"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630000221.7268653/events.out.tfevents.1630000221.8e89bd551565.924.271 +3 -0
- model-bin/finetune/base/log/1630000701.8113801/events.out.tfevents.1630000703.8e89bd551565.924.273 +3 -0
- model-bin/finetune/base/log/1630001150.9827878/events.out.tfevents.1630001150.8e89bd551565.924.275 +3 -0
- model-bin/finetune/base/log/1630001578.948476/events.out.tfevents.1630001578.8e89bd551565.924.277 +3 -0
- model-bin/finetune/base/log/1630002016.2967117/events.out.tfevents.1630002017.8e89bd551565.924.279 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630000221.8e89bd551565.924.270 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630000701.8e89bd551565.924.272 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630001150.8e89bd551565.924.274 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630001578.8e89bd551565.924.276 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630002016.8e89bd551565.924.278 +3 -0
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/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 sha256:1cc67d59982335b5f1594cea912639620a1bfd368cd0183861691a5cfd69624e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/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 sha256:5ebe3fe7df35047174bb8f3acfcd7995c87c450609cee5cf7925cd456f6818e3
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b006148e32c67717ff92ba33b683fa617305fe46e06c05a7c312a81b4993938
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/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 sha256:4e88b585ceb1b439cdbd75313bebd4c1732cfce1d2afa12ab658b7c33ad04a04
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/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:c329732efc46436918507eab48149ab7f31c0afb01f76cf6a5fa9d633b3314e2
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -233163,11 +233163,800 @@
|
|
| 233163 |
"eval_steps_per_second": 0.665,
|
| 233164 |
"eval_wer": 0.18248653370214005,
|
| 233165 |
"step": 105160
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 233166 |
}
|
| 233167 |
],
|
| 233168 |
-
"max_steps":
|
| 233169 |
"num_train_epochs": 5000,
|
| 233170 |
-
"total_flos": 2.
|
| 233171 |
"trial_name": null,
|
| 233172 |
"trial_params": null
|
| 233173 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 846.0,
|
| 5 |
+
"global_step": 105783,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 233163 |
"eval_steps_per_second": 0.665,
|
| 233164 |
"eval_wer": 0.18248653370214005,
|
| 233165 |
"step": 105160
|
| 233166 |
+
},
|
| 233167 |
+
{
|
| 233168 |
+
"epoch": 848.04,
|
| 233169 |
+
"learning_rate": 8.331346153846156e-06,
|
| 233170 |
+
"loss": 0.3863,
|
| 233171 |
+
"step": 105165
|
| 233172 |
+
},
|
| 233173 |
+
{
|
| 233174 |
+
"epoch": 848.08,
|
| 233175 |
+
"learning_rate": 8.331266025641025e-06,
|
| 233176 |
+
"loss": 0.266,
|
| 233177 |
+
"step": 105170
|
| 233178 |
+
},
|
| 233179 |
+
{
|
| 233180 |
+
"epoch": 848.12,
|
| 233181 |
+
"learning_rate": 8.331185897435898e-06,
|
| 233182 |
+
"loss": 0.3381,
|
| 233183 |
+
"step": 105175
|
| 233184 |
+
},
|
| 233185 |
+
{
|
| 233186 |
+
"epoch": 848.16,
|
| 233187 |
+
"learning_rate": 8.33110576923077e-06,
|
| 233188 |
+
"loss": 0.4272,
|
| 233189 |
+
"step": 105180
|
| 233190 |
+
},
|
| 233191 |
+
{
|
| 233192 |
+
"epoch": 848.2,
|
| 233193 |
+
"learning_rate": 8.331025641025641e-06,
|
| 233194 |
+
"loss": 1.2679,
|
| 233195 |
+
"step": 105185
|
| 233196 |
+
},
|
| 233197 |
+
{
|
| 233198 |
+
"epoch": 848.24,
|
| 233199 |
+
"learning_rate": 8.330945512820512e-06,
|
| 233200 |
+
"loss": 0.3238,
|
| 233201 |
+
"step": 105190
|
| 233202 |
+
},
|
| 233203 |
+
{
|
| 233204 |
