"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629680139.8226664/events.out.tfevents.1629680139.74272264b15c.932.3 +3 -0
- model-bin/finetune/base/log/1629680780.9958236/events.out.tfevents.1629680780.74272264b15c.932.5 +3 -0
- model-bin/finetune/base/log/1629681427.7007234/events.out.tfevents.1629681427.74272264b15c.932.7 +3 -0
- model-bin/finetune/base/log/1629682093.6511927/events.out.tfevents.1629682093.74272264b15c.932.9 +3 -0
- model-bin/finetune/base/log/1629682736.1880279/events.out.tfevents.1629682736.74272264b15c.932.11 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629680139.74272264b15c.932.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629680780.74272264b15c.932.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629681427.74272264b15c.932.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629682093.74272264b15c.932.8 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629682736.74272264b15c.932.10 +3 -0
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/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 sha256:822638e01a87cd770d6f6839186be2fc4b03d6b493ded86c346043a3998b4854
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/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:e672e755ab6f3247a912e80195f727806aee27380d58d38495af59037433c06c
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/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 sha256:3d77b0580d22cbbfed1d4f8264a1eea8ff650b29d6bcff5208e45aa4314872ae
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/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:d79845aa663507cb641579ac08c63c5b4e38cb856208f89318263ecf59d68d89
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/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 sha256:532b99c9ff992c908f957657ae13aa205a5a5b662826c0078ada5f250aa9dd90
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -148128,11 +148128,806 @@
|
|
| 148128 |
"eval_steps_per_second": 0.711,
|
| 148129 |
"eval_wer": 0.20721445818394674,
|
| 148130 |
"step": 38327
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 148131 |
}
|
| 148132 |
],
|
| 148133 |
"max_steps": 625000,
|
| 148134 |
"num_train_epochs": 5000,
|
| 148135 |
-
"total_flos": 1.
|
| 148136 |
"trial_name": null,
|
| 148137 |
"trial_params": null
|
| 148138 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 311.0,
|
| 5 |
+
"global_step": 38950,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 148128 |
"eval_steps_per_second": 0.711,
|
| 148129 |
"eval_wer": 0.20721445818394674,
|
| 148130 |
"step": 38327
|
| 148131 |
+
},
|
| 148132 |
+
{
|
| 148133 |
+
"epoch": 306.02,
|
| 148134 |
+
"learning_rate": 9.402003205128206e-06,
|
| 148135 |
+
"loss": 0.4559,
|
| 148136 |
+
"step": 38330
|
| 148137 |
+
},
|
| 148138 |
+
{
|
| 148139 |
+
"epoch": 306.06,
|
| 148140 |
+
"learning_rate": 9.401923076923077e-06,
|
| 148141 |
+
"loss": 0.3937,
|
| 148142 |
+
"step": 38335
|
| 148143 |
+
},
|
| 148144 |
+
{
|
| 148145 |
+
"epoch": 306.1,
|
| 148146 |
+
"learning_rate": 9.40184294871795e-06,
|
| 148147 |
+
"loss": 0.347,
|
| 148148 |
+
"step": 38340
|
| 148149 |
+
},
|
| 148150 |
+
{
|
| 148151 |
+
"epoch": 306.14,
|
| 148152 |
+
"learning_rate": 9.401762820512821e-06,
|
| 148153 |
+
"loss": 0.4612,
|
| 148154 |
+
"step": 38345
|
| 148155 |
+
},
|
| 148156 |
+
{
|
| 148157 |
+
"epoch": 306.18,
|
| 148158 |
+
"learning_rate": 9.401682692307693e-06,
|
| 148159 |
+
"loss": 0.7785,
|
| 148160 |
+
"step": 38350
|
| 148161 |
+
},
|
| 148162 |
+
{
|
| 148163 |
+
"epoch": 306.22,
|
| 148164 |
+
"learning_rate": 9.401602564102564e-06,
|
| 148165 |
+
"loss": 1.0325,
|
| 148166 |
+
"step": 38355
|
| 148167 |
+
},
