate": 8.70599353796446e-06,
|
| 202582 |
+
"loss": 0.296,
|
| 202583 |
+
"step": 81130
|
| 202584 |
+
},
|
| 202585 |
+
{
|
| 202586 |
+
"epoch": 648.95,
|
| 202587 |
+
"learning_rate": 8.705912762520195e-06,
|
| 202588 |
+
"loss": 0.501,
|
| 202589 |
+
"step": 81135
|
| 202590 |
+
},
|
| 202591 |
+
{
|
| 202592 |
+
"epoch": 648.99,
|
| 202593 |
+
"learning_rate": 8.70583198707593e-06,
|
| 202594 |
+
"loss": 0.9077,
|
| 202595 |
+
"step": 81140
|
| 202596 |
+
},
|
| 202597 |
+
{
|
| 202598 |
+
"epoch": 649.0,
|
| 202599 |
+
"eval_loss": 0.393308162689209,
|
| 202600 |
+
"eval_runtime": 43.1416,
|
| 202601 |
+
"eval_samples_per_second": 19.494,
|
| 202602 |
+
"eval_steps_per_second": 0.626,
|
| 202603 |
+
"eval_wer": 0.18834796488427774,
|
| 202604 |
+
"step": 81141
|
| 202605 |
+
},
|
| 202606 |
+
{
|
| 202607 |
+
"epoch": 654.03,
|
| 202608 |
+
"learning_rate": 8.705751211631665e-06,
|
| 202609 |
+
"loss": 0.3191,
|
| 202610 |
+
"step": 81145
|
| 202611 |
+
},
|
| 202612 |
+
{
|
| 202613 |
+
"epoch": 654.07,
|
| 202614 |
+
"learning_rate": 8.7056704361874e-06,
|
| 202615 |
+
"loss": 0.3338,
|
| 202616 |
+
"step": 81150
|
| 202617 |
+
},
|
| 202618 |
+
{
|
| 202619 |
+
"epoch": 654.11,
|
| 202620 |
+
"learning_rate": 8.705589660743135e-06,
|
| 202621 |
+
"loss": 0.3336,
|
| 202622 |
+
"step": 81155
|
| 202623 |
+
},
|
| 202624 |
+
{
|
| 202625 |
+
"epoch": 654.15,
|
| 202626 |
+
"learning_rate": 8.70550888529887e-06,
|
| 202627 |
+
"loss": 0.3999,
|
| 202628 |
+
"step": 81160
|
| 202629 |
+
},
|
| 202630 |
+
{
|
| 202631 |
+
"epoch": 654.19,
|
| 202632 |
+
"learning_rate": 8.705428109854604e-06,
|
| 202633 |
+
"loss": 0.9108,
|
| 202634 |
+
"step": 81165
|
| 202635 |
+
},
|
| 202636 |
+
{
|
| 202637 |
+
"epoch": 654.23,
|
| 202638 |
+
"learning_rate": 8.70534733441034e-06,
|
| 202639 |
+
"loss": 0.6993,
|
| 202640 |
+
"step": 81170
|
| 202641 |
+
},
|
| 202642 |
+
{
|
| 202643 |
+
"epoch": 654.27,
|
| 202644 |
+
"learning_rate": 8.705266558966074e-06,
|
| 202645 |
+
"loss": 0.2722,
|
| 202646 |
+
"step": 81175
|
| 202647 |
+
},
|
| 202648 |
+
{
|
| 202649 |
+
"epoch": 654.31,
|
| 202650 |
+
"learning_rate": 8.70518578352181e-06,
|
| 202651 |
+
"loss": 0.3614,
|
| 202652 |
+
"step": 81180
|
| 202653 |
+
},
|
| 202654 |
+
{
|
| 202655 |
+
"epoch": 654.35,
|
| 202656 |
+
"learning_rate": 8.705105008077546e-06,
|
| 202657 |
+
"loss": 0.4454,
|
| 202658 |
+
"step": 81185
|
| 202659 |
+
},
|
| 202660 |
+
{
|
| 202661 |
+
"epoch": 654.39,
|
| 202662 |
+
"learning_rate": 8.70502423263328e-06,
