"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/optimizer.pt +2 -2
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629824464.0486395/events.out.tfevents.1629824464.c435e1c5ee04.920.181 +3 -0
- model-bin/finetune/base/log/1629825098.0807111/events.out.tfevents.1629825098.c435e1c5ee04.920.183 +3 -0
- model-bin/finetune/base/log/1629825728.7070112/events.out.tfevents.1629825728.c435e1c5ee04.920.185 +3 -0
- model-bin/finetune/base/log/1629826356.1797123/events.out.tfevents.1629826356.c435e1c5ee04.920.187 +3 -0
- model-bin/finetune/base/log/1629826982.8940318/events.out.tfevents.1629826982.c435e1c5ee04.920.189 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629824464.c435e1c5ee04.920.180 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629825098.c435e1c5ee04.920.182 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629825728.c435e1c5ee04.920.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629826356.c435e1c5ee04.920.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629826982.c435e1c5ee04.920.188 +3 -0
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/optimizer.pt
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:dd06fe640e23974d0f6a0907f379f9ed0b9d1d39aa9e878b427221d68e27fb0c
|
| 3 |
+
size 722165393
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/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 sha256:9ca37dc64921b379bc597984828ccb1ffca37d2bfba4dc65a0bf832d7fb196fc
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/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 sha256:21a9d2cdf11457adf20e12a9253bdd90f36c56e3655c2a2506c52983a83e6230
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/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:d7219f96e098583795db3c9787cf815cf9738bb70a1274c9695498cf865dfb3c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/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 sha256:7ba2fa47ca503d1240381ecfce0ecd5428d683eb16b4f83b6f890ff832b6f62f
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -182169,11 +182169,806 @@
|
|
| 182169 |
"eval_steps_per_second": 0.67,
|
| 182170 |
"eval_wer": 0.19669648009232546,
|
| 182171 |
"step": 65082
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182172 |
}
|
| 182173 |
],
|
| 182174 |
-
"max_steps":
|
| 182175 |
"num_train_epochs": 5000,
|
| 182176 |
-
"total_flos": 1.
|
| 182177 |
"trial_name": null,
|
| 182178 |
"trial_params": null
|
| 182179 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 525.0,
|
| 5 |
+
"global_step": 65705,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 182169 |
"eval_steps_per_second": 0.67,
|
| 182170 |
"eval_wer": 0.19669648009232546,
|
| 182171 |
"step": 65082
|
| 182172 |
+
},
|
| 182173 |
+
{
|
| 182174 |
+
"epoch": 520.02,
|
| 182175 |
+
"learning_rate": 8.973397435897437e-06,
|
| 182176 |
+
"loss": 0.3888,
|
| 182177 |
+
"step": 65085
|
| 182178 |
+
},
|
| 182179 |
+
{
|
| 182180 |
+
"epoch": 520.06,
|
| 182181 |
+
