"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629966687.6381752/events.out.tfevents.1629966687.8e89bd551565.924.121 +3 -0
- model-bin/finetune/base/log/1629967137.0971546/events.out.tfevents.1629967137.8e89bd551565.924.123 +3 -0
- model-bin/finetune/base/log/1629967569.9641695/events.out.tfevents.1629967571.8e89bd551565.924.125 +3 -0
- model-bin/finetune/base/log/1629968006.3075402/events.out.tfevents.1629968006.8e89bd551565.924.127 +3 -0
- model-bin/finetune/base/log/1629968435.5634236/events.out.tfevents.1629968435.8e89bd551565.924.129 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629966687.8e89bd551565.924.120 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629967137.8e89bd551565.924.122 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629967569.8e89bd551565.924.124 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629968006.8e89bd551565.924.126 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629968435.8e89bd551565.924.128 +3 -0
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/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:acde8f80998d241c3fd9d081c6f91b6e98270c037f38af9034e7ff5df3580a2c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/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 sha256:4cab998fc5062a9c6beed2d2792d74072443e26e134f86834a0ac7d3a1b6de36
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/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:fbe536b03ba18cda40806ec7653fdcdef2be789c5c7db90a9e5d23b6cfabeba4
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/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 sha256:28918aa80005fea706d5161a6170bf2c6c0107d5f472fafa7d72c070b9e34f4e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/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:bfe8763e085a67b9c2b154c35e5469a64be0e1422490055f194b53dc60704d80
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch": 771.
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -221280,11 +221280,806 @@
|
|
| 221280 |
"eval_steps_per_second": 0.698,
|
| 221281 |
"eval_wer": 0.1937148929710035,
|
| 221282 |
"step": 95824
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 221283 |
}
|
| 221284 |
],
|
| 221285 |
-
"max_steps":
|
| 221286 |
"num_train_epochs": 5000,
|
| 221287 |
-
"total_flos": 2.
|
| 221288 |
"trial_name": null,
|
| 221289 |
"trial_params": null
|
| 221290 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 771.0,
|
| 5 |
+
"global_step": 96447,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 221280 |
"eval_steps_per_second": 0.698,
|
| 221281 |
"eval_wer": 0.1937148929710035,
|
| 221282 |
"step": 95824
|
| 221283 |
+
},
|
| 221284 |
+
{
|
| 221285 |
+
"epoch": 766.01,
|
| 221286 |
+
"learning_rate": 8.480977564102565e-06,
|
| 221287 |
+
"loss": 0.3802,
|
| 221288 |
+
"step": 95825
|
| 221289 |
+
},
|
| 221290 |
+
{
|
| 221291 |
+
"epoch": 766.05,
|
| 221292 |
+
"learning_rate": 8.480897435897436e-06,
|
| 221293 |
+
"loss": 0.3188,
|
| 221294 |
+
"step": 95830
|
| 221295 |
+
},
|
| 221296 |
+
{
|
| 221297 |
+
"epoch": 766.09,
|
| 221298 |
+
"learning_rate": 8.48081730769231e-06,
|
| 221299 |
+
"loss": 0.2795,
|
| 221300 |
+
"step": 95835
|
| 221301 |
+
},
|
| 221302 |
+
{
|
| 221303 |
+
"epoch": 766.13,
|
| 221304 |
+
"learning_rate": 8.48073717948718e-06,
|
| 221305 |
+
"loss": 0.3266,
|
| 221306 |
+
"step": 95840
|
| 221307 |
+
},
|
| 221308 |
+
{
|
| 221309 |
+
"epoch": 766.17,
|
| 221310 |
+
"learning_rate": 8.480657051282052e-06,
|
| 221311 |
+
"loss": 0.5135,
|
| 221312 |
+
"step": 95845
|
| 221313 |
+
},
|
| 221314 |
+
{
|
| 221315 |
+
"epoch": 766.21,
|
| 221316 |
+
"learning_rate": 8.480576923076925e-06,
|
| 221317 |
+
"loss": 1.23,
