"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629775137.1840463/events.out.tfevents.1629775137.c435e1c5ee04.920.31 +3 -0
- model-bin/finetune/base/log/1629775789.2149596/events.out.tfevents.1629775789.c435e1c5ee04.920.33 +3 -0
- model-bin/finetune/base/log/1629776456.5037358/events.out.tfevents.1629776456.c435e1c5ee04.920.35 +3 -0
- model-bin/finetune/base/log/1629777113.4720376/events.out.tfevents.1629777113.c435e1c5ee04.920.37 +3 -0
- model-bin/finetune/base/log/1629777774.3360224/events.out.tfevents.1629777774.c435e1c5ee04.920.39 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629775137.c435e1c5ee04.920.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629775789.c435e1c5ee04.920.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629776456.c435e1c5ee04.920.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629777113.c435e1c5ee04.920.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629777774.c435e1c5ee04.920.38 +3 -0
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9588d1cdc17839a81d3b9833fa479a88e833d5e7e4445b67e88528d92457cd28
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/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 sha256:0bdd0211a727cf8c3094eb7dd0798468d83876617b84e685844e1e82c9ef74b5
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/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 sha256:886ff01e12101694ba90eee9a9b7d639459bf4fb95e67d4c35cce050d60d612e
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/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 sha256:93631ed3f39e4cb03d22fbc0af7e360abac36a33350849fe88351d54a6afd41c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/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 sha256:90699fcee6c06088cbb78a3b5245a2e25cdf72d6b45818c2c1fb1b391e326908
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -170298,11 +170298,800 @@
|
|
| 170298 |
"eval_steps_per_second": 0.674,
|
| 170299 |
"eval_wer": 0.2049155834453907,
|
| 170300 |
"step": 55750
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 170301 |
}
|
| 170302 |
],
|
| 170303 |
"max_steps": 620000,
|
| 170304 |
"num_train_epochs": 5000,
|
| 170305 |
-
"total_flos": 1.
|
| 170306 |
"trial_name": null,
|
| 170307 |
"trial_params": null
|
| 170308 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 453.99598393574297,
|
| 5 |
+
"global_step": 56372,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 170298 |
"eval_steps_per_second": 0.674,
|
| 170299 |
"eval_wer": 0.2049155834453907,
|
| 170300 |
"step": 55750
|
| 170301 |
+
},
|
| 170302 |
+
{
|
| 170303 |
+
"epoch": 449.04,
|
| 170304 |
+
"learning_rate": 9.122852564102564e-06,
|
| 170305 |
+
"loss": 0.3862,
|
| 170306 |
+
"step": 55755
|
| 170307 |
+
},
|
| 170308 |
+
{
|
| 170309 |
+
"epoch": 449.08,
|
| 170310 |
+
"learning_rate": 9.122772435897438e-06,
|
| 170311 |
+
"loss": 0.3152,
|
| 170312 |
+
"step": 55760
|
| 170313 |
+
},
|
| 170314 |
+
{
|
| 170315 |
+
"epoch": 449.12,
|
| 170316 |
+
"learning_rate": 9.122692307692307e-06,
|
| 170317 |
+
"loss": 0.3497,
|
| 170318 |
+
"step": 55765
|
| 170319 |
+
},
|
| 170320 |
+
{
|
| 170321 |
+
"epoch": 449.16,
|
| 170322 |
+
"learning_rate": 9.12261217948718e-06,
|
| 170323 |
+
"loss": 0.599,
|
| 170324 |
+
"step": 55770
|
| 170325 |
+
},
|
| 170326 |
+
{
|
| 170327 |
+
"epoch": 449.2,
|
| 170328 |
+
"learning_rate": 9.122532051282053e-06,
|
| 170329 |
+
"loss": 1.2486,
|
| 170330 |
+
"step": 55775
|
| 170331 |
+
},
|
| 170332 |
+
{
|
| 170333 |
+
"epoch": 449.24,
|
| 170334 |
+
"learning_rate": 9.122451923076923e-06,
|
| 170335 |
+
"loss": 0.3865,
|
| 170336 |
+
"step": 55780
|
| 170337 |
+
},
|
| 170338 |