"learning_rate": 8.973317307692308e-06,
|
| 182182 |
+
"loss": 0.3305,
|
| 182183 |
+
"step": 65090
|
| 182184 |
+
},
|
| 182185 |
+
{
|
| 182186 |
+
"epoch": 520.1,
|
| 182187 |
+
"learning_rate": 8.97323717948718e-06,
|
| 182188 |
+
"loss": 0.3337,
|
| 182189 |
+
"step": 65095
|
| 182190 |
+
},
|
| 182191 |
+
{
|
| 182192 |
+
"epoch": 520.14,
|
| 182193 |
+
"learning_rate": 8.973157051282053e-06,
|
| 182194 |
+
"loss": 0.3415,
|
| 182195 |
+
"step": 65100
|
| 182196 |
+
},
|
| 182197 |
+
{
|
| 182198 |
+
"epoch": 520.18,
|
| 182199 |
+
"learning_rate": 8.973076923076924e-06,
|
| 182200 |
+
"loss": 0.7668,
|
| 182201 |
+
"step": 65105
|
| 182202 |
+
},
|
| 182203 |
+
{
|
| 182204 |
+
"epoch": 520.22,
|
| 182205 |
+
"learning_rate": 8.972996794871795e-06,
|
| 182206 |
+
"loss": 0.823,
|
| 182207 |
+
"step": 65110
|
| 182208 |
+
},
|
| 182209 |
+
{
|
| 182210 |
+
"epoch": 520.26,
|
| 182211 |
+
"learning_rate": 8.972916666666668e-06,
|
| 182212 |
+
"loss": 0.2872,
|
| 182213 |
+
"step": 65115
|
| 182214 |
+
},
|
| 182215 |
+
{
|
| 182216 |
+
"epoch": 520.3,
|
| 182217 |
+
"learning_rate": 8.97283653846154e-06,
|
| 182218 |
+
"loss": 0.3716,
|
| 182219 |
+
"step": 65120
|
| 182220 |
+
},
|
| 182221 |
+
{
|
| 182222 |
+
"epoch": 520.34,
|
| 182223 |
+
"learning_rate": 8.972756410256411e-06,
|
| 182224 |
+
"loss": 0.3787,
|
| 182225 |
+
"step": 65125
|
| 182226 |
+
},
|
| 182227 |
+
{
|
| 182228 |
+
"epoch": 520.38,
|
| 182229 |
+
"learning_rate": 8.972676282051283e-06,
|
| 182230 |
+
"loss": 0.7423,
|
| 182231 |
+
"step": 65130
|
| 182232 |
+
},
|
| 182233 |
+
{
|
| 182234 |
+
"epoch": 520.42,
|
| 182235 |
+
"learning_rate": 8.972596153846156e-06,
|
| 182236 |
+
"loss": 0.9309,
|
| 182237 |
+
"step": 65135
|
| 182238 |
+
},
|
| 182239 |
+
{
|
| 182240 |
+
"epoch": 520.46,
|
| 182241 |
+
"learning_rate": 8.972516025641025e-06,
|
| 182242 |
+
"loss": 0.3168,
|
| 182243 |
+
"step": 65140
|
| 182244 |
+
},
|
| 182245 |
+
{
|
| 182246 |
+
"epoch": 520.5,
|
| 182247 |
+
"learning_rate": 8.972435897435898e-06,
|
| 182248 |
+
"loss": 0.3513,
|
| 182249 |
+
"step": 65145
|
| 182250 |
+
},
|
| 182251 |
+
{
|
| 182252 |
+
"epoch": 520.54,
|
| 182253 |
+
"learning_rate": 8.97235576923077e-06,
|
| 182254 |
+
"loss": 0.4328,
|
| 182255 |
+
"step": 65150
|
| 182256 |
+
},
|
| 182257 |
+
{
|
| 182258 |
+
"epoch": 520.58,
|
| 182259 |
+
"learning_rate": 8.972275641025641e-06,
|
| 182260 |
+
"loss": 0.8416,
|
| 182261 |
+
"step": 65155
|
| 182262 |
+
},
|
| 182263 |
+
{
|
| 182264 |
+
"epoch": 520.62,
|
| 182265 |
+
"learning_rate": 8.972195512820514e-06,
|
| 182266 |
+
"loss": 0.8876,
|
| 182267 |
+
"step": 65160
|
| 182268 |
+
},
|
| 182269 |
+
{
|
| 182270 |
+
"epoch": 520.66,
|
| 182271 |
+
"learning_rate": 8.972115384615385e-06,
|
| 182272 |
+
"loss": 0.3056,
|
| 182273 |
+
"step": 65165
|
| 182274 |
+
},
|
| 182275 |
+
{
|
| 182276 |
+
"epoch": 520.7,
|
| 182277 |
+
"learning_rate": 8.972035256410257e-06,
|
| 182278 |
+
"loss": 0.2643,
|