|
| 148168 |
+
{
|
| 148169 |
+
"epoch": 306.26,
|
| 148170 |
+
"learning_rate": 9.401522435897437e-06,
|
| 148171 |
+
"loss": 0.3868,
|
| 148172 |
+
"step": 38360
|
| 148173 |
+
},
|
| 148174 |
+
{
|
| 148175 |
+
"epoch": 306.3,
|
| 148176 |
+
"learning_rate": 9.401442307692309e-06,
|
| 148177 |
+
"loss": 0.3405,
|
| 148178 |
+
"step": 38365
|
| 148179 |
+
},
|
| 148180 |
+
{
|
| 148181 |
+
"epoch": 306.34,
|
| 148182 |
+
"learning_rate": 9.40136217948718e-06,
|
| 148183 |
+
"loss": 0.463,
|
| 148184 |
+
"step": 38370
|
| 148185 |
+
},
|
| 148186 |
+
{
|
| 148187 |
+
"epoch": 306.38,
|
| 148188 |
+
"learning_rate": 9.401282051282053e-06,
|
| 148189 |
+
"loss": 0.9342,
|
| 148190 |
+
"step": 38375
|
| 148191 |
+
},
|
| 148192 |
+
{
|
| 148193 |
+
"epoch": 306.42,
|
| 148194 |
+
"learning_rate": 9.401201923076924e-06,
|
| 148195 |
+
"loss": 0.9508,
|
| 148196 |
+
"step": 38380
|
| 148197 |
+
},
|
| 148198 |
+
{
|
| 148199 |
+
"epoch": 306.46,
|
| 148200 |
+
"learning_rate": 9.401121794871796e-06,
|
| 148201 |
+
"loss": 0.9046,
|
| 148202 |
+
"step": 38385
|
| 148203 |
+
},
|
| 148204 |
+
{
|
| 148205 |
+
"epoch": 306.5,
|
| 148206 |
+
"learning_rate": 9.401041666666667e-06,
|
| 148207 |
+
"loss": 0.3943,
|
| 148208 |
+
"step": 38390
|
| 148209 |
+
},
|
| 148210 |
+
{
|
| 148211 |
+
"epoch": 306.54,
|
| 148212 |
+
"learning_rate": 9.40096153846154e-06,
|
| 148213 |
+
"loss": 0.4688,
|
| 148214 |
+
"step": 38395
|
| 148215 |
+
},
|
| 148216 |
+
{
|
| 148217 |
+
"epoch": 306.58,
|
| 148218 |
+
"learning_rate": 9.40088141025641e-06,
|
| 148219 |
+
"loss": 0.7631,
|
| 148220 |
+
"step": 38400
|
| 148221 |
+
},
|
| 148222 |
+
{
|
| 148223 |
+
"epoch": 306.62,
|
| 148224 |
+
"learning_rate": 9.400801282051283e-06,
|
| 148225 |
+
"loss": 1.0508,
|
| 148226 |
+
"step": 38405
|
| 148227 |
+
},
|
| 148228 |
+
{
|
| 148229 |
+
"epoch": 306.66,
|
| 148230 |
+
"learning_rate": 9.400721153846156e-06,
|
| 148231 |
+
"loss": 0.4309,
|
| 148232 |
+
"step": 38410
|
| 148233 |
+
},
|
| 148234 |
+
{
|
| 148235 |
+
"epoch": 306.7,
|
| 148236 |
+
"learning_rate": 9.400641025641026e-06,
|
| 148237 |
+
"loss": 0.3868,
|
| 148238 |
+
"step": 38415
|
| 148239 |
+
},
|
| 148240 |
+
{
|
| 148241 |
+
"epoch": 306.74,
|
| 148242 |
+
"learning_rate": 9.400560897435899e-06,
|
| 148243 |
+
"loss": 0.4163,
|
| 148244 |
+
"step": 38420
|
| 148245 |
+
},
|
| 148246 |
+
{
|
| 148247 |
+
"epoch": 306.78,
|
| 148248 |
+
"learning_rate": 9.40048076923077e-06,
|
| 148249 |
+
"loss": 1.0068,
|
| 148250 |
+
"step": 38425
|
| 148251 |
+
},
|
| 148252 |
+
{
|
| 148253 |
+
"epoch": 306.82,
|
| 148254 |
+
"learning_rate": 9.400400641025641e-06,
|
| 148255 |
+
"loss": 0.9118,
|
| 148256 |
+
"step": 38430
|
| 148257 |
+
},
|
| 148258 |
+
{
|
| 148259 |
+
"epoch": 306.86,
|
| 148260 |
+
"learning_rate": 9.400320512820513e-06,
|
| 148261 |
+
"loss": 0.3929,
|
| 148262 |
+
"step": 38435
|
| 148263 |
+
},
|
| 148264 |
+
{
|
| 148265 |
+
"epoch": 306.9,
|
| 148266 |
+
"learning_rate": 9.400240384615386e-06,
|
| 148267 |
+
"loss": 0.4526,
|
| 148268 |
+
"step": 38440
|
| 148269 |
+
},
|
| 148270 |
+
{
|
| 148271 |
+
"epoch": 306.94,
|
| 148272 |
+
"learning_rate": 9.400160256410257e-06,
|
| 148273 |
+
"loss": 0.44,
|
| 148274 |
+
"step": 38445
|
| 148275 |
+
},
|
| 148276 |
+
{
|
| 148277 |
+
"epoch": 306.98,
|
| 148278 |
+
"learning_rate": 9.400080128205128e-06,
|
| 148279 |
+
"loss": 0.9921,
|
| 148280 |
+
"step": 38450
|
| 148281 |
+
},
|
| 148282 |
+
{
|
| 148283 |
+
"epoch": 307.0,
|
| 148284 |
+
"eval_loss": 0.5049299597740173,
|
| 148285 |
+
"eval_runtime": 39.2095,
|
| 148286 |
+
"eval_samples_per_second": 21.423,
|
| 148287 |
+
"eval_steps_per_second": 0.689,