|
| 221318 |
+
"step": 95850
|
| 221319 |
+
},
|
| 221320 |
+
{
|
| 221321 |
+
"epoch": 766.25,
|
| 221322 |
+
"learning_rate": 8.480496794871795e-06,
|
| 221323 |
+
"loss": 0.2848,
|
| 221324 |
+
"step": 95855
|
| 221325 |
+
},
|
| 221326 |
+
{
|
| 221327 |
+
"epoch": 766.29,
|
| 221328 |
+
"learning_rate": 8.480416666666668e-06,
|
| 221329 |
+
"loss": 0.2998,
|
| 221330 |
+
"step": 95860
|
| 221331 |
+
},
|
| 221332 |
+
{
|
| 221333 |
+
"epoch": 766.33,
|
| 221334 |
+
"learning_rate": 8.480336538461539e-06,
|
| 221335 |
+
"loss": 0.3392,
|
| 221336 |
+
"step": 95865
|
| 221337 |
+
},
|
| 221338 |
+
{
|
| 221339 |
+
"epoch": 766.37,
|
| 221340 |
+
"learning_rate": 8.48025641025641e-06,
|
| 221341 |
+
"loss": 0.6364,
|
| 221342 |
+
"step": 95870
|
| 221343 |
+
},
|
| 221344 |
+
{
|
| 221345 |
+
"epoch": 766.41,
|
| 221346 |
+
"learning_rate": 8.480176282051282e-06,
|
| 221347 |
+
"loss": 1.1482,
|
| 221348 |
+
"step": 95875
|
| 221349 |
+
},
|
| 221350 |
+
{
|
| 221351 |
+
"epoch": 766.45,
|
| 221352 |
+
"learning_rate": 8.480096153846155e-06,
|
| 221353 |
+
"loss": 0.3259,
|
| 221354 |
+
"step": 95880
|
| 221355 |
+
},
|
| 221356 |
+
{
|
| 221357 |
+
"epoch": 766.49,
|
| 221358 |
+
"learning_rate": 8.480016025641026e-06,
|
| 221359 |
+
"loss": 0.2939,
|
| 221360 |
+
"step": 95885
|
| 221361 |
+
},
|
| 221362 |
+
{
|
| 221363 |
+
"epoch": 766.53,
|
| 221364 |
+
"learning_rate": 8.479935897435898e-06,
|
| 221365 |
+
"loss": 0.3331,
|
| 221366 |
+
"step": 95890
|
| 221367 |
+
},
|
| 221368 |
+
{
|
| 221369 |
+
"epoch": 766.57,
|
| 221370 |
+
"learning_rate": 8.47985576923077e-06,
|
| 221371 |
+
"loss": 0.5985,
|
| 221372 |
+
"step": 95895
|
| 221373 |
+
},
|
| 221374 |
+
{
|
| 221375 |
+
"epoch": 766.61,
|
| 221376 |
+
"learning_rate": 8.479775641025642e-06,
|
| 221377 |
+
"loss": 1.2024,
|
| 221378 |
+
"step": 95900
|
| 221379 |
+
},
|
| 221380 |
+
{
|
| 221381 |
+
"epoch": 766.65,
|
| 221382 |
+
"learning_rate": 8.479695512820513e-06,
|
| 221383 |
+
"loss": 0.3309,
|
| 221384 |
+
"step": 95905
|
| 221385 |
+
},
|
| 221386 |
+
{
|
| 221387 |
+
"epoch": 766.69,
|
| 221388 |
+
"learning_rate": 8.479615384615385e-06,
|
| 221389 |
+
"loss": 0.2882,
|
| 221390 |
+
"step": 95910
|
| 221391 |
+
},
|
| 221392 |
+
{
|
| 221393 |
+
"epoch": 766.73,
|
| 221394 |
+
"learning_rate": 8.479535256410258e-06,
|
| 221395 |
+
"loss": 0.394,
|
| 221396 |
+
"step": 95915
|
| 221397 |
+
},
|
| 221398 |
+
{
|
| 221399 |
+
"epoch": 766.77,
|
| 221400 |
+
"learning_rate": 8.479455128205129e-06,
|
| 221401 |
+
"loss": 0.595,
|
| 221402 |
+
"step": 95920
|
| 221403 |
+
},
|
| 221404 |
+
{
|
| 221405 |
+
"epoch": 766.81,
|
| 221406 |
+
"learning_rate": 8.479375e-06,
|
| 221407 |
+
"loss": 1.1719,
|
| 221408 |
+
"step": 95925
|
| 221409 |
+
},
|
| 221410 |
+
{
|
| 221411 |
+
"epoch": 766.85,
|
| 221412 |
+
"learning_rate": 8.479294871794872e-06,
|
| 221413 |
+
"loss": 0.308,
|
| 221414 |
+
"step": 95930
|
| 221415 |
+
},
|
| 221416 |
+
{
|
| 221417 |
+
"epoch": 766.89,
|
| 221418 |
+
"learning_rate": 8.479214743589745e-06,
|
| 221419 |
+
"loss": 0.2688,
|
| 221420 |
+
"step": 95935
|
| 221421 |
+
},
|
| 221422 |
+
{
|
| 221423 |
+
"epoch": 766.93,
|
| 221424 |
+
"learning_rate": 8.479134615384616e-06,
|
| 221425 |
+
"loss": 0.3931,
|
| 221426 |
+
"step": 95940
|
| 221427 |
+
},
|
| 221428 |
+
{
|
| 221429 |
+
"epoch": 766.97,
|
| 221430 |
+
"learning_rate": 8.479054487179488e-06,
|
| 221431 |
+
"loss": 0.5052,
|
| 221432 |
+
"step": 95945
|
| 221433 |
+
},
|
| 221434 |
+
{
|
| 221435 |
+
"epoch": 767.0,
|
| 221436 |
+
"eval_loss": 0.35029274225234985,
|
| 221437 |
+
"eval_runtime": 39.3042,
|
| 221438 |
+
"eval_samples_per_second": 21.372,
|