| 182279 |
+
"step": 65170
|
| 182280 |
+
},
|
| 182281 |
+
{
|
| 182282 |
+
"epoch": 520.74,
|
| 182283 |
+
"learning_rate": 8.971955128205128e-06,
|
| 182284 |
+
"loss": 0.3669,
|
| 182285 |
+
"step": 65175
|
| 182286 |
+
},
|
| 182287 |
+
{
|
| 182288 |
+
"epoch": 520.78,
|
| 182289 |
+
"learning_rate": 8.971875000000001e-06,
|
| 182290 |
+
"loss": 0.8501,
|
| 182291 |
+
"step": 65180
|
| 182292 |
+
},
|
| 182293 |
+
{
|
| 182294 |
+
"epoch": 520.82,
|
| 182295 |
+
"learning_rate": 8.971794871794873e-06,
|
| 182296 |
+
"loss": 0.9203,
|
| 182297 |
+
"step": 65185
|
| 182298 |
+
},
|
| 182299 |
+
{
|
| 182300 |
+
"epoch": 520.86,
|
| 182301 |
+
"learning_rate": 8.971714743589744e-06,
|
| 182302 |
+
"loss": 0.3056,
|
| 182303 |
+
"step": 65190
|
| 182304 |
+
},
|
| 182305 |
+
{
|
| 182306 |
+
"epoch": 520.9,
|
| 182307 |
+
"learning_rate": 8.971634615384615e-06,
|
| 182308 |
+
"loss": 0.3321,
|
| 182309 |
+
"step": 65195
|
| 182310 |
+
},
|
| 182311 |
+
{
|
| 182312 |
+
"epoch": 520.94,
|
| 182313 |
+
"learning_rate": 8.971554487179488e-06,
|
| 182314 |
+
"loss": 0.6396,
|
| 182315 |
+
"step": 65200
|
| 182316 |
+
},
|
| 182317 |
+
{
|
| 182318 |
+
"epoch": 520.98,
|
| 182319 |
+
"learning_rate": 8.97147435897436e-06,
|
| 182320 |
+
"loss": 0.7963,
|
| 182321 |
+
"step": 65205
|
| 182322 |
+
},
|
| 182323 |
+
{
|
| 182324 |
+
"epoch": 521.0,
|
| 182325 |
+
"eval_loss": 0.4146020710468292,
|
| 182326 |
+
"eval_runtime": 39.2579,
|
| 182327 |
+
"eval_samples_per_second": 21.321,
|
| 182328 |
+
"eval_steps_per_second": 0.688,
|
| 182329 |
+
"eval_wer": 0.19308674592690445,
|
| 182330 |
+
"step": 65207
|
| 182331 |
+
},
|
| 182332 |
+
{
|
| 182333 |
+
"epoch": 525.02,
|
| 182334 |
+
"learning_rate": 8.971394230769231e-06,
|
| 182335 |
+
"loss": 0.3125,
|
| 182336 |
+
"step": 65210
|
| 182337 |
+
},
|
| 182338 |
+
{
|
| 182339 |
+
"epoch": 525.06,
|
| 182340 |
+
"learning_rate": 8.971314102564104e-06,
|
| 182341 |
+
"loss": 0.3018,
|
| 182342 |
+
"step": 65215
|
| 182343 |
+
},
|
| 182344 |
+
{
|
| 182345 |
+
"epoch": 525.1,
|
| 182346 |
+
"learning_rate": 8.971233974358975e-06,
|
| 182347 |
+
"loss": 0.301,
|
| 182348 |
+
"step": 65220
|
| 182349 |
+
},
|
| 182350 |
+
{
|
| 182351 |
+
"epoch": 525.14,
|
| 182352 |
+
"learning_rate": 8.971153846153847e-06,
|
| 182353 |
+
"loss": 0.427,
|
| 182354 |
+
"step": 65225
|
| 182355 |
+
},
|
| 182356 |
+
{
|
| 182357 |
+
"epoch": 525.18,
|
| 182358 |
+
"learning_rate": 8.971073717948718e-06,
|
| 182359 |
+
"loss": 0.9105,
|
| 182360 |
+
"step": 65230
|
| 182361 |
+
},
|
| 182362 |
+
{
|
| 182363 |
+
"epoch": 525.22,
|
| 182364 |
+
"learning_rate": 8.970993589743591e-06,
|
| 182365 |
+
"loss": 0.8828,
|
| 182366 |
+
"step": 65235
|
| 182367 |
+
},
|
| 182368 |
+
{
|
| 182369 |
+
"epoch": 525.27,
|
| 182370 |
+
"learning_rate": 8.970913461538463e-06,
|
| 182371 |
+
"loss": 0.3953,
|
| 182372 |
+
"step": 65240
|
| 182373 |
+
},
|
| 182374 |
+
{
|
| 182375 |
+
"epoch": 525.31,
|