|
| 148288 |
+
"eval_wer": 0.21908471275559882,
|
| 148289 |
+
"step": 38452
|
| 148290 |
+
},
|
| 148291 |
+
{
|
| 148292 |
+
"epoch": 307.02,
|
| 148293 |
+
"learning_rate": 9.4e-06,
|
| 148294 |
+
"loss": 0.3909,
|
| 148295 |
+
"step": 38455
|
| 148296 |
+
},
|
| 148297 |
+
{
|
| 148298 |
+
"epoch": 307.06,
|
| 148299 |
+
"learning_rate": 9.399919871794873e-06,
|
| 148300 |
+
"loss": 0.3909,
|
| 148301 |
+
"step": 38460
|
| 148302 |
+
},
|
| 148303 |
+
{
|
| 148304 |
+
"epoch": 307.1,
|
| 148305 |
+
"learning_rate": 9.399839743589744e-06,
|
| 148306 |
+
"loss": 0.3718,
|
| 148307 |
+
"step": 38465
|
| 148308 |
+
},
|
| 148309 |
+
{
|
| 148310 |
+
"epoch": 307.14,
|
| 148311 |
+
"learning_rate": 9.399759615384616e-06,
|
| 148312 |
+
"loss": 0.4126,
|
| 148313 |
+
"step": 38470
|
| 148314 |
+
},
|
| 148315 |
+
{
|
| 148316 |
+
"epoch": 307.18,
|
| 148317 |
+
"learning_rate": 9.399679487179489e-06,
|
| 148318 |
+
"loss": 0.8313,
|
| 148319 |
+
"step": 38475
|
| 148320 |
+
},
|
| 148321 |
+
{
|
| 148322 |
+
"epoch": 307.22,
|
| 148323 |
+
"learning_rate": 9.39959935897436e-06,
|
| 148324 |
+
"loss": 1.0761,
|
| 148325 |
+
"step": 38480
|
| 148326 |
+
},
|
| 148327 |
+
{
|
| 148328 |
+
"epoch": 307.26,
|
| 148329 |
+
"learning_rate": 9.399519230769231e-06,
|
| 148330 |
+
"loss": 0.3499,
|
| 148331 |
+
"step": 38485
|
| 148332 |
+
},
|
| 148333 |
+
{
|
| 148334 |
+
"epoch": 307.3,
|
| 148335 |
+
"learning_rate": 9.399439102564103e-06,
|
| 148336 |
+
"loss": 0.3935,
|
| 148337 |
+
"step": 38490
|
| 148338 |
+
},
|
| 148339 |
+
{
|
| 148340 |
+
"epoch": 307.34,
|
| 148341 |
+
"learning_rate": 9.399358974358976e-06,
|
| 148342 |
+
"loss": 0.5196,
|
| 148343 |
+
"step": 38495
|
| 148344 |
+
},
|
| 148345 |
+
{
|
| 148346 |
+
"epoch": 307.38,
|
| 148347 |
+
"learning_rate": 9.399278846153847e-06,
|
| 148348 |
+
"loss": 0.992,
|
| 148349 |
+
"step": 38500
|
| 148350 |
+
},
|
| 148351 |
+
{
|
| 148352 |
+
"epoch": 307.42,
|
| 148353 |
+
"learning_rate": 9.399198717948718e-06,
|
| 148354 |
+
"loss": 0.9763,
|
| 148355 |
+
"step": 38505
|
| 148356 |
+
},
|
| 148357 |
+
{
|
| 148358 |
+
"epoch": 307.46,
|
| 148359 |
+
"learning_rate": 9.399118589743592e-06,
|
| 148360 |
+
"loss": 0.3615,
|
| 148361 |
+
"step": 38510
|
| 148362 |
+
},
|
| 148363 |
+
{
|
| 148364 |
+
"epoch": 307.5,
|
| 148365 |
+
"learning_rate": 9.399038461538463e-06,
|
| 148366 |
+
"loss": 0.3285,
|
| 148367 |
+
"step": 38515
|
| 148368 |
+
},
|
| 148369 |
+
{
|
| 148370 |
+
"epoch": 307.54,
|
| 148371 |
+
"learning_rate": 9.398958333333334e-06,
|
| 148372 |
+
"loss": 0.5792,
|
| 148373 |
+
"step": 38520
|
| 148374 |
+
},
|
| 148375 |
+
{
|
| 148376 |
+
"epoch": 307.58,
|
| 148377 |
+
"learning_rate": 9.398878205128206e-06,
|
| 148378 |
+
"loss": 0.8291,
|
| 148379 |
+
"step": 38525
|
| 148380 |
+
},
|
| 148381 |
+
{
|
| 148382 |
+
"epoch": 307.62,
|
| 148383 |
+
"learning_rate": 9.398798076923079e-06,
|
| 148384 |
+
"loss": 0.9546,
|
| 148385 |
+
"step": 38530
|
| 148386 |
+
},
|
| 148387 |
+
{
|
| 148388 |
+
"epoch": 307.66,
|
| 148389 |
+
"learning_rate": 9.39871794871795e-06,
|
| 148390 |
+
"loss": 0.3678,
|
| 148391 |
+
"step": 38535
|
| 148392 |
+
},
|
| 148393 |
+
{
|
| 148394 |
+
"epoch": 307.7,
|
| 148395 |
+
"learning_rate": 9.398637820512821e-06,
|
| 148396 |
+
"loss": 0.355,
|
| 148397 |
+
"step": 38540
|
| 148398 |
+
},
|
| 148399 |
+
{
|
| 148400 |
+
"epoch": 307.74,
|
| 148401 |
+
"learning_rate": 9.398557692307693e-06,
|
| 148402 |
+
"loss": 0.4718,
|
| 148403 |
+
"step": 38545
|
| 148404 |
+
},
|
| 148405 |
+
{
|
| 148406 |
+
"epoch": 307.78,
|
| 148407 |
+
"learning_rate": 9.398477564102566e-06,
|
| 148408 |
+
"loss": 1.0717,