| 221439 |
+
"eval_steps_per_second": 0.687,
|
| 221440 |
+
"eval_wer": 0.1812636165577342,
|
| 221441 |
+
"step": 95949
|
| 221442 |
+
},
|
| 221443 |
+
{
|
| 221444 |
+
"epoch": 773.01,
|
| 221445 |
+
"learning_rate": 8.47897435897436e-06,
|
| 221446 |
+
"loss": 0.5014,
|
| 221447 |
+
"step": 95950
|
| 221448 |
+
},
|
| 221449 |
+
{
|
| 221450 |
+
"epoch": 773.05,
|
| 221451 |
+
"learning_rate": 8.478894230769232e-06,
|
| 221452 |
+
"loss": 0.3029,
|
| 221453 |
+
"step": 95955
|
| 221454 |
+
},
|
| 221455 |
+
{
|
| 221456 |
+
"epoch": 773.09,
|
| 221457 |
+
"learning_rate": 8.478814102564103e-06,
|
| 221458 |
+
"loss": 0.3354,
|
| 221459 |
+
"step": 95960
|
| 221460 |
+
},
|
| 221461 |
+
{
|
| 221462 |
+
"epoch": 773.13,
|
| 221463 |
+
"learning_rate": 8.478733974358975e-06,
|
| 221464 |
+
"loss": 0.3095,
|
| 221465 |
+
"step": 95965
|
| 221466 |
+
},
|
| 221467 |
+
{
|
| 221468 |
+
"epoch": 773.17,
|
| 221469 |
+
"learning_rate": 8.478653846153848e-06,
|
| 221470 |
+
"loss": 0.55,
|
| 221471 |
+
"step": 95970
|
| 221472 |
+
},
|
| 221473 |
+
{
|
| 221474 |
+
"epoch": 773.21,
|
| 221475 |
+
"learning_rate": 8.478573717948717e-06,
|
| 221476 |
+
"loss": 1.0675,
|
| 221477 |
+
"step": 95975
|
| 221478 |
+
},
|
| 221479 |
+
{
|
| 221480 |
+
"epoch": 773.25,
|
| 221481 |
+
"learning_rate": 8.47849358974359e-06,
|
| 221482 |
+
"loss": 0.3399,
|
| 221483 |
+
"step": 95980
|
| 221484 |
+
},
|
| 221485 |
+
{
|
| 221486 |
+
"epoch": 773.29,
|
| 221487 |
+
"learning_rate": 8.478413461538464e-06,
|
| 221488 |
+
"loss": 0.3358,
|
| 221489 |
+
"step": 95985
|
| 221490 |
+
},
|
| 221491 |
+
{
|
| 221492 |
+
"epoch": 773.33,
|
| 221493 |
+
"learning_rate": 8.478333333333333e-06,
|
| 221494 |
+
"loss": 0.315,
|
| 221495 |
+
"step": 95990
|
| 221496 |
+
},
|
| 221497 |
+
{
|
| 221498 |
+
"epoch": 773.37,
|
| 221499 |
+
"learning_rate": 8.478253205128206e-06,
|
| 221500 |
+
"loss": 0.553,
|
| 221501 |
+
"step": 95995
|
| 221502 |
+
},
|
| 221503 |
+
{
|
| 221504 |
+
"epoch": 773.41,
|
| 221505 |
+
"learning_rate": 8.478173076923078e-06,
|
| 221506 |
+
"loss": 1.1838,
|
| 221507 |
+
"step": 96000
|
| 221508 |
+
},
|
| 221509 |
+
{
|
| 221510 |
+
"epoch": 773.45,
|
| 221511 |
+
"learning_rate": 8.478092948717949e-06,
|
| 221512 |
+
"loss": 0.2905,
|
| 221513 |
+
"step": 96005
|
| 221514 |
+
},
|
| 221515 |
+
{
|
| 221516 |
+
"epoch": 773.49,
|
| 221517 |
+
"learning_rate": 8.47801282051282e-06,
|
| 221518 |
+
"loss": 0.2957,
|
| 221519 |
+
"step": 96010
|
| 221520 |
+
},
|
| 221521 |
+
{
|
| 221522 |
+
"epoch": 773.53,
|
| 221523 |
+
"learning_rate": 8.477932692307693e-06,
|
| 221524 |
+
"loss": 0.3731,
|
| 221525 |
+
"step": 96015
|
| 221526 |
+
},
|
| 221527 |
+
{
|
| 221528 |
+
"epoch": 773.57,
|
| 221529 |
+
"learning_rate": 8.477852564102565e-06,
|
| 221530 |
+
"loss": 0.5858,
|
| 221531 |
+
"step": 96020
|
| 221532 |
+
},
|
| 221533 |
+
{
|
| 221534 |
+
"epoch": 773.61,
|
| 221535 |
+
"learning_rate": 8.477772435897436e-06,
|
| 221536 |
+
"loss": 1.1793,
|
| 221537 |
+
"step": 96025
|
| 221538 |
+
},
|
| 221539 |
+
{
|
| 221540 |
+
"epoch": 773.65,
|
| 221541 |
+
"learning_rate": 8.477692307692307e-06,
|
| 221542 |
+
"loss": 0.3217,
|
| 221543 |
+
"step": 96030
|
| 221544 |
+
},
|
| 221545 |
+
{
|
| 221546 |
+
"epoch": 773.69,
|
| 221547 |
+
"learning_rate": 8.47761217948718e-06,
|
| 221548 |
+
"loss": 0.3186,
|
| 221549 |
+
"step": 96035
|
| 221550 |
+
},
|
| 221551 |
+
{
|
| 221552 |
+
"epoch": 773.73,
|
| 221553 |
+
"learning_rate": 8.477532051282052e-06,
|
| 221554 |
+
"loss": 0.4021,
|
| 221555 |
+
"step": 96040
|
| 221556 |
+
},
|
| 221557 |
+
{
|
| 221558 |
+
"epoch": 773.77,
|
| 221559 |
+
"learning_rate": 8.477451923076923e-06,