| 182376 |
+
"learning_rate": 8.970833333333334e-06,
|
| 182377 |
+
"loss": 0.3561,
|
| 182378 |
+
"step": 65245
|
| 182379 |
+
},
|
| 182380 |
+
{
|
| 182381 |
+
"epoch": 525.35,
|
| 182382 |
+
"learning_rate": 8.970753205128205e-06,
|
| 182383 |
+
"loss": 0.3945,
|
| 182384 |
+
"step": 65250
|
| 182385 |
+
},
|
| 182386 |
+
{
|
| 182387 |
+
"epoch": 525.39,
|
| 182388 |
+
"learning_rate": 8.970673076923078e-06,
|
| 182389 |
+
"loss": 0.859,
|
| 182390 |
+
"step": 65255
|
| 182391 |
+
},
|
| 182392 |
+
{
|
| 182393 |
+
"epoch": 525.43,
|
| 182394 |
+
"learning_rate": 8.97059294871795e-06,
|
| 182395 |
+
"loss": 0.9469,
|
| 182396 |
+
"step": 65260
|
| 182397 |
+
},
|
| 182398 |
+
{
|
| 182399 |
+
"epoch": 525.47,
|
| 182400 |
+
"learning_rate": 8.970512820512821e-06,
|
| 182401 |
+
"loss": 0.3447,
|
| 182402 |
+
"step": 65265
|
| 182403 |
+
},
|
| 182404 |
+
{
|
| 182405 |
+
"epoch": 525.51,
|
| 182406 |
+
"learning_rate": 8.970432692307694e-06,
|
| 182407 |
+
"loss": 0.3317,
|
| 182408 |
+
"step": 65270
|
| 182409 |
+
},
|
| 182410 |
+
{
|
| 182411 |
+
"epoch": 525.55,
|
| 182412 |
+
"learning_rate": 8.970352564102564e-06,
|
| 182413 |
+
"loss": 0.4336,
|
| 182414 |
+
"step": 65275
|
| 182415 |
+
},
|
| 182416 |
+
{
|
| 182417 |
+
"epoch": 525.59,
|
| 182418 |
+
"learning_rate": 8.970272435897437e-06,
|
| 182419 |
+
"loss": 0.8079,
|
| 182420 |
+
"step": 65280
|
| 182421 |
+
},
|
| 182422 |
+
{
|
| 182423 |
+
"epoch": 525.63,
|
| 182424 |
+
"learning_rate": 8.970192307692308e-06,
|
| 182425 |
+
"loss": 1.0053,
|
| 182426 |
+
"step": 65285
|
| 182427 |
+
},
|
| 182428 |
+
{
|
| 182429 |
+
"epoch": 525.67,
|
| 182430 |
+
"learning_rate": 8.97011217948718e-06,
|
| 182431 |
+
"loss": 0.3516,
|
| 182432 |
+
"step": 65290
|
| 182433 |
+
},
|
| 182434 |
+
{
|
| 182435 |
+
"epoch": 525.71,
|
| 182436 |
+
"learning_rate": 8.970032051282051e-06,
|
| 182437 |
+
"loss": 0.3336,
|
| 182438 |
+
"step": 65295
|
| 182439 |
+
},
|
| 182440 |
+
{
|
| 182441 |
+
"epoch": 525.75,
|
| 182442 |
+
"learning_rate": 8.969951923076924e-06,
|
| 182443 |
+
"loss": 0.3766,
|
| 182444 |
+
"step": 65300
|
| 182445 |
+
},
|
| 182446 |
+
{
|
| 182447 |
+
"epoch": 525.79,
|
| 182448 |
+
"learning_rate": 8.969871794871795e-06,
|
| 182449 |
+
"loss": 0.9008,
|
| 182450 |
+
"step": 65305
|
| 182451 |
+
},
|
| 182452 |
+
{
|
| 182453 |
+
"epoch": 525.83,
|
| 182454 |
+
"learning_rate": 8.969791666666667e-06,
|
| 182455 |
+
"loss": 0.8085,
|
| 182456 |
+
"step": 65310
|
| 182457 |
+
},
|
| 182458 |
+
{
|
| 182459 |
+
"epoch": 525.87,
|
| 182460 |
+
"learning_rate": 8.96971153846154e-06,
|
| 182461 |
+
"loss": 0.3121,
|
| 182462 |
+
"step": 65315
|
| 182463 |
+
},
|
| 182464 |
+
{
|
| 182465 |
+
"epoch": 525.91,
|
| 182466 |
+
"learning_rate": 8.969631410256411e-06,
|
| 182467 |
+
"loss": 0.4201,
|
| 182468 |
+
"step": 65320
|
| 182469 |
+
},
|
| 182470 |
+
{
|
| 182471 |
+
"epoch": 525.95,
|
| 182472 |
+
"learning_rate": 8.969551282051282e-06,
|