+
"epoch": 848.28,
|
| 233205 |
+
"learning_rate": 8.330865384615385e-06,
|
| 233206 |
+
"loss": 0.2744,
|
| 233207 |
+
"step": 105195
|
| 233208 |
+
},
|
| 233209 |
+
{
|
| 233210 |
+
"epoch": 848.32,
|
| 233211 |
+
"learning_rate": 8.330785256410257e-06,
|
| 233212 |
+
"loss": 0.3612,
|
| 233213 |
+
"step": 105200
|
| 233214 |
+
},
|
| 233215 |
+
{
|
| 233216 |
+
"epoch": 848.36,
|
| 233217 |
+
"learning_rate": 8.330705128205128e-06,
|
| 233218 |
+
"loss": 0.4517,
|
| 233219 |
+
"step": 105205
|
| 233220 |
+
},
|
| 233221 |
+
{
|
| 233222 |
+
"epoch": 848.4,
|
| 233223 |
+
"learning_rate": 8.330625000000001e-06,
|
| 233224 |
+
"loss": 1.3312,
|
| 233225 |
+
"step": 105210
|
| 233226 |
+
},
|
| 233227 |
+
{
|
| 233228 |
+
"epoch": 848.44,
|
| 233229 |
+
"learning_rate": 8.330544871794873e-06,
|
| 233230 |
+
"loss": 0.3226,
|
| 233231 |
+
"step": 105215
|
| 233232 |
+
},
|
| 233233 |
+
{
|
| 233234 |
+
"epoch": 848.48,
|
| 233235 |
+
"learning_rate": 8.330464743589744e-06,
|
| 233236 |
+
"loss": 0.2646,
|
| 233237 |
+
"step": 105220
|
| 233238 |
+
},
|
| 233239 |
+
{
|
| 233240 |
+
"epoch": 848.52,
|
| 233241 |
+
"learning_rate": 8.330384615384615e-06,
|
| 233242 |
+
"loss": 0.3282,
|
| 233243 |
+
"step": 105225
|
| 233244 |
+
},
|
| 233245 |
+
{
|
| 233246 |
+
"epoch": 848.56,
|
| 233247 |
+
"learning_rate": 8.330304487179488e-06,
|
| 233248 |
+
"loss": 0.4587,
|
| 233249 |
+
"step": 105230
|
| 233250 |
+
},
|
| 233251 |
+
{
|
| 233252 |
+
"epoch": 848.6,
|
| 233253 |
+
"learning_rate": 8.33022435897436e-06,
|
| 233254 |
+
"loss": 1.2765,
|
| 233255 |
+
"step": 105235
|
| 233256 |
+
},
|
| 233257 |
+
{
|
| 233258 |
+
"epoch": 848.64,
|
| 233259 |
+
"learning_rate": 8.330144230769231e-06,
|
| 233260 |
+
"loss": 0.2903,
|
| 233261 |
+
"step": 105240
|
| 233262 |
+
},
|
| 233263 |
+
{
|
| 233264 |
+
"epoch": 848.68,
|
| 233265 |
+
"learning_rate": 8.330064102564102e-06,
|
| 233266 |
+
"loss": 0.2608,
|
| 233267 |
+
"step": 105245
|
| 233268 |
+
},
|
| 233269 |
+
{
|
| 233270 |
+
"epoch": 848.72,
|
| 233271 |
+
"learning_rate": 8.329983974358975e-06,
|
| 233272 |
+
"loss": 0.3429,
|
| 233273 |
+
"step": 105250
|
| 233274 |
+
},
|
| 233275 |
+
{
|
| 233276 |
+
"epoch": 848.76,
|
| 233277 |
+
"learning_rate": 8.329903846153847e-06,
|
| 233278 |
+
"loss": 0.4259,
|
| 233279 |
+
"step": 105255
|
| 233280 |
+
},
|
| 233281 |
+
{
|
| 233282 |
+
"epoch": 848.8,
|
| 233283 |
+
"learning_rate": 8.329823717948718e-06,
|
| 233284 |
+
"loss": 1.0882,
|
| 233285 |
+
"step": 105260
|
| 233286 |
+
},
|
| 233287 |
+
{
|
| 233288 |
+
"epoch": 848.84,
|
| 233289 |
+
"learning_rate": 8.329743589743591e-06,
|
| 233290 |
+
"loss": 0.3339,
|
| 233291 |
+
"step": 105265
|
| 233292 |
+
},
|
| 233293 |
+
{
|
| 233294 |
+
"epoch": 848.88,
|
| 233295 |
+
"learning_rate": 8.329663461538463e-06,
|
| 233296 |
+
"loss": 0.2745,
|
| 233297 |
+
"step": 105270
|
| 233298 |
+
},
|
| 233299 |
+
{
|
| 233300 |
+
"epoch": 848.92,
|
| 233301 |
+
"learning_rate": 8.329583333333334e-06,
|
| 233302 |
+
"loss": 0.2822,
|
| 233303 |
+
"step": 105275
|
| 233304 |
+
},
|
| 233305 |
+
{
|
| 233306 |
+
"epoch": 848.96,
|
| 233307 |
+
"learning_rate": 8.329503205128205e-06,
|
| 233308 |
+
"loss": 0.4802,
|
| 233309 |
+
"step": 105280
|
| 233310 |
+
},
|
| 233311 |
+
{
|
| 233312 |
+
"epoch": 849.0,
|
| 233313 |
+
"eval_loss": 0.381322979927063,
|
| 233314 |
+
"eval_runtime": 40.0317,
|
| 233315 |
+
"eval_samples_per_second": 20.933,
|
| 233316 |
+
"eval_steps_per_second": 0.674,
|
| 233317 |
+
"eval_wer": 0.18333577390540343,
|
| 233318 |
+
"step": 105284
|
| 233319 |
+
},
|
| 233320 |
+
{
|
| 233321 |
+
"epoch": 842.01,
|
| 233322 |
+