|
| 221560 |
+
"loss": 0.508,
|
| 221561 |
+
"step": 96045
|
| 221562 |
+
},
|
| 221563 |
+
{
|
| 221564 |
+
"epoch": 773.81,
|
| 221565 |
+
"learning_rate": 8.477371794871796e-06,
|
| 221566 |
+
"loss": 1.1391,
|
| 221567 |
+
"step": 96050
|
| 221568 |
+
},
|
| 221569 |
+
{
|
| 221570 |
+
"epoch": 773.85,
|
| 221571 |
+
"learning_rate": 8.477291666666668e-06,
|
| 221572 |
+
"loss": 0.3956,
|
| 221573 |
+
"step": 96055
|
| 221574 |
+
},
|
| 221575 |
+
{
|
| 221576 |
+
"epoch": 773.89,
|
| 221577 |
+
"learning_rate": 8.477211538461539e-06,
|
| 221578 |
+
"loss": 0.267,
|
| 221579 |
+
"step": 96060
|
| 221580 |
+
},
|
| 221581 |
+
{
|
| 221582 |
+
"epoch": 773.93,
|
| 221583 |
+
"learning_rate": 8.47713141025641e-06,
|
| 221584 |
+
"loss": 0.3235,
|
| 221585 |
+
"step": 96065
|
| 221586 |
+
},
|
| 221587 |
+
{
|
| 221588 |
+
"epoch": 773.97,
|
| 221589 |
+
"learning_rate": 8.477051282051283e-06,
|
| 221590 |
+
"loss": 0.6326,
|
| 221591 |
+
"step": 96070
|
| 221592 |
+
},
|
| 221593 |
+
{
|
| 221594 |
+
"epoch": 774.0,
|
| 221595 |
+
"eval_loss": 0.37618476152420044,
|
| 221596 |
+
"eval_runtime": 38.1696,
|
| 221597 |
+
"eval_samples_per_second": 22.007,
|
| 221598 |
+
"eval_steps_per_second": 0.707,
|
| 221599 |
+
"eval_wer": 0.18894791205548178,
|
| 221600 |
+
"step": 96073
|
| 221601 |
+
},
|
| 221602 |
+
{
|
| 221603 |
+
"epoch": 774.02,
|
| 221604 |
+
"learning_rate": 8.476971153846155e-06,
|
| 221605 |
+
"loss": 0.3748,
|
| 221606 |
+
"step": 96075
|
| 221607 |
+
},
|
| 221608 |
+
{
|
| 221609 |
+
"epoch": 774.06,
|
| 221610 |
+
"learning_rate": 8.476891025641026e-06,
|
| 221611 |
+
"loss": 0.3832,
|
| 221612 |
+
"step": 96080
|
| 221613 |
+
},
|
| 221614 |
+
{
|
| 221615 |
+
"epoch": 774.1,
|
| 221616 |
+
"learning_rate": 8.476810897435897e-06,
|
| 221617 |
+
"loss": 0.3015,
|
| 221618 |
+
"step": 96085
|
| 221619 |
+
},
|
| 221620 |
+
{
|
| 221621 |
+
"epoch": 774.14,
|
| 221622 |
+
"learning_rate": 8.47673076923077e-06,
|
| 221623 |
+
"loss": 0.3606,
|
| 221624 |
+
"step": 96090
|
| 221625 |
+
},
|
| 221626 |
+
{
|
| 221627 |
+
"epoch": 774.18,
|
| 221628 |
+
"learning_rate": 8.476650641025642e-06,
|
| 221629 |
+
"loss": 0.6886,
|
| 221630 |
+
"step": 96095
|
| 221631 |
+
},
|
| 221632 |
+
{
|
| 221633 |
+
"epoch": 774.22,
|
| 221634 |
+
"learning_rate": 8.476570512820513e-06,
|
| 221635 |
+
"loss": 0.9608,
|
| 221636 |
+
"step": 96100
|
| 221637 |
+
},
|
| 221638 |
+
{
|
| 221639 |
+
"epoch": 774.26,
|
| 221640 |
+
"learning_rate": 8.476490384615386e-06,
|
| 221641 |
+
"loss": 0.2642,
|
| 221642 |
+
"step": 96105
|
| 221643 |
+
},
|
| 221644 |
+
{
|
| 221645 |
+
"epoch": 774.3,
|
| 221646 |
+
"learning_rate": 8.476410256410258e-06,
|
| 221647 |
+
"loss": 0.4536,
|
| 221648 |
+
"step": 96110
|
| 221649 |
+
},
|
| 221650 |
+
{
|
| 221651 |
+
"epoch": 774.34,
|
| 221652 |
+
"learning_rate": 8.476330128205129e-06,
|
| 221653 |
+
"loss": 0.3945,
|
| 221654 |
+
"step": 96115
|
| 221655 |
+
},
|
| 221656 |
+
{
|
| 221657 |
+
"epoch": 774.38,
|
| 221658 |
+
"learning_rate": 8.47625e-06,
|
| 221659 |
+
"loss": 0.631,
|
| 221660 |
+
"step": 96120
|
| 221661 |
+
},
|
| 221662 |
+
{
|
| 221663 |
+
"epoch": 774.42,
|
| 221664 |
+
"learning_rate": 8.476169871794873e-06,
|
| 221665 |
+
"loss": 0.9518,
|
| 221666 |
+
"step": 96125
|
| 221667 |
+
},
|
| 221668 |
+
{
|
| 221669 |
+
"epoch": 774.46,
|
| 221670 |
+
"learning_rate": 8.476089743589743e-06,
|
| 221671 |
+
"loss": 0.2784,
|
| 221672 |
+
"step": 96130
|
| 221673 |
+
},
|
| 221674 |
+
{
|
| 221675 |
+
"epoch": 774.5,
|
| 221676 |
+
"learning_rate": 8.476009615384616e-06,
|
| 221677 |
+
"loss": 0.3067,
|
| 221678 |
+
"step": 96135
|
| 221679 |
+
},
|
| 221680 |