|
| 148409 |
+
"step": 38550
|
| 148410 |
+
},
|
| 148411 |
+
{
|
| 148412 |
+
"epoch": 307.82,
|
| 148413 |
+
"learning_rate": 9.398397435897435e-06,
|
| 148414 |
+
"loss": 0.9421,
|
| 148415 |
+
"step": 38555
|
| 148416 |
+
},
|
| 148417 |
+
{
|
| 148418 |
+
"epoch": 307.86,
|
| 148419 |
+
"learning_rate": 9.398317307692308e-06,
|
| 148420 |
+
"loss": 0.3596,
|
| 148421 |
+
"step": 38560
|
| 148422 |
+
},
|
| 148423 |
+
{
|
| 148424 |
+
"epoch": 307.9,
|
| 148425 |
+
"learning_rate": 9.398237179487182e-06,
|
| 148426 |
+
"loss": 0.356,
|
| 148427 |
+
"step": 38565
|
| 148428 |
+
},
|
| 148429 |
+
{
|
| 148430 |
+
"epoch": 307.94,
|
| 148431 |
+
"learning_rate": 9.398157051282051e-06,
|
| 148432 |
+
"loss": 0.4294,
|
| 148433 |
+
"step": 38570
|
| 148434 |
+
},
|
| 148435 |
+
{
|
| 148436 |
+
"epoch": 307.98,
|
| 148437 |
+
"learning_rate": 9.398076923076924e-06,
|
| 148438 |
+
"loss": 0.8726,
|
| 148439 |
+
"step": 38575
|
| 148440 |
+
},
|
| 148441 |
+
{
|
| 148442 |
+
"epoch": 308.0,
|
| 148443 |
+
"eval_loss": 0.375550776720047,
|
| 148444 |
+
"eval_runtime": 39.839,
|
| 148445 |
+
"eval_samples_per_second": 21.085,
|
| 148446 |
+
"eval_steps_per_second": 0.678,
|
| 148447 |
+
"eval_wer": 0.20714589536367503,
|
| 148448 |
+
"step": 38577
|
| 148449 |
+
},
|
| 148450 |
+
{
|
| 148451 |
+
"epoch": 311.02,
|
| 148452 |
+
"learning_rate": 9.397996794871796e-06,
|
| 148453 |
+
"loss": 0.4375,
|
| 148454 |
+
"step": 38580
|
| 148455 |
+
},
|
| 148456 |
+
{
|
| 148457 |
+
"epoch": 311.06,
|
| 148458 |
+
"learning_rate": 9.397916666666667e-06,
|
| 148459 |
+
"loss": 0.3489,
|
| 148460 |
+
"step": 38585
|
| 148461 |
+
},
|
| 148462 |
+
{
|
| 148463 |
+
"epoch": 311.1,
|
| 148464 |
+
"learning_rate": 9.397836538461538e-06,
|
| 148465 |
+
"loss": 0.3713,
|
| 148466 |
+
"step": 38590
|
| 148467 |
+
},
|
| 148468 |
+
{
|
| 148469 |
+
"epoch": 311.14,
|
| 148470 |
+
"learning_rate": 9.397756410256411e-06,
|
| 148471 |
+
"loss": 0.387,
|
| 148472 |
+
"step": 38595
|
| 148473 |
+
},
|
| 148474 |
+
{
|
| 148475 |
+
"epoch": 311.18,
|
| 148476 |
+
"learning_rate": 9.397676282051283e-06,
|
| 148477 |
+
"loss": 0.8715,
|
| 148478 |
+
"step": 38600
|
| 148479 |
+
},
|
| 148480 |
+
{
|
| 148481 |
+
"epoch": 311.22,
|
| 148482 |
+
"learning_rate": 9.397596153846154e-06,
|
| 148483 |
+
"loss": 0.8938,
|
| 148484 |
+
"step": 38605
|
| 148485 |
+
},
|
| 148486 |
+
{
|
| 148487 |
+
"epoch": 311.27,
|
| 148488 |
+
"learning_rate": 9.397516025641027e-06,
|
| 148489 |
+
"loss": 0.3631,
|
| 148490 |
+
"step": 38610
|
| 148491 |
+
},
|
| 148492 |
+
{
|
| 148493 |
+
"epoch": 311.31,
|
| 148494 |
+
"learning_rate": 9.397435897435899e-06,
|
| 148495 |
+
"loss": 0.3896,
|
| 148496 |
+
"step": 38615
|
| 148497 |
+
},
|
| 148498 |
+
{
|
| 148499 |
+
"epoch": 311.35,
|
| 148500 |
+
"learning_rate": 9.39735576923077e-06,
|
| 148501 |
+
"loss": 0.4737,
|
| 148502 |
+
"step": 38620
|
| 148503 |
+
},
|
| 148504 |
+
{
|
| 148505 |
+
"epoch": 311.39,
|
| 148506 |
+
"learning_rate": 9.397275641025641e-06,
|
| 148507 |
+
"loss": 1.0259,
|
| 148508 |
+
"step": 38625
|
| 148509 |
+
},
|
| 148510 |
+
{
|
| 148511 |
+
"epoch": 311.43,
|
| 148512 |
+
"learning_rate": 9.397195512820514e-06,
|
| 148513 |
+
"loss": 1.0104,
|
| 148514 |
+
"step": 38630
|
| 148515 |
+
},
|
| 148516 |
+
{
|
| 148517 |
+
"epoch": 311.47,
|
| 148518 |
+
"learning_rate": 9.397115384615386e-06,
|
| 148519 |
+
"loss": 0.3973,
|
| 148520 |
+
"step": 38635
|
| 148521 |
+
},
|
| 148522 |
+
{
|
| 148523 |
+
"epoch": 311.51,
|
| 148524 |
+
"learning_rate": 9.397035256410257e-06,
|
| 148525 |
+
"loss": 0.345,
|
| 148526 |
+
"step": 38640
|
| 148527 |
+
},
|
| 148528 |
+
{
|