| 182473 |
+
"loss": 0.4964,
|
| 182474 |
+
"step": 65325
|
| 182475 |
+
},
|
| 182476 |
+
{
|
| 182477 |
+
"epoch": 525.99,
|
| 182478 |
+
"learning_rate": 8.969471153846154e-06,
|
| 182479 |
+
"loss": 0.9138,
|
| 182480 |
+
"step": 65330
|
| 182481 |
+
},
|
| 182482 |
+
{
|
| 182483 |
+
"epoch": 526.0,
|
| 182484 |
+
"eval_loss": 0.3816920518875122,
|
| 182485 |
+
"eval_runtime": 39.9137,
|
| 182486 |
+
"eval_samples_per_second": 20.97,
|
| 182487 |
+
"eval_steps_per_second": 0.676,
|
| 182488 |
+
"eval_wer": 0.19690751868679465,
|
| 182489 |
+
"step": 65331
|
| 182490 |
+
},
|
| 182491 |
+
{
|
| 182492 |
+
"epoch": 522.03,
|
| 182493 |
+
"learning_rate": 8.969391025641027e-06,
|
| 182494 |
+
"loss": 1.4717,
|
| 182495 |
+
"step": 65335
|
| 182496 |
+
},
|
| 182497 |
+
{
|
| 182498 |
+
"epoch": 522.07,
|
| 182499 |
+
"learning_rate": 8.969310897435898e-06,
|
| 182500 |
+
"loss": 0.3719,
|
| 182501 |
+
"step": 65340
|
| 182502 |
+
},
|
| 182503 |
+
{
|
| 182504 |
+
"epoch": 522.11,
|
| 182505 |
+
"learning_rate": 8.96923076923077e-06,
|
| 182506 |
+
"loss": 0.3129,
|
| 182507 |
+
"step": 65345
|
| 182508 |
+
},
|
| 182509 |
+
{
|
| 182510 |
+
"epoch": 522.15,
|
| 182511 |
+
"learning_rate": 8.969150641025641e-06,
|
| 182512 |
+
"loss": 0.3886,
|
| 182513 |
+
"step": 65350
|
| 182514 |
+
},
|
| 182515 |
+
{
|
| 182516 |
+
"epoch": 522.19,
|
| 182517 |
+
"learning_rate": 8.969070512820514e-06,
|
| 182518 |
+
"loss": 1.072,
|
| 182519 |
+
"step": 65355
|
| 182520 |
+
},
|
| 182521 |
+
{
|
| 182522 |
+
"epoch": 522.23,
|
| 182523 |
+
"learning_rate": 8.968990384615385e-06,
|
| 182524 |
+
"loss": 0.7247,
|
| 182525 |
+
"step": 65360
|
| 182526 |
+
},
|
| 182527 |
+
{
|
| 182528 |
+
"epoch": 522.27,
|
| 182529 |
+
"learning_rate": 8.968910256410257e-06,
|
| 182530 |
+
"loss": 0.3319,
|
| 182531 |
+
"step": 65365
|
| 182532 |
+
},
|
| 182533 |
+
{
|
| 182534 |
+
"epoch": 522.31,
|
| 182535 |
+
"learning_rate": 8.96883012820513e-06,
|
| 182536 |
+
"loss": 0.375,
|
| 182537 |
+
"step": 65370
|
| 182538 |
+
},
|
| 182539 |
+
{
|
| 182540 |
+
"epoch": 522.35,
|
| 182541 |
+
"learning_rate": 8.968750000000001e-06,
|
| 182542 |
+
"loss": 0.4314,
|
| 182543 |
+
"step": 65375
|
| 182544 |
+
},
|
| 182545 |
+
{
|
| 182546 |
+
"epoch": 522.39,
|
| 182547 |
+
"learning_rate": 8.968669871794873e-06,
|
| 182548 |
+
"loss": 0.9989,
|
| 182549 |
+
"step": 65380
|
| 182550 |
+
},
|
| 182551 |
+
{
|
| 182552 |
+
"epoch": 522.43,
|
| 182553 |
+
"learning_rate": 8.968589743589744e-06,
|
| 182554 |
+
"loss": 0.8225,
|
| 182555 |
+
"step": 65385
|
| 182556 |
+
},
|
| 182557 |
+
{
|
| 182558 |
+
"epoch": 522.47,
|
| 182559 |
+
"learning_rate": 8.968509615384617e-06,
|
| 182560 |
+
"loss": 1.302,
|
| 182561 |
+
"step": 65390
|
| 182562 |
+
},
|
| 182563 |
+
{
|
| 182564 |
+
"epoch": 522.51,
|
| 182565 |
+
"learning_rate": 8.968429487179488e-06,
|
| 182566 |
+
"loss": 0.4039,
|
| 182567 |
+
"step": 65395
|
| 182568 |
+
},
|
| 182569 |
+
{