|
| 202663 |
+
"loss": 0.765,
|
| 202664 |
+
"step": 81190
|
| 202665 |
+
},
|
| 202666 |
+
{
|
| 202667 |
+
"epoch": 654.43,
|
| 202668 |
+
"learning_rate": 8.704943457189016e-06,
|
| 202669 |
+
"loss": 0.6223,
|
| 202670 |
+
"step": 81195
|
| 202671 |
+
},
|
| 202672 |
+
{
|
| 202673 |
+
"epoch": 654.47,
|
| 202674 |
+
"learning_rate": 8.70486268174475e-06,
|
| 202675 |
+
"loss": 0.2822,
|
| 202676 |
+
"step": 81200
|
| 202677 |
+
},
|
| 202678 |
+
{
|
| 202679 |
+
"epoch": 654.51,
|
| 202680 |
+
"learning_rate": 8.704781906300486e-06,
|
| 202681 |
+
"loss": 0.2984,
|
| 202682 |
+
"step": 81205
|
| 202683 |
+
},
|
| 202684 |
+
{
|
| 202685 |
+
"epoch": 654.55,
|
| 202686 |
+
"learning_rate": 8.70470113085622e-06,
|
| 202687 |
+
"loss": 0.3864,
|
| 202688 |
+
"step": 81210
|
| 202689 |
+
},
|
| 202690 |
+
{
|
| 202691 |
+
"epoch": 654.59,
|
| 202692 |
+
"learning_rate": 8.704620355411956e-06,
|
| 202693 |
+
"loss": 1.0032,
|
| 202694 |
+
"step": 81215
|
| 202695 |
+
},
|
| 202696 |
+
{
|
| 202697 |
+
"epoch": 654.63,
|
| 202698 |
+
"learning_rate": 8.70453957996769e-06,
|
| 202699 |
+
"loss": 0.7351,
|
| 202700 |
+
"step": 81220
|
| 202701 |
+
},
|
| 202702 |
+
{
|
| 202703 |
+
"epoch": 654.67,
|
| 202704 |
+
"learning_rate": 8.704458804523426e-06,
|
| 202705 |
+
"loss": 0.356,
|
| 202706 |
+
"step": 81225
|
| 202707 |
+
},
|
| 202708 |
+
{
|
| 202709 |
+
"epoch": 654.71,
|
| 202710 |
+
"learning_rate": 8.70437802907916e-06,
|
| 202711 |
+
"loss": 0.3058,
|
| 202712 |
+
"step": 81230
|
| 202713 |
+
},
|
| 202714 |
+
{
|
| 202715 |
+
"epoch": 654.76,
|
| 202716 |
+
"learning_rate": 8.704297253634896e-06,
|
| 202717 |
+
"loss": 0.469,
|
| 202718 |
+
"step": 81235
|
| 202719 |
+
},
|
| 202720 |
+
{
|
| 202721 |
+
"epoch": 654.8,
|
| 202722 |
+
"learning_rate": 8.704216478190632e-06,
|
| 202723 |
+
"loss": 1.0174,
|
| 202724 |
+
"step": 81240
|
| 202725 |
+
},
|
| 202726 |
+
{
|
| 202727 |
+
"epoch": 654.84,
|
| 202728 |
+
"learning_rate": 8.704135702746366e-06,
|
| 202729 |
+
"loss": 0.6536,
|
| 202730 |
+
"step": 81245
|
| 202731 |
+
},
|
| 202732 |
+
{
|
| 202733 |
+
"epoch": 654.88,
|
| 202734 |
+
"learning_rate": 8.704054927302102e-06,
|
| 202735 |
+
"loss": 0.3065,
|
| 202736 |
+
"step": 81250
|
| 202737 |
+
},
|
| 202738 |
+
{
|
| 202739 |
+
"epoch": 654.92,
|
| 202740 |
+
"learning_rate": 8.703974151857836e-06,
|
| 202741 |
+
"loss": 0.3085,
|
| 202742 |
+
"step": 81255
|
| 202743 |
+
},
|
| 202744 |
+
{
|
| 202745 |
+
"epoch": 654.96,
|
| 202746 |
+
"learning_rate": 8.703893376413572e-06,