+
{
|
| 170339 |
+
"epoch": 449.28,
|
| 170340 |
+
"learning_rate": 9.122371794871796e-06,
|
| 170341 |
+
"loss": 0.2894,
|
| 170342 |
+
"step": 55785
|
| 170343 |
+
},
|
| 170344 |
+
{
|
| 170345 |
+
"epoch": 449.32,
|
| 170346 |
+
"learning_rate": 9.122291666666667e-06,
|
| 170347 |
+
"loss": 0.3429,
|
| 170348 |
+
"step": 55790
|
| 170349 |
+
},
|
| 170350 |
+
{
|
| 170351 |
+
"epoch": 449.36,
|
| 170352 |
+
"learning_rate": 9.122211538461539e-06,
|
| 170353 |
+
"loss": 0.8315,
|
| 170354 |
+
"step": 55795
|
| 170355 |
+
},
|
| 170356 |
+
{
|
| 170357 |
+
"epoch": 449.4,
|
| 170358 |
+
"learning_rate": 9.12213141025641e-06,
|
| 170359 |
+
"loss": 1.2594,
|
| 170360 |
+
"step": 55800
|
| 170361 |
+
},
|
| 170362 |
+
{
|
| 170363 |
+
"epoch": 449.44,
|
| 170364 |
+
"learning_rate": 9.122051282051283e-06,
|
| 170365 |
+
"loss": 0.3777,
|
| 170366 |
+
"step": 55805
|
| 170367 |
+
},
|
| 170368 |
+
{
|
| 170369 |
+
"epoch": 449.48,
|
| 170370 |
+
"learning_rate": 9.121971153846155e-06,
|
| 170371 |
+
"loss": 0.3153,
|
| 170372 |
+
"step": 55810
|
| 170373 |
+
},
|
| 170374 |
+
{
|
| 170375 |
+
"epoch": 449.52,
|
| 170376 |
+
"learning_rate": 9.121891025641026e-06,
|
| 170377 |
+
"loss": 0.3825,
|
| 170378 |
+
"step": 55815
|
| 170379 |
+
},
|
| 170380 |
+
{
|
| 170381 |
+
"epoch": 449.56,
|
| 170382 |
+
"learning_rate": 9.121810897435897e-06,
|
| 170383 |
+
"loss": 0.4814,
|
| 170384 |
+
"step": 55820
|
| 170385 |
+
},
|
| 170386 |
+
{
|
| 170387 |
+
"epoch": 449.6,
|
| 170388 |
+
"learning_rate": 9.12173076923077e-06,
|
| 170389 |
+
"loss": 1.2004,
|
| 170390 |
+
"step": 55825
|
| 170391 |
+
},
|
| 170392 |
+
{
|
| 170393 |
+
"epoch": 449.64,
|
| 170394 |
+
"learning_rate": 9.121650641025642e-06,
|
| 170395 |
+
"loss": 0.3668,
|
| 170396 |
+
"step": 55830
|
| 170397 |
+
},
|
| 170398 |
+
{
|
| 170399 |
+
"epoch": 449.68,
|
| 170400 |
+
"learning_rate": 9.121570512820513e-06,
|
| 170401 |
+
"loss": 0.3436,
|
| 170402 |
+
"step": 55835
|
| 170403 |
+
},
|
| 170404 |
+
{
|
| 170405 |
+
"epoch": 449.72,
|
| 170406 |
+
"learning_rate": 9.121490384615386e-06,
|
| 170407 |
+
"loss": 0.3016,
|
| 170408 |
+
"step": 55840
|
| 170409 |
+
},
|
| 170410 |
+
{
|
| 170411 |
+
"epoch": 449.76,
|
| 170412 |
+
"learning_rate": 9.121410256410257e-06,
|
| 170413 |
+
"loss": 0.5931,
|
| 170414 |
+
"step": 55845
|
| 170415 |
+
},
|
| 170416 |
+
{
|
| 170417 |
+
"epoch": 449.8,
|
| 170418 |
+
"learning_rate": 9.121330128205129e-06,
|
| 170419 |
+
"loss": 1.3251,
|
| 170420 |
+
"step": 55850
|
| 170421 |
+
},
|
| 170422 |
+
{
|
| 170423 |
+
"epoch": 449.84,
|
| 170424 |
+
"learning_rate": 9.12125e-06,
|
| 170425 |
+
"loss": 0.3472,
|
| 170426 |
+
"step": 55855
|
| 170427 |
+
},
|
| 170428 |
+
{
|
| 170429 |
+
"epoch": 449.88,
|
| 170430 |
+
"learning_rate": 9.121169871794873e-06,
|
| 170431 |
+
"loss": 0.3114,
|
| 170432 |
+
"step": 55860
|
| 170433 |
+
},
|
| 170434 |
+
{
|
| 170435 |
+
"epoch": 449.92,
|
| 170436 |
+
"learning_rate": 9.121089743589745e-06,
|
| 170437 |
+
"loss": 0.4166,
|
| 170438 |
+
"step": 55865
|
| 170439 |
+
},
|
| 170440 |
+
{
|
| 170441 |
+
"epoch": 449.96,
|
| 170442 |
+
"learning_rate": 9.121009615384616e-06,
|
| 170443 |
+
"loss": 0.627,
|
| 170444 |
+
"step": 55870
|
| 170445 |
+
},
|
| 170446 |
+
{
|
| 170447 |
+
"epoch": 450.0,
|
| 170448 |
+
"eval_loss": 0.4388013780117035,
|
| 170449 |
+
"eval_runtime": 40.1754,
|
| 170450 |
+
"eval_samples_per_second": 20.883,
|
| 170451 |
+
"eval_steps_per_second": 0.672,
|
| 170452 |
+
"eval_wer": 0.20151898734177215,
|
| 170453 |
+
"step": 55874
|
| 170454 |
+
},
|
| 170455 |
+
{
|
| 170456 |
+
"epoch": 446.01,
|
| 170457 |
+