|
| 182570 |
+
"epoch": 522.55,
|
| 182571 |
+
"learning_rate": 8.96834935897436e-06,
|
| 182572 |
+
"loss": 0.4623,
|
| 182573 |
+
"step": 65400
|
| 182574 |
+
},
|
| 182575 |
+
{
|
| 182576 |
+
"epoch": 522.59,
|
| 182577 |
+
"learning_rate": 8.968269230769231e-06,
|
| 182578 |
+
"loss": 0.8863,
|
| 182579 |
+
"step": 65405
|
| 182580 |
+
},
|
| 182581 |
+
{
|
| 182582 |
+
"epoch": 522.63,
|
| 182583 |
+
"learning_rate": 8.968189102564104e-06,
|
| 182584 |
+
"loss": 0.796,
|
| 182585 |
+
"step": 65410
|
| 182586 |
+
},
|
| 182587 |
+
{
|
| 182588 |
+
"epoch": 522.67,
|
| 182589 |
+
"learning_rate": 8.968108974358975e-06,
|
| 182590 |
+
"loss": 0.3117,
|
| 182591 |
+
"step": 65415
|
| 182592 |
+
},
|
| 182593 |
+
{
|
| 182594 |
+
"epoch": 522.71,
|
| 182595 |
+
"learning_rate": 8.968028846153847e-06,
|
| 182596 |
+
"loss": 0.3329,
|
| 182597 |
+
"step": 65420
|
| 182598 |
+
},
|
| 182599 |
+
{
|
| 182600 |
+
"epoch": 522.75,
|
| 182601 |
+
"learning_rate": 8.96794871794872e-06,
|
| 182602 |
+
"loss": 0.4391,
|
| 182603 |
+
"step": 65425
|
| 182604 |
+
},
|
| 182605 |
+
{
|
| 182606 |
+
"epoch": 522.79,
|
| 182607 |
+
"learning_rate": 8.96786858974359e-06,
|
| 182608 |
+
"loss": 0.9478,
|
| 182609 |
+
"step": 65430
|
| 182610 |
+
},
|
| 182611 |
+
{
|
| 182612 |
+
"epoch": 522.83,
|
| 182613 |
+
"learning_rate": 8.967788461538463e-06,
|
| 182614 |
+
"loss": 0.7507,
|
| 182615 |
+
"step": 65435
|
| 182616 |
+
},
|
| 182617 |
+
{
|
| 182618 |
+
"epoch": 522.87,
|
| 182619 |
+
"learning_rate": 8.967708333333334e-06,
|
| 182620 |
+
"loss": 0.3015,
|
| 182621 |
+
"step": 65440
|
| 182622 |
+
},
|
| 182623 |
+
{
|
| 182624 |
+
"epoch": 522.91,
|
| 182625 |
+
"learning_rate": 8.967628205128205e-06,
|
| 182626 |
+
"loss": 0.3645,
|
| 182627 |
+
"step": 65445
|
| 182628 |
+
},
|
| 182629 |
+
{
|
| 182630 |
+
"epoch": 522.95,
|
| 182631 |
+
"learning_rate": 8.967548076923077e-06,
|
| 182632 |
+
"loss": 0.3587,
|
| 182633 |
+
"step": 65450
|
| 182634 |
+
},
|
| 182635 |
+
{
|
| 182636 |
+
"epoch": 522.99,
|
| 182637 |
+
"learning_rate": 8.96746794871795e-06,
|
| 182638 |
+
"loss": 1.0812,
|
| 182639 |
+
"step": 65455
|
| 182640 |
+
},
|
| 182641 |
+
{
|
| 182642 |
+
"epoch": 523.0,
|
| 182643 |
+
"eval_loss": 0.38995593786239624,
|
| 182644 |
+
"eval_runtime": 40.0163,
|
| 182645 |
+
"eval_samples_per_second": 20.916,
|
| 182646 |
+
"eval_steps_per_second": 0.675,
|
| 182647 |
+
"eval_wer": 0.18948449072376622,
|
| 182648 |
+
"step": 65456
|
| 182649 |
+
},
|
| 182650 |
+
{
|
| 182651 |
+
"epoch": 527.03,
|
| 182652 |
+
"learning_rate": 8.967387820512821e-06,
|
| 182653 |
+
"loss": 0.3841,
|
| 182654 |
+
"step": 65460
|
| 182655 |
+
},
|
| 182656 |
+
{
|
| 182657 |
+
"epoch": 527.07,
|
| 182658 |
+
"learning_rate": 8.967307692307692e-06,
|
| 182659 |
+
"loss": 0.2641,
|
| 182660 |
+
"step": 65465
|
| 182661 |
+
},
|
| 182662 |
+
{
|
| 182663 |
+
"epoch": 527.11,
|
| 182664 |
+
"learning_rate": 8.967227564102565e-06,
|