| 148529 |
+
"epoch": 311.55,
|
| 148530 |
+
"learning_rate": 9.396955128205128e-06,
|
| 148531 |
+
"loss": 0.4207,
|
| 148532 |
+
"step": 38645
|
| 148533 |
+
},
|
| 148534 |
+
{
|
| 148535 |
+
"epoch": 311.59,
|
| 148536 |
+
"learning_rate": 9.396875000000001e-06,
|
| 148537 |
+
"loss": 0.8608,
|
| 148538 |
+
"step": 38650
|
| 148539 |
+
},
|
| 148540 |
+
{
|
| 148541 |
+
"epoch": 311.63,
|
| 148542 |
+
"learning_rate": 9.396794871794873e-06,
|
| 148543 |
+
"loss": 1.1728,
|
| 148544 |
+
"step": 38655
|
| 148545 |
+
},
|
| 148546 |
+
{
|
| 148547 |
+
"epoch": 311.67,
|
| 148548 |
+
"learning_rate": 9.396714743589744e-06,
|
| 148549 |
+
"loss": 0.4259,
|
| 148550 |
+
"step": 38660
|
| 148551 |
+
},
|
| 148552 |
+
{
|
| 148553 |
+
"epoch": 311.71,
|
| 148554 |
+
"learning_rate": 9.396634615384617e-06,
|
| 148555 |
+
"loss": 0.3846,
|
| 148556 |
+
"step": 38665
|
| 148557 |
+
},
|
| 148558 |
+
{
|
| 148559 |
+
"epoch": 311.75,
|
| 148560 |
+
"learning_rate": 9.396554487179489e-06,
|
| 148561 |
+
"loss": 0.4341,
|
| 148562 |
+
"step": 38670
|
| 148563 |
+
},
|
| 148564 |
+
{
|
| 148565 |
+
"epoch": 311.79,
|
| 148566 |
+
"learning_rate": 9.39647435897436e-06,
|
| 148567 |
+
"loss": 0.892,
|
| 148568 |
+
"step": 38675
|
| 148569 |
+
},
|
| 148570 |
+
{
|
| 148571 |
+
"epoch": 311.83,
|
| 148572 |
+
"learning_rate": 9.396394230769231e-06,
|
| 148573 |
+
"loss": 0.8414,
|
| 148574 |
+
"step": 38680
|
| 148575 |
+
},
|
| 148576 |
+
{
|
| 148577 |
+
"epoch": 311.87,
|
| 148578 |
+
"learning_rate": 9.396314102564104e-06,
|
| 148579 |
+
"loss": 0.4098,
|
| 148580 |
+
"step": 38685
|
| 148581 |
+
},
|
| 148582 |
+
{
|
| 148583 |
+
"epoch": 311.91,
|
| 148584 |
+
"learning_rate": 9.396233974358974e-06,
|
| 148585 |
+
"loss": 0.423,
|
| 148586 |
+
"step": 38690
|
| 148587 |
+
},
|
| 148588 |
+
{
|
| 148589 |
+
"epoch": 311.95,
|
| 148590 |
+
"learning_rate": 9.396153846153847e-06,
|
| 148591 |
+
"loss": 0.4517,
|
| 148592 |
+
"step": 38695
|
| 148593 |
+
},
|
| 148594 |
+
{
|
| 148595 |
+
"epoch": 311.99,
|
| 148596 |
+
"learning_rate": 9.396073717948718e-06,
|
| 148597 |
+
"loss": 0.924,
|
| 148598 |
+
"step": 38700
|
| 148599 |
+
},
|
| 148600 |
+
{
|
| 148601 |
+
"epoch": 312.0,
|
| 148602 |
+
"eval_loss": 0.42753782868385315,
|
| 148603 |
+
"eval_runtime": 40.9125,
|
| 148604 |
+
"eval_samples_per_second": 20.532,
|
| 148605 |
+
"eval_steps_per_second": 0.66,
|
| 148606 |
+
"eval_wer": 0.2139098279049922,
|
| 148607 |
+
"step": 38701
|
| 148608 |
+
},
|
| 148609 |
+
{
|
| 148610 |
+
"epoch": 312.03,
|
| 148611 |
+
"learning_rate": 9.39599358974359e-06,
|
| 148612 |
+
"loss": 0.3239,
|
| 148613 |
+
"step": 38705
|
| 148614 |
+
},
|
| 148615 |
+
{
|
| 148616 |
+
"epoch": 312.07,
|
| 148617 |
+
"learning_rate": 9.395913461538463e-06,
|
| 148618 |
+
"loss": 0.4137,
|
| 148619 |
+
"step": 38710
|
| 148620 |
+
},
|
| 148621 |
+
{
|
| 148622 |
+
"epoch": 312.11,
|
| 148623 |
+
"learning_rate": 9.395833333333334e-06,
|
| 148624 |
+
"loss": 0.4107,
|
| 148625 |
+
"step": 38715
|
| 148626 |
+
},
|
| 148627 |
+
{
|
| 148628 |
+
"epoch": 312.15,
|
| 148629 |
+
"learning_rate": 9.395753205128206e-06,
|
| 148630 |
+
"loss": 0.4368,
|
| 148631 |
+
"step": 38720
|
| 148632 |
+
},
|
| 148633 |
+
{
|
| 148634 |
+
"epoch": 312.19,
|
| 148635 |
+
"learning_rate": 9.395673076923077e-06,
|
| 148636 |
+
"loss": 1.1172,
|
| 148637 |
+
"step": 38725
|
| 148638 |
+
},
|
| 148639 |
+
{
|
| 148640 |
+
"epoch": 312.23,
|
| 148641 |
+
"learning_rate": 9.39559294871795e-06,
|
| 148642 |
+
"loss": 0.7152,
|
| 148643 |
+
"step": 38730
|
| 148644 |
+
},
|
| 148645 |
+
{
|
| 148646 |
+
"epoch": 312.27,
|
| 148647 |
+
"learning_rate": 9.395512820512821e-06,
