"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629745852.2460718/events.out.tfevents.1629745852.74272264b15c.932.203 +3 -0
- model-bin/finetune/base/log/1629746486.2610247/events.out.tfevents.1629746486.74272264b15c.932.205 +3 -0
- model-bin/finetune/base/log/1629747145.1602602/events.out.tfevents.1629747145.74272264b15c.932.207 +3 -0
- model-bin/finetune/base/log/1629747777.952989/events.out.tfevents.1629747777.74272264b15c.932.209 +3 -0
- model-bin/finetune/base/log/1629748416.137271/events.out.tfevents.1629748416.74272264b15c.932.211 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629745852.74272264b15c.932.202 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629746486.74272264b15c.932.204 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629747145.74272264b15c.932.206 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629747777.74272264b15c.932.208 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629748416.74272264b15c.932.210 +3 -0
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/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 sha256:06357c6412f787d1cb53abb44ac11367920860ca953a27d567a355683c83f40c
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/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:d8236250452d040754b78ad4a49c587ffd85ead29462c399fe3781b5eaaa42d2
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/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:fe613188f7fbc369c8166b21459dcb518deb34fb4d7eaa14885d283a7296aa45
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/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:b9f5ed619f6a869313295261e9822307694ee63dfd9e9bdf19fc810404339399
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/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 sha256:3618fa6df8466c4af8ada6a62766d36d94c6ba824ea935774cf8bd8195de9789
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -163962,11 +163962,806 @@
|
|
| 163962 |
"eval_steps_per_second": 0.671,
|
| 163963 |
"eval_wer": 0.20255368543238536,
|
| 163964 |
"step": 50773
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163965 |
}
|
| 163966 |
],
|
| 163967 |
"max_steps": 620000,
|
| 163968 |
"num_train_epochs": 5000,
|
| 163969 |
-
"total_flos": 1.
|
| 163970 |
"trial_name": null,
|
| 163971 |
"trial_params": null
|
| 163972 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 413.99598393574297,
|
| 5 |
+
"global_step": 51395,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 163962 |
"eval_steps_per_second": 0.671,
|
| 163963 |
"eval_wer": 0.20255368543238536,
|
| 163964 |
"step": 50773
|
| 163965 |
+
},
|
| 163966 |
+
{
|
| 163967 |
+
"epoch": 409.02,
|
| 163968 |
+
"learning_rate": 9.202628205128206e-06,
|
| 163969 |
+
"loss": 0.4735,
|
| 163970 |
+
"step": 50775
|
| 163971 |
+
},
|
| 163972 |
+
{
|
| 163973 |
+
"epoch": 409.06,
|
| 163974 |
+
"learning_rate": 9.202548076923079e-06,
|
| 163975 |
+
"loss": 0.6031,
|
| 163976 |
+
"step": 50780
|
| 163977 |
+
},
|
| 163978 |
+
{
|
| 163979 |
+
"epoch": 409.1,
|
| 163980 |
+
"learning_rate": 9.20246794871795e-06,
|
| 163981 |
+
"loss": 0.3739,
|
| 163982 |
+
"step": 50785
|
| 163983 |
+
},
|
| 163984 |
+
{
|
| 163985 |
+
"epoch": 409.14,
|
| 163986 |
+
"learning_rate": 9.202387820512821e-06,
|
| 163987 |
+
"loss": 0.369,
|
| 163988 |
+
"step": 50790
|
| 163989 |
+
},
|
| 163990 |
+
{
|
| 163991 |
+
"epoch": 409.18,
|
| 163992 |
+
"learning_rate": 9.202307692307693e-06,
|
| 163993 |
+
"loss": 0.6963,
|
| 163994 |
+
"step": 50795
|
| 163995 |
+
},
|
| 163996 |
+
{
|
| 163997 |
+
"epoch": 409.22,
|
| 163998 |
+
"learning_rate": 9.202227564102566e-06,
|
| 163999 |
+
"loss": 1.1502,
|
| 164000 |
+
"step": 50800
|