| 182665 |
+
"loss": 0.4179,
|
| 182666 |
+
"step": 65470
|
| 182667 |
+
},
|
| 182668 |
+
{
|
| 182669 |
+
"epoch": 527.15,
|
| 182670 |
+
"learning_rate": 8.967147435897437e-06,
|
| 182671 |
+
"loss": 0.386,
|
| 182672 |
+
"step": 65475
|
| 182673 |
+
},
|
| 182674 |
+
{
|
| 182675 |
+
"epoch": 527.19,
|
| 182676 |
+
"learning_rate": 8.967067307692308e-06,
|
| 182677 |
+
"loss": 1.1192,
|
| 182678 |
+
"step": 65480
|
| 182679 |
+
},
|
| 182680 |
+
{
|
| 182681 |
+
"epoch": 527.23,
|
| 182682 |
+
"learning_rate": 8.96698717948718e-06,
|
| 182683 |
+
"loss": 0.7657,
|
| 182684 |
+
"step": 65485
|
| 182685 |
+
},
|
| 182686 |
+
{
|
| 182687 |
+
"epoch": 527.27,
|
| 182688 |
+
"learning_rate": 8.966907051282053e-06,
|
| 182689 |
+
"loss": 0.3058,
|
| 182690 |
+
"step": 65490
|
| 182691 |
+
},
|
| 182692 |
+
{
|
| 182693 |
+
"epoch": 527.31,
|
| 182694 |
+
"learning_rate": 8.966826923076924e-06,
|
| 182695 |
+
"loss": 0.3247,
|
| 182696 |
+
"step": 65495
|
| 182697 |
+
},
|
| 182698 |
+
{
|
| 182699 |
+
"epoch": 527.35,
|
| 182700 |
+
"learning_rate": 8.966746794871795e-06,
|
| 182701 |
+
"loss": 0.4941,
|
| 182702 |
+
"step": 65500
|
| 182703 |
+
},
|
| 182704 |
+
{
|
| 182705 |
+
"epoch": 527.39,
|
| 182706 |
+
"learning_rate": 8.966666666666667e-06,
|
| 182707 |
+
"loss": 0.9982,
|
| 182708 |
+
"step": 65505
|
| 182709 |
+
},
|
| 182710 |
+
{
|
| 182711 |
+
"epoch": 527.43,
|
| 182712 |
+
"learning_rate": 8.96658653846154e-06,
|
| 182713 |
+
"loss": 0.5945,
|
| 182714 |
+
"step": 65510
|
| 182715 |
+
},
|
| 182716 |
+
{
|
| 182717 |
+
"epoch": 527.47,
|
| 182718 |
+
"learning_rate": 8.966506410256411e-06,
|
| 182719 |
+
"loss": 0.3769,
|
| 182720 |
+
"step": 65515
|
| 182721 |
+
},
|
| 182722 |
+
{
|
| 182723 |
+
"epoch": 527.51,
|
| 182724 |
+
"learning_rate": 8.966426282051282e-06,
|
| 182725 |
+
"loss": 0.2999,
|
| 182726 |
+
"step": 65520
|
| 182727 |
+
},
|
| 182728 |
+
{
|
| 182729 |
+
"epoch": 527.55,
|
| 182730 |
+
"learning_rate": 8.966346153846155e-06,
|
| 182731 |
+
"loss": 0.4129,
|
| 182732 |
+
"step": 65525
|
| 182733 |
+
},
|
| 182734 |
+
{
|
| 182735 |
+
"epoch": 527.59,
|
| 182736 |
+
"learning_rate": 8.966266025641027e-06,
|
| 182737 |
+
"loss": 0.8935,
|
| 182738 |
+
"step": 65530
|
| 182739 |
+
},
|
| 182740 |
+
{
|
| 182741 |
+
"epoch": 527.63,
|
| 182742 |
+
"learning_rate": 8.966185897435898e-06,
|
| 182743 |
+
"loss": 0.6514,
|
| 182744 |
+
"step": 65535
|
| 182745 |
+
},
|
| 182746 |
+
{
|
| 182747 |
+
"epoch": 527.67,
|
| 182748 |
+
"learning_rate": 8.96610576923077e-06,
|
| 182749 |
+
"loss": 0.2591,
|
| 182750 |
+
"step": 65540
|
| 182751 |
+
},
|
| 182752 |
+
{
|
| 182753 |
+
"epoch": 527.71,
|
| 182754 |
+
"learning_rate": 8.966025641025643e-06,
|
| 182755 |
+
"loss": 0.3275,
|
| 182756 |
+
"step": 65545
|
| 182757 |
+
},
|
| 182758 |
+
{
|
| 182759 |
+
"epoch": 527.76,
|
| 182760 |
+
"learning_rate": 8.965945512820512e-06,
|
| 182761 |
+
"loss": 0.3632,
|
| 182762 |
+
"step": 65550
