"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629933091.5883334/events.out.tfevents.1629933091.7e498afd5545.7645.185 +3 -0
- model-bin/finetune/base/log/1629933593.7463934/events.out.tfevents.1629933593.7e498afd5545.7645.187 +3 -0
- model-bin/finetune/base/log/1629934073.4499795/events.out.tfevents.1629934073.7e498afd5545.7645.189 +3 -0
- model-bin/finetune/base/log/1629934533.9362981/events.out.tfevents.1629934535.7e498afd5545.7645.191 +3 -0
- model-bin/finetune/base/log/1629934991.9355497/events.out.tfevents.1629934991.7e498afd5545.7645.193 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629933091.7e498afd5545.7645.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629933593.7e498afd5545.7645.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629934072.7e498afd5545.7645.188 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629934533.7e498afd5545.7645.190 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629934991.7e498afd5545.7645.192 +3 -0
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/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 sha256:02b5f5d438dd7750ddcbd5e03a06bf5da10647a0694f2c80b916bb08f63d65f3
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/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 sha256:1bfbc399f7e6345794f7e9c2f12f77dc0abb92b9ddca2b8f4a0d98a910f1596a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/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 sha256:967ade1e5ebd4405f020ecc7951719de991516fbdf76457cd993be421aabfdd1
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/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:da825b21776e1bdfcebabe3acb018ec3e5d206390ffcfc17db0f26cbc8bed857
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/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 sha256:08370db5ce158b578b395d894b742358ae8d2c0dfd355a9ab2b4606ae3610978
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -210993,11 +210993,800 @@
|
|
| 210993 |
"eval_steps_per_second": 0.646,
|
| 210994 |
"eval_wer": 0.19399109792284866,
|
| 210995 |
"step": 87737
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 210996 |
}
|
| 210997 |
],
|
| 210998 |
"max_steps": 620000,
|
| 210999 |
"num_train_epochs": 5000,
|
| 211000 |
-
"total_flos": 2.
|
| 211001 |
"trial_name": null,
|
| 211002 |
"trial_params": null
|
| 211003 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 711.995983935743,
|
| 5 |
+
"global_step": 88359,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 210993 |
"eval_steps_per_second": 0.646,
|
| 210994 |
"eval_wer": 0.19399109792284866,
|
| 210995 |
"step": 87737
|
| 210996 |
+
},
|
| 210997 |
+
{
|
| 210998 |
+
"epoch": 701.02,
|
| 210999 |
+
"learning_rate": 8.59928917609047e-06,
|
| 211000 |
+
"loss": 0.3252,
|
| 211001 |
+
"step": 87740
|
| 211002 |
+
},
|
| 211003 |
+
{
|
| 211004 |
+
"epoch": 701.06,
|
| 211005 |
+
"learning_rate": 8.599208400646204e-06,
|
| 211006 |
+
"loss": 0.2929,
|
| 211007 |
+
"step": 87745
|
| 211008 |
+
},
|
| 211009 |
+
{
|
| 211010 |
+
"epoch": 701.1,
|
| 211011 |
+
"learning_rate": 8.59912762520194e-06,
|
| 211012 |
+
"loss": 0.303,
|
| 211013 |
+
"step": 87750
|
| 211014 |
+
},
|
| 211015 |
+
{
|
| 211016 |
+
"epoch": 701.14,
|
| 211017 |
+
"learning_rate": 8.599046849757674e-06,
|
| 211018 |
+
"loss": 0.3789,
|
| 211019 |
+
"step": 87755
|
| 211020 |
+
},
|
| 211021 |
+
{
|
| 211022 |
+
"epoch": 701.18,
|
| 211023 |
+
"learning_rate": 8.59896607431341e-06,
|
| 211024 |
+
"loss": 0.7483,
|
| 211025 |
+
"step": 87760
|
| 211026 |
+
},
|
| 211027 |
+
{
|
| 211028 |
+
"epoch": 701.22,
|
| 211029 |
+
"learning_rate": 8.598885298869144e-06,
|
| 211030 |
+
"loss": 0.977,