| 164001 |
+
},
|
| 164002 |
+
{
|
| 164003 |
+
"epoch": 409.26,
|
| 164004 |
+
"learning_rate": 9.202147435897437e-06,
|
| 164005 |
+
"loss": 0.3254,
|
| 164006 |
+
"step": 50805
|
| 164007 |
+
},
|
| 164008 |
+
{
|
| 164009 |
+
"epoch": 409.3,
|
| 164010 |
+
"learning_rate": 9.202067307692308e-06,
|
| 164011 |
+
"loss": 0.3354,
|
| 164012 |
+
"step": 50810
|
| 164013 |
+
},
|
| 164014 |
+
{
|
| 164015 |
+
"epoch": 409.34,
|
| 164016 |
+
"learning_rate": 9.201987179487182e-06,
|
| 164017 |
+
"loss": 0.391,
|
| 164018 |
+
"step": 50815
|
| 164019 |
+
},
|
| 164020 |
+
{
|
| 164021 |
+
"epoch": 409.38,
|
| 164022 |
+
"learning_rate": 9.201907051282051e-06,
|
| 164023 |
+
"loss": 0.7122,
|
| 164024 |
+
"step": 50820
|
| 164025 |
+
},
|
| 164026 |
+
{
|
| 164027 |
+
"epoch": 409.42,
|
| 164028 |
+
"learning_rate": 9.201826923076924e-06,
|
| 164029 |
+
"loss": 1.0584,
|
| 164030 |
+
"step": 50825
|
| 164031 |
+
},
|
| 164032 |
+
{
|
| 164033 |
+
"epoch": 409.46,
|
| 164034 |
+
"learning_rate": 9.201746794871796e-06,
|
| 164035 |
+
"loss": 0.3459,
|
| 164036 |
+
"step": 50830
|
| 164037 |
+
},
|
| 164038 |
+
{
|
| 164039 |
+
"epoch": 409.5,
|
| 164040 |
+
"learning_rate": 9.201666666666667e-06,
|
| 164041 |
+
"loss": 0.3254,
|
| 164042 |
+
"step": 50835
|
| 164043 |
+
},
|
| 164044 |
+
{
|
| 164045 |
+
"epoch": 409.54,
|
| 164046 |
+
"learning_rate": 9.201586538461538e-06,
|
| 164047 |
+
"loss": 0.3937,
|
| 164048 |
+
"step": 50840
|
| 164049 |
+
},
|
| 164050 |
+
{
|
| 164051 |
+
"epoch": 409.58,
|
| 164052 |
+
"learning_rate": 9.201506410256411e-06,
|
| 164053 |
+
"loss": 0.7381,
|
| 164054 |
+
"step": 50845
|
| 164055 |
+
},
|
| 164056 |
+
{
|
| 164057 |
+
"epoch": 409.62,
|
| 164058 |
+
"learning_rate": 9.201426282051283e-06,
|
| 164059 |
+
"loss": 1.013,
|
| 164060 |
+
"step": 50850
|
| 164061 |
+
},
|
| 164062 |
+
{
|
| 164063 |
+
"epoch": 409.66,
|
| 164064 |
+
"learning_rate": 9.201346153846154e-06,
|
| 164065 |
+
"loss": 0.3073,
|
| 164066 |
+
"step": 50855
|
| 164067 |
+
},
|
| 164068 |
+
{
|
| 164069 |
+
"epoch": 409.7,
|
| 164070 |
+
"learning_rate": 9.201266025641027e-06,
|
| 164071 |
+
"loss": 0.367,
|
| 164072 |
+
"step": 50860
|
| 164073 |
+
},
|
| 164074 |
+
{
|
| 164075 |
+
"epoch": 409.74,
|
| 164076 |
+
"learning_rate": 9.201185897435898e-06,
|
| 164077 |
+
"loss": 0.379,
|
| 164078 |
+
"step": 50865
|
| 164079 |
+
},
|
| 164080 |
+
{
|
| 164081 |
+
"epoch": 409.78,
|
| 164082 |
+
"learning_rate": 9.20110576923077e-06,
|
| 164083 |
+
"loss": 0.6617,
|
| 164084 |
+
"step": 50870
|
| 164085 |
+
},
|
| 164086 |
+
{
|
| 164087 |
+
"epoch": 409.82,
|
| 164088 |
+
"learning_rate": 9.201025641025641e-06,
|
| 164089 |
+
"loss": 1.0443,
|
| 164090 |
+
"step": 50875
|
| 164091 |
+
},
|
| 164092 |
+
{
|
| 164093 |
+
"epoch": 409.86,
|
| 164094 |
+
"learning_rate": 9.200945512820514e-06,
|
| 164095 |
+
"loss": 0.354,
|
| 164096 |
+
"step": 50880
|
| 164097 |
+
},
|
| 164098 |
+
{
|
| 164099 |
+
"epoch": 409.9,
|
| 164100 |
+
"learning_rate": 9.200865384615386e-06,
|
| 164101 |
+
"loss": 0.316,
|
| 164102 |
+
"step": 50885
|
| 164103 |
+
},
|
| 164104 |
+
{
|
| 164105 |
+
"epoch": 409.94,
|
| 164106 |
+
"learning_rate": 9.200785256410257e-06,
|
| 164107 |
+
"loss": 0.5285,
|
| 164108 |
+
"step": 50890
|
| 164109 |
+
},
|
| 164110 |
+
{
|
| 164111 |
+
"epoch": 409.98,
|
| 164112 |
+
"learning_rate": 9.200705128205128e-06,
|
| 164113 |
+
"loss": 0.7452,
|
| 164114 |
+
"step": 50895
|
| 164115 |
+
},
|
| 164116 |
+
{
|
| 164117 |
+
"epoch": 410.0,
|
| 164118 |
+
"eval_loss": 0.40997835993766785,
|
| 164119 |
+
"eval_runtime": 40.0349,
|
| 164120 |
+
"eval_samples_per_second": 20.957,
|
| 164121 |
+
"eval_steps_per_second": 0.674,