"learning_rate": 9.120929487179489e-06,
|
| 170458 |
+
"loss": 0.4029,
|
| 170459 |
+
"step": 55875
|
| 170460 |
+
},
|
| 170461 |
+
{
|
| 170462 |
+
"epoch": 446.05,
|
| 170463 |
+
"learning_rate": 9.12084935897436e-06,
|
| 170464 |
+
"loss": 0.3038,
|
| 170465 |
+
"step": 55880
|
| 170466 |
+
},
|
| 170467 |
+
{
|
| 170468 |
+
"epoch": 446.09,
|
| 170469 |
+
"learning_rate": 9.120769230769232e-06,
|
| 170470 |
+
"loss": 0.3231,
|
| 170471 |
+
"step": 55885
|
| 170472 |
+
},
|
| 170473 |
+
{
|
| 170474 |
+
"epoch": 446.13,
|
| 170475 |
+
"learning_rate": 9.120689102564103e-06,
|
| 170476 |
+
"loss": 0.4361,
|
| 170477 |
+
"step": 55890
|
| 170478 |
+
},
|
| 170479 |
+
{
|
| 170480 |
+
"epoch": 446.17,
|
| 170481 |
+
"learning_rate": 9.120608974358976e-06,
|
| 170482 |
+
"loss": 0.5979,
|
| 170483 |
+
"step": 55895
|
| 170484 |
+
},
|
| 170485 |
+
{
|
| 170486 |
+
"epoch": 446.21,
|
| 170487 |
+
"learning_rate": 9.120528846153847e-06,
|
| 170488 |
+
"loss": 1.0902,
|
| 170489 |
+
"step": 55900
|
| 170490 |
+
},
|
| 170491 |
+
{
|
| 170492 |
+
"epoch": 446.25,
|
| 170493 |
+
"learning_rate": 9.120448717948719e-06,
|
| 170494 |
+
"loss": 0.3394,
|
| 170495 |
+
"step": 55905
|
| 170496 |
+
},
|
| 170497 |
+
{
|
| 170498 |
+
"epoch": 446.29,
|
| 170499 |
+
"learning_rate": 9.12036858974359e-06,
|
| 170500 |
+
"loss": 0.3205,
|
| 170501 |
+
"step": 55910
|
| 170502 |
+
},
|
| 170503 |
+
{
|
| 170504 |
+
"epoch": 446.33,
|
| 170505 |
+
"learning_rate": 9.120288461538463e-06,
|
| 170506 |
+
"loss": 0.465,
|
| 170507 |
+
"step": 55915
|
| 170508 |
+
},
|
| 170509 |
+
{
|
| 170510 |
+
"epoch": 446.37,
|
| 170511 |
+
"learning_rate": 9.120208333333333e-06,
|
| 170512 |
+
"loss": 0.6148,
|
| 170513 |
+
"step": 55920
|
| 170514 |
+
},
|
| 170515 |
+
{
|
| 170516 |
+
"epoch": 446.41,
|
| 170517 |
+
"learning_rate": 9.120128205128206e-06,
|
| 170518 |
+
"loss": 1.3384,
|
| 170519 |
+
"step": 55925
|
| 170520 |
+
},
|
| 170521 |
+
{
|
| 170522 |
+
"epoch": 446.45,
|
| 170523 |
+
"learning_rate": 9.120048076923079e-06,
|
| 170524 |
+
"loss": 0.368,
|
| 170525 |
+
"step": 55930
|
| 170526 |
+
},
|
| 170527 |
+
{
|
| 170528 |
+
"epoch": 446.49,
|
| 170529 |
+
"learning_rate": 9.119967948717949e-06,
|
| 170530 |
+
"loss": 0.3318,
|
| 170531 |
+
"step": 55935
|
| 170532 |
+
},
|
| 170533 |
+
{
|
| 170534 |
+
"epoch": 446.53,
|
| 170535 |
+
"learning_rate": 9.119887820512822e-06,
|
| 170536 |
+
"loss": 0.3432,
|
| 170537 |
+
"step": 55940
|
| 170538 |
+
},
|
| 170539 |
+
{
|
| 170540 |
+
"epoch": 446.57,
|
| 170541 |
+
"learning_rate": 9.119807692307693e-06,
|
| 170542 |
+
"loss": 0.577,
|
| 170543 |
+
"step": 55945
|
| 170544 |
+
},
|
| 170545 |
+
{
|
| 170546 |
+
"epoch": 446.61,
|
| 170547 |
+
"learning_rate": 9.119727564102564e-06,
|
| 170548 |
+
"loss": 1.2361,
|
| 170549 |
+
"step": 55950
|
| 170550 |
+
},
|
| 170551 |
+
{
|
| 170552 |
+
"epoch": 446.65,
|
| 170553 |
+
"learning_rate": 9.119647435897436e-06,
|
| 170554 |
+
"loss": 0.3011,
|
| 170555 |
+
"step": 55955
|
| 170556 |
+
},
|
| 170557 |
+
{
|
| 170558 |
+
"epoch": 446.69,
|
| 170559 |
+
"learning_rate": 9.119567307692309e-06,
|
| 170560 |
+
"loss": 0.301,
|
| 170561 |
+
"step": 55960
|
| 170562 |
+
},
|
| 170563 |
+
{
|
| 170564 |
+
"epoch": 446.73,
|
| 170565 |
+
"learning_rate": 9.11948717948718e-06,
|
| 170566 |
+
"loss": 0.3025,
|
| 170567 |
+
"step": 55965
|
| 170568 |
+
},
|
| 170569 |
+
{
|
| 170570 |
+
"epoch": 446.77,
|
| 170571 |
+
"learning_rate": 9.119407051282052e-06,
|
| 170572 |
+
"loss": 0.57,
|
| 170573 |
+
"step": 55970
|
| 170574 |
+
},
|
| 170575 |
+
{
|
| 170576 |
+
"epoch": 446.81,
|
| 170577 |
+