+
{
|
| 221681 |
+
"epoch": 774.54,
|
| 221682 |
+
"learning_rate": 8.47592948717949e-06,
|
| 221683 |
+
"loss": 0.3015,
|
| 221684 |
+
"step": 96140
|
| 221685 |
+
},
|
| 221686 |
+
{
|
| 221687 |
+
"epoch": 774.58,
|
| 221688 |
+
"learning_rate": 8.475849358974359e-06,
|
| 221689 |
+
"loss": 0.6307,
|
| 221690 |
+
"step": 96145
|
| 221691 |
+
},
|
| 221692 |
+
{
|
| 221693 |
+
"epoch": 774.62,
|
| 221694 |
+
"learning_rate": 8.475769230769232e-06,
|
| 221695 |
+
"loss": 1.0663,
|
| 221696 |
+
"step": 96150
|
| 221697 |
+
},
|
| 221698 |
+
{
|
| 221699 |
+
"epoch": 774.66,
|
| 221700 |
+
"learning_rate": 8.475689102564103e-06,
|
| 221701 |
+
"loss": 0.2521,
|
| 221702 |
+
"step": 96155
|
| 221703 |
+
},
|
| 221704 |
+
{
|
| 221705 |
+
"epoch": 774.7,
|
| 221706 |
+
"learning_rate": 8.475608974358975e-06,
|
| 221707 |
+
"loss": 0.3085,
|
| 221708 |
+
"step": 96160
|
| 221709 |
+
},
|
| 221710 |
+
{
|
| 221711 |
+
"epoch": 774.74,
|
| 221712 |
+
"learning_rate": 8.475528846153846e-06,
|
| 221713 |
+
"loss": 0.3783,
|
| 221714 |
+
"step": 96165
|
| 221715 |
+
},
|
| 221716 |
+
{
|
| 221717 |
+
"epoch": 774.78,
|
| 221718 |
+
"learning_rate": 8.475448717948719e-06,
|
| 221719 |
+
"loss": 0.6273,
|
| 221720 |
+
"step": 96170
|
| 221721 |
+
},
|
| 221722 |
+
{
|
| 221723 |
+
"epoch": 774.82,
|
| 221724 |
+
"learning_rate": 8.47536858974359e-06,
|
| 221725 |
+
"loss": 1.0402,
|
| 221726 |
+
"step": 96175
|
| 221727 |
+
},
|
| 221728 |
+
{
|
| 221729 |
+
"epoch": 774.86,
|
| 221730 |
+
"learning_rate": 8.475288461538462e-06,
|
| 221731 |
+
"loss": 0.3043,
|
| 221732 |
+
"step": 96180
|
| 221733 |
+
},
|
| 221734 |
+
{
|
| 221735 |
+
"epoch": 774.9,
|
| 221736 |
+
"learning_rate": 8.475208333333333e-06,
|
| 221737 |
+
"loss": 0.3552,
|
| 221738 |
+
"step": 96185
|
| 221739 |
+
},
|
| 221740 |
+
{
|
| 221741 |
+
"epoch": 774.94,
|
| 221742 |
+
"learning_rate": 8.475128205128206e-06,
|
| 221743 |
+
"loss": 0.3561,
|
| 221744 |
+
"step": 96190
|
| 221745 |
+
},
|
| 221746 |
+
{
|
| 221747 |
+
"epoch": 774.98,
|
| 221748 |
+
"learning_rate": 8.475048076923078e-06,
|
| 221749 |
+
"loss": 0.7199,
|
| 221750 |
+
"step": 96195
|
| 221751 |
+
},
|
| 221752 |
+
{
|
| 221753 |
+
"epoch": 775.0,
|
| 221754 |
+
"eval_loss": 0.38370317220687866,
|
| 221755 |
+
"eval_runtime": 39.4417,
|
| 221756 |
+
"eval_samples_per_second": 21.297,
|
| 221757 |
+
"eval_steps_per_second": 0.685,
|
| 221758 |
+
"eval_wer": 0.1848883800801374,
|
| 221759 |
+
"step": 96197
|
| 221760 |
+
},
|
| 221761 |
+
{
|
| 221762 |
+
"epoch": 769.02,
|
| 221763 |
+
"learning_rate": 8.474967948717949e-06,
|
| 221764 |
+
"loss": 0.3397,
|
| 221765 |
+
"step": 96200
|
| 221766 |
+
},
|
| 221767 |
+
{
|
| 221768 |
+
"epoch": 769.06,
|
| 221769 |
+
"learning_rate": 8.474887820512822e-06,
|
| 221770 |
+
"loss": 0.2854,
|
| 221771 |
+
"step": 96205
|
| 221772 |
+
},
|
| 221773 |
+
{
|
| 221774 |
+
"epoch": 769.1,
|
| 221775 |
+
"learning_rate": 8.474807692307693e-06,
|
| 221776 |
+
"loss": 0.335,
|
| 221777 |
+
"step": 96210
|
| 221778 |
+
},
|
| 221779 |
+
{
|
| 221780 |
+
"epoch": 769.14,
|
| 221781 |
+
"learning_rate": 8.474727564102565e-06,
|
| 221782 |
+
"loss": 0.3452,
|
| 221783 |
+
"step": 96215
|
| 221784 |
+
},
|
| 221785 |
+
{
|
| 221786 |
+
"epoch": 769.18,
|
| 221787 |
+
"learning_rate": 8.474647435897436e-06,
|
| 221788 |
+
"loss": 0.7299,
|
| 221789 |
+
"step": 96220
|
| 221790 |
+
},
|
| 221791 |
+
{
|
| 221792 |
+
"epoch": 769.22,
|
| 221793 |
+
"learning_rate": 8.474567307692309e-06,
|
| 221794 |
+
"loss": 0.8785,
|
| 221795 |
+
"step": 96225
|
| 221796 |
+
},
|
| 221797 |
+
{
|
| 221798 |
+
"epoch": 769.26,
|
| 221799 |
+
"learning_rate": 8.47448717948718e-06,