|
| 211031 |
+
"step": 87765
|
| 211032 |
+
},
|
| 211033 |
+
{
|
| 211034 |
+
"epoch": 701.26,
|
| 211035 |
+
"learning_rate": 8.59880452342488e-06,
|
| 211036 |
+
"loss": 0.3296,
|
| 211037 |
+
"step": 87770
|
| 211038 |
+
},
|
| 211039 |
+
{
|
| 211040 |
+
"epoch": 701.3,
|
| 211041 |
+
"learning_rate": 8.598723747980614e-06,
|
| 211042 |
+
"loss": 0.3069,
|
| 211043 |
+
"step": 87775
|
| 211044 |
+
},
|
| 211045 |
+
{
|
| 211046 |
+
"epoch": 701.34,
|
| 211047 |
+
"learning_rate": 8.59864297253635e-06,
|
| 211048 |
+
"loss": 0.3334,
|
| 211049 |
+
"step": 87780
|
| 211050 |
+
},
|
| 211051 |
+
{
|
| 211052 |
+
"epoch": 701.38,
|
| 211053 |
+
"learning_rate": 8.598562197092085e-06,
|
| 211054 |
+
"loss": 1.5383,
|
| 211055 |
+
"step": 87785
|
| 211056 |
+
},
|
| 211057 |
+
{
|
| 211058 |
+
"epoch": 701.42,
|
| 211059 |
+
"learning_rate": 8.59848142164782e-06,
|
| 211060 |
+
"loss": 0.8475,
|
| 211061 |
+
"step": 87790
|
| 211062 |
+
},
|
| 211063 |
+
{
|
| 211064 |
+
"epoch": 701.46,
|
| 211065 |
+
"learning_rate": 8.598400646203555e-06,
|
| 211066 |
+
"loss": 0.2889,
|
| 211067 |
+
"step": 87795
|
| 211068 |
+
},
|
| 211069 |
+
{
|
| 211070 |
+
"epoch": 701.5,
|
| 211071 |
+
"learning_rate": 8.59831987075929e-06,
|
| 211072 |
+
"loss": 0.3578,
|
| 211073 |
+
"step": 87800
|
| 211074 |
+
},
|
| 211075 |
+
{
|
| 211076 |
+
"epoch": 701.54,
|
| 211077 |
+
"learning_rate": 8.598239095315025e-06,
|
| 211078 |
+
"loss": 0.3767,
|
| 211079 |
+
"step": 87805
|
| 211080 |
+
},
|
| 211081 |
+
{
|
| 211082 |
+
"epoch": 701.58,
|
| 211083 |
+
"learning_rate": 8.59815831987076e-06,
|
| 211084 |
+
"loss": 0.6988,
|
| 211085 |
+
"step": 87810
|
| 211086 |
+
},
|
| 211087 |
+
{
|
| 211088 |
+
"epoch": 701.62,
|
| 211089 |
+
"learning_rate": 8.598077544426495e-06,
|
| 211090 |
+
"loss": 0.8645,
|
| 211091 |
+
"step": 87815
|
| 211092 |
+
},
|
| 211093 |
+
{
|
| 211094 |
+
"epoch": 701.66,
|
| 211095 |
+
"learning_rate": 8.59799676898223e-06,
|
| 211096 |
+
"loss": 0.2969,
|
| 211097 |
+
"step": 87820
|
| 211098 |
+
},
|
| 211099 |
+
{
|
| 211100 |
+
"epoch": 701.7,
|
| 211101 |
+
"learning_rate": 8.597915993537965e-06,
|
| 211102 |
+
"loss": 0.3161,
|
| 211103 |
+
"step": 87825
|
| 211104 |
+
},
|
| 211105 |
+
{
|
| 211106 |
+
"epoch": 701.74,
|
| 211107 |
+
"learning_rate": 8.5978352180937e-06,
|
| 211108 |
+
"loss": 0.3573,
|
| 211109 |
+
"step": 87830
|
| 211110 |
+
},
|
| 211111 |
+
{
|
| 211112 |
+
"epoch": 701.78,
|
| 211113 |
+
"learning_rate": 8.597754442649435e-06,
|
| 211114 |
+
"loss": 0.7298,
|
| 211115 |
+
"step": 87835
|
| 211116 |
+
},
|
| 211117 |
+
{
|
| 211118 |
+
"epoch": 701.82,
|
| 211119 |
+
"learning_rate": 8.597673667205171e-06,
|
| 211120 |
+
"loss": 0.8887,
|
| 211121 |
+
"step": 87840
|
| 211122 |
+
},
|
| 211123 |
+
{
|
| 211124 |
+
"epoch": 701.86,
|
| 211125 |
+
"learning_rate": 8.597592891760905e-06,
|
| 211126 |
+
"loss": 0.2905,
|
| 211127 |
+
"step": 87845
|
| 211128 |
+
},
|
| 211129 |
+
{
|
| 211130 |
+
"epoch": 701.9,
|
| 211131 |
+
"learning_rate": 8.597512116316641e-06,
|
| 211132 |
+
"loss": 0.3144,
|
| 211133 |
+
"step": 87850
|
| 211134 |
+
},
|
| 211135 |
+
{
|
| 211136 |
+
"epoch": 701.94,
|
| 211137 |
+
"learning_rate": 8.597431340872375e-06,
|
| 211138 |
+
"loss": 0.4287,
|
| 211139 |
+
"step": 87855
|
| 211140 |
+
},
|
| 211141 |
+
{
|
| 211142 |
+
"epoch": 701.98,
|
| 211143 |
+
"learning_rate": 8.597350565428111e-06,
|
| 211144 |
+
"loss": 0.7736,
|
| 211145 |
+
"step": 87860
|
| 211146 |
+
},
|
| 211147 |
+
{
|
| 211148 |
+
"epoch": 702.0,
|
| 211149 |
+
"eval_loss": 0.41614237427711487,
|
| 211150 |
+
"eval_runtime": 42.5764,
|
| 211151 |
+