"learning_rate": 9.119326923076925e-06,
|
| 170578 |
+
"loss": 1.174,
|
| 170579 |
+
"step": 55975
|
| 170580 |
+
},
|
| 170581 |
+
{
|
| 170582 |
+
"epoch": 446.85,
|
| 170583 |
+
"learning_rate": 9.119246794871796e-06,
|
| 170584 |
+
"loss": 0.3125,
|
| 170585 |
+
"step": 55980
|
| 170586 |
+
},
|
| 170587 |
+
{
|
| 170588 |
+
"epoch": 446.89,
|
| 170589 |
+
"learning_rate": 9.119166666666667e-06,
|
| 170590 |
+
"loss": 0.3804,
|
| 170591 |
+
"step": 55985
|
| 170592 |
+
},
|
| 170593 |
+
{
|
| 170594 |
+
"epoch": 446.93,
|
| 170595 |
+
"learning_rate": 9.119086538461539e-06,
|
| 170596 |
+
"loss": 0.4235,
|
| 170597 |
+
"step": 55990
|
| 170598 |
+
},
|
| 170599 |
+
{
|
| 170600 |
+
"epoch": 446.97,
|
| 170601 |
+
"learning_rate": 9.119006410256412e-06,
|
| 170602 |
+
"loss": 0.5852,
|
| 170603 |
+
"step": 55995
|
| 170604 |
+
},
|
| 170605 |
+
{
|
| 170606 |
+
"epoch": 447.0,
|
| 170607 |
+
"eval_loss": 0.36926010251045227,
|
| 170608 |
+
"eval_runtime": 40.6525,
|
| 170609 |
+
"eval_samples_per_second": 20.638,
|
| 170610 |
+
"eval_steps_per_second": 0.664,
|
| 170611 |
+
"eval_wer": 0.20080704328686722,
|
| 170612 |
+
"step": 55999
|
| 170613 |
+
},
|
| 170614 |
+
{
|
| 170615 |
+
"epoch": 447.01,
|
| 170616 |
+
"learning_rate": 9.118926282051283e-06,
|
| 170617 |
+
"loss": 0.5495,
|
| 170618 |
+
"step": 56000
|
| 170619 |
+
},
|
| 170620 |
+
{
|
| 170621 |
+
"epoch": 447.05,
|
| 170622 |
+
"learning_rate": 9.118846153846154e-06,
|
| 170623 |
+
"loss": 0.3405,
|
| 170624 |
+
"step": 56005
|
| 170625 |
+
},
|
| 170626 |
+
{
|
| 170627 |
+
"epoch": 447.09,
|
| 170628 |
+
"learning_rate": 9.118766025641026e-06,
|
| 170629 |
+
"loss": 0.3531,
|
| 170630 |
+
"step": 56010
|
| 170631 |
+
},
|
| 170632 |
+
{
|
| 170633 |
+
"epoch": 447.13,
|
| 170634 |
+
"learning_rate": 9.118685897435899e-06,
|
| 170635 |
+
"loss": 0.3044,
|
| 170636 |
+
"step": 56015
|
| 170637 |
+
},
|
| 170638 |
+
{
|
| 170639 |
+
"epoch": 447.17,
|
| 170640 |
+
"learning_rate": 9.11860576923077e-06,
|
| 170641 |
+
"loss": 0.6015,
|
| 170642 |
+
"step": 56020
|
| 170643 |
+
},
|
| 170644 |
+
{
|
| 170645 |
+
"epoch": 447.21,
|
| 170646 |
+
"learning_rate": 9.118525641025642e-06,
|
| 170647 |
+
"loss": 1.2181,
|
| 170648 |
+
"step": 56025
|
| 170649 |
+
},
|
| 170650 |
+
{
|
| 170651 |
+
"epoch": 447.25,
|
| 170652 |
+
"learning_rate": 9.118445512820515e-06,
|
| 170653 |
+
"loss": 0.3428,
|
| 170654 |
+
"step": 56030
|
| 170655 |
+
},
|
| 170656 |
+
{
|
| 170657 |
+
"epoch": 447.29,
|
| 170658 |
+
"learning_rate": 9.118365384615386e-06,
|
| 170659 |
+
"loss": 0.3112,
|
| 170660 |
+
"step": 56035
|
| 170661 |
+
},
|
| 170662 |
+
{
|
| 170663 |
+
"epoch": 447.33,
|
| 170664 |
+
"learning_rate": 9.118285256410257e-06,
|
| 170665 |
+
"loss": 0.3775,
|
| 170666 |
+
"step": 56040
|
| 170667 |
+
},
|
| 170668 |
+
{
|
| 170669 |
+
"epoch": 447.37,
|
| 170670 |
+
"learning_rate": 9.118205128205129e-06,
|
| 170671 |
+
"loss": 0.6077,
|
| 170672 |
+
"step": 56045
|
| 170673 |
+
},
|
| 170674 |
+
{
|
| 170675 |
+
"epoch": 447.41,
|
| 170676 |
+
"learning_rate": 9.118125000000002e-06,
|
| 170677 |
+
"loss": 1.1451,
|
| 170678 |
+
"step": 56050
|
| 170679 |
+
},
|
| 170680 |
+
{
|
| 170681 |
+
"epoch": 447.45,
|
| 170682 |
+
"learning_rate": 9.118044871794871e-06,
|
| 170683 |
+
"loss": 0.3721,
|
| 170684 |
+
"step": 56055
|
| 170685 |
+
},
|
| 170686 |
+
{
|
| 170687 |
+
"epoch": 447.49,
|
| 170688 |
+
"learning_rate": 9.117964743589744e-06,
|
| 170689 |
+
"loss": 0.3278,
|
| 170690 |
+
"step": 56060
|
| 170691 |
+
},
|
| 170692 |
+
{
|
| 170693 |
+
"epoch": 447.53,
|
| 170694 |
+
"learning_rate": 9.117884615384616e-06,
|
| 170695 |
+