|
| 164122 |
+
"eval_wer": 0.20446797707985784,
|
| 164123 |
+
"step": 50897
|
| 164124 |
+
},
|
| 164125 |
+
{
|
| 164126 |
+
"epoch": 407.02,
|
| 164127 |
+
"learning_rate": 9.200625000000001e-06,
|
| 164128 |
+
"loss": 0.3863,
|
| 164129 |
+
"step": 50900
|
| 164130 |
+
},
|
| 164131 |
+
{
|
| 164132 |
+
"epoch": 407.06,
|
| 164133 |
+
"learning_rate": 9.200544871794873e-06,
|
| 164134 |
+
"loss": 0.3885,
|
| 164135 |
+
"step": 50905
|
| 164136 |
+
},
|
| 164137 |
+
{
|
| 164138 |
+
"epoch": 407.1,
|
| 164139 |
+
"learning_rate": 9.200464743589744e-06,
|
| 164140 |
+
"loss": 0.3331,
|
| 164141 |
+
"step": 50910
|
| 164142 |
+
},
|
| 164143 |
+
{
|
| 164144 |
+
"epoch": 407.14,
|
| 164145 |
+
"learning_rate": 9.200384615384617e-06,
|
| 164146 |
+
"loss": 0.3941,
|
| 164147 |
+
"step": 50915
|
| 164148 |
+
},
|
| 164149 |
+
{
|
| 164150 |
+
"epoch": 407.18,
|
| 164151 |
+
"learning_rate": 9.200304487179489e-06,
|
| 164152 |
+
"loss": 0.9034,
|
| 164153 |
+
"step": 50920
|
| 164154 |
+
},
|
| 164155 |
+
{
|
| 164156 |
+
"epoch": 407.22,
|
| 164157 |
+
"learning_rate": 9.20022435897436e-06,
|
| 164158 |
+
"loss": 1.0837,
|
| 164159 |
+
"step": 50925
|
| 164160 |
+
},
|
| 164161 |
+
{
|
| 164162 |
+
"epoch": 407.26,
|
| 164163 |
+
"learning_rate": 9.200144230769231e-06,
|
| 164164 |
+
"loss": 0.4027,
|
| 164165 |
+
"step": 50930
|
| 164166 |
+
},
|
| 164167 |
+
{
|
| 164168 |
+
"epoch": 407.3,
|
| 164169 |
+
"learning_rate": 9.200064102564104e-06,
|
| 164170 |
+
"loss": 0.399,
|
| 164171 |
+
"step": 50935
|
| 164172 |
+
},
|
| 164173 |
+
{
|
| 164174 |
+
"epoch": 407.34,
|
| 164175 |
+
"learning_rate": 9.199983974358974e-06,
|
| 164176 |
+
"loss": 0.4672,
|
| 164177 |
+
"step": 50940
|
| 164178 |
+
},
|
| 164179 |
+
{
|
| 164180 |
+
"epoch": 407.38,
|
| 164181 |
+
"learning_rate": 9.199903846153847e-06,
|
| 164182 |
+
"loss": 0.9135,
|
| 164183 |
+
"step": 50945
|
| 164184 |
+
},
|
| 164185 |
+
{
|
| 164186 |
+
"epoch": 407.42,
|
| 164187 |
+
"learning_rate": 9.199823717948718e-06,
|
| 164188 |
+
"loss": 1.1382,
|
| 164189 |
+
"step": 50950
|
| 164190 |
+
},
|
| 164191 |
+
{
|
| 164192 |
+
"epoch": 407.46,
|
| 164193 |
+
"learning_rate": 9.19974358974359e-06,
|
| 164194 |
+
"loss": 0.3736,
|
| 164195 |
+
"step": 50955
|
| 164196 |
+
},
|
| 164197 |
+
{
|
| 164198 |
+
"epoch": 407.5,
|
| 164199 |
+
"learning_rate": 9.199663461538463e-06,
|
| 164200 |
+
"loss": 0.4124,
|
| 164201 |
+
"step": 50960
|
| 164202 |
+
},
|
| 164203 |
+
{
|
| 164204 |
+
"epoch": 407.54,
|
| 164205 |
+
"learning_rate": 9.199583333333334e-06,
|
| 164206 |
+
"loss": 0.3823,
|
| 164207 |
+
"step": 50965
|
| 164208 |
+
},
|
| 164209 |
+
{
|
| 164210 |
+
"epoch": 407.58,
|
| 164211 |
+
"learning_rate": 9.199503205128205e-06,
|
| 164212 |
+
"loss": 0.7784,
|
| 164213 |
+
"step": 50970
|
| 164214 |
+
},
|
| 164215 |
+
{
|
| 164216 |
+
"epoch": 407.62,
|
| 164217 |
+
"learning_rate": 9.199423076923077e-06,
|
| 164218 |
+
"loss": 0.9183,
|
| 164219 |
+
"step": 50975
|
| 164220 |
+
},
|
| 164221 |
+
{
|
| 164222 |
+
"epoch": 407.66,
|
| 164223 |
+
"learning_rate": 9.19934294871795e-06,
|
| 164224 |
+
"loss": 0.3183,
|
| 164225 |
+
"step": 50980
|
| 164226 |
+
},
|
| 164227 |
+
{
|
| 164228 |
+
"epoch": 407.7,
|
| 164229 |
+
"learning_rate": 9.199262820512821e-06,
|
| 164230 |
+
"loss": 0.3618,
|
| 164231 |
+
"step": 50985
|
| 164232 |
+
},
|
| 164233 |
+
{
|
| 164234 |
+
"epoch": 407.74,
|
| 164235 |
+
"learning_rate": 9.199182692307693e-06,
|
| 164236 |
+
"loss": 0.4161,
|
| 164237 |
+
"step": 50990
|
| 164238 |
+
},
|
| 164239 |
+
{
|
| 164240 |
+
"epoch": 407.78,
|
| 164241 |
+
"learning_rate": 9.199102564102564e-06,
|
| 164242 |
+