"learning_rate": 8.329423076923078e-06,
|
| 233323 |
+
"loss": 0.3944,
|
| 233324 |
+
"step": 105285
|
| 233325 |
+
},
|
| 233326 |
+
{
|
| 233327 |
+
"epoch": 842.05,
|
| 233328 |
+
"learning_rate": 8.32934294871795e-06,
|
| 233329 |
+
"loss": 0.2897,
|
| 233330 |
+
"step": 105290
|
| 233331 |
+
},
|
| 233332 |
+
{
|
| 233333 |
+
"epoch": 842.09,
|
| 233334 |
+
"learning_rate": 8.329262820512821e-06,
|
| 233335 |
+
"loss": 0.2899,
|
| 233336 |
+
"step": 105295
|
| 233337 |
+
},
|
| 233338 |
+
{
|
| 233339 |
+
"epoch": 842.13,
|
| 233340 |
+
"learning_rate": 8.329182692307694e-06,
|
| 233341 |
+
"loss": 0.2785,
|
| 233342 |
+
"step": 105300
|
| 233343 |
+
},
|
| 233344 |
+
{
|
| 233345 |
+
"epoch": 842.17,
|
| 233346 |
+
"learning_rate": 8.329102564102566e-06,
|
| 233347 |
+
"loss": 0.4343,
|
| 233348 |
+
"step": 105305
|
| 233349 |
+
},
|
| 233350 |
+
{
|
| 233351 |
+
"epoch": 842.21,
|
| 233352 |
+
"learning_rate": 8.329022435897437e-06,
|
| 233353 |
+
"loss": 1.112,
|
| 233354 |
+
"step": 105310
|
| 233355 |
+
},
|
| 233356 |
+
{
|
| 233357 |
+
"epoch": 842.25,
|
| 233358 |
+
"learning_rate": 8.328942307692308e-06,
|
| 233359 |
+
"loss": 0.3307,
|
| 233360 |
+
"step": 105315
|
| 233361 |
+
},
|
| 233362 |
+
{
|
| 233363 |
+
"epoch": 842.29,
|
| 233364 |
+
"learning_rate": 8.328862179487181e-06,
|
| 233365 |
+
"loss": 0.2525,
|
| 233366 |
+
"step": 105320
|
| 233367 |
+
},
|
| 233368 |
+
{
|
| 233369 |
+
"epoch": 842.33,
|
| 233370 |
+
"learning_rate": 8.328782051282051e-06,
|
| 233371 |
+
"loss": 0.3191,
|
| 233372 |
+
"step": 105325
|
| 233373 |
+
},
|
| 233374 |
+
{
|
| 233375 |
+
"epoch": 842.37,
|
| 233376 |
+
"learning_rate": 8.328701923076924e-06,
|
| 233377 |
+
"loss": 0.4257,
|
| 233378 |
+
"step": 105330
|
| 233379 |
+
},
|
| 233380 |
+
{
|
| 233381 |
+
"epoch": 842.41,
|
| 233382 |
+
"learning_rate": 8.328621794871795e-06,
|
| 233383 |
+
"loss": 1.0987,
|
| 233384 |
+
"step": 105335
|
| 233385 |
+
},
|
| 233386 |
+
{
|
| 233387 |
+
"epoch": 842.45,
|
| 233388 |
+
"learning_rate": 8.328541666666667e-06,
|
| 233389 |
+
"loss": 0.3385,
|
| 233390 |
+
"step": 105340
|
| 233391 |
+
},
|
| 233392 |
+
{
|
| 233393 |
+
"epoch": 842.49,
|
| 233394 |
+
"learning_rate": 8.328461538461538e-06,
|
| 233395 |
+
"loss": 0.2685,
|
| 233396 |
+
"step": 105345
|
| 233397 |
+
},
|
| 233398 |
+
{
|
| 233399 |
+
"epoch": 842.53,
|
| 233400 |
+
"learning_rate": 8.328381410256411e-06,
|
| 233401 |
+
"loss": 0.4699,
|
| 233402 |
+
"step": 105350
|
| 233403 |
+
},
|
| 233404 |
+
{
|
| 233405 |
+
"epoch": 842.57,
|
| 233406 |
+
"learning_rate": 8.328301282051282e-06,
|
| 233407 |
+
"loss": 0.5122,
|
| 233408 |
+
"step": 105355
|
| 233409 |
+
},
|
| 233410 |
+
{
|
| 233411 |
+
"epoch": 842.61,
|
| 233412 |
+
"learning_rate": 8.328221153846154e-06,
|
| 233413 |
+
"loss": 1.2801,
|
| 233414 |
+
"step": 105360
|
| 233415 |
+
},
|
| 233416 |
+
{
|
| 233417 |
+
"epoch": 842.65,
|
| 233418 |
+
"learning_rate": 8.328141025641027e-06,
|
| 233419 |
+
"loss": 0.3043,
|
| 233420 |
+
"step": 105365
|
| 233421 |
+
},
|
| 233422 |
+
{
|
| 233423 |
+
"epoch": 842.69,
|
| 233424 |
+
"learning_rate": 8.328060897435898e-06,
|
| 233425 |
+
"loss": 0.2906,
|
| 233426 |
+
"step": 105370
|
| 233427 |
+
},
|
| 233428 |
+
{
|
| 233429 |
+
"epoch": 842.73,
|
| 233430 |
+
"learning_rate": 8.32798076923077e-06,
|
| 233431 |
+
"loss": 0.2943,
|
| 233432 |
+
"step": 105375
|
| 233433 |
+
},
|
| 233434 |
+
{
|
| 233435 |
+
"epoch": 842.77,
|
| 233436 |
+
"learning_rate": 8.327900641025641e-06,
|
| 233437 |
+
"loss": 0.5976,
|
| 233438 |
+
"step": 105380
|
| 233439 |
+
},
|
| 233440 |
+
{
|
| 233441 |
+
"epoch": 842.81,
|
| 233442 |
+
"learning_rate": 8.327820512820514e-06,