|
| 202747 |
+
"loss": 0.512,
|
| 202748 |
+
"step": 81260
|
| 202749 |
+
},
|
| 202750 |
+
{
|
| 202751 |
+
"epoch": 655.0,
|
| 202752 |
+
"learning_rate": 8.703812600969306e-06,
|
| 202753 |
+
"loss": 1.0805,
|
| 202754 |
+
"step": 81265
|
| 202755 |
+
},
|
| 202756 |
+
{
|
| 202757 |
+
"epoch": 655.0,
|
| 202758 |
+
"eval_loss": 0.3865291178226471,
|
| 202759 |
+
"eval_runtime": 40.8909,
|
| 202760 |
+
"eval_samples_per_second": 20.567,
|
| 202761 |
+
"eval_steps_per_second": 0.66,
|
| 202762 |
+
"eval_wer": 0.19052887807036956,
|
| 202763 |
+
"step": 81265
|
| 202764 |
+
},
|
| 202765 |
+
{
|
| 202766 |
+
"epoch": 655.04,
|
| 202767 |
+
"learning_rate": 8.703731825525042e-06,
|
| 202768 |
+
"loss": 0.3231,
|
| 202769 |
+
"step": 81270
|
| 202770 |
+
},
|
| 202771 |
+
{
|
| 202772 |
+
"epoch": 655.08,
|
| 202773 |
+
"learning_rate": 8.703651050080776e-06,
|
| 202774 |
+
"loss": 0.2945,
|
| 202775 |
+
"step": 81275
|
| 202776 |
+
},
|
| 202777 |
+
{
|
| 202778 |
+
"epoch": 655.12,
|
| 202779 |
+
"learning_rate": 8.703570274636512e-06,
|
| 202780 |
+
"loss": 0.3442,
|
| 202781 |
+
"step": 81280
|
| 202782 |
+
},
|
| 202783 |
+
{
|
| 202784 |
+
"epoch": 655.16,
|
| 202785 |
+
"learning_rate": 8.703489499192246e-06,
|
| 202786 |
+
"loss": 0.4286,
|
| 202787 |
+
"step": 81285
|
| 202788 |
+
},
|
| 202789 |
+
{
|
| 202790 |
+
"epoch": 655.2,
|
| 202791 |
+
"learning_rate": 8.703408723747982e-06,
|
| 202792 |
+
"loss": 1.2732,
|
| 202793 |
+
"step": 81290
|
| 202794 |
+
},
|
| 202795 |
+
{
|
| 202796 |
+
"epoch": 655.24,
|
| 202797 |
+
"learning_rate": 8.703327948303716e-06,
|
| 202798 |
+
"loss": 0.3129,
|
| 202799 |
+
"step": 81295
|
| 202800 |
+
},
|
| 202801 |
+
{
|
| 202802 |
+
"epoch": 655.28,
|
| 202803 |
+
"learning_rate": 8.703247172859452e-06,
|
| 202804 |
+
"loss": 0.373,
|
| 202805 |
+
"step": 81300
|
| 202806 |
+
},
|
| 202807 |
+
{
|
| 202808 |
+
"epoch": 655.32,
|
| 202809 |
+
"learning_rate": 8.703166397415188e-06,
|
| 202810 |
+
"loss": 0.5552,
|
| 202811 |
+
"step": 81305
|
| 202812 |
+
},
|
| 202813 |
+
{
|
| 202814 |
+
"epoch": 655.36,
|
| 202815 |
+
"learning_rate": 8.703085621970922e-06,
|
| 202816 |
+
"loss": 0.4064,
|
| 202817 |
+
"step": 81310
|
| 202818 |
+
},
|
| 202819 |
+
{
|
| 202820 |
+
"epoch": 655.4,
|
| 202821 |
+
"learning_rate": 8.703004846526657e-06,
|
| 202822 |
+
"loss": 1.1142,
|
| 202823 |
+
"step": 81315
|
| 202824 |
+
},
|
| 202825 |
+
{
|
| 202826 |
+
"epoch": 655.44,
|
| 202827 |
+
"learning_rate": 8.702924071082392e-06,
|
| 202828 |
+