|
| 221800 |
+
"loss": 0.2532,
|
| 221801 |
+
"step": 96230
|
| 221802 |
+
},
|
| 221803 |
+
{
|
| 221804 |
+
"epoch": 769.3,
|
| 221805 |
+
"learning_rate": 8.474407051282052e-06,
|
| 221806 |
+
"loss": 0.3313,
|
| 221807 |
+
"step": 96235
|
| 221808 |
+
},
|
| 221809 |
+
{
|
| 221810 |
+
"epoch": 769.34,
|
| 221811 |
+
"learning_rate": 8.474326923076925e-06,
|
| 221812 |
+
"loss": 0.3875,
|
| 221813 |
+
"step": 96240
|
| 221814 |
+
},
|
| 221815 |
+
{
|
| 221816 |
+
"epoch": 769.38,
|
| 221817 |
+
"learning_rate": 8.474246794871796e-06,
|
| 221818 |
+
"loss": 0.7963,
|
| 221819 |
+
"step": 96245
|
| 221820 |
+
},
|
| 221821 |
+
{
|
| 221822 |
+
"epoch": 769.42,
|
| 221823 |
+
"learning_rate": 8.474166666666668e-06,
|
| 221824 |
+
"loss": 0.7468,
|
| 221825 |
+
"step": 96250
|
| 221826 |
+
},
|
| 221827 |
+
{
|
| 221828 |
+
"epoch": 769.46,
|
| 221829 |
+
"learning_rate": 8.474086538461539e-06,
|
| 221830 |
+
"loss": 0.3469,
|
| 221831 |
+
"step": 96255
|
| 221832 |
+
},
|
| 221833 |
+
{
|
| 221834 |
+
"epoch": 769.5,
|
| 221835 |
+
"learning_rate": 8.474006410256412e-06,
|
| 221836 |
+
"loss": 0.3035,
|
| 221837 |
+
"step": 96260
|
| 221838 |
+
},
|
| 221839 |
+
{
|
| 221840 |
+
"epoch": 769.54,
|
| 221841 |
+
"learning_rate": 8.473926282051282e-06,
|
| 221842 |
+
"loss": 0.3875,
|
| 221843 |
+
"step": 96265
|
| 221844 |
+
},
|
| 221845 |
+
{
|
| 221846 |
+
"epoch": 769.58,
|
| 221847 |
+
"learning_rate": 8.473846153846155e-06,
|
| 221848 |
+
"loss": 0.8018,
|
| 221849 |
+
"step": 96270
|
| 221850 |
+
},
|
| 221851 |
+
{
|
| 221852 |
+
"epoch": 769.62,
|
| 221853 |
+
"learning_rate": 8.473766025641026e-06,
|
| 221854 |
+
"loss": 0.9109,
|
| 221855 |
+
"step": 96275
|
| 221856 |
+
},
|
| 221857 |
+
{
|
| 221858 |
+
"epoch": 769.66,
|
| 221859 |
+
"learning_rate": 8.473685897435897e-06,
|
| 221860 |
+
"loss": 0.2836,
|
| 221861 |
+
"step": 96280
|
| 221862 |
+
},
|
| 221863 |
+
{
|
| 221864 |
+
"epoch": 769.7,
|
| 221865 |
+
"learning_rate": 8.473605769230769e-06,
|
| 221866 |
+
"loss": 0.3713,
|
| 221867 |
+
"step": 96285
|
| 221868 |
+
},
|
| 221869 |
+
{
|
| 221870 |
+
"epoch": 769.74,
|
| 221871 |
+
"learning_rate": 8.473525641025642e-06,
|
| 221872 |
+
"loss": 0.4423,
|
| 221873 |
+
"step": 96290
|
| 221874 |
+
},
|
| 221875 |
+
{
|
| 221876 |
+
"epoch": 769.78,
|
| 221877 |
+
"learning_rate": 8.473445512820513e-06,
|
| 221878 |
+
"loss": 0.8542,
|
| 221879 |
+
"step": 96295
|
| 221880 |
+
},
|
| 221881 |
+
{
|
| 221882 |
+
"epoch": 769.82,
|
| 221883 |
+
"learning_rate": 8.473365384615385e-06,
|
| 221884 |
+
"loss": 0.9054,
|
| 221885 |
+
"step": 96300
|
| 221886 |
+
},
|
| 221887 |
+
{
|
| 221888 |
+
"epoch": 769.86,
|
| 221889 |
+
"learning_rate": 8.473285256410258e-06,
|
| 221890 |
+
"loss": 0.2686,
|
| 221891 |
+
"step": 96305
|
| 221892 |
+
},
|
| 221893 |
+
{
|
| 221894 |
+
"epoch": 769.9,
|
| 221895 |
+
"learning_rate": 8.473205128205129e-06,
|
| 221896 |
+
"loss": 0.3756,
|
| 221897 |
+
"step": 96310
|
| 221898 |
+
},
|
| 221899 |
+
{
|
| 221900 |
+
"epoch": 769.94,
|
| 221901 |
+
"learning_rate": 8.473125e-06,
|
| 221902 |
+
"loss": 0.3431,
|
| 221903 |
+
"step": 96315
|
| 221904 |
+
},
|
| 221905 |
+
{
|
| 221906 |
+
"epoch": 769.98,
|
| 221907 |
+
"learning_rate": 8.473044871794872e-06,
|
| 221908 |
+
"loss": 0.8294,
|
| 221909 |
+
"step": 96320
|
| 221910 |
+
},
|
| 221911 |
+
{
|
| 221912 |
+
"epoch": 770.0,
|
| 221913 |
+
"eval_loss": 0.3914641737937927,
|
| 221914 |
+
"eval_runtime": 38.8111,
|
| 221915 |
+
"eval_samples_per_second": 21.643,
|
| 221916 |
+
"eval_steps_per_second": 0.696,
|
| 221917 |
+
"eval_wer": 0.19241309678349922,
|
| 221918 |
+
"step": 96322
|
| 221919 |
+
},
|
| 221920 |