"loss": 0.447,
|
| 170696 |
+
"step": 56065
|
| 170697 |
+
},
|
| 170698 |
+
{
|
| 170699 |
+
"epoch": 447.57,
|
| 170700 |
+
"learning_rate": 9.117804487179487e-06,
|
| 170701 |
+
"loss": 0.5954,
|
| 170702 |
+
"step": 56070
|
| 170703 |
+
},
|
| 170704 |
+
{
|
| 170705 |
+
"epoch": 447.61,
|
| 170706 |
+
"learning_rate": 9.11772435897436e-06,
|
| 170707 |
+
"loss": 1.3239,
|
| 170708 |
+
"step": 56075
|
| 170709 |
+
},
|
| 170710 |
+
{
|
| 170711 |
+
"epoch": 447.65,
|
| 170712 |
+
"learning_rate": 9.117644230769232e-06,
|
| 170713 |
+
"loss": 0.3252,
|
| 170714 |
+
"step": 56080
|
| 170715 |
+
},
|
| 170716 |
+
{
|
| 170717 |
+
"epoch": 447.69,
|
| 170718 |
+
"learning_rate": 9.117564102564103e-06,
|
| 170719 |
+
"loss": 0.4033,
|
| 170720 |
+
"step": 56085
|
| 170721 |
+
},
|
| 170722 |
+
{
|
| 170723 |
+
"epoch": 447.73,
|
| 170724 |
+
"learning_rate": 9.117483974358974e-06,
|
| 170725 |
+
"loss": 0.4424,
|
| 170726 |
+
"step": 56090
|
| 170727 |
+
},
|
| 170728 |
+
{
|
| 170729 |
+
"epoch": 447.77,
|
| 170730 |
+
"learning_rate": 9.117403846153847e-06,
|
| 170731 |
+
"loss": 0.6376,
|
| 170732 |
+
"step": 56095
|
| 170733 |
+
},
|
| 170734 |
+
{
|
| 170735 |
+
"epoch": 447.81,
|
| 170736 |
+
"learning_rate": 9.117323717948719e-06,
|
| 170737 |
+
"loss": 1.1042,
|
| 170738 |
+
"step": 56100
|
| 170739 |
+
},
|
| 170740 |
+
{
|
| 170741 |
+
"epoch": 447.85,
|
| 170742 |
+
"learning_rate": 9.11724358974359e-06,
|
| 170743 |
+
"loss": 0.3842,
|
| 170744 |
+
"step": 56105
|
| 170745 |
+
},
|
| 170746 |
+
{
|
| 170747 |
+
"epoch": 447.89,
|
| 170748 |
+
"learning_rate": 9.117163461538461e-06,
|
| 170749 |
+
"loss": 0.3331,
|
| 170750 |
+
"step": 56110
|
| 170751 |
+
},
|
| 170752 |
+
{
|
| 170753 |
+
"epoch": 447.93,
|
| 170754 |
+
"learning_rate": 9.117083333333334e-06,
|
| 170755 |
+
"loss": 0.3574,
|
| 170756 |
+
"step": 56115
|
| 170757 |
+
},
|
| 170758 |
+
{
|
| 170759 |
+
"epoch": 447.97,
|
| 170760 |
+
"learning_rate": 9.117003205128206e-06,
|
| 170761 |
+
"loss": 0.6026,
|
| 170762 |
+
"step": 56120
|
| 170763 |
+
},
|
| 170764 |
+
{
|
| 170765 |
+
"epoch": 448.0,
|
| 170766 |
+
"eval_loss": 0.4421532154083252,
|
| 170767 |
+
"eval_runtime": 39.0754,
|
| 170768 |
+
"eval_samples_per_second": 21.497,
|
| 170769 |
+
"eval_steps_per_second": 0.691,
|
| 170770 |
+
"eval_wer": 0.1973405421202601,
|
| 170771 |
+
"step": 56124
|
| 170772 |
+
},
|
| 170773 |
+
{
|
| 170774 |
+
"epoch": 452.01,
|
| 170775 |
+
"learning_rate": 9.116923076923077e-06,
|
| 170776 |
+
"loss": 0.3693,
|
| 170777 |
+
"step": 56125
|
| 170778 |
+
},
|
| 170779 |
+
{
|
| 170780 |
+
"epoch": 452.05,
|
| 170781 |
+
"learning_rate": 9.11684294871795e-06,
|
| 170782 |
+
"loss": 0.41,
|
| 170783 |
+
"step": 56130
|
| 170784 |
+
},
|
| 170785 |
+
{
|
| 170786 |
+
"epoch": 452.09,
|
| 170787 |
+
"learning_rate": 9.116762820512822e-06,
|
| 170788 |
+
"loss": 0.3313,
|
| 170789 |
+
"step": 56135
|
| 170790 |
+
},
|
| 170791 |
+
{
|
| 170792 |
+
"epoch": 452.13,
|
| 170793 |
+
"learning_rate": 9.116682692307693e-06,
|
| 170794 |
+
"loss": 0.3218,
|
| 170795 |
+
"step": 56140
|
| 170796 |
+
},
|
| 170797 |
+
{
|
| 170798 |
+
"epoch": 452.17,
|
| 170799 |
+
"learning_rate": 9.116602564102564e-06,
|
| 170800 |
+
"loss": 0.5579,
|
| 170801 |
+
"step": 56145
|
| 170802 |
+
},
|
| 170803 |
+
{
|
| 170804 |
+
"epoch": 452.21,
|
| 170805 |
+
"learning_rate": 9.116522435897437e-06,
|
| 170806 |
+
"loss": 1.2605,
|
| 170807 |
+
"step": 56150
|
| 170808 |
+
},
|
| 170809 |
+
{
|
| 170810 |
+
"epoch": 452.25,
|
| 170811 |
+
"learning_rate": 9.116442307692309e-06,
|
| 170812 |
+
"loss": 0.4012,
|
| 170813 |
+
"step": 56155
|
| 170814 |