|
| 233443 |
+
"loss": 1.1249,
|
| 233444 |
+
"step": 105385
|
| 233445 |
+
},
|
| 233446 |
+
{
|
| 233447 |
+
"epoch": 842.85,
|
| 233448 |
+
"learning_rate": 8.327740384615385e-06,
|
| 233449 |
+
"loss": 0.3051,
|
| 233450 |
+
"step": 105390
|
| 233451 |
+
},
|
| 233452 |
+
{
|
| 233453 |
+
"epoch": 842.89,
|
| 233454 |
+
"learning_rate": 8.327660256410257e-06,
|
| 233455 |
+
"loss": 0.3528,
|
| 233456 |
+
"step": 105395
|
| 233457 |
+
},
|
| 233458 |
+
{
|
| 233459 |
+
"epoch": 842.93,
|
| 233460 |
+
"learning_rate": 8.32758012820513e-06,
|
| 233461 |
+
"loss": 0.361,
|
| 233462 |
+
"step": 105400
|
| 233463 |
+
},
|
| 233464 |
+
{
|
| 233465 |
+
"epoch": 842.97,
|
| 233466 |
+
"learning_rate": 8.327500000000001e-06,
|
| 233467 |
+
"loss": 1.0755,
|
| 233468 |
+
"step": 105405
|
| 233469 |
+
},
|
| 233470 |
+
{
|
| 233471 |
+
"epoch": 843.0,
|
| 233472 |
+
"eval_loss": 0.35919445753097534,
|
| 233473 |
+
"eval_runtime": 38.3904,
|
| 233474 |
+
"eval_samples_per_second": 21.802,
|
| 233475 |
+
"eval_steps_per_second": 0.703,
|
| 233476 |
+
"eval_wer": 0.18651912923989297,
|
| 233477 |
+
"step": 105409
|
| 233478 |
+
},
|
| 233479 |
+
{
|
| 233480 |
+
"epoch": 850.01,
|
| 233481 |
+
"learning_rate": 8.327419871794873e-06,
|
| 233482 |
+
"loss": 0.4187,
|
| 233483 |
+
"step": 105410
|
| 233484 |
+
},
|
| 233485 |
+
{
|
| 233486 |
+
"epoch": 850.05,
|
| 233487 |
+
"learning_rate": 8.327339743589744e-06,
|
| 233488 |
+
"loss": 0.2872,
|
| 233489 |
+
"step": 105415
|
| 233490 |
+
},
|
| 233491 |
+
{
|
| 233492 |
+
"epoch": 850.09,
|
| 233493 |
+
"learning_rate": 8.327259615384617e-06,
|
| 233494 |
+
"loss": 0.3013,
|
| 233495 |
+
"step": 105420
|
| 233496 |
+
},
|
| 233497 |
+
{
|
| 233498 |
+
"epoch": 850.13,
|
| 233499 |
+
"learning_rate": 8.327179487179488e-06,
|
| 233500 |
+
"loss": 0.3807,
|
| 233501 |
+
"step": 105425
|
| 233502 |
+
},
|
| 233503 |
+
{
|
| 233504 |
+
"epoch": 850.17,
|
| 233505 |
+
"learning_rate": 8.32709935897436e-06,
|
| 233506 |
+
"loss": 0.5397,
|
| 233507 |
+
"step": 105430
|
| 233508 |
+
},
|
| 233509 |
+
{
|
| 233510 |
+
"epoch": 850.21,
|
| 233511 |
+
"learning_rate": 8.327019230769231e-06,
|
| 233512 |
+
"loss": 1.0682,
|
| 233513 |
+
"step": 105435
|
| 233514 |
+
},
|
| 233515 |
+
{
|
| 233516 |
+
"epoch": 850.25,
|
| 233517 |
+
"learning_rate": 8.326939102564104e-06,
|
| 233518 |
+
"loss": 0.2983,
|
| 233519 |
+
"step": 105440
|
| 233520 |
+
},
|
| 233521 |
+
{
|
| 233522 |
+
"epoch": 850.29,
|
| 233523 |
+
"learning_rate": 8.326858974358974e-06,
|
| 233524 |
+
"loss": 0.3082,
|
| 233525 |
+
"step": 105445
|
| 233526 |
+
},
|
| 233527 |
+
{
|
| 233528 |
+
"epoch": 850.33,
|
| 233529 |
+
"learning_rate": 8.326778846153847e-06,
|
| 233530 |
+
"loss": 0.2716,
|
| 233531 |
+
"step": 105450
|
| 233532 |
+
},
|
| 233533 |
+
{
|
| 233534 |
+
"epoch": 850.37,
|
| 233535 |
+
"learning_rate": 8.32669871794872e-06,
|
| 233536 |
+
"loss": 0.594,
|
| 233537 |
+
"step": 105455
|
| 233538 |
+
},
|
| 233539 |
+
{
|
| 233540 |
+
"epoch": 850.41,
|
| 233541 |
+
"learning_rate": 8.32661858974359e-06,
|
| 233542 |
+
"loss": 1.1799,
|
| 233543 |
+
"step": 105460
|
| 233544 |
+
},
|
| 233545 |
+
{
|
| 233546 |
+
"epoch": 850.45,
|
| 233547 |
+
"learning_rate": 8.326538461538463e-06,
|
| 233548 |
+
"loss": 0.2721,
|
| 233549 |
+
"step": 105465
|
| 233550 |
+
},
|
| 233551 |
+
{
|
| 233552 |
+
"epoch": 850.49,
|
| 233553 |
+
"learning_rate": 8.326458333333334e-06,
|
| 233554 |
+
"loss": 0.3039,
|
| 233555 |
+
"step": 105470
|
| 233556 |
+
},
|
| 233557 |
+
{
|
| 233558 |
+
"epoch": 850.53,
|
| 233559 |
+
"learning_rate": 8.326378205128205e-06,
|
| 233560 |
+
"loss": 0.3523,
|
| 233561 |