"loss": 0.3153,
|
| 202829 |
+
"step": 81320
|
| 202830 |
+
},
|
| 202831 |
+
{
|
| 202832 |
+
"epoch": 655.48,
|
| 202833 |
+
"learning_rate": 8.702843295638127e-06,
|
| 202834 |
+
"loss": 0.2903,
|
| 202835 |
+
"step": 81325
|
| 202836 |
+
},
|
| 202837 |
+
{
|
| 202838 |
+
"epoch": 655.52,
|
| 202839 |
+
"learning_rate": 8.702762520193862e-06,
|
| 202840 |
+
"loss": 0.3914,
|
| 202841 |
+
"step": 81330
|
| 202842 |
+
},
|
| 202843 |
+
{
|
| 202844 |
+
"epoch": 655.56,
|
| 202845 |
+
"learning_rate": 8.702681744749597e-06,
|
| 202846 |
+
"loss": 0.5081,
|
| 202847 |
+
"step": 81335
|
| 202848 |
+
},
|
| 202849 |
+
{
|
| 202850 |
+
"epoch": 655.6,
|
| 202851 |
+
"learning_rate": 8.702600969305332e-06,
|
| 202852 |
+
"loss": 1.1017,
|
| 202853 |
+
"step": 81340
|
| 202854 |
+
},
|
| 202855 |
+
{
|
| 202856 |
+
"epoch": 655.64,
|
| 202857 |
+
"learning_rate": 8.702520193861067e-06,
|
| 202858 |
+
"loss": 0.5924,
|
| 202859 |
+
"step": 81345
|
| 202860 |
+
},
|
| 202861 |
+
{
|
| 202862 |
+
"epoch": 655.68,
|
| 202863 |
+
"learning_rate": 8.702439418416802e-06,
|
| 202864 |
+
"loss": 0.284,
|
| 202865 |
+
"step": 81350
|
| 202866 |
+
},
|
| 202867 |
+
{
|
| 202868 |
+
"epoch": 655.72,
|
| 202869 |
+
"learning_rate": 8.702358642972537e-06,
|
| 202870 |
+
"loss": 0.2982,
|
| 202871 |
+
"step": 81355
|
| 202872 |
+
},
|
| 202873 |
+
{
|
| 202874 |
+
"epoch": 655.76,
|
| 202875 |
+
"learning_rate": 8.702277867528273e-06,
|
| 202876 |
+
"loss": 0.4095,
|
| 202877 |
+
"step": 81360
|
| 202878 |
+
},
|
| 202879 |
+
{
|
| 202880 |
+
"epoch": 655.8,
|
| 202881 |
+
"learning_rate": 8.702197092084007e-06,
|
| 202882 |
+
"loss": 1.2279,
|
| 202883 |
+
"step": 81365
|
| 202884 |
+
},
|
| 202885 |
+
{
|
| 202886 |
+
"epoch": 655.84,
|
| 202887 |
+
"learning_rate": 8.702116316639743e-06,
|
| 202888 |
+
"loss": 0.3089,
|
| 202889 |
+
"step": 81370
|
| 202890 |
+
},
|
| 202891 |
+
{
|
| 202892 |
+
"epoch": 655.88,
|
| 202893 |
+
"learning_rate": 8.702035541195477e-06,
|
| 202894 |
+
"loss": 0.3333,
|
| 202895 |
+
"step": 81375
|
| 202896 |
+
},
|
| 202897 |
+
{
|
| 202898 |
+
"epoch": 655.92,
|
| 202899 |
+
"learning_rate": 8.701954765751213e-06,
|
| 202900 |
+
"loss": 0.3419,
|
| 202901 |
+
"step": 81380
|
| 202902 |
+
},
|
| 202903 |
+
{
|
| 202904 |
+
"epoch": 655.96,
|
| 202905 |
+
"learning_rate": 8.701873990306947e-06,
|
| 202906 |
+
"loss": 0.5394,
|
| 202907 |
+
"step": 81385
|
| 202908 |
+
},
|
| 202909 |
+
{
|
| 202910 |
+
"epoch": 656.0,
|
| 202911 |
+
"eval_loss": 0.4789983630180359,
|
| 202912 |