+
},
|
| 170815 |
+
{
|
| 170816 |
+
"epoch": 452.29,
|
| 170817 |
+
"learning_rate": 9.11636217948718e-06,
|
| 170818 |
+
"loss": 0.3052,
|
| 170819 |
+
"step": 56160
|
| 170820 |
+
},
|
| 170821 |
+
{
|
| 170822 |
+
"epoch": 452.33,
|
| 170823 |
+
"learning_rate": 9.116282051282051e-06,
|
| 170824 |
+
"loss": 0.3319,
|
| 170825 |
+
"step": 56165
|
| 170826 |
+
},
|
| 170827 |
+
{
|
| 170828 |
+
"epoch": 452.37,
|
| 170829 |
+
"learning_rate": 9.116201923076925e-06,
|
| 170830 |
+
"loss": 0.5104,
|
| 170831 |
+
"step": 56170
|
| 170832 |
+
},
|
| 170833 |
+
{
|
| 170834 |
+
"epoch": 452.41,
|
| 170835 |
+
"learning_rate": 9.116121794871796e-06,
|
| 170836 |
+
"loss": 1.1352,
|
| 170837 |
+
"step": 56175
|
| 170838 |
+
},
|
| 170839 |
+
{
|
| 170840 |
+
"epoch": 452.45,
|
| 170841 |
+
"learning_rate": 9.116041666666667e-06,
|
| 170842 |
+
"loss": 0.3218,
|
| 170843 |
+
"step": 56180
|
| 170844 |
+
},
|
| 170845 |
+
{
|
| 170846 |
+
"epoch": 452.49,
|
| 170847 |
+
"learning_rate": 9.11596153846154e-06,
|
| 170848 |
+
"loss": 0.2944,
|
| 170849 |
+
"step": 56185
|
| 170850 |
+
},
|
| 170851 |
+
{
|
| 170852 |
+
"epoch": 452.53,
|
| 170853 |
+
"learning_rate": 9.115881410256412e-06,
|
| 170854 |
+
"loss": 0.3214,
|
| 170855 |
+
"step": 56190
|
| 170856 |
+
},
|
| 170857 |
+
{
|
| 170858 |
+
"epoch": 452.57,
|
| 170859 |
+
"learning_rate": 9.115801282051283e-06,
|
| 170860 |
+
"loss": 0.5675,
|
| 170861 |
+
"step": 56195
|
| 170862 |
+
},
|
| 170863 |
+
{
|
| 170864 |
+
"epoch": 452.61,
|
| 170865 |
+
"learning_rate": 9.115721153846154e-06,
|
| 170866 |
+
"loss": 1.0955,
|
| 170867 |
+
"step": 56200
|
| 170868 |
+
},
|
| 170869 |
+
{
|
| 170870 |
+
"epoch": 452.65,
|
| 170871 |
+
"learning_rate": 9.115641025641027e-06,
|
| 170872 |
+
"loss": 0.3456,
|
| 170873 |
+
"step": 56205
|
| 170874 |
+
},
|
| 170875 |
+
{
|
| 170876 |
+
"epoch": 452.69,
|
| 170877 |
+
"learning_rate": 9.115560897435897e-06,
|
| 170878 |
+
"loss": 0.3618,
|
| 170879 |
+
"step": 56210
|
| 170880 |
+
},
|
| 170881 |
+
{
|
| 170882 |
+
"epoch": 452.73,
|
| 170883 |
+
"learning_rate": 9.11548076923077e-06,
|
| 170884 |
+
"loss": 0.3901,
|
| 170885 |
+
"step": 56215
|
| 170886 |
+
},
|
| 170887 |
+
{
|
| 170888 |
+
"epoch": 452.77,
|
| 170889 |
+
"learning_rate": 9.115400641025641e-06,
|
| 170890 |
+
"loss": 0.531,
|
| 170891 |
+
"step": 56220
|
| 170892 |
+
},
|
| 170893 |
+
{
|
| 170894 |
+
"epoch": 452.81,
|
| 170895 |
+
"learning_rate": 9.115320512820513e-06,
|
| 170896 |
+
"loss": 1.2777,
|
| 170897 |
+
"step": 56225
|
| 170898 |
+
},
|
| 170899 |
+
{
|
| 170900 |
+
"epoch": 452.85,
|
| 170901 |
+
"learning_rate": 9.115240384615386e-06,
|
| 170902 |
+
"loss": 0.3329,
|
| 170903 |
+
"step": 56230
|
| 170904 |
+
},
|
| 170905 |
+
{
|
| 170906 |
+
"epoch": 452.89,
|
| 170907 |
+
"learning_rate": 9.115160256410257e-06,
|
| 170908 |
+
"loss": 0.3679,
|
| 170909 |
+
"step": 56235
|
| 170910 |
+
},
|
| 170911 |
+
{
|
| 170912 |
+
"epoch": 452.93,
|
| 170913 |
+
"learning_rate": 9.115080128205129e-06,
|
| 170914 |
+
"loss": 0.4268,
|
| 170915 |
+
"step": 56240
|
| 170916 |
+
},
|
| 170917 |
+
{
|
| 170918 |
+
"epoch": 452.97,
|
| 170919 |
+
"learning_rate": 9.115e-06,
|
| 170920 |
+
"loss": 0.5382,
|
| 170921 |
+
"step": 56245
|
| 170922 |
+
},
|
| 170923 |
+
{
|
| 170924 |
+
"epoch": 453.0,
|
| 170925 |
+
"eval_loss": 0.39150935411453247,
|
| 170926 |
+
"eval_runtime": 39.6345,
|
| 170927 |
+
"eval_samples_per_second": 21.194,
|
| 170928 |
+
"eval_steps_per_second": 0.681,
|
| 170929 |
+
"eval_wer": 0.1960477255779269,
|
| 170930 |
+
"step": 56248
|
| 170931 |
+
},
|
| 170932 |
+
{
|
| 170933 |
+
"epoch": 453.02,