"loss": 0.6897,
|
| 164243 |
+
"step": 50995
|
| 164244 |
+
},
|
| 164245 |
+
{
|
| 164246 |
+
"epoch": 407.82,
|
| 164247 |
+
"learning_rate": 9.199022435897437e-06,
|
| 164248 |
+
"loss": 1.0322,
|
| 164249 |
+
"step": 51000
|
| 164250 |
+
},
|
| 164251 |
+
{
|
| 164252 |
+
"epoch": 407.86,
|
| 164253 |
+
"learning_rate": 9.198942307692308e-06,
|
| 164254 |
+
"loss": 0.3193,
|
| 164255 |
+
"step": 51005
|
| 164256 |
+
},
|
| 164257 |
+
{
|
| 164258 |
+
"epoch": 407.9,
|
| 164259 |
+
"learning_rate": 9.19886217948718e-06,
|
| 164260 |
+
"loss": 0.3782,
|
| 164261 |
+
"step": 51010
|
| 164262 |
+
},
|
| 164263 |
+
{
|
| 164264 |
+
"epoch": 407.94,
|
| 164265 |
+
"learning_rate": 9.198782051282053e-06,
|
| 164266 |
+
"loss": 0.4674,
|
| 164267 |
+
"step": 51015
|
| 164268 |
+
},
|
| 164269 |
+
{
|
| 164270 |
+
"epoch": 407.98,
|
| 164271 |
+
"learning_rate": 9.198701923076924e-06,
|
| 164272 |
+
"loss": 0.8146,
|
| 164273 |
+
"step": 51020
|
| 164274 |
+
},
|
| 164275 |
+
{
|
| 164276 |
+
"epoch": 408.0,
|
| 164277 |
+
"eval_loss": 0.3701570928096771,
|
| 164278 |
+
"eval_runtime": 40.3169,
|
| 164279 |
+
"eval_samples_per_second": 20.81,
|
| 164280 |
+
"eval_steps_per_second": 0.67,
|
| 164281 |
+
"eval_wer": 0.20420333066686058,
|
| 164282 |
+
"step": 51022
|
| 164283 |
+
},
|
| 164284 |
+
{
|
| 164285 |
+
"epoch": 411.02,
|
| 164286 |
+
"learning_rate": 9.198621794871796e-06,
|
| 164287 |
+
"loss": 0.3877,
|
| 164288 |
+
"step": 51025
|
| 164289 |
+
},
|
| 164290 |
+
{
|
| 164291 |
+
"epoch": 411.06,
|
| 164292 |
+
"learning_rate": 9.198541666666667e-06,
|
| 164293 |
+
"loss": 0.3139,
|
| 164294 |
+
"step": 51030
|
| 164295 |
+
},
|
| 164296 |
+
{
|
| 164297 |
+
"epoch": 411.1,
|
| 164298 |
+
"learning_rate": 9.19846153846154e-06,
|
| 164299 |
+
"loss": 0.33,
|
| 164300 |
+
"step": 51035
|
| 164301 |
+
},
|
| 164302 |
+
{
|
| 164303 |
+
"epoch": 411.14,
|
| 164304 |
+
"learning_rate": 9.198381410256411e-06,
|
| 164305 |
+
"loss": 0.375,
|
| 164306 |
+
"step": 51040
|
| 164307 |
+
},
|
| 164308 |
+
{
|
| 164309 |
+
"epoch": 411.18,
|
| 164310 |
+
"learning_rate": 9.198301282051283e-06,
|
| 164311 |
+
"loss": 0.8677,
|
| 164312 |
+
"step": 51045
|
| 164313 |
+
},
|
| 164314 |
+
{
|
| 164315 |
+
"epoch": 411.22,
|
| 164316 |
+
"learning_rate": 9.198221153846154e-06,
|
| 164317 |
+
"loss": 0.9139,
|
| 164318 |
+
"step": 51050
|
| 164319 |
+
},
|
| 164320 |
+
{
|
| 164321 |
+
"epoch": 411.27,
|
| 164322 |
+
"learning_rate": 9.198141025641027e-06,
|
| 164323 |
+
"loss": 0.2915,
|
| 164324 |
+
"step": 51055
|
| 164325 |
+
},
|
| 164326 |
+
{
|
| 164327 |
+
"epoch": 411.31,
|
| 164328 |
+
"learning_rate": 9.198060897435898e-06,
|
| 164329 |
+
"loss": 0.3343,
|
| 164330 |
+
"step": 51060
|
| 164331 |
+
},
|
| 164332 |
+
{
|
| 164333 |
+
"epoch": 411.35,
|
| 164334 |
+
"learning_rate": 9.19798076923077e-06,
|
| 164335 |
+
"loss": 0.4676,
|
| 164336 |
+
"step": 51065
|
| 164337 |
+
},
|
| 164338 |
+
{
|
| 164339 |
+
"epoch": 411.39,
|
| 164340 |
+
"learning_rate": 9.197900641025643e-06,
|
| 164341 |
+
"loss": 0.8988,
|
| 164342 |
+
"step": 51070
|
| 164343 |
+
},
|
| 164344 |
+
{
|
| 164345 |
+
"epoch": 411.43,
|
| 164346 |
+
"learning_rate": 9.197820512820514e-06,
|
| 164347 |
+
"loss": 1.1219,
|
| 164348 |
+
"step": 51075
|
| 164349 |
+
},
|
| 164350 |
+
{
|
| 164351 |
+
"epoch": 411.47,
|
| 164352 |
+
"learning_rate": 9.197740384615386e-06,
|
| 164353 |
+
"loss": 0.4589,
|
| 164354 |
+
"step": 51080
|
| 164355 |
+
},
|
| 164356 |
+
{
|
| 164357 |
+
"epoch": 411.51,
|
| 164358 |
+
"learning_rate": 9.197660256410257e-06,
|
| 164359 |
+
"loss": 0.3311,
|
| 164360 |
+
"step": 51085
|
| 164361 |
+
},
|
| 164362 |
+
{
|
| 164363 |