+
"eval_runtime": 42.2511,
|
| 202913 |
+
"eval_samples_per_second": 19.905,
|
| 202914 |
+
"eval_steps_per_second": 0.639,
|
| 202915 |
+
"eval_wer": 0.1910178322448081,
|
| 202916 |
+
"step": 81389
|
| 202917 |
+
},
|
| 202918 |
+
{
|
| 202919 |
+
"epoch": 656.01,
|
| 202920 |
+
"learning_rate": 8.701793214862683e-06,
|
| 202921 |
+
"loss": 0.3377,
|
| 202922 |
+
"step": 81390
|
| 202923 |
+
},
|
| 202924 |
+
{
|
| 202925 |
+
"epoch": 656.05,
|
| 202926 |
+
"learning_rate": 8.701712439418417e-06,
|
| 202927 |
+
"loss": 0.3423,
|
| 202928 |
+
"step": 81395
|
| 202929 |
+
},
|
| 202930 |
+
{
|
| 202931 |
+
"epoch": 656.09,
|
| 202932 |
+
"learning_rate": 8.701631663974153e-06,
|
| 202933 |
+
"loss": 0.3462,
|
| 202934 |
+
"step": 81400
|
| 202935 |
+
},
|
| 202936 |
+
{
|
| 202937 |
+
"epoch": 656.13,
|
| 202938 |
+
"learning_rate": 8.701550888529887e-06,
|
| 202939 |
+
"loss": 0.3524,
|
| 202940 |
+
"step": 81405
|
| 202941 |
+
},
|
| 202942 |
+
{
|
| 202943 |
+
"epoch": 656.17,
|
| 202944 |
+
"learning_rate": 8.701470113085623e-06,
|
| 202945 |
+
"loss": 0.5533,
|
| 202946 |
+
"step": 81410
|
| 202947 |
+
},
|
| 202948 |
+
{
|
| 202949 |
+
"epoch": 656.21,
|
| 202950 |
+
"learning_rate": 8.701389337641359e-06,
|
| 202951 |
+
"loss": 1.0987,
|
| 202952 |
+
"step": 81415
|
| 202953 |
+
},
|
| 202954 |
+
{
|
| 202955 |
+
"epoch": 656.25,
|
| 202956 |
+
"learning_rate": 8.701308562197093e-06,
|
| 202957 |
+
"loss": 0.2836,
|
| 202958 |
+
"step": 81420
|
| 202959 |
+
},
|
| 202960 |
+
{
|
| 202961 |
+
"epoch": 656.29,
|
| 202962 |
+
"learning_rate": 8.701227786752829e-06,
|
| 202963 |
+
"loss": 0.2713,
|
| 202964 |
+
"step": 81425
|
| 202965 |
+
},
|
| 202966 |
+
{
|
| 202967 |
+
"epoch": 656.33,
|
| 202968 |
+
"learning_rate": 8.701147011308563e-06,
|
| 202969 |
+
"loss": 0.4014,
|
| 202970 |
+
"step": 81430
|
| 202971 |
+
},
|
| 202972 |
+
{
|
| 202973 |
+
"epoch": 656.37,
|
| 202974 |
+
"learning_rate": 8.701066235864299e-06,
|
| 202975 |
+
"loss": 0.5932,
|
| 202976 |
+
"step": 81435
|
| 202977 |
+
},
|
| 202978 |
+
{
|
| 202979 |
+
"epoch": 656.41,
|
| 202980 |
+
"learning_rate": 8.700985460420033e-06,
|
| 202981 |
+
"loss": 1.1982,
|
| 202982 |
+
"step": 81440
|
| 202983 |
+
},
|
| 202984 |
+
{
|
| 202985 |
+
"epoch": 656.45,
|
| 202986 |
+
"learning_rate": 8.700904684975769e-06,
|
| 202987 |
+
"loss": 0.4703,
|
| 202988 |
+
"step": 81445
|
| 202989 |
+
},
|
| 202990 |
+
{
|
| 202991 |
+
"epoch": 656.49,
|
| 202992 |
+
"learning_rate": 8.700823909531503e-06,
|
| 202993 |
+