|
| 148648 |
+
"loss": 0.3885,
|
| 148649 |
+
"step": 38735
|
| 148650 |
+
},
|
| 148651 |
+
{
|
| 148652 |
+
"epoch": 312.31,
|
| 148653 |
+
"learning_rate": 9.395432692307693e-06,
|
| 148654 |
+
"loss": 0.4226,
|
| 148655 |
+
"step": 38740
|
| 148656 |
+
},
|
| 148657 |
+
{
|
| 148658 |
+
"epoch": 312.35,
|
| 148659 |
+
"learning_rate": 9.395352564102564e-06,
|
| 148660 |
+
"loss": 0.5501,
|
| 148661 |
+
"step": 38745
|
| 148662 |
+
},
|
| 148663 |
+
{
|
| 148664 |
+
"epoch": 312.39,
|
| 148665 |
+
"learning_rate": 9.395272435897437e-06,
|
| 148666 |
+
"loss": 1.1345,
|
| 148667 |
+
"step": 38750
|
| 148668 |
+
},
|
| 148669 |
+
{
|
| 148670 |
+
"epoch": 312.43,
|
| 148671 |
+
"learning_rate": 9.395192307692308e-06,
|
| 148672 |
+
"loss": 0.7812,
|
| 148673 |
+
"step": 38755
|
| 148674 |
+
},
|
| 148675 |
+
{
|
| 148676 |
+
"epoch": 312.47,
|
| 148677 |
+
"learning_rate": 9.39511217948718e-06,
|
| 148678 |
+
"loss": 0.313,
|
| 148679 |
+
"step": 38760
|
| 148680 |
+
},
|
| 148681 |
+
{
|
| 148682 |
+
"epoch": 312.51,
|
| 148683 |
+
"learning_rate": 9.395032051282053e-06,
|
| 148684 |
+
"loss": 0.349,
|
| 148685 |
+
"step": 38765
|
| 148686 |
+
},
|
| 148687 |
+
{
|
| 148688 |
+
"epoch": 312.55,
|
| 148689 |
+
"learning_rate": 9.394951923076924e-06,
|
| 148690 |
+
"loss": 0.5412,
|
| 148691 |
+
"step": 38770
|
| 148692 |
+
},
|
| 148693 |
+
{
|
| 148694 |
+
"epoch": 312.59,
|
| 148695 |
+
"learning_rate": 9.394871794871796e-06,
|
| 148696 |
+
"loss": 1.0449,
|
| 148697 |
+
"step": 38775
|
| 148698 |
+
},
|
| 148699 |
+
{
|
| 148700 |
+
"epoch": 312.63,
|
| 148701 |
+
"learning_rate": 9.394791666666667e-06,
|
| 148702 |
+
"loss": 1.0005,
|
| 148703 |
+
"step": 38780
|
| 148704 |
+
},
|
| 148705 |
+
{
|
| 148706 |
+
"epoch": 312.67,
|
| 148707 |
+
"learning_rate": 9.39471153846154e-06,
|
| 148708 |
+
"loss": 0.3296,
|
| 148709 |
+
"step": 38785
|
| 148710 |
+
},
|
| 148711 |
+
{
|
| 148712 |
+
"epoch": 312.71,
|
| 148713 |
+
"learning_rate": 9.394631410256411e-06,
|
| 148714 |
+
"loss": 0.4059,
|
| 148715 |
+
"step": 38790
|
| 148716 |
+
},
|
| 148717 |
+
{
|
| 148718 |
+
"epoch": 312.76,
|
| 148719 |
+
"learning_rate": 9.394551282051283e-06,
|
| 148720 |
+
"loss": 0.5688,
|
| 148721 |
+
"step": 38795
|
| 148722 |
+
},
|
| 148723 |
+
{
|
| 148724 |
+
"epoch": 312.8,
|
| 148725 |
+
"learning_rate": 9.394471153846154e-06,
|
| 148726 |
+
"loss": 1.0171,
|
| 148727 |
+
"step": 38800
|
| 148728 |
+
},
|
| 148729 |
+
{
|
| 148730 |
+
"epoch": 312.84,
|
| 148731 |
+
"learning_rate": 9.394391025641027e-06,
|
| 148732 |
+
"loss": 0.8489,
|
| 148733 |
+
"step": 38805
|
| 148734 |
+
},
|
| 148735 |
+
{
|
| 148736 |
+
"epoch": 312.88,
|
| 148737 |
+
"learning_rate": 9.394310897435898e-06,
|
| 148738 |
+
"loss": 0.4251,
|
| 148739 |
+
"step": 38810
|
| 148740 |
+
},
|
| 148741 |
+
{
|
| 148742 |
+
"epoch": 312.92,
|
| 148743 |
+
"learning_rate": 9.39423076923077e-06,
|
| 148744 |
+
"loss": 0.4275,
|
| 148745 |
+
"step": 38815
|
| 148746 |
+
},
|
| 148747 |
+
{
|
| 148748 |
+
"epoch": 312.96,
|
| 148749 |
+
"learning_rate": 9.394150641025643e-06,
|
| 148750 |
+
"loss": 0.5801,
|
| 148751 |
+
"step": 38820
|
| 148752 |
+
},
|
| 148753 |
+
{
|
| 148754 |
+
"epoch": 313.0,
|
| 148755 |
+
"learning_rate": 9.394070512820514e-06,
|
| 148756 |
+
"loss": 1.4023,
|
| 148757 |
+
"step": 38825
|
| 148758 |
+
},
|
| 148759 |
+
{
|
| 148760 |
+
"epoch": 313.0,
|
| 148761 |
+
"eval_loss": 0.5195120573043823,
|
| 148762 |
+
"eval_runtime": 41.0821,
|
| 148763 |
+
"eval_samples_per_second": 20.447,
|
| 148764 |
+
"eval_steps_per_second": 0.657,
|
| 148765 |
+
"eval_wer": 0.21609358383551933,
|
| 148766 |
+
"step": 38825
|