+
"epoch": 411.55,
|
| 164364 |
+
"learning_rate": 9.19758012820513e-06,
|
| 164365 |
+
"loss": 0.4065,
|
| 164366 |
+
"step": 51090
|
| 164367 |
+
},
|
| 164368 |
+
{
|
| 164369 |
+
"epoch": 411.59,
|
| 164370 |
+
"learning_rate": 9.1975e-06,
|
| 164371 |
+
"loss": 0.8172,
|
| 164372 |
+
"step": 51095
|
| 164373 |
+
},
|
| 164374 |
+
{
|
| 164375 |
+
"epoch": 411.63,
|
| 164376 |
+
"learning_rate": 9.197419871794873e-06,
|
| 164377 |
+
"loss": 0.9767,
|
| 164378 |
+
"step": 51100
|
| 164379 |
+
},
|
| 164380 |
+
{
|
| 164381 |
+
"epoch": 411.67,
|
| 164382 |
+
"learning_rate": 9.197339743589744e-06,
|
| 164383 |
+
"loss": 0.285,
|
| 164384 |
+
"step": 51105
|
| 164385 |
+
},
|
| 164386 |
+
{
|
| 164387 |
+
"epoch": 411.71,
|
| 164388 |
+
"learning_rate": 9.197259615384615e-06,
|
| 164389 |
+
"loss": 0.3242,
|
| 164390 |
+
"step": 51110
|
| 164391 |
+
},
|
| 164392 |
+
{
|
| 164393 |
+
"epoch": 411.75,
|
| 164394 |
+
"learning_rate": 9.197179487179488e-06,
|
| 164395 |
+
"loss": 0.4426,
|
| 164396 |
+
"step": 51115
|
| 164397 |
+
},
|
| 164398 |
+
{
|
| 164399 |
+
"epoch": 411.79,
|
| 164400 |
+
"learning_rate": 9.19709935897436e-06,
|
| 164401 |
+
"loss": 0.8639,
|
| 164402 |
+
"step": 51120
|
| 164403 |
+
},
|
| 164404 |
+
{
|
| 164405 |
+
"epoch": 411.83,
|
| 164406 |
+
"learning_rate": 9.197019230769231e-06,
|
| 164407 |
+
"loss": 0.9292,
|
| 164408 |
+
"step": 51125
|
| 164409 |
+
},
|
| 164410 |
+
{
|
| 164411 |
+
"epoch": 411.87,
|
| 164412 |
+
"learning_rate": 9.196939102564103e-06,
|
| 164413 |
+
"loss": 0.3492,
|
| 164414 |
+
"step": 51130
|
| 164415 |
+
},
|
| 164416 |
+
{
|
| 164417 |
+
"epoch": 411.91,
|
| 164418 |
+
"learning_rate": 9.196858974358976e-06,
|
| 164419 |
+
"loss": 0.3472,
|
| 164420 |
+
"step": 51135
|
| 164421 |
+
},
|
| 164422 |
+
{
|
| 164423 |
+
"epoch": 411.95,
|
| 164424 |
+
"learning_rate": 9.196778846153847e-06,
|
| 164425 |
+
"loss": 0.3816,
|
| 164426 |
+
"step": 51140
|
| 164427 |
+
},
|
| 164428 |
+
{
|
| 164429 |
+
"epoch": 411.99,
|
| 164430 |
+
"learning_rate": 9.196698717948718e-06,
|
| 164431 |
+
"loss": 0.8373,
|
| 164432 |
+
"step": 51145
|
| 164433 |
+
},
|
| 164434 |
+
{
|
| 164435 |
+
"epoch": 412.0,
|
| 164436 |
+
"eval_loss": 0.4409707188606262,
|
| 164437 |
+
"eval_runtime": 39.5471,
|
| 164438 |
+
"eval_samples_per_second": 21.24,
|
| 164439 |
+
"eval_steps_per_second": 0.683,
|
| 164440 |
+
"eval_wer": 0.19903090815652302,
|
| 164441 |
+
"step": 51146
|
| 164442 |
+
},
|
| 164443 |
+
{
|
| 164444 |
+
"epoch": 409.03,
|
| 164445 |
+
"learning_rate": 9.19661858974359e-06,
|
| 164446 |
+
"loss": 0.387,
|
| 164447 |
+
"step": 51150
|
| 164448 |
+
},
|
| 164449 |
+
{
|
| 164450 |
+
"epoch": 409.07,
|
| 164451 |
+
"learning_rate": 9.196538461538463e-06,
|
| 164452 |
+
"loss": 0.2921,
|
| 164453 |
+
"step": 51155
|
| 164454 |
+
},
|
| 164455 |
+
{
|
| 164456 |
+
"epoch": 409.11,
|
| 164457 |
+
"learning_rate": 9.196458333333334e-06,
|
| 164458 |
+
"loss": 0.3159,
|
| 164459 |
+
"step": 51160
|
| 164460 |
+
},
|
| 164461 |
+
{
|
| 164462 |
+
"epoch": 409.15,
|
| 164463 |
+
"learning_rate": 9.196378205128205e-06,
|
| 164464 |
+
"loss": 0.5311,
|
| 164465 |
+
"step": 51165
|
| 164466 |
+
},
|
| 164467 |
+
{
|
| 164468 |
+
"epoch": 409.19,
|
| 164469 |
+
"learning_rate": 9.196298076923078e-06,
|
| 164470 |
+
"loss": 1.1234,
|
| 164471 |
+
"step": 51170
|
| 164472 |
+
},
|
| 164473 |
+
{
|
| 164474 |
+
"epoch": 409.23,
|
| 164475 |
+
"learning_rate": 9.19621794871795e-06,
|
| 164476 |
+
"loss": 0.7092,
|
| 164477 |
+
"step": 51175
|
| 164478 |
+
},
|
| 164479 |
+
{
|
| 164480 |
+
"epoch": 409.27,
|
| 164481 |
+
"learning_rate": 9.196137820512821e-06,
|
| 164482 |
+