"loss": 0.3237,
|
| 202994 |
+
"step": 81450
|
| 202995 |
+
},
|
| 202996 |
+
{
|
| 202997 |
+
"epoch": 656.53,
|
| 202998 |
+
"learning_rate": 8.700743134087239e-06,
|
| 202999 |
+
"loss": 0.3723,
|
| 203000 |
+
"step": 81455
|
| 203001 |
+
},
|
| 203002 |
+
{
|
| 203003 |
+
"epoch": 656.57,
|
| 203004 |
+
"learning_rate": 8.700662358642973e-06,
|
| 203005 |
+
"loss": 0.5361,
|
| 203006 |
+
"step": 81460
|
| 203007 |
+
},
|
| 203008 |
+
{
|
| 203009 |
+
"epoch": 656.61,
|
| 203010 |
+
"learning_rate": 8.700581583198709e-06,
|
| 203011 |
+
"loss": 1.2484,
|
| 203012 |
+
"step": 81465
|
| 203013 |
+
},
|
| 203014 |
+
{
|
| 203015 |
+
"epoch": 656.65,
|
| 203016 |
+
"learning_rate": 8.700500807754443e-06,
|
| 203017 |
+
"loss": 0.3541,
|
| 203018 |
+
"step": 81470
|
| 203019 |
+
},
|
| 203020 |
+
{
|
| 203021 |
+
"epoch": 656.69,
|
| 203022 |
+
"learning_rate": 8.700420032310179e-06,
|
| 203023 |
+
"loss": 0.3066,
|
| 203024 |
+
"step": 81475
|
| 203025 |
+
},
|
| 203026 |
+
{
|
| 203027 |
+
"epoch": 656.73,
|
| 203028 |
+
"learning_rate": 8.700339256865915e-06,
|
| 203029 |
+
"loss": 0.3067,
|
| 203030 |
+
"step": 81480
|
| 203031 |
+
},
|
| 203032 |
+
{
|
| 203033 |
+
"epoch": 656.77,
|
| 203034 |
+
"learning_rate": 8.700258481421649e-06,
|
| 203035 |
+
"loss": 0.5809,
|
| 203036 |
+
"step": 81485
|
| 203037 |
+
},
|
| 203038 |
+
{
|
| 203039 |
+
"epoch": 656.81,
|
| 203040 |
+
"learning_rate": 8.700177705977385e-06,
|
| 203041 |
+
"loss": 1.1829,
|
| 203042 |
+
"step": 81490
|
| 203043 |
+
},
|
| 203044 |
+
{
|
| 203045 |
+
"epoch": 656.85,
|
| 203046 |
+
"learning_rate": 8.700096930533119e-06,
|
| 203047 |
+
"loss": 0.3181,
|
| 203048 |
+
"step": 81495
|
| 203049 |
+
},
|
| 203050 |
+
{
|
| 203051 |
+
"epoch": 656.89,
|
| 203052 |
+
"learning_rate": 8.700016155088855e-06,
|
| 203053 |
+
"loss": 0.2962,
|
| 203054 |
+
"step": 81500
|
| 203055 |
+
},
|
| 203056 |
+
{
|
| 203057 |
+
"epoch": 656.93,
|
| 203058 |
+
"learning_rate": 8.699935379644589e-06,
|
| 203059 |
+
"loss": 0.2912,
|
| 203060 |
+
"step": 81505
|
| 203061 |
+
},
|
| 203062 |
+
{
|
| 203063 |
+
"epoch": 656.97,
|
| 203064 |
+
"learning_rate": 8.699854604200325e-06,
|
| 203065 |
+
"loss": 0.6958,
|
| 203066 |
+
"step": 81510
|
| 203067 |
+
},
|
| 203068 |
+
{
|
| 203069 |
+
"epoch": 657.0,
|
| 203070 |
+
"eval_loss": 0.3742680847644806,
|
| 203071 |
+
"eval_runtime": 45.4213,
|
| 203072 |
+
"eval_samples_per_second": 18.494,
|
| 203073 |
+
"eval_steps_per_second": 0.594,
|
| 203074 |
+
"eval_wer": 0.1944121915820029,
|