"eval_samples_per_second": 19.729,
|
| 211152 |
+
"eval_steps_per_second": 0.634,
|
| 211153 |
+
"eval_wer": 0.18335028330669767,
|
| 211154 |
+
"step": 87862
|
| 211155 |
+
},
|
| 211156 |
+
{
|
| 211157 |
+
"epoch": 708.02,
|
| 211158 |
+
"learning_rate": 8.597269789983845e-06,
|
| 211159 |
+
"loss": 0.4847,
|
| 211160 |
+
"step": 87865
|
| 211161 |
+
},
|
| 211162 |
+
{
|
| 211163 |
+
"epoch": 708.06,
|
| 211164 |
+
"learning_rate": 8.597189014539581e-06,
|
| 211165 |
+
"loss": 0.2719,
|
| 211166 |
+
"step": 87870
|
| 211167 |
+
},
|
| 211168 |
+
{
|
| 211169 |
+
"epoch": 708.1,
|
| 211170 |
+
"learning_rate": 8.597108239095315e-06,
|
| 211171 |
+
"loss": 0.2781,
|
| 211172 |
+
"step": 87875
|
| 211173 |
+
},
|
| 211174 |
+
{
|
| 211175 |
+
"epoch": 708.15,
|
| 211176 |
+
"learning_rate": 8.597027463651051e-06,
|
| 211177 |
+
"loss": 0.3696,
|
| 211178 |
+
"step": 87880
|
| 211179 |
+
},
|
| 211180 |
+
{
|
| 211181 |
+
"epoch": 708.19,
|
| 211182 |
+
"learning_rate": 8.596946688206785e-06,
|
| 211183 |
+
"loss": 0.7388,
|
| 211184 |
+
"step": 87885
|
| 211185 |
+
},
|
| 211186 |
+
{
|
| 211187 |
+
"epoch": 708.23,
|
| 211188 |
+
"learning_rate": 8.596865912762521e-06,
|
| 211189 |
+
"loss": 0.8541,
|
| 211190 |
+
"step": 87890
|
| 211191 |
+
},
|
| 211192 |
+
{
|
| 211193 |
+
"epoch": 708.27,
|
| 211194 |
+
"learning_rate": 8.596785137318257e-06,
|
| 211195 |
+
"loss": 0.2591,
|
| 211196 |
+
"step": 87895
|
| 211197 |
+
},
|
| 211198 |
+
{
|
| 211199 |
+
"epoch": 708.31,
|
| 211200 |
+
"learning_rate": 8.596704361873991e-06,
|
| 211201 |
+
"loss": 0.3406,
|
| 211202 |
+
"step": 87900
|
| 211203 |
+
},
|
| 211204 |
+
{
|
| 211205 |
+
"epoch": 708.35,
|
| 211206 |
+
"learning_rate": 8.596623586429727e-06,
|
| 211207 |
+
"loss": 0.3502,
|
| 211208 |
+
"step": 87905
|
| 211209 |
+
},
|
| 211210 |
+
{
|
| 211211 |
+
"epoch": 708.39,
|
| 211212 |
+
"learning_rate": 8.59654281098546e-06,
|
| 211213 |
+
"loss": 0.7739,
|
| 211214 |
+
"step": 87910
|
| 211215 |
+
},
|
| 211216 |
+
{
|
| 211217 |
+
"epoch": 708.43,
|
| 211218 |
+
"learning_rate": 8.596462035541197e-06,
|
| 211219 |
+
"loss": 0.7964,
|
| 211220 |
+
"step": 87915
|
| 211221 |
+
},
|
| 211222 |
+
{
|
| 211223 |
+
"epoch": 708.47,
|
| 211224 |
+
"learning_rate": 8.59638126009693e-06,
|
| 211225 |
+
"loss": 0.2839,
|
| 211226 |
+
"step": 87920
|
| 211227 |
+
},
|
| 211228 |
+
{
|
| 211229 |
+
"epoch": 708.51,
|
| 211230 |
+
"learning_rate": 8.596300484652667e-06,
|
| 211231 |
+
"loss": 0.3608,
|
| 211232 |
+
"step": 87925
|
| 211233 |
+
},
|
| 211234 |
+
{
|
| 211235 |
+
"epoch": 708.55,
|
| 211236 |
+
"learning_rate": 8.5962197092084e-06,
|
| 211237 |
+
"loss": 0.3517,
|
| 211238 |
+
"step": 87930
|
| 211239 |
+
},
|
| 211240 |
+
{
|
| 211241 |
+
"epoch": 708.59,
|
| 211242 |
+
"learning_rate": 8.596138933764137e-06,
|
| 211243 |
+
"loss": 0.6862,
|
| 211244 |
+
"step": 87935
|
| 211245 |
+
},
|
| 211246 |
+
{
|
| 211247 |
+
"epoch": 708.63,
|
| 211248 |
+
"learning_rate": 8.59605815831987e-06,
|
| 211249 |
+
"loss": 0.8938,
|
| 211250 |
+
"step": 87940
|
| 211251 |
+
},
|
| 211252 |
+
{
|
| 211253 |
+
"epoch": 708.67,
|
| 211254 |
+
"learning_rate": 8.595977382875607e-06,
|
| 211255 |
+
"loss": 0.2593,
|
| 211256 |
+
"step": 87945
|
| 211257 |
+
},
|
| 211258 |
+
{
|
| 211259 |
+
"epoch": 708.71,
|
| 211260 |
+
"learning_rate": 8.59589660743134e-06,
|
| 211261 |
+
"loss": 0.2619,
|
| 211262 |
+
"step": 87950
|
| 211263 |
+
},
|
| 211264 |
+
{
|
| 211265 |
+
"epoch": 708.75,
|
| 211266 |
+
"learning_rate": 8.595815831987077e-06,
|
| 211267 |
+
"loss": 0.403,
|
| 211268 |
+
"step": 87955
|
| 211269 |
+
},
|
| 211270 |
+
{
|
| 211271 |
+