"loss": 0.3405,
|
| 164483 |
+
"step": 51180
|
| 164484 |
+
},
|
| 164485 |
+
{
|
| 164486 |
+
"epoch": 409.31,
|
| 164487 |
+
"learning_rate": 9.196057692307693e-06,
|
| 164488 |
+
"loss": 0.3519,
|
| 164489 |
+
"step": 51185
|
| 164490 |
+
},
|
| 164491 |
+
{
|
| 164492 |
+
"epoch": 409.35,
|
| 164493 |
+
"learning_rate": 9.195977564102566e-06,
|
| 164494 |
+
"loss": 0.4383,
|
| 164495 |
+
"step": 51190
|
| 164496 |
+
},
|
| 164497 |
+
{
|
| 164498 |
+
"epoch": 409.39,
|
| 164499 |
+
"learning_rate": 9.195897435897437e-06,
|
| 164500 |
+
"loss": 1.0161,
|
| 164501 |
+
"step": 51195
|
| 164502 |
+
},
|
| 164503 |
+
{
|
| 164504 |
+
"epoch": 409.43,
|
| 164505 |
+
"learning_rate": 9.195817307692308e-06,
|
| 164506 |
+
"loss": 0.7352,
|
| 164507 |
+
"step": 51200
|
| 164508 |
+
},
|
| 164509 |
+
{
|
| 164510 |
+
"epoch": 409.47,
|
| 164511 |
+
"learning_rate": 9.19573717948718e-06,
|
| 164512 |
+
"loss": 0.3026,
|
| 164513 |
+
"step": 51205
|
| 164514 |
+
},
|
| 164515 |
+
{
|
| 164516 |
+
"epoch": 409.51,
|
| 164517 |
+
"learning_rate": 9.195657051282053e-06,
|
| 164518 |
+
"loss": 0.3251,
|
| 164519 |
+
"step": 51210
|
| 164520 |
+
},
|
| 164521 |
+
{
|
| 164522 |
+
"epoch": 409.55,
|
| 164523 |
+
"learning_rate": 9.195576923076924e-06,
|
| 164524 |
+
"loss": 0.4582,
|
| 164525 |
+
"step": 51215
|
| 164526 |
+
},
|
| 164527 |
+
{
|
| 164528 |
+
"epoch": 409.59,
|
| 164529 |
+
"learning_rate": 9.195496794871795e-06,
|
| 164530 |
+
"loss": 0.968,
|
| 164531 |
+
"step": 51220
|
| 164532 |
+
},
|
| 164533 |
+
{
|
| 164534 |
+
"epoch": 409.63,
|
| 164535 |
+
"learning_rate": 9.195416666666668e-06,
|
| 164536 |
+
"loss": 0.7348,
|
| 164537 |
+
"step": 51225
|
| 164538 |
+
},
|
| 164539 |
+
{
|
| 164540 |
+
"epoch": 409.67,
|
| 164541 |
+
"learning_rate": 9.195336538461538e-06,
|
| 164542 |
+
"loss": 0.2916,
|
| 164543 |
+
"step": 51230
|
| 164544 |
+
},
|
| 164545 |
+
{
|
| 164546 |
+
"epoch": 409.71,
|
| 164547 |
+
"learning_rate": 9.195256410256411e-06,
|
| 164548 |
+
"loss": 0.3603,
|
| 164549 |
+
"step": 51235
|
| 164550 |
+
},
|
| 164551 |
+
{
|
| 164552 |
+
"epoch": 409.75,
|
| 164553 |
+
"learning_rate": 9.195176282051283e-06,
|
| 164554 |
+
"loss": 0.4094,
|
| 164555 |
+
"step": 51240
|
| 164556 |
+
},
|
| 164557 |
+
{
|
| 164558 |
+
"epoch": 409.79,
|
| 164559 |
+
"learning_rate": 9.195096153846154e-06,
|
| 164560 |
+
"loss": 1.0895,
|
| 164561 |
+
"step": 51245
|
| 164562 |
+
},
|
| 164563 |
+
{
|
| 164564 |
+
"epoch": 409.83,
|
| 164565 |
+
"learning_rate": 9.195016025641025e-06,
|
| 164566 |
+
"loss": 0.6658,
|
| 164567 |
+
"step": 51250
|
| 164568 |
+
},
|
| 164569 |
+
{
|
| 164570 |
+
"epoch": 409.87,
|
| 164571 |
+
"learning_rate": 9.194935897435898e-06,
|
| 164572 |
+
"loss": 0.3517,
|
| 164573 |
+
"step": 51255
|
| 164574 |
+
},
|
| 164575 |
+
{
|
| 164576 |
+
"epoch": 409.91,
|
| 164577 |
+
"learning_rate": 9.19485576923077e-06,
|
| 164578 |
+
"loss": 0.4797,
|
| 164579 |
+
"step": 51260
|
| 164580 |
+
},
|
| 164581 |
+
{
|
| 164582 |
+
"epoch": 409.95,
|
| 164583 |
+
"learning_rate": 9.194775641025641e-06,
|
| 164584 |
+
"loss": 0.439,
|
| 164585 |
+
"step": 51265
|
| 164586 |
+
},
|
| 164587 |
+
{
|
| 164588 |
+
"epoch": 409.99,
|
| 164589 |
+
"learning_rate": 9.194695512820514e-06,
|
| 164590 |
+
"loss": 1.0348,
|
| 164591 |
+
"step": 51270
|
| 164592 |
+
},
|
| 164593 |
+
{
|
| 164594 |
+
"epoch": 410.0,
|
| 164595 |
+
"eval_loss": 0.41439250111579895,
|
| 164596 |
+
"eval_runtime": 39.2725,
|
| 164597 |
+
"eval_samples_per_second": 21.389,
|
| 164598 |
+
"eval_steps_per_second": 0.688,
|
| 164599 |
+
"eval_wer": 0.20278344505974935,
|
| 164600 |
+
"step": 51271
|
| 164601 |
+
},
|
| 164602 |
+
{
|