| 203075 |
+
"step": 81513
|
| 203076 |
}
|
| 203077 |
],
|
| 203078 |
"max_steps": 620000,
|
| 203079 |
"num_train_epochs": 5000,
|
| 203080 |
+
"total_flos": 2.2939365245267496e+20,
|
| 203081 |
"trial_name": null,
|
| 203082 |
"trial_params": null
|
| 203083 |
}
|
model-bin/finetune/base/{checkpoint-80891 → checkpoint-81513}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629906505.5518122/events.out.tfevents.1629906505.7e498afd5545.7645.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b970b57a5430a0c15b0472e5f25afdb0320e18202ed228c6d9c2e4cc20346ca9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629906984.4172366/events.out.tfevents.1629906984.7e498afd5545.7645.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ee7c0148e95f76a241655beae8239d7c40ad57b525b8875958b66f16bf9f960e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629907453.1797047/events.out.tfevents.1629907453.7e498afd5545.7645.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be67e72630bba267ce6875e9c2386d8c1938f97ccb1267d8a2ee75a9e4cfab84
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629907922.5338476/events.out.tfevents.1629907922.7e498afd5545.7645.81
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b218eec03b0794c046442f4a78b23629860df03e1adfba3fa9af5a8d1fb4c5e4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629908391.3240607/events.out.tfevents.1629908391.7e498afd5545.7645.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e190c775f17a636aaf7bf274b3dfd4d44de8184f7736f80ff537cbbee9f27cab
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629906505.7e498afd5545.7645.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:51c2da05d36c93beb130a76a0a71932e25169367fbcae1f37b42f906926bdebd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629906983.7e498afd5545.7645.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4f2d305af42d00e80abd67bf90d6b1b2d4d049ae01053b5351ad0b73d28bc03
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629907453.7e498afd5545.7645.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c86766569d14a1aaf7bf3b6d89bf0ac3473d1dec14cc9163a6d67ebc73215a4d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629907922.7e498afd5545.7645.80
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b95acbd7f47bb7ec44916fd3c32661fe47bfb60be8cc2a6730184b0be2b2917f
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629908391.7e498afd5545.7645.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1b44d321b78bbdeffada43bffe6ac0fffefc07858d9327518de1dcc62f73c6a
|
| 3 |
+
size 8622
|