"epoch": 708.79,
|
| 211272 |
+
"learning_rate": 8.595735056542812e-06,
|
| 211273 |
+
"loss": 0.816,
|
| 211274 |
+
"step": 87960
|
| 211275 |
+
},
|
| 211276 |
+
{
|
| 211277 |
+
"epoch": 708.83,
|
| 211278 |
+
"learning_rate": 8.595654281098547e-06,
|
| 211279 |
+
"loss": 0.9535,
|
| 211280 |
+
"step": 87965
|
| 211281 |
+
},
|
| 211282 |
+
{
|
| 211283 |
+
"epoch": 708.87,
|
| 211284 |
+
"learning_rate": 8.595573505654282e-06,
|
| 211285 |
+
"loss": 0.3413,
|
| 211286 |
+
"step": 87970
|
| 211287 |
+
},
|
| 211288 |
+
{
|
| 211289 |
+
"epoch": 708.91,
|
| 211290 |
+
"learning_rate": 8.595492730210017e-06,
|
| 211291 |
+
"loss": 0.3064,
|
| 211292 |
+
"step": 87975
|
| 211293 |
+
},
|
| 211294 |
+
{
|
| 211295 |
+
"epoch": 708.95,
|
| 211296 |
+
"learning_rate": 8.595411954765752e-06,
|
| 211297 |
+
"loss": 0.4452,
|
| 211298 |
+
"step": 87980
|
| 211299 |
+
},
|
| 211300 |
+
{
|
| 211301 |
+
"epoch": 708.99,
|
| 211302 |
+
"learning_rate": 8.595331179321487e-06,
|
| 211303 |
+
"loss": 0.9466,
|
| 211304 |
+
"step": 87985
|
| 211305 |
+
},
|
| 211306 |
+
{
|
| 211307 |
+
"epoch": 709.0,
|
| 211308 |
+
"eval_loss": 0.33699852228164673,
|
| 211309 |
+
"eval_runtime": 41.8576,
|
| 211310 |
+
"eval_samples_per_second": 20.068,
|
| 211311 |
+
"eval_steps_per_second": 0.645,
|
| 211312 |
+
"eval_wer": 0.18305133147711858,
|
| 211313 |
+
"step": 87986
|
| 211314 |
+
},
|
| 211315 |
+
{
|
| 211316 |
+
"epoch": 703.03,
|
| 211317 |
+
"learning_rate": 8.595250403877222e-06,
|
| 211318 |
+
"loss": 0.3072,
|
| 211319 |
+
"step": 87990
|
| 211320 |
+
},
|
| 211321 |
+
{
|
| 211322 |
+
"epoch": 703.07,
|
| 211323 |
+
"learning_rate": 8.595169628432956e-06,
|
| 211324 |
+
"loss": 0.3646,
|
| 211325 |
+
"step": 87995
|
| 211326 |
+
},
|
| 211327 |
+
{
|
| 211328 |
+
"epoch": 703.11,
|
| 211329 |
+
"learning_rate": 8.595088852988692e-06,
|
| 211330 |
+
"loss": 0.3119,
|
| 211331 |
+
"step": 88000
|
| 211332 |
+
},
|
| 211333 |
+
{
|
| 211334 |
+
"epoch": 703.15,
|
| 211335 |
+
"learning_rate": 8.595008077544426e-06,
|
| 211336 |
+
"loss": 0.4022,
|
| 211337 |
+
"step": 88005
|
| 211338 |
+
},
|
| 211339 |
+
{
|
| 211340 |
+
"epoch": 703.19,
|
| 211341 |
+
"learning_rate": 8.594927302100162e-06,
|
| 211342 |
+
"loss": 1.075,
|
| 211343 |
+
"step": 88010
|
| 211344 |
+
},
|
| 211345 |
+
{
|
| 211346 |
+
"epoch": 703.23,
|
| 211347 |
+
"learning_rate": 8.594846526655898e-06,
|
| 211348 |
+
"loss": 0.6486,
|
| 211349 |
+
"step": 88015
|
| 211350 |
+
},
|
| 211351 |
+
{
|
| 211352 |
+
"epoch": 703.27,
|
| 211353 |
+
"learning_rate": 8.594765751211632e-06,
|
| 211354 |
+
"loss": 0.2992,
|
| 211355 |
+
"step": 88020
|
| 211356 |
+
},
|
| 211357 |
+
{
|
| 211358 |
+
"epoch": 703.31,
|
| 211359 |
+
"learning_rate": 8.594684975767368e-06,
|
| 211360 |
+
"loss": 0.3578,
|
| 211361 |
+
"step": 88025
|
| 211362 |
+
},
|
| 211363 |
+
{
|
| 211364 |
+
"epoch": 703.35,
|
| 211365 |
+
"learning_rate": 8.594604200323102e-06,
|
| 211366 |
+
"loss": 0.4427,
|
| 211367 |
+
"step": 88030
|
| 211368 |
+
},
|
| 211369 |
+
{
|
| 211370 |
+
"epoch": 703.39,
|
| 211371 |
+
"learning_rate": 8.594523424878838e-06,
|
| 211372 |
+
"loss": 0.8995,
|
| 211373 |
+
"step": 88035
|
| 211374 |
+
},
|
| 211375 |
+
{
|
| 211376 |
+
"epoch": 703.43,
|
| 211377 |
+
"learning_rate": 8.594442649434572e-06,
|
| 211378 |
+
"loss": 0.6611,
|
| 211379 |
+
"step": 88040
|
| 211380 |
+
},
|
| 211381 |
+
{
|
| 211382 |
+
"epoch": 703.47,
|
| 211383 |
+
"learning_rate": 8.594361873990308e-06,
|
| 211384 |
+
"loss": 0.3466,
|
| 211385 |
+
"step": 88045
|
| 211386 |
+
},
|
| 211387 |
+
{
|
| 211388 |
+
"epoch": 703.51,
|
| 211389 |
+
"learning_rate": 8.594281098546042e-06,