| 148767 |
+
},
|
| 148768 |
+
{
|
| 148769 |
+
"epoch": 310.04,
|
| 148770 |
+
"learning_rate": 9.393990384615386e-06,
|
| 148771 |
+
"loss": 0.4123,
|
| 148772 |
+
"step": 38830
|
| 148773 |
+
},
|
| 148774 |
+
{
|
| 148775 |
+
"epoch": 310.08,
|
| 148776 |
+
"learning_rate": 9.393910256410257e-06,
|
| 148777 |
+
"loss": 0.4073,
|
| 148778 |
+
"step": 38835
|
| 148779 |
+
},
|
| 148780 |
+
{
|
| 148781 |
+
"epoch": 310.12,
|
| 148782 |
+
"learning_rate": 9.39383012820513e-06,
|
| 148783 |
+
"loss": 0.3939,
|
| 148784 |
+
"step": 38840
|
| 148785 |
+
},
|
| 148786 |
+
{
|
| 148787 |
+
"epoch": 310.16,
|
| 148788 |
+
"learning_rate": 9.39375e-06,
|
| 148789 |
+
"loss": 0.615,
|
| 148790 |
+
"step": 38845
|
| 148791 |
+
},
|
| 148792 |
+
{
|
| 148793 |
+
"epoch": 310.2,
|
| 148794 |
+
"learning_rate": 9.393669871794873e-06,
|
| 148795 |
+
"loss": 1.2968,
|
| 148796 |
+
"step": 38850
|
| 148797 |
+
},
|
| 148798 |
+
{
|
| 148799 |
+
"epoch": 310.24,
|
| 148800 |
+
"learning_rate": 9.393589743589744e-06,
|
| 148801 |
+
"loss": 0.4979,
|
| 148802 |
+
"step": 38855
|
| 148803 |
+
},
|
| 148804 |
+
{
|
| 148805 |
+
"epoch": 310.28,
|
| 148806 |
+
"learning_rate": 9.393509615384615e-06,
|
| 148807 |
+
"loss": 0.3586,
|
| 148808 |
+
"step": 38860
|
| 148809 |
+
},
|
| 148810 |
+
{
|
| 148811 |
+
"epoch": 310.32,
|
| 148812 |
+
"learning_rate": 9.393429487179488e-06,
|
| 148813 |
+
"loss": 0.3935,
|
| 148814 |
+
"step": 38865
|
| 148815 |
+
},
|
| 148816 |
+
{
|
| 148817 |
+
"epoch": 310.36,
|
| 148818 |
+
"learning_rate": 9.39334935897436e-06,
|
| 148819 |
+
"loss": 0.6846,
|
| 148820 |
+
"step": 38870
|
| 148821 |
+
},
|
| 148822 |
+
{
|
| 148823 |
+
"epoch": 310.4,
|
| 148824 |
+
"learning_rate": 9.393269230769231e-06,
|
| 148825 |
+
"loss": 1.4301,
|
| 148826 |
+
"step": 38875
|
| 148827 |
+
},
|
| 148828 |
+
{
|
| 148829 |
+
"epoch": 310.44,
|
| 148830 |
+
"learning_rate": 9.393189102564103e-06,
|
| 148831 |
+
"loss": 0.4133,
|
| 148832 |
+
"step": 38880
|
| 148833 |
+
},
|
| 148834 |
+
{
|
| 148835 |
+
"epoch": 310.48,
|
| 148836 |
+
"learning_rate": 9.393108974358976e-06,
|
| 148837 |
+
"loss": 0.363,
|
| 148838 |
+
"step": 38885
|
| 148839 |
+
},
|
| 148840 |
+
{
|
| 148841 |
+
"epoch": 310.52,
|
| 148842 |
+
"learning_rate": 9.393028846153847e-06,
|
| 148843 |
+
"loss": 0.4275,
|
| 148844 |
+
"step": 38890
|
| 148845 |
+
},
|
| 148846 |
+
{
|
| 148847 |
+
"epoch": 310.56,
|
| 148848 |
+
"learning_rate": 9.392948717948718e-06,
|
| 148849 |
+
"loss": 0.5738,
|
| 148850 |
+
"step": 38895
|
| 148851 |
+
},
|
| 148852 |
+
{
|
| 148853 |
+
"epoch": 310.6,
|
| 148854 |
+
"learning_rate": 9.39286858974359e-06,
|
| 148855 |
+
"loss": 1.3592,
|
| 148856 |
+
"step": 38900
|
| 148857 |
+
},
|
| 148858 |
+
{
|
| 148859 |
+
"epoch": 310.64,
|
| 148860 |
+
"learning_rate": 9.392788461538463e-06,
|
| 148861 |
+
"loss": 0.3815,
|
| 148862 |
+
"step": 38905
|
| 148863 |
+
},
|
| 148864 |
+
{
|
| 148865 |
+
"epoch": 310.68,
|
| 148866 |
+
"learning_rate": 9.392708333333334e-06,
|
| 148867 |
+
"loss": 0.3217,
|
| 148868 |
+
"step": 38910
|
| 148869 |
+
},
|
| 148870 |
+
{
|
| 148871 |
+
"epoch": 310.72,
|
| 148872 |
+
"learning_rate": 9.392628205128205e-06,
|
| 148873 |
+
"loss": 0.3792,
|
| 148874 |
+
"step": 38915
|
| 148875 |
+
},
|
| 148876 |
+
{
|
| 148877 |
+
"epoch": 310.76,
|
| 148878 |
+
"learning_rate": 9.392548076923078e-06,
|
| 148879 |
+
"loss": 0.5473,
|
| 148880 |
+
"step": 38920
|
| 148881 |
+
},
|
| 148882 |
+
{
|
| 148883 |
+
"epoch": 310.8,
|
| 148884 |
+
"learning_rate": 9.39246794871795e-06,
|
| 148885 |
+
"loss": 1.1989,
|
| 148886 |
+
"step": 38925
|
| 148887 |
+
},
|
| 148888 |
+
{
|
| 148889 |
+