| 164603 |
+
"epoch": 413.03,
|
| 164604 |
+
"learning_rate": 9.194615384615385e-06,
|
| 164605 |
+
"loss": 0.3525,
|
| 164606 |
+
"step": 51275
|
| 164607 |
+
},
|
| 164608 |
+
{
|
| 164609 |
+
"epoch": 413.07,
|
| 164610 |
+
"learning_rate": 9.194535256410257e-06,
|
| 164611 |
+
"loss": 0.3379,
|
| 164612 |
+
"step": 51280
|
| 164613 |
+
},
|
| 164614 |
+
{
|
| 164615 |
+
"epoch": 413.11,
|
| 164616 |
+
"learning_rate": 9.194455128205128e-06,
|
| 164617 |
+
"loss": 0.326,
|
| 164618 |
+
"step": 51285
|
| 164619 |
+
},
|
| 164620 |
+
{
|
| 164621 |
+
"epoch": 413.15,
|
| 164622 |
+
"learning_rate": 9.194375000000001e-06,
|
| 164623 |
+
"loss": 0.4398,
|
| 164624 |
+
"step": 51290
|
| 164625 |
+
},
|
| 164626 |
+
{
|
| 164627 |
+
"epoch": 413.19,
|
| 164628 |
+
"learning_rate": 9.194294871794873e-06,
|
| 164629 |
+
"loss": 0.9764,
|
| 164630 |
+
"step": 51295
|
| 164631 |
+
},
|
| 164632 |
+
{
|
| 164633 |
+
"epoch": 413.23,
|
| 164634 |
+
"learning_rate": 9.194214743589744e-06,
|
| 164635 |
+
"loss": 0.7491,
|
| 164636 |
+
"step": 51300
|
| 164637 |
+
},
|
| 164638 |
+
{
|
| 164639 |
+
"epoch": 413.27,
|
| 164640 |
+
"learning_rate": 9.194134615384615e-06,
|
| 164641 |
+
"loss": 0.3822,
|
| 164642 |
+
"step": 51305
|
| 164643 |
+
},
|
| 164644 |
+
{
|
| 164645 |
+
"epoch": 413.31,
|
| 164646 |
+
"learning_rate": 9.194054487179488e-06,
|
| 164647 |
+
"loss": 0.3469,
|
| 164648 |
+
"step": 51310
|
| 164649 |
+
},
|
| 164650 |
+
{
|
| 164651 |
+
"epoch": 413.35,
|
| 164652 |
+
"learning_rate": 9.19397435897436e-06,
|
| 164653 |
+
"loss": 0.4459,
|
| 164654 |
+
"step": 51315
|
| 164655 |
+
},
|
| 164656 |
+
{
|
| 164657 |
+
"epoch": 413.39,
|
| 164658 |
+
"learning_rate": 9.193894230769231e-06,
|
| 164659 |
+
"loss": 1.0217,
|
| 164660 |
+
"step": 51320
|
| 164661 |
+
},
|
| 164662 |
+
{
|
| 164663 |
+
"epoch": 413.43,
|
| 164664 |
+
"learning_rate": 9.193814102564104e-06,
|
| 164665 |
+
"loss": 0.7432,
|
| 164666 |
+
"step": 51325
|
| 164667 |
+
},
|
| 164668 |
+
{
|
| 164669 |
+
"epoch": 413.47,
|
| 164670 |
+
"learning_rate": 9.193733974358975e-06,
|
| 164671 |
+
"loss": 0.3442,
|
| 164672 |
+
"step": 51330
|
| 164673 |
+
},
|
| 164674 |
+
{
|
| 164675 |
+
"epoch": 413.51,
|
| 164676 |
+
"learning_rate": 9.193653846153847e-06,
|
| 164677 |
+
"loss": 0.3479,
|
| 164678 |
+
"step": 51335
|
| 164679 |
+
},
|
| 164680 |
+
{
|
| 164681 |
+
"epoch": 413.55,
|
| 164682 |
+
"learning_rate": 9.193573717948718e-06,
|
| 164683 |
+
"loss": 0.4101,
|
| 164684 |
+
"step": 51340
|
| 164685 |
+
},
|
| 164686 |
+
{
|
| 164687 |
+
"epoch": 413.59,
|
| 164688 |
+
"learning_rate": 9.193493589743591e-06,
|
| 164689 |
+
"loss": 1.0136,
|
| 164690 |
+
"step": 51345
|
| 164691 |
+
},
|
| 164692 |
+
{
|
| 164693 |
+
"epoch": 413.63,
|
| 164694 |
+
"learning_rate": 9.193413461538463e-06,
|
| 164695 |
+
"loss": 0.8008,
|
| 164696 |
+
"step": 51350
|
| 164697 |
+
},
|
| 164698 |
+
{
|
| 164699 |
+
"epoch": 413.67,
|
| 164700 |
+
"learning_rate": 9.193333333333334e-06,
|
| 164701 |
+
"loss": 0.3218,
|
| 164702 |
+
"step": 51355
|
| 164703 |
+
},
|
| 164704 |
+
{
|
| 164705 |
+
"epoch": 413.71,
|
| 164706 |
+
"learning_rate": 9.193253205128207e-06,
|
| 164707 |
+
"loss": 0.6458,
|
| 164708 |
+
"step": 51360
|
| 164709 |
+
},
|
| 164710 |
+
{
|
| 164711 |
+
"epoch": 413.76,
|
| 164712 |
+
"learning_rate": 9.193173076923078e-06,
|
| 164713 |
+
"loss": 0.4622,
|
| 164714 |
+
"step": 51365
|
| 164715 |
+
},
|
| 164716 |
+
{
|
| 164717 |
+
"epoch": 413.8,
|
| 164718 |
+
"learning_rate": 9.19309294871795e-06,
|
| 164719 |
+
"loss": 1.0731,
|
| 164720 |
+
"step": 51370
|
| 164721 |
+
},
|
| 164722 |
+
{
|
| 164723 |
+
"epoch": 413.84,
|
| 164724 |
+