|
| 170934 |
+
"learning_rate": 9.114919871794873e-06,
|
| 170935 |
+
"loss": 0.411,
|
| 170936 |
+
"step": 56250
|
| 170937 |
+
},
|
| 170938 |
+
{
|
| 170939 |
+
"epoch": 453.06,
|
| 170940 |
+
"learning_rate": 9.114839743589744e-06,
|
| 170941 |
+
"loss": 0.3452,
|
| 170942 |
+
"step": 56255
|
| 170943 |
+
},
|
| 170944 |
+
{
|
| 170945 |
+
"epoch": 453.1,
|
| 170946 |
+
"learning_rate": 9.114759615384616e-06,
|
| 170947 |
+
"loss": 0.4267,
|
| 170948 |
+
"step": 56260
|
| 170949 |
+
},
|
| 170950 |
+
{
|
| 170951 |
+
"epoch": 453.14,
|
| 170952 |
+
"learning_rate": 9.114679487179487e-06,
|
| 170953 |
+
"loss": 0.3528,
|
| 170954 |
+
"step": 56265
|
| 170955 |
+
},
|
| 170956 |
+
{
|
| 170957 |
+
"epoch": 453.18,
|
| 170958 |
+
"learning_rate": 9.11459935897436e-06,
|
| 170959 |
+
"loss": 0.6615,
|
| 170960 |
+
"step": 56270
|
| 170961 |
+
},
|
| 170962 |
+
{
|
| 170963 |
+
"epoch": 453.22,
|
| 170964 |
+
"learning_rate": 9.114519230769232e-06,
|
| 170965 |
+
"loss": 1.1156,
|
| 170966 |
+
"step": 56275
|
| 170967 |
+
},
|
| 170968 |
+
{
|
| 170969 |
+
"epoch": 453.26,
|
| 170970 |
+
"learning_rate": 9.114439102564103e-06,
|
| 170971 |
+
"loss": 0.2969,
|
| 170972 |
+
"step": 56280
|
| 170973 |
+
},
|
| 170974 |
+
{
|
| 170975 |
+
"epoch": 453.3,
|
| 170976 |
+
"learning_rate": 9.114358974358976e-06,
|
| 170977 |
+
"loss": 0.2923,
|
| 170978 |
+
"step": 56285
|
| 170979 |
+
},
|
| 170980 |
+
{
|
| 170981 |
+
"epoch": 453.34,
|
| 170982 |
+
"learning_rate": 9.114278846153847e-06,
|
| 170983 |
+
"loss": 0.4115,
|
| 170984 |
+
"step": 56290
|
| 170985 |
+
},
|
| 170986 |
+
{
|
| 170987 |
+
"epoch": 453.38,
|
| 170988 |
+
"learning_rate": 9.114198717948719e-06,
|
| 170989 |
+
"loss": 0.7736,
|
| 170990 |
+
"step": 56295
|
| 170991 |
+
},
|
| 170992 |
+
{
|
| 170993 |
+
"epoch": 453.42,
|
| 170994 |
+
"learning_rate": 9.11411858974359e-06,
|
| 170995 |
+
"loss": 1.0942,
|
| 170996 |
+
"step": 56300
|
| 170997 |
+
},
|
| 170998 |
+
{
|
| 170999 |
+
"epoch": 453.46,
|
| 171000 |
+
"learning_rate": 9.114038461538463e-06,
|
| 171001 |
+
"loss": 0.4294,
|
| 171002 |
+
"step": 56305
|
| 171003 |
+
},
|
| 171004 |
+
{
|
| 171005 |
+
"epoch": 453.5,
|
| 171006 |
+
"learning_rate": 9.113958333333334e-06,
|
| 171007 |
+
"loss": 0.3271,
|
| 171008 |
+
"step": 56310
|
| 171009 |
+
},
|
| 171010 |
+
{
|
| 171011 |
+
"epoch": 453.54,
|
| 171012 |
+
"learning_rate": 9.113878205128206e-06,
|
| 171013 |
+
"loss": 0.3859,
|
| 171014 |
+
"step": 56315
|
| 171015 |
+
},
|
| 171016 |
+
{
|
| 171017 |
+
"epoch": 453.58,
|
| 171018 |
+
"learning_rate": 9.113798076923077e-06,
|
| 171019 |
+
"loss": 0.9001,
|
| 171020 |
+
"step": 56320
|
| 171021 |
+
},
|
| 171022 |
+
{
|
| 171023 |
+
"epoch": 453.62,
|
| 171024 |
+
"learning_rate": 9.11371794871795e-06,
|
| 171025 |
+
"loss": 1.0293,
|
| 171026 |
+
"step": 56325
|
| 171027 |
+
},
|
| 171028 |
+
{
|
| 171029 |
+
"epoch": 453.66,
|
| 171030 |
+
"learning_rate": 9.113637820512822e-06,
|
| 171031 |
+
"loss": 0.3261,
|
| 171032 |
+
"step": 56330
|
| 171033 |
+
},
|
| 171034 |
+
{
|
| 171035 |
+
"epoch": 453.7,
|
| 171036 |
+
"learning_rate": 9.113557692307693e-06,
|
| 171037 |
+
"loss": 0.4001,
|
| 171038 |
+
"step": 56335
|
| 171039 |
+
},
|
| 171040 |
+
{
|
| 171041 |
+
"epoch": 453.74,
|
| 171042 |
+
"learning_rate": 9.113477564102566e-06,
|
| 171043 |
+
"loss": 0.4428,
|
| 171044 |
+
"step": 56340
|
| 171045 |
+
},
|
| 171046 |
+
{
|
| 171047 |
+
"epoch": 453.78,
|
| 171048 |
+
"learning_rate": 9.113397435897436e-06,
|
| 171049 |
+
"loss": 0.6543,
|
| 171050 |
+
"step": 56345
|
| 171051 |
+
},
|
| 171052 |
+
{
|
| 171053 |
+
"epoch": 453.82,
|
| 171054 |