+
{
|
| 221921 |
+
"epoch": 770.02,
|
| 221922 |
+
"learning_rate": 8.472964743589745e-06,
|
| 221923 |
+
"loss": 0.3872,
|
| 221924 |
+
"step": 96325
|
| 221925 |
+
},
|
| 221926 |
+
{
|
| 221927 |
+
"epoch": 770.06,
|
| 221928 |
+
"learning_rate": 8.472884615384616e-06,
|
| 221929 |
+
"loss": 0.2868,
|
| 221930 |
+
"step": 96330
|
| 221931 |
+
},
|
| 221932 |
+
{
|
| 221933 |
+
"epoch": 770.1,
|
| 221934 |
+
"learning_rate": 8.472804487179487e-06,
|
| 221935 |
+
"loss": 0.3157,
|
| 221936 |
+
"step": 96335
|
| 221937 |
+
},
|
| 221938 |
+
{
|
| 221939 |
+
"epoch": 770.14,
|
| 221940 |
+
"learning_rate": 8.47272435897436e-06,
|
| 221941 |
+
"loss": 0.3176,
|
| 221942 |
+
"step": 96340
|
| 221943 |
+
},
|
| 221944 |
+
{
|
| 221945 |
+
"epoch": 770.18,
|
| 221946 |
+
"learning_rate": 8.472644230769232e-06,
|
| 221947 |
+
"loss": 0.7519,
|
| 221948 |
+
"step": 96345
|
| 221949 |
+
},
|
| 221950 |
+
{
|
| 221951 |
+
"epoch": 770.22,
|
| 221952 |
+
"learning_rate": 8.472564102564103e-06,
|
| 221953 |
+
"loss": 0.8266,
|
| 221954 |
+
"step": 96350
|
| 221955 |
+
},
|
| 221956 |
+
{
|
| 221957 |
+
"epoch": 770.26,
|
| 221958 |
+
"learning_rate": 8.472483974358975e-06,
|
| 221959 |
+
"loss": 0.2878,
|
| 221960 |
+
"step": 96355
|
| 221961 |
+
},
|
| 221962 |
+
{
|
| 221963 |
+
"epoch": 770.3,
|
| 221964 |
+
"learning_rate": 8.472403846153848e-06,
|
| 221965 |
+
"loss": 0.3173,
|
| 221966 |
+
"step": 96360
|
| 221967 |
+
},
|
| 221968 |
+
{
|
| 221969 |
+
"epoch": 770.34,
|
| 221970 |
+
"learning_rate": 8.472323717948719e-06,
|
| 221971 |
+
"loss": 0.3258,
|
| 221972 |
+
"step": 96365
|
| 221973 |
+
},
|
| 221974 |
+
{
|
| 221975 |
+
"epoch": 770.38,
|
| 221976 |
+
"learning_rate": 8.47224358974359e-06,
|
| 221977 |
+
"loss": 0.9034,
|
| 221978 |
+
"step": 96370
|
| 221979 |
+
},
|
| 221980 |
+
{
|
| 221981 |
+
"epoch": 770.42,
|
| 221982 |
+
"learning_rate": 8.472163461538462e-06,
|
| 221983 |
+
"loss": 0.9744,
|
| 221984 |
+
"step": 96375
|
| 221985 |
+
},
|
| 221986 |
+
{
|
| 221987 |
+
"epoch": 770.46,
|
| 221988 |
+
"learning_rate": 8.472083333333335e-06,
|
| 221989 |
+
"loss": 0.2759,
|
| 221990 |
+
"step": 96380
|
| 221991 |
+
},
|
| 221992 |
+
{
|
| 221993 |
+
"epoch": 770.5,
|
| 221994 |
+
"learning_rate": 8.472003205128206e-06,
|
| 221995 |
+
"loss": 0.3071,
|
| 221996 |
+
"step": 96385
|
| 221997 |
+
},
|
| 221998 |
+
{
|
| 221999 |
+
"epoch": 770.54,
|
| 222000 |
+
"learning_rate": 8.471923076923077e-06,
|
| 222001 |
+
"loss": 0.4087,
|
| 222002 |
+
"step": 96390
|
| 222003 |
+
},
|
| 222004 |
+
{
|
| 222005 |
+
"epoch": 770.58,
|
| 222006 |
+
"learning_rate": 8.47184294871795e-06,
|
| 222007 |
+
"loss": 0.7003,
|
| 222008 |
+
"step": 96395
|
| 222009 |
+
},
|
| 222010 |
+
{
|
| 222011 |
+
"epoch": 770.62,
|
| 222012 |
+
"learning_rate": 8.471762820512822e-06,
|
| 222013 |
+
"loss": 0.8226,
|
| 222014 |
+
"step": 96400
|
| 222015 |
+
},
|
| 222016 |
+
{
|
| 222017 |
+
"epoch": 770.66,
|
| 222018 |
+
"learning_rate": 8.471682692307693e-06,
|
| 222019 |
+
"loss": 0.3137,
|
| 222020 |
+
"step": 96405
|
| 222021 |
+
},
|
| 222022 |
+
{
|
| 222023 |
+
"epoch": 770.7,
|
| 222024 |
+
"learning_rate": 8.471602564102565e-06,
|
| 222025 |
+
"loss": 0.3283,
|
| 222026 |
+
"step": 96410
|
| 222027 |
+
},
|
| 222028 |
+
{
|
| 222029 |
+
"epoch": 770.74,
|
| 222030 |
+
"learning_rate": 8.471522435897438e-06,
|
| 222031 |
+
"loss": 0.3588,
|
| 222032 |
+
"step": 96415
|
| 222033 |
+
},
|
| 222034 |
+
{
|
| 222035 |
+
"epoch": 770.78,
|
| 222036 |
+
"learning_rate": 8.471442307692307e-06,
|
| 222037 |
+
"loss": 0.7525,
|
| 222038 |
+
"step": 96420
|
| 222039 |
+
},
|
| 222040 |
+
{
|
| 222041 |
+
"epoch": 770.82,
|
| 222042 |
+