"learning_rate": 9.193012820512821e-06,
|
| 164725 |
+
"loss": 0.7366,
|
| 164726 |
+
"step": 51375
|
| 164727 |
+
},
|
| 164728 |
+
{
|
| 164729 |
+
"epoch": 413.88,
|
| 164730 |
+
"learning_rate": 9.192932692307694e-06,
|
| 164731 |
+
"loss": 0.4681,
|
| 164732 |
+
"step": 51380
|
| 164733 |
+
},
|
| 164734 |
+
{
|
| 164735 |
+
"epoch": 413.92,
|
| 164736 |
+
"learning_rate": 9.192852564102564e-06,
|
| 164737 |
+
"loss": 0.356,
|
| 164738 |
+
"step": 51385
|
| 164739 |
+
},
|
| 164740 |
+
{
|
| 164741 |
+
"epoch": 413.96,
|
| 164742 |
+
"learning_rate": 9.192772435897437e-06,
|
| 164743 |
+
"loss": 0.4918,
|
| 164744 |
+
"step": 51390
|
| 164745 |
+
},
|
| 164746 |
+
{
|
| 164747 |
+
"epoch": 414.0,
|
| 164748 |
+
"learning_rate": 9.192692307692308e-06,
|
| 164749 |
+
"loss": 1.2729,
|
| 164750 |
+
"step": 51395
|
| 164751 |
+
},
|
| 164752 |
+
{
|
| 164753 |
+
"epoch": 414.0,
|
| 164754 |
+
"eval_loss": 0.3700979948043823,
|
| 164755 |
+
"eval_runtime": 39.239,
|
| 164756 |
+
"eval_samples_per_second": 21.407,
|
| 164757 |
+
"eval_steps_per_second": 0.688,
|
| 164758 |
+
"eval_wer": 0.2009889470622455,
|
| 164759 |
+
"step": 51395
|
| 164760 |
}
|
| 164761 |
],
|
| 164762 |
"max_steps": 620000,
|
| 164763 |
"num_train_epochs": 5000,
|
| 164764 |
+
"total_flos": 1.4461660908124747e+20,
|
| 164765 |
"trial_name": null,
|
| 164766 |
"trial_params": null
|
| 164767 |
}
|
model-bin/finetune/base/{checkpoint-50773 → checkpoint-51395}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629745852.2460718/events.out.tfevents.1629745852.74272264b15c.932.203
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:672c4a8ecc3bff9c63c73d0ce8b7812cdcd56d35555c124174d11a88f2838611
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629746486.2610247/events.out.tfevents.1629746486.74272264b15c.932.205
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f2e79c62a1cc12d1824cec9d50683eded7fc89b50a3ef5e5134e446000dd95a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629747145.1602602/events.out.tfevents.1629747145.74272264b15c.932.207
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8ef6e096da7b4cc64f0be005c69a95b7044b5898f019e3a541f26d3104975cc9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629747777.952989/events.out.tfevents.1629747777.74272264b15c.932.209
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0eb1a43407e9b92ff98f91138d40254a50a7d987b56a879f7ee39a63ab4be61
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629748416.137271/events.out.tfevents.1629748416.74272264b15c.932.211
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:315169cf391addca7fc362318e4a729e38c20522c3a451bb82c8d1ece47e0850
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629745852.74272264b15c.932.202
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6edfbb324f6bf21a1984bb3bb827fbe3e469e27a694e1c99e27747eb4409db1b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629746486.74272264b15c.932.204
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6db7c890c6bf3c3625005cb70f0af998cacd512b4596a5a77f4aeb47c2fa5da3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629747145.74272264b15c.932.206
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9194c31969ee17584d90d63a76eb07b7b880890f7c2c24d44456940d2e9fa505
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629747777.74272264b15c.932.208
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8cde6beb7d58f91166affa25a89133281cc9f3a7cd803b17ecc16a2bfe683b0f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629748416.74272264b15c.932.210
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b04aae8af6d72f9169b765263742e303741080461a354c223ce4d0a5a3f857d3
|
| 3 |
+
size 8622
|