|
| 211390 |
+
"loss": 0.3477,
|
| 211391 |
+
"step": 88050
|
| 211392 |
+
},
|
| 211393 |
+
{
|
| 211394 |
+
"epoch": 703.55,
|
| 211395 |
+
"learning_rate": 8.594200323101778e-06,
|
| 211396 |
+
"loss": 0.47,
|
| 211397 |
+
"step": 88055
|
| 211398 |
+
},
|
| 211399 |
+
{
|
| 211400 |
+
"epoch": 703.59,
|
| 211401 |
+
"learning_rate": 8.594119547657512e-06,
|
| 211402 |
+
"loss": 0.9951,
|
| 211403 |
+
"step": 88060
|
| 211404 |
+
},
|
| 211405 |
+
{
|
| 211406 |
+
"epoch": 703.63,
|
| 211407 |
+
"learning_rate": 8.594038772213248e-06,
|
| 211408 |
+
"loss": 0.7673,
|
| 211409 |
+
"step": 88065
|
| 211410 |
+
},
|
| 211411 |
+
{
|
| 211412 |
+
"epoch": 703.67,
|
| 211413 |
+
"learning_rate": 8.593957996768984e-06,
|
| 211414 |
+
"loss": 0.343,
|
| 211415 |
+
"step": 88070
|
| 211416 |
+
},
|
| 211417 |
+
{
|
| 211418 |
+
"epoch": 703.71,
|
| 211419 |
+
"learning_rate": 8.593877221324718e-06,
|
| 211420 |
+
"loss": 0.333,
|
| 211421 |
+
"step": 88075
|
| 211422 |
+
},
|
| 211423 |
+
{
|
| 211424 |
+
"epoch": 703.75,
|
| 211425 |
+
"learning_rate": 8.593796445880454e-06,
|
| 211426 |
+
"loss": 0.3409,
|
| 211427 |
+
"step": 88080
|
| 211428 |
+
},
|
| 211429 |
+
{
|
| 211430 |
+
"epoch": 703.79,
|
| 211431 |
+
"learning_rate": 8.593715670436188e-06,
|
| 211432 |
+
"loss": 1.0327,
|
| 211433 |
+
"step": 88085
|
| 211434 |
+
},
|
| 211435 |
+
{
|
| 211436 |
+
"epoch": 703.83,
|
| 211437 |
+
"learning_rate": 8.593634894991924e-06,
|
| 211438 |
+
"loss": 0.6733,
|
| 211439 |
+
"step": 88090
|
| 211440 |
+
},
|
| 211441 |
+
{
|
| 211442 |
+
"epoch": 703.87,
|
| 211443 |
+
"learning_rate": 8.593554119547658e-06,
|
| 211444 |
+
"loss": 0.2969,
|
| 211445 |
+
"step": 88095
|
| 211446 |
+
},
|
| 211447 |
+
{
|
| 211448 |
+
"epoch": 703.91,
|
| 211449 |
+
"learning_rate": 8.593473344103394e-06,
|
| 211450 |
+
"loss": 0.3121,
|
| 211451 |
+
"step": 88100
|
| 211452 |
+
},
|
| 211453 |
+
{
|
| 211454 |
+
"epoch": 703.95,
|
| 211455 |
+
"learning_rate": 8.593392568659128e-06,
|
| 211456 |
+
"loss": 0.4129,
|
| 211457 |
+
"step": 88105
|
| 211458 |
+
},
|
| 211459 |
+
{
|
| 211460 |
+
"epoch": 703.99,
|
| 211461 |
+
"learning_rate": 8.593311793214864e-06,
|
| 211462 |
+
"loss": 1.1018,
|
| 211463 |
+
"step": 88110
|
| 211464 |
+
},
|
| 211465 |
+
{
|
| 211466 |
+
"epoch": 704.0,
|
| 211467 |
+
"eval_loss": 0.326661616563797,
|
| 211468 |
+
"eval_runtime": 41.1502,
|
| 211469 |
+
"eval_samples_per_second": 20.389,
|
| 211470 |
+
"eval_steps_per_second": 0.656,
|
| 211471 |
+
"eval_wer": 0.1878322504430006,
|
| 211472 |
+
"step": 88111
|
| 211473 |
+
},
|
| 211474 |
+
{
|
| 211475 |
+
"epoch": 710.03,
|
| 211476 |
+
"learning_rate": 8.593231017770598e-06,
|
| 211477 |
+
"loss": 0.3615,
|
| 211478 |
+
"step": 88115
|
| 211479 |
+
},
|
| 211480 |
+
{
|
| 211481 |
+
"epoch": 710.07,
|
| 211482 |
+
"learning_rate": 8.593150242326334e-06,
|
| 211483 |
+
"loss": 0.2882,
|
| 211484 |
+
"step": 88120
|
| 211485 |
+
},
|
| 211486 |
+
{
|
| 211487 |
+
"epoch": 710.11,
|
| 211488 |
+
"learning_rate": 8.593069466882068e-06,
|
| 211489 |
+
"loss": 0.2709,
|
| 211490 |
+
"step": 88125
|
| 211491 |
+
},
|
| 211492 |
+
{
|
| 211493 |
+
"epoch": 710.15,
|
| 211494 |
+
"learning_rate": 8.592988691437804e-06,
|
| 211495 |
+
"loss": 0.4074,
|
| 211496 |
+
"step": 88130
|
| 211497 |
+
},
|
| 211498 |
+
{
|
| 211499 |
+
"epoch": 710.19,
|
| 211500 |
+
"learning_rate": 8.59290791599354e-06,
|
| 211501 |
+
"loss": 0.8489,
|
| 211502 |
+
"step": 88135
|
| 211503 |
+
},
|
| 211504 |
+
{
|
| 211505 |
+
"epoch": 710.23,
|
| 211506 |
+
"learning_rate": 8.592827140549274e-06,
|
| 211507 |
+
"loss": 0.6687,
|
| 211508 |