"learning_rate": 8.47136217948718e-06,
|
| 222043 |
+
"loss": 0.9263,
|
| 222044 |
+
"step": 96425
|
| 222045 |
+
},
|
| 222046 |
+
{
|
| 222047 |
+
"epoch": 770.86,
|
| 222048 |
+
"learning_rate": 8.471282051282052e-06,
|
| 222049 |
+
"loss": 0.2465,
|
| 222050 |
+
"step": 96430
|
| 222051 |
+
},
|
| 222052 |
+
{
|
| 222053 |
+
"epoch": 770.9,
|
| 222054 |
+
"learning_rate": 8.471201923076923e-06,
|
| 222055 |
+
"loss": 0.2789,
|
| 222056 |
+
"step": 96435
|
| 222057 |
+
},
|
| 222058 |
+
{
|
| 222059 |
+
"epoch": 770.94,
|
| 222060 |
+
"learning_rate": 8.471121794871796e-06,
|
| 222061 |
+
"loss": 0.3788,
|
| 222062 |
+
"step": 96440
|
| 222063 |
+
},
|
| 222064 |
+
{
|
| 222065 |
+
"epoch": 770.98,
|
| 222066 |
+
"learning_rate": 8.471041666666667e-06,
|
| 222067 |
+
"loss": 0.8567,
|
| 222068 |
+
"step": 96445
|
| 222069 |
+
},
|
| 222070 |
+
{
|
| 222071 |
+
"epoch": 771.0,
|
| 222072 |
+
"eval_loss": 0.33281025290489197,
|
| 222073 |
+
"eval_runtime": 40.656,
|
| 222074 |
+
"eval_samples_per_second": 20.661,
|
| 222075 |
+
"eval_steps_per_second": 0.664,
|
| 222076 |
+
"eval_wer": 0.18247860174063152,
|
| 222077 |
+
"step": 96447
|
| 222078 |
}
|
| 222079 |
],
|
| 222080 |
+
"max_steps": 625000,
|
| 222081 |
"num_train_epochs": 5000,
|
| 222082 |
+
"total_flos": 2.7141352016257637e+20,
|
| 222083 |
"trial_name": null,
|
| 222084 |
"trial_params": null
|
| 222085 |
}
|
model-bin/finetune/base/{checkpoint-95824 → checkpoint-96447}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629966687.6381752/events.out.tfevents.1629966687.8e89bd551565.924.121
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f28ac8dccd430849d225aea988e79c803fbad2fd121ad2fc7a28369f8a9cdbe
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629967137.0971546/events.out.tfevents.1629967137.8e89bd551565.924.123
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:904d9e0c77b30c606211f273603ea3b678aff40f1f2a22945e12b284a2e770f5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629967569.9641695/events.out.tfevents.1629967571.8e89bd551565.924.125
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c8dca8139b449735ff41a76c78917168cabbdc383c3e696315d085093a6c732a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629968006.3075402/events.out.tfevents.1629968006.8e89bd551565.924.127
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:989262e143c1abc5dc3b39e884339a7e93c0d179bd90b4e3633b1ee9f322a79c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629968435.5634236/events.out.tfevents.1629968435.8e89bd551565.924.129
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0d4e89399e6b9c6d85c00b7e5bf1e99aa6fc065cf43a58ea95a99ad60497398
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629966687.8e89bd551565.924.120
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5aa5463a6b87a42b6803c1fde7610ed1ba888696f34661a7e3d7cc51f9680e26
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629967137.8e89bd551565.924.122
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a91a14e75dadc682e8aa417ab116f834a3aa2e239407ad458f65b3ef8369650f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629967569.8e89bd551565.924.124
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa1ab52697cfd3c012c7683c6e3fd068f28af9c63f9490e9f08f6921579fddcf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629968006.8e89bd551565.924.126
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca7b761cf77c4b77c8531df627560881205692eec612463448460161658378e4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629968435.8e89bd551565.924.128
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:23ac6c809eac6881b1609659f7ee5b388d12896dadb31762315c7e0d8cf46078
|
| 3 |
+
size 8622
|