+
"learning_rate": 9.113317307692309e-06,
|
| 171055 |
+
"loss": 1.0776,
|
| 171056 |
+
"step": 56350
|
| 171057 |
+
},
|
| 171058 |
+
{
|
| 171059 |
+
"epoch": 453.86,
|
| 171060 |
+
"learning_rate": 9.11323717948718e-06,
|
| 171061 |
+
"loss": 0.3333,
|
| 171062 |
+
"step": 56355
|
| 171063 |
+
},
|
| 171064 |
+
{
|
| 171065 |
+
"epoch": 453.9,
|
| 171066 |
+
"learning_rate": 9.113157051282051e-06,
|
| 171067 |
+
"loss": 0.3174,
|
| 171068 |
+
"step": 56360
|
| 171069 |
+
},
|
| 171070 |
+
{
|
| 171071 |
+
"epoch": 453.94,
|
| 171072 |
+
"learning_rate": 9.113076923076923e-06,
|
| 171073 |
+
"loss": 0.3947,
|
| 171074 |
+
"step": 56365
|
| 171075 |
+
},
|
| 171076 |
+
{
|
| 171077 |
+
"epoch": 453.98,
|
| 171078 |
+
"learning_rate": 9.112996794871796e-06,
|
| 171079 |
+
"loss": 0.7021,
|
| 171080 |
+
"step": 56370
|
| 171081 |
+
},
|
| 171082 |
+
{
|
| 171083 |
+
"epoch": 454.0,
|
| 171084 |
+
"eval_loss": 0.4025454521179199,
|
| 171085 |
+
"eval_runtime": 40.2845,
|
| 171086 |
+
"eval_samples_per_second": 20.852,
|
| 171087 |
+
"eval_steps_per_second": 0.67,
|
| 171088 |
+
"eval_wer": 0.20751998847511344,
|
| 171089 |
+
"step": 56372
|
| 171090 |
}
|
| 171091 |
],
|
| 171092 |
"max_steps": 620000,
|
| 171093 |
"num_train_epochs": 5000,
|
| 171094 |
+
"total_flos": 1.5864105616332263e+20,
|
| 171095 |
"trial_name": null,
|
| 171096 |
"trial_params": null
|
| 171097 |
}
|
model-bin/finetune/base/{checkpoint-55750 → checkpoint-56372}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629775137.1840463/events.out.tfevents.1629775137.c435e1c5ee04.920.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a7574162fb7c30e1b3c389e52c7525a90f5699679ee1d3f4f4df94c0bb3f2a5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629775789.2149596/events.out.tfevents.1629775789.c435e1c5ee04.920.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1af167a43b471c0f369df5626336392d931cadfeee95c276355fb9a951f3e04
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629776456.5037358/events.out.tfevents.1629776456.c435e1c5ee04.920.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13a62020d4ac48825c4a349110ba3a40cda2ad94d67530a99db0f7668c15538d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629777113.4720376/events.out.tfevents.1629777113.c435e1c5ee04.920.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b79c93f5ed8859b9cbafbd76dd145be1429a1e8eb1514c0dd789d3c765ff05eb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629777774.3360224/events.out.tfevents.1629777774.c435e1c5ee04.920.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68ba56ea0f7687aeb172864c9a8961a40224b2f11396ddc8d7f65f0a7f199605
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629775137.c435e1c5ee04.920.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2f3ed99958c45f39b3552933ad050036896623429b16f65793029056a6bcb4ff
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629775789.c435e1c5ee04.920.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f1bf96bea57641a96755716e3a2dc8b16528873b08f0b9f4fe3556043771244
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629776456.c435e1c5ee04.920.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cee341bde82e285d4160b79f98ec74de6b1c5418a9f678ac4d8614dcbb4fad2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629777113.c435e1c5ee04.920.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d6512008c981cb1efc6836f57510e5fc565b8b6ac88bcbac3e1744513b977886
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629777774.c435e1c5ee04.920.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e122e54ef57f5f01fd46e35147c917c9f63d63ddac918281925c0550f47177cf
|
| 3 |
+
size 8622
|