+
"step": 88140
|
| 211509 |
+
},
|
| 211510 |
+
{
|
| 211511 |
+
"epoch": 710.27,
|
| 211512 |
+
"learning_rate": 8.59274636510501e-06,
|
| 211513 |
+
"loss": 0.3755,
|
| 211514 |
+
"step": 88145
|
| 211515 |
+
},
|
| 211516 |
+
{
|
| 211517 |
+
"epoch": 710.31,
|
| 211518 |
+
"learning_rate": 8.592665589660744e-06,
|
| 211519 |
+
"loss": 0.2988,
|
| 211520 |
+
"step": 88150
|
| 211521 |
+
},
|
| 211522 |
+
{
|
| 211523 |
+
"epoch": 710.35,
|
| 211524 |
+
"learning_rate": 8.59258481421648e-06,
|
| 211525 |
+
"loss": 0.464,
|
| 211526 |
+
"step": 88155
|
| 211527 |
+
},
|
| 211528 |
+
{
|
| 211529 |
+
"epoch": 710.4,
|
| 211530 |
+
"learning_rate": 8.592504038772214e-06,
|
| 211531 |
+
"loss": 1.1016,
|
| 211532 |
+
"step": 88160
|
| 211533 |
+
},
|
| 211534 |
+
{
|
| 211535 |
+
"epoch": 710.44,
|
| 211536 |
+
"learning_rate": 8.59242326332795e-06,
|
| 211537 |
+
"loss": 0.6133,
|
| 211538 |
+
"step": 88165
|
| 211539 |
+
},
|
| 211540 |
+
{
|
| 211541 |
+
"epoch": 710.48,
|
| 211542 |
+
"learning_rate": 8.592342487883684e-06,
|
| 211543 |
+
"loss": 0.2555,
|
| 211544 |
+
"step": 88170
|
| 211545 |
+
},
|
| 211546 |
+
{
|
| 211547 |
+
"epoch": 710.52,
|
| 211548 |
+
"learning_rate": 8.59226171243942e-06,
|
| 211549 |
+
"loss": 0.3477,
|
| 211550 |
+
"step": 88175
|
| 211551 |
+
},
|
| 211552 |
+
{
|
| 211553 |
+
"epoch": 710.56,
|
| 211554 |
+
"learning_rate": 8.592180936995154e-06,
|
| 211555 |
+
"loss": 0.3306,
|
| 211556 |
+
"step": 88180
|
| 211557 |
+
},
|
| 211558 |
+
{
|
| 211559 |
+
"epoch": 710.6,
|
| 211560 |
+
"learning_rate": 8.59210016155089e-06,
|
| 211561 |
+
"loss": 1.0426,
|
| 211562 |
+
"step": 88185
|
| 211563 |
+
},
|
| 211564 |
+
{
|
| 211565 |
+
"epoch": 710.64,
|
| 211566 |
+
"learning_rate": 8.592019386106625e-06,
|
| 211567 |
+
"loss": 0.6553,
|
| 211568 |
+
"step": 88190
|
| 211569 |
+
},
|
| 211570 |
+
{
|
| 211571 |
+
"epoch": 710.68,
|
| 211572 |
+
"learning_rate": 8.59193861066236e-06,
|
| 211573 |
+
"loss": 0.284,
|
| 211574 |
+
"step": 88195
|
| 211575 |
+
},
|
| 211576 |
+
{
|
| 211577 |
+
"epoch": 710.72,
|
| 211578 |
+
"learning_rate": 8.591857835218095e-06,
|
| 211579 |
+
"loss": 0.4124,
|
| 211580 |
+
"step": 88200
|
| 211581 |
+
},
|
| 211582 |
+
{
|
| 211583 |
+
"epoch": 710.76,
|
| 211584 |
+
"learning_rate": 8.59177705977383e-06,
|
| 211585 |
+
"loss": 0.3958,
|
| 211586 |
+
"step": 88205
|
| 211587 |
+
},
|
| 211588 |
+
{
|
| 211589 |
+
"epoch": 710.8,
|
| 211590 |
+
"learning_rate": 8.591696284329565e-06,
|
| 211591 |
+
"loss": 0.9705,
|
| 211592 |
+
"step": 88210
|
| 211593 |
+
},
|
| 211594 |
+
{
|
| 211595 |
+
"epoch": 710.84,
|
| 211596 |
+
"learning_rate": 8.5916155088853e-06,
|
| 211597 |
+
"loss": 0.625,
|
| 211598 |
+
"step": 88215
|
| 211599 |
+
},
|
| 211600 |
+
{
|
| 211601 |
+
"epoch": 710.88,
|
| 211602 |
+
"learning_rate": 8.591534733441035e-06,
|
| 211603 |
+
"loss": 0.2431,
|
| 211604 |
+
"step": 88220
|
| 211605 |
+
},
|
| 211606 |
+
{
|
| 211607 |
+
"epoch": 710.92,
|
| 211608 |
+
"learning_rate": 8.59145395799677e-06,
|
| 211609 |
+
"loss": 0.2842,
|
| 211610 |
+
"step": 88225
|
| 211611 |
+
},
|
| 211612 |
+
{
|
| 211613 |
+
"epoch": 710.96,
|
| 211614 |
+
"learning_rate": 8.591373182552505e-06,
|
| 211615 |
+
"loss": 0.4265,
|
| 211616 |
+
"step": 88230
|
| 211617 |
+
},
|
| 211618 |
+
{
|
| 211619 |
+
"epoch": 711.0,
|
| 211620 |
+
"learning_rate": 8.59129240710824e-06,
|
| 211621 |
+
"loss": 1.0673,
|
| 211622 |
+
"step": 88235
|
| 211623 |
+
},
|
| 211624 |
+
{
|
| 211625 |
+
"epoch": 711.0,
|
| 211626 |
+
"eval_loss": 0.34067657589912415,
|
| 211627 |
+
"eval_runtime": 40.9052,
|
| 211628 |
+
"eval_samples_per_second": 20.511,