+
"step": 105475
|
| 233562 |
+
},
|
| 233563 |
+
{
|
| 233564 |
+
"epoch": 850.57,
|
| 233565 |
+
"learning_rate": 8.326298076923077e-06,
|
| 233566 |
+
"loss": 0.482,
|
| 233567 |
+
"step": 105480
|
| 233568 |
+
},
|
| 233569 |
+
{
|
| 233570 |
+
"epoch": 850.61,
|
| 233571 |
+
"learning_rate": 8.32621794871795e-06,
|
| 233572 |
+
"loss": 1.1859,
|
| 233573 |
+
"step": 105485
|
| 233574 |
+
},
|
| 233575 |
+
{
|
| 233576 |
+
"epoch": 850.65,
|
| 233577 |
+
"learning_rate": 8.326137820512821e-06,
|
| 233578 |
+
"loss": 0.3675,
|
| 233579 |
+
"step": 105490
|
| 233580 |
+
},
|
| 233581 |
+
{
|
| 233582 |
+
"epoch": 850.69,
|
| 233583 |
+
"learning_rate": 8.326057692307692e-06,
|
| 233584 |
+
"loss": 0.2998,
|
| 233585 |
+
"step": 105495
|
| 233586 |
+
},
|
| 233587 |
+
{
|
| 233588 |
+
"epoch": 850.73,
|
| 233589 |
+
"learning_rate": 8.325977564102565e-06,
|
| 233590 |
+
"loss": 0.3144,
|
| 233591 |
+
"step": 105500
|
| 233592 |
+
},
|
| 233593 |
+
{
|
| 233594 |
+
"epoch": 850.77,
|
| 233595 |
+
"learning_rate": 8.325897435897437e-06,
|
| 233596 |
+
"loss": 0.5713,
|
| 233597 |
+
"step": 105505
|
| 233598 |
+
},
|
| 233599 |
+
{
|
| 233600 |
+
"epoch": 850.81,
|
| 233601 |
+
"learning_rate": 8.325817307692308e-06,
|
| 233602 |
+
"loss": 1.1369,
|
| 233603 |
+
"step": 105510
|
| 233604 |
+
},
|
| 233605 |
+
{
|
| 233606 |
+
"epoch": 850.85,
|
| 233607 |
+
"learning_rate": 8.32573717948718e-06,
|
| 233608 |
+
"loss": 0.3177,
|
| 233609 |
+
"step": 105515
|
| 233610 |
+
},
|
| 233611 |
+
{
|
| 233612 |
+
"epoch": 850.89,
|
| 233613 |
+
"learning_rate": 8.325657051282053e-06,
|
| 233614 |
+
"loss": 0.2842,
|
| 233615 |
+
"step": 105520
|
| 233616 |
+
},
|
| 233617 |
+
{
|
| 233618 |
+
"epoch": 850.93,
|
| 233619 |
+
"learning_rate": 8.325576923076924e-06,
|
| 233620 |
+
"loss": 0.3687,
|
| 233621 |
+
"step": 105525
|
| 233622 |
+
},
|
| 233623 |
+
{
|
| 233624 |
+
"epoch": 850.97,
|
| 233625 |
+
"learning_rate": 8.325496794871795e-06,
|
| 233626 |
+
"loss": 0.5373,
|
| 233627 |
+
"step": 105530
|
| 233628 |
+
},
|
| 233629 |
+
{
|
| 233630 |
+
"epoch": 851.0,
|
| 233631 |
+
"eval_loss": 0.3748731315135956,
|
| 233632 |
+
"eval_runtime": 38.0116,
|
| 233633 |
+
"eval_samples_per_second": 22.046,
|
| 233634 |
+
"eval_steps_per_second": 0.71,
|
| 233635 |
+
"eval_wer": 0.18251511130768097,
|
| 233636 |
+
"step": 105533
|
| 233637 |
+
},
|
| 233638 |
+
{
|
| 233639 |
+
"epoch": 844.02,
|
| 233640 |
+
"learning_rate": 8.325416666666667e-06,
|
| 233641 |
+
"loss": 0.4471,
|
| 233642 |
+
"step": 105535
|
| 233643 |
+
},
|
| 233644 |
+
{
|
| 233645 |
+
"epoch": 844.06,
|
| 233646 |
+
"learning_rate": 8.32533653846154e-06,
|
| 233647 |
+
"loss": 0.3063,
|
| 233648 |
+
"step": 105540
|
| 233649 |
+
},
|
| 233650 |
+
{
|
| 233651 |
+
"epoch": 844.1,
|
| 233652 |
+
"learning_rate": 8.325256410256411e-06,
|
| 233653 |
+
"loss": 0.3365,
|
| 233654 |
+
"step": 105545
|
| 233655 |
+
},
|
| 233656 |
+
{
|
| 233657 |
+
"epoch": 844.14,
|
| 233658 |
+
"learning_rate": 8.325176282051282e-06,
|
| 233659 |
+
"loss": 0.309,
|
| 233660 |
+
"step": 105550
|
| 233661 |
+
},
|
| 233662 |
+
{
|
| 233663 |
+
"epoch": 844.18,
|
| 233664 |
+
"learning_rate": 8.325096153846155e-06,
|
| 233665 |
+
"loss": 0.6159,
|
| 233666 |
+
"step": 105555
|
| 233667 |
+
},
|
| 233668 |
+
{
|
| 233669 |
+
"epoch": 844.22,
|
| 233670 |
+
"learning_rate": 8.325016025641027e-06,
|
| 233671 |
+
"loss": 1.0953,
|
| 233672 |
+
"step": 105560
|
| 233673 |
+
},
|
| 233674 |
+
{
|
| 233675 |
+
"epoch": 844.25,
|
| 233676 |
+
"learning_rate": 8.324935897435898e-06,
|
| 233677 |
+
"loss": 0.2684,
|
| 233678 |
+
"step": 105565
|
| 233679 |
+
},
|
| 233680 |
+
{
|