"epoch": 310.84,
|
| 148890 |
+
"learning_rate": 9.392387820512821e-06,
|
| 148891 |
+
"loss": 0.4254,
|
| 148892 |
+
"step": 38930
|
| 148893 |
+
},
|
| 148894 |
+
{
|
| 148895 |
+
"epoch": 310.88,
|
| 148896 |
+
"learning_rate": 9.392307692307693e-06,
|
| 148897 |
+
"loss": 0.3243,
|
| 148898 |
+
"step": 38935
|
| 148899 |
+
},
|
| 148900 |
+
{
|
| 148901 |
+
"epoch": 310.92,
|
| 148902 |
+
"learning_rate": 9.392227564102566e-06,
|
| 148903 |
+
"loss": 0.4262,
|
| 148904 |
+
"step": 38940
|
| 148905 |
+
},
|
| 148906 |
+
{
|
| 148907 |
+
"epoch": 310.96,
|
| 148908 |
+
"learning_rate": 9.392147435897437e-06,
|
| 148909 |
+
"loss": 0.5739,
|
| 148910 |
+
"step": 38945
|
| 148911 |
+
},
|
| 148912 |
+
{
|
| 148913 |
+
"epoch": 311.0,
|
| 148914 |
+
"learning_rate": 9.392067307692308e-06,
|
| 148915 |
+
"loss": 1.4891,
|
| 148916 |
+
"step": 38950
|
| 148917 |
+
},
|
| 148918 |
+
{
|
| 148919 |
+
"epoch": 311.0,
|
| 148920 |
+
"eval_loss": 0.442674458026886,
|
| 148921 |
+
"eval_runtime": 39.4417,
|
| 148922 |
+
"eval_samples_per_second": 21.272,
|
| 148923 |
+
"eval_steps_per_second": 0.685,
|
| 148924 |
+
"eval_wer": 0.20816059349770893,
|
| 148925 |
+
"step": 38950
|
| 148926 |
}
|
| 148927 |
],
|
| 148928 |
"max_steps": 625000,
|
| 148929 |
"num_train_epochs": 5000,
|
| 148930 |
+
"total_flos": 1.0960553907242182e+20,
|
| 148931 |
"trial_name": null,
|
| 148932 |
"trial_params": null
|
| 148933 |
}
|
model-bin/finetune/base/{checkpoint-38327 → checkpoint-38950}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629680139.8226664/events.out.tfevents.1629680139.74272264b15c.932.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05fce1b12b1f99b33c1bbec9df8618772b77eafe9921a9f041d4f3b19c4d26a4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629680780.9958236/events.out.tfevents.1629680780.74272264b15c.932.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1194bfed491701f77ee6e12814c7b111d4b4e512d770410db7d51ecdf6212fe5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629681427.7007234/events.out.tfevents.1629681427.74272264b15c.932.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d2ad955e4c062f18b19ab3b7588ca3430077807300038352af21423ea28de8f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629682093.6511927/events.out.tfevents.1629682093.74272264b15c.932.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de36c06c517cc2c3c1d5282197e75312eac6dbac914fa021f3a4a9beb8556ab5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629682736.1880279/events.out.tfevents.1629682736.74272264b15c.932.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6b7ec0dd200b2f6be6354e56b1686b006ff968d1af45efe7f3d22ffe2ba4d0c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629680139.74272264b15c.932.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:241536188f4bcdbaf5fde8f91200aaef683b60549eda7368c1234bb6646bfa8b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629680780.74272264b15c.932.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3cfb04b079ace78936590a5666fb8941af9febecd1346eaf462a6a5c70bde1da
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629681427.74272264b15c.932.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17f6b392608733baf1de8bfe2fb6859da93001748dfc5f93b205a20507678462
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629682093.74272264b15c.932.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a13b88c41c998386f4b0e4611aeeb5740238d7102d5a66019126a91dbd1b0b84
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629682736.74272264b15c.932.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1fbf881a5942ca0252801a6e573cf2ee10c9c18c7d0da077ea780655f37caf2e
|
| 3 |
+
size 8622
|