|
| 211629 |
+
"eval_steps_per_second": 0.66,
|
| 211630 |
+
"eval_wer": 0.18349353183493533,
|
| 211631 |
+
"step": 88235
|
| 211632 |
+
},
|
| 211633 |
+
{
|
| 211634 |
+
"epoch": 711.04,
|
| 211635 |
+
"learning_rate": 8.591211631663975e-06,
|
| 211636 |
+
"loss": 0.3905,
|
| 211637 |
+
"step": 88240
|
| 211638 |
+
},
|
| 211639 |
+
{
|
| 211640 |
+
"epoch": 711.08,
|
| 211641 |
+
"learning_rate": 8.591130856219711e-06,
|
| 211642 |
+
"loss": 0.298,
|
| 211643 |
+
"step": 88245
|
| 211644 |
+
},
|
| 211645 |
+
{
|
| 211646 |
+
"epoch": 711.12,
|
| 211647 |
+
"learning_rate": 8.591050080775445e-06,
|
| 211648 |
+
"loss": 0.2672,
|
| 211649 |
+
"step": 88250
|
| 211650 |
+
},
|
| 211651 |
+
{
|
| 211652 |
+
"epoch": 711.16,
|
| 211653 |
+
"learning_rate": 8.590969305331181e-06,
|
| 211654 |
+
"loss": 0.5272,
|
| 211655 |
+
"step": 88255
|
| 211656 |
+
},
|
| 211657 |
+
{
|
| 211658 |
+
"epoch": 711.2,
|
| 211659 |
+
"learning_rate": 8.590888529886915e-06,
|
| 211660 |
+
"loss": 1.2542,
|
| 211661 |
+
"step": 88260
|
| 211662 |
+
},
|
| 211663 |
+
{
|
| 211664 |
+
"epoch": 711.24,
|
| 211665 |
+
"learning_rate": 8.590807754442651e-06,
|
| 211666 |
+
"loss": 0.3318,
|
| 211667 |
+
"step": 88265
|
| 211668 |
+
},
|
| 211669 |
+
{
|
| 211670 |
+
"epoch": 711.28,
|
| 211671 |
+
"learning_rate": 8.590726978998385e-06,
|
| 211672 |
+
"loss": 0.3141,
|
| 211673 |
+
"step": 88270
|
| 211674 |
+
},
|
| 211675 |
+
{
|
| 211676 |
+
"epoch": 711.32,
|
| 211677 |
+
"learning_rate": 8.590646203554121e-06,
|
| 211678 |
+
"loss": 0.3081,
|
| 211679 |
+
"step": 88275
|
| 211680 |
+
},
|
| 211681 |
+
{
|
| 211682 |
+
"epoch": 711.36,
|
| 211683 |
+
"learning_rate": 8.590565428109855e-06,
|
| 211684 |
+
"loss": 0.492,
|
| 211685 |
+
"step": 88280
|
| 211686 |
+
},
|
| 211687 |
+
{
|
| 211688 |
+
"epoch": 711.4,
|
| 211689 |
+
"learning_rate": 8.59048465266559e-06,
|
| 211690 |
+
"loss": 1.1349,
|
| 211691 |
+
"step": 88285
|
| 211692 |
+
},
|
| 211693 |
+
{
|
| 211694 |
+
"epoch": 711.44,
|
| 211695 |
+
"learning_rate": 8.590403877221325e-06,
|
| 211696 |
+
"loss": 0.3078,
|
| 211697 |
+
"step": 88290
|
| 211698 |
+
},
|
| 211699 |
+
{
|
| 211700 |
+
"epoch": 711.48,
|
| 211701 |
+
"learning_rate": 8.59032310177706e-06,
|
| 211702 |
+
"loss": 0.3303,
|
| 211703 |
+
"step": 88295
|
| 211704 |
+
},
|
| 211705 |
+
{
|
| 211706 |
+
"epoch": 711.52,
|
| 211707 |
+
"learning_rate": 8.590242326332795e-06,
|
| 211708 |
+
"loss": 0.3825,
|
| 211709 |
+
"step": 88300
|
| 211710 |
+
},
|
| 211711 |
+
{
|
| 211712 |
+
"epoch": 711.56,
|
| 211713 |
+
"learning_rate": 8.59016155088853e-06,
|
| 211714 |
+
"loss": 0.5498,
|
| 211715 |
+
"step": 88305
|
| 211716 |
+
},
|
| 211717 |
+
{
|
| 211718 |
+
"epoch": 711.6,
|
| 211719 |
+
"learning_rate": 8.590080775444267e-06,
|
| 211720 |
+
"loss": 1.1221,
|
| 211721 |
+
"step": 88310
|
| 211722 |
+
},
|
| 211723 |
+
{
|
| 211724 |
+
"epoch": 711.64,
|
| 211725 |
+
"learning_rate": 8.59e-06,
|
| 211726 |
+
"loss": 0.3378,
|
| 211727 |
+
"step": 88315
|
| 211728 |
+
},
|
| 211729 |
+
{
|
| 211730 |
+
"epoch": 711.68,
|
| 211731 |
+
"learning_rate": 8.589919224555737e-06,
|
| 211732 |
+
"loss": 0.3251,
|
| 211733 |
+
"step": 88320
|
| 211734 |
+
},
|
| 211735 |
+
{
|
| 211736 |
+
"epoch": 711.72,
|
| 211737 |
+
"learning_rate": 8.58983844911147e-06,
|
| 211738 |
+
"loss": 0.2951,
|
| 211739 |
+
"step": 88325
|
| 211740 |
+
},
|
| 211741 |
+
{
|
| 211742 |
+
"epoch": 711.76,
|
| 211743 |
+
"learning_rate": 8.589757673667207e-06,
|
| 211744 |
+
"loss": 0.5104,
|
| 211745 |
+
"step": 88330
|
| 211746 |
+
},
|
| 211747 |
+
{
|
| 211748 |
+
"epoch": 711.8,
|
| 211749 |
+