| 233681 |
+
"epoch": 844.29,
|
| 233682 |
+
"learning_rate": 8.32485576923077e-06,
|
| 233683 |
+
"loss": 0.3235,
|
| 233684 |
+
"step": 105570
|
| 233685 |
+
},
|
| 233686 |
+
{
|
| 233687 |
+
"epoch": 844.33,
|
| 233688 |
+
"learning_rate": 8.324775641025643e-06,
|
| 233689 |
+
"loss": 0.2988,
|
| 233690 |
+
"step": 105575
|
| 233691 |
+
},
|
| 233692 |
+
{
|
| 233693 |
+
"epoch": 844.37,
|
| 233694 |
+
"learning_rate": 8.324695512820514e-06,
|
| 233695 |
+
"loss": 0.6459,
|
| 233696 |
+
"step": 105580
|
| 233697 |
+
},
|
| 233698 |
+
{
|
| 233699 |
+
"epoch": 844.41,
|
| 233700 |
+
"learning_rate": 8.324615384615385e-06,
|
| 233701 |
+
"loss": 0.9401,
|
| 233702 |
+
"step": 105585
|
| 233703 |
+
},
|
| 233704 |
+
{
|
| 233705 |
+
"epoch": 844.45,
|
| 233706 |
+
"learning_rate": 8.324535256410257e-06,
|
| 233707 |
+
"loss": 0.2557,
|
| 233708 |
+
"step": 105590
|
| 233709 |
+
},
|
| 233710 |
+
{
|
| 233711 |
+
"epoch": 844.49,
|
| 233712 |
+
"learning_rate": 8.32445512820513e-06,
|
| 233713 |
+
"loss": 0.2755,
|
| 233714 |
+
"step": 105595
|
| 233715 |
+
},
|
| 233716 |
+
{
|
| 233717 |
+
"epoch": 844.53,
|
| 233718 |
+
"learning_rate": 8.324375000000001e-06,
|
| 233719 |
+
"loss": 0.3172,
|
| 233720 |
+
"step": 105600
|
| 233721 |
+
},
|
| 233722 |
+
{
|
| 233723 |
+
"epoch": 844.57,
|
| 233724 |
+
"learning_rate": 8.324294871794872e-06,
|
| 233725 |
+
"loss": 0.7529,
|
| 233726 |
+
"step": 105605
|
| 233727 |
+
},
|
| 233728 |
+
{
|
| 233729 |
+
"epoch": 844.61,
|
| 233730 |
+
"learning_rate": 8.324214743589745e-06,
|
| 233731 |
+
"loss": 0.8829,
|
| 233732 |
+
"step": 105610
|
| 233733 |
+
},
|
| 233734 |
+
{
|
| 233735 |
+
"epoch": 844.65,
|
| 233736 |
+
"learning_rate": 8.324134615384615e-06,
|
| 233737 |
+
"loss": 0.2851,
|
| 233738 |
+
"step": 105615
|
| 233739 |
+
},
|
| 233740 |
+
{
|
| 233741 |
+
"epoch": 844.69,
|
| 233742 |
+
"learning_rate": 8.324054487179488e-06,
|
| 233743 |
+
"loss": 0.3347,
|
| 233744 |
+
"step": 105620
|
| 233745 |
+
},
|
| 233746 |
+
{
|
| 233747 |
+
"epoch": 844.73,
|
| 233748 |
+
"learning_rate": 8.32397435897436e-06,
|
| 233749 |
+
"loss": 0.4114,
|
| 233750 |
+
"step": 105625
|
| 233751 |
+
},
|
| 233752 |
+
{
|
| 233753 |
+
"epoch": 844.77,
|
| 233754 |
+
"learning_rate": 8.323894230769231e-06,
|
| 233755 |
+
"loss": 0.563,
|
| 233756 |
+
"step": 105630
|
| 233757 |
+
},
|
| 233758 |
+
{
|
| 233759 |
+
"epoch": 844.81,
|
| 233760 |
+
"learning_rate": 8.323814102564102e-06,
|
| 233761 |
+
"loss": 1.0647,
|
| 233762 |
+
"step": 105635
|
| 233763 |
+
},
|
| 233764 |
+
{
|
| 233765 |
+
"epoch": 844.85,
|
| 233766 |
+
"learning_rate": 8.323733974358975e-06,
|
| 233767 |
+
"loss": 0.2941,
|
| 233768 |
+
"step": 105640
|
| 233769 |
+
},
|
| 233770 |
+
{
|
| 233771 |
+
"epoch": 844.89,
|
| 233772 |
+
"learning_rate": 8.323653846153847e-06,
|
| 233773 |
+
"loss": 0.2757,
|
| 233774 |
+
"step": 105645
|
| 233775 |
+
},
|
| 233776 |
+
{
|
| 233777 |
+
"epoch": 844.93,
|
| 233778 |
+
"learning_rate": 8.323573717948718e-06,
|
| 233779 |
+
"loss": 0.3461,
|
| 233780 |
+
"step": 105650
|
| 233781 |
+
},
|
| 233782 |
+
{
|
| 233783 |
+
"epoch": 844.97,
|
| 233784 |
+
"learning_rate": 8.323493589743591e-06,
|
| 233785 |
+
"loss": 0.704,
|
| 233786 |
+
"step": 105655
|
| 233787 |
+
},
|
| 233788 |
+
{
|
| 233789 |
+
"epoch": 845.0,
|
| 233790 |
+
"eval_loss": 0.4049234092235565,
|
| 233791 |
+
"eval_runtime": 38.6181,
|
| 233792 |
+
"eval_samples_per_second": 21.7,
|
| 233793 |
+
"eval_steps_per_second": 0.699,
|
| 233794 |
+
"eval_wer": 0.18607667915825887,
|
| 233795 |
+
"step": 105658
|
| 233796 |
+
},
|
| 233797 |
+
{
|
| 233798 |
+
"epoch": 845.02,
|
| 233799 |
+
"learning_rate": 8.323413461538462e-06,
|
| 233800 |
+
"loss": 0.5315,
|
| 233801 |
+
"step": 105660
|
| 233802 |
+
},
|
| 233803 |
+
{
|
| 233804 |
+
"epoch": 845.06,
|
| 233805 |
+
"learning_rate": 8.323333333333334e-06,
|
| 233806 |
+
"loss": 0.3053,
|
| 233807 |
+
"step": 105665
|
| 233808 |
+
},
|
| 233809 |
+
{
|
| 233810 |
+
"epoch": 845.1,
|
| 233811 |
+
"learning_rate": 8.323253205128205e-06,
|
| 233812 |
+
"loss": 0.3493,
|
| 233813 |
+
"step": 105670
|
| 233814 |
+
},
|
| 233815 |
+
{
|
| 233816 |
+
"epoch": 845.14,
|
| 233817 |
+
"learning_rate": 8.323173076923078e-06,
|
| 233818 |
+
"loss": 0.4111,
|
| 233819 |
+
"step": 105675
|
| 233820 |
+
},
|
| 233821 |
+
{
|
| 233822 |
+
"epoch": 845.18,
|
| 233823 |
+
"learning_rate": 8.32309294871795e-06,
|
| 233824 |
+
"loss": 0.6999,
|
| 233825 |
+
"step": 105680
|
| 233826 |
+
},
|
| 233827 |
+
{
|
| 233828 |
+
"epoch": 845.22,
|
| 233829 |
+
"learning_rate": 8.323012820512821e-06,
|
| 233830 |
+
"loss": 1.0329,
|
| 233831 |
+
"step": 105685
|
| 233832 |
+
},
|
| 233833 |
+
{
|
| 233834 |
+
"epoch": 845.26,
|
| 233835 |
+
"learning_rate": 8.322932692307692e-06,
|
| 233836 |
+
"loss": 0.2908,
|
| 233837 |
+
"step": 105690
|
| 233838 |
+
},
|
| 233839 |
+
{
|
| 233840 |
+
"epoch": 845.3,
|
| 233841 |
+
"learning_rate": 8.322852564102565e-06,
|
| 233842 |
+
"loss": 0.3494,
|
| 233843 |
+
"step": 105695
|
| 233844 |
+
},
|
| 233845 |
+
{
|
| 233846 |
+
"epoch": 845.34,
|
| 233847 |
+
"learning_rate": 8.322772435897437e-06,
|
| 233848 |
+
"loss": 0.3988,
|
| 233849 |
+
"step": 105700
|
| 233850 |
+
},
|
| 233851 |
+
{
|
| 233852 |
+
"epoch": 845.38,
|
| 233853 |
+
"learning_rate": 8.322692307692308e-06,
|
| 233854 |
+
"loss": 0.6452,
|
| 233855 |
+
"step": 105705
|
| 233856 |
+
},
|
| 233857 |
+
{
|
| 233858 |
+
"epoch": 845.42,
|
| 233859 |
+
"learning_rate": 8.322612179487181e-06,
|
| 233860 |
+
"loss": 1.091,
|
| 233861 |
+
"step": 105710
|
| 233862 |
+
},
|
| 233863 |
+
{
|
| 233864 |
+
"epoch": 845.46,
|
| 233865 |
+
"learning_rate": 8.322532051282052e-06,
|
| 233866 |
+
"loss": 0.2905,
|
| 233867 |
+
"step": 105715
|
| 233868 |
+
},
|
| 233869 |
+
{
|
| 233870 |
+
"epoch": 845.5,
|
| 233871 |
+
"learning_rate": 8.322451923076924e-06,
|
| 233872 |
+
"loss": 0.3066,
|
| 233873 |
+
"step": 105720
|
| 233874 |
+
},
|
| 233875 |
+
{
|
| 233876 |
+
"epoch": 845.54,
|
| 233877 |
+
"learning_rate": 8.322371794871795e-06,
|
| 233878 |
+
"loss": 0.4366,
|
| 233879 |
+
"step": 105725
|
| 233880 |
+
},
|
| 233881 |
+
{
|
| 233882 |
+
"epoch": 845.58,
|
| 233883 |
+
"learning_rate": 8.322291666666668e-06,
|
| 233884 |
+
"loss": 0.7364,
|
| 233885 |
+
"step": 105730
|
| 233886 |
+
},
|
| 233887 |
+
{
|
| 233888 |
+
"epoch": 845.62,
|
| 233889 |
+
"learning_rate": 8.322211538461538e-06,
|
| 233890 |
+
"loss": 1.1455,
|
| 233891 |
+
"step": 105735
|
| 233892 |
+
},
|
| 233893 |
+
{
|
| 233894 |
+
"epoch": 845.66,
|
| 233895 |
+
"learning_rate": 8.322131410256411e-06,
|
| 233896 |
+
"loss": 0.3331,
|
| 233897 |
+
"step": 105740
|
| 233898 |
+
},
|
| 233899 |
+
{
|
| 233900 |
+
"epoch": 845.7,
|
| 233901 |
+
"learning_rate": 8.322051282051282e-06,
|
| 233902 |
+
"loss": 0.2754,
|
| 233903 |
+
"step": 105745
|
| 233904 |
+
},
|
| 233905 |
+
{
|
| 233906 |
+
"epoch": 845.74,
|
| 233907 |
+
"learning_rate": 8.321971153846154e-06,
|
| 233908 |
+
"loss": 0.3058,
|
| 233909 |
+
"step": 105750
|
| 233910 |
+
},
|
| 233911 |
+
{
|
| 233912 |
+
"epoch": 845.78,
|
| 233913 |
+
"learning_rate": 8.321891025641027e-06,
|
| 233914 |
+
"loss": 0.5798,
|
| 233915 |
+
"step": 105755
|
| 233916 |
+
},
|
| 233917 |
+
{
|
| 233918 |
+
"epoch": 845.82,
|
| 233919 |
+
"learning_rate": 8.321810897435898e-06,