"learning_rate": 8.58967689822294e-06,
|
| 211750 |
+
"loss": 1.2468,
|
| 211751 |
+
"step": 88335
|
| 211752 |
+
},
|
| 211753 |
+
{
|
| 211754 |
+
"epoch": 711.84,
|
| 211755 |
+
"learning_rate": 8.589596122778677e-06,
|
| 211756 |
+
"loss": 0.3304,
|
| 211757 |
+
"step": 88340
|
| 211758 |
+
},
|
| 211759 |
+
{
|
| 211760 |
+
"epoch": 711.88,
|
| 211761 |
+
"learning_rate": 8.58951534733441e-06,
|
| 211762 |
+
"loss": 0.2819,
|
| 211763 |
+
"step": 88345
|
| 211764 |
+
},
|
| 211765 |
+
{
|
| 211766 |
+
"epoch": 711.92,
|
| 211767 |
+
"learning_rate": 8.589434571890147e-06,
|
| 211768 |
+
"loss": 0.3221,
|
| 211769 |
+
"step": 88350
|
| 211770 |
+
},
|
| 211771 |
+
{
|
| 211772 |
+
"epoch": 711.96,
|
| 211773 |
+
"learning_rate": 8.58935379644588e-06,
|
| 211774 |
+
"loss": 0.5135,
|
| 211775 |
+
"step": 88355
|
| 211776 |
+
},
|
| 211777 |
+
{
|
| 211778 |
+
"epoch": 712.0,
|
| 211779 |
+
"eval_loss": 0.4080963432788849,
|
| 211780 |
+
"eval_runtime": 41.1853,
|
| 211781 |
+
"eval_samples_per_second": 20.371,
|
| 211782 |
+
"eval_steps_per_second": 0.656,
|
| 211783 |
+
"eval_wer": 0.189195053156867,
|
| 211784 |
+
"step": 88359
|
| 211785 |
}
|
| 211786 |
],
|
| 211787 |
"max_steps": 620000,
|
| 211788 |
"num_train_epochs": 5000,
|
| 211789 |
+
"total_flos": 2.486625232987812e+20,
|
| 211790 |
"trial_name": null,
|
| 211791 |
"trial_params": null
|
| 211792 |
}
|
model-bin/finetune/base/{checkpoint-87737 → checkpoint-88359}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629933091.5883334/events.out.tfevents.1629933091.7e498afd5545.7645.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f80e805d66a0c42e268f1374c0033838d9e818a3a9ae9013e8f5edd45fc424a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629933593.7463934/events.out.tfevents.1629933593.7e498afd5545.7645.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9e461bc668a8a8e6cd57d1c56e118d77e725e060e72e3c33bd2d66529c5e7fbd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629934073.4499795/events.out.tfevents.1629934073.7e498afd5545.7645.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a963d7a5e8671893ba22ec24ebe5a131ab44ac1563ea41adbbea3ac13f078b93
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629934533.9362981/events.out.tfevents.1629934535.7e498afd5545.7645.191
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9344b5f64bd06a6e0708334790dff4c9b1c7aef33c879a993c5a6959437989d6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629934991.9355497/events.out.tfevents.1629934991.7e498afd5545.7645.193
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bad719d5a364796215a916cc33940bb4d804a93ece6ecbab574f6d0d96785337
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629933091.7e498afd5545.7645.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c98f83bd8f2bb15e8596d0ea6582f9ceee23c0d6f77f570365b263f7cb6b238a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629933593.7e498afd5545.7645.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95ee969fc8f9ed6499751c1ff0fed13290b20646aa0c45fede7bddcf9c64e1a2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629934072.7e498afd5545.7645.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2453f95e192a2ad043142f0191a1e8bebe9c69cad8ea8f78f10f20ce184c0683
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629934533.7e498afd5545.7645.190
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:80cc3707e4daea38efbac5c798f35ac8c6abe6fd326a36c168c6222dc3f8cdc0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629934991.7e498afd5545.7645.192
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b997b28ed9e78ec5ae13b8cac450aa5041425a47698410f1f929242c6be552f
|
| 3 |
+
size 8462
|