|
| 233920 |
+
"loss": 1.0449,
|
| 233921 |
+
"step": 105760
|
| 233922 |
+
},
|
| 233923 |
+
{
|
| 233924 |
+
"epoch": 845.86,
|
| 233925 |
+
"learning_rate": 8.32173076923077e-06,
|
| 233926 |
+
"loss": 0.2771,
|
| 233927 |
+
"step": 105765
|
| 233928 |
+
},
|
| 233929 |
+
{
|
| 233930 |
+
"epoch": 845.9,
|
| 233931 |
+
"learning_rate": 8.32165064102564e-06,
|
| 233932 |
+
"loss": 0.3172,
|
| 233933 |
+
"step": 105770
|
| 233934 |
+
},
|
| 233935 |
+
{
|
| 233936 |
+
"epoch": 845.94,
|
| 233937 |
+
"learning_rate": 8.321570512820514e-06,
|
| 233938 |
+
"loss": 0.3653,
|
| 233939 |
+
"step": 105775
|
| 233940 |
+
},
|
| 233941 |
+
{
|
| 233942 |
+
"epoch": 845.98,
|
| 233943 |
+
"learning_rate": 8.321490384615385e-06,
|
| 233944 |
+
"loss": 0.6452,
|
| 233945 |
+
"step": 105780
|
| 233946 |
+
},
|
| 233947 |
+
{
|
| 233948 |
+
"epoch": 846.0,
|
| 233949 |
+
"eval_loss": 0.3681880235671997,
|
| 233950 |
+
"eval_runtime": 39.1408,
|
| 233951 |
+
"eval_samples_per_second": 21.41,
|
| 233952 |
+
"eval_steps_per_second": 0.69,
|
| 233953 |
+
"eval_wer": 0.18124235226372992,
|
| 233954 |
+
"step": 105783
|
| 233955 |
}
|
| 233956 |
],
|
| 233957 |
+
"max_steps": 625000,
|
| 233958 |
"num_train_epochs": 5000,
|
| 233959 |
+
"total_flos": 2.9770259394097886e+20,
|
| 233960 |
"trial_name": null,
|
| 233961 |
"trial_params": null
|
| 233962 |
}
|
model-bin/finetune/base/{checkpoint-105160 → checkpoint-105783}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630000221.7268653/events.out.tfevents.1630000221.8e89bd551565.924.271
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de8590cf1a60b7b983dfbebe36cbe9c89f7d652eb90fa9feb3fd922007e1eef2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630000701.8113801/events.out.tfevents.1630000703.8e89bd551565.924.273
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:edfc1f26e93416e2536a1f571524d918b2b52615fc6271822bb7999495890b6f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630001150.9827878/events.out.tfevents.1630001150.8e89bd551565.924.275
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17b6c77d04ff1c1a252edbd0ec5f579bf649f651409a2abda40fff0eea615d3a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630001578.948476/events.out.tfevents.1630001578.8e89bd551565.924.277
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aae2c414788bf0284c9058553afd246cf9cc918b365a089b3c88a139ccd97883
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630002016.2967117/events.out.tfevents.1630002017.8e89bd551565.924.279
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad563be89574cfe2763bde0cb81ab070e2b400a78dcd5362b097c8233dec9fd2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630000221.8e89bd551565.924.270
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca9e391cde92bdf6ffd0f5f9eed7ca06c03194c5d4352cea56a32032868b5e81
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630000701.8e89bd551565.924.272
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b45e2b3b95ddb5700e22470219f5fe9b474a6a134bdafe0139d92286d8e3a65
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630001150.8e89bd551565.924.274
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:493ba05db50fbdc6e71e71d89cdd966ab45117bad4e094c433ebddd76dc13020
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630001578.8e89bd551565.924.276
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f359ad2658b2ce5030209ab99330c89226381e665ab83c75b530a53fde81f996
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630002016.8e89bd551565.924.278
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e413bf04f73efa1c15e7facc960859f5e0d7ee0b37f0f738e56c6f20ab97a1d1
|
| 3 |
+
size 8622
|