|
| 182763 |
+
},
|
| 182764 |
+
{
|
| 182765 |
+
"epoch": 527.8,
|
| 182766 |
+
"learning_rate": 8.965865384615385e-06,
|
| 182767 |
+
"loss": 0.843,
|
| 182768 |
+
"step": 65555
|
| 182769 |
+
},
|
| 182770 |
+
{
|
| 182771 |
+
"epoch": 527.84,
|
| 182772 |
+
"learning_rate": 8.965785256410258e-06,
|
| 182773 |
+
"loss": 0.7772,
|
| 182774 |
+
"step": 65560
|
| 182775 |
+
},
|
| 182776 |
+
{
|
| 182777 |
+
"epoch": 527.88,
|
| 182778 |
+
"learning_rate": 8.965705128205128e-06,
|
| 182779 |
+
"loss": 0.4167,
|
| 182780 |
+
"step": 65565
|
| 182781 |
+
},
|
| 182782 |
+
{
|
| 182783 |
+
"epoch": 527.92,
|
| 182784 |
+
"learning_rate": 8.965625000000001e-06,
|
| 182785 |
+
"loss": 0.3216,
|
| 182786 |
+
"step": 65570
|
| 182787 |
+
},
|
| 182788 |
+
{
|
| 182789 |
+
"epoch": 527.96,
|
| 182790 |
+
"learning_rate": 8.965544871794872e-06,
|
| 182791 |
+
"loss": 0.4614,
|
| 182792 |
+
"step": 65575
|
| 182793 |
+
},
|
| 182794 |
+
{
|
| 182795 |
+
"epoch": 528.0,
|
| 182796 |
+
"learning_rate": 8.965464743589744e-06,
|
| 182797 |
+
"loss": 1.073,
|
| 182798 |
+
"step": 65580
|
| 182799 |
+
},
|
| 182800 |
+
{
|
| 182801 |
+
"epoch": 528.0,
|
| 182802 |
+
"eval_loss": 0.42701783776283264,
|
| 182803 |
+
"eval_runtime": 39.5917,
|
| 182804 |
+
"eval_samples_per_second": 21.141,
|
| 182805 |
+
"eval_steps_per_second": 0.682,
|
| 182806 |
+
"eval_wer": 0.20242253102474547,
|
| 182807 |
+
"step": 65580
|
| 182808 |
+
},
|
| 182809 |
+
{
|
| 182810 |
+
"epoch": 524.04,
|
| 182811 |
+
"learning_rate": 8.965384615384615e-06,
|
| 182812 |
+
"loss": 0.3231,
|
| 182813 |
+
"step": 65585
|
| 182814 |
+
},
|
| 182815 |
+
{
|
| 182816 |
+
"epoch": 524.08,
|
| 182817 |
+
"learning_rate": 8.965304487179488e-06,
|
| 182818 |
+
"loss": 0.3152,
|
| 182819 |
+
"step": 65590
|
| 182820 |
+
},
|
| 182821 |
+
{
|
| 182822 |
+
"epoch": 524.12,
|
| 182823 |
+
"learning_rate": 8.96522435897436e-06,
|
| 182824 |
+
"loss": 0.3409,
|
| 182825 |
+
"step": 65595
|
| 182826 |
+
},
|
| 182827 |
+
{
|
| 182828 |
+
"epoch": 524.16,
|
| 182829 |
+
"learning_rate": 8.965144230769231e-06,
|
| 182830 |
+
"loss": 0.5746,
|
| 182831 |
+
"step": 65600
|
| 182832 |
+
},
|
| 182833 |
+
{
|
| 182834 |
+
"epoch": 524.2,
|
| 182835 |
+
"learning_rate": 8.965064102564102e-06,
|
| 182836 |
+
"loss": 1.276,
|
| 182837 |
+
"step": 65605
|
| 182838 |
+
},
|
| 182839 |
+
{
|
| 182840 |
+
"epoch": 524.24,
|
| 182841 |
+
"learning_rate": 8.964983974358975e-06,
|
| 182842 |
+
"loss": 0.3679,
|
| 182843 |
+
"step": 65610
|
| 182844 |
+
},
|
| 182845 |
+
{
|
| 182846 |
+
"epoch": 524.28,
|
| 182847 |
+
"learning_rate": 8.964903846153847e-06,
|
| 182848 |
+
"loss": 0.3402,
|
| 182849 |
+
"step": 65615
|
| 182850 |
+
},
|
| 182851 |
+
{
|
| 182852 |
+
"epoch": 524.32,
|
| 182853 |
+
"learning_rate": 8.964823717948718e-06,
|
| 182854 |
+
"loss": 0.3618,
|
| 182855 |
+
"step": 65620
|
| 182856 |
+
},
|
| 182857 |
+
{
|
| 182858 |
+
"epoch": 524.36,
|
| 182859 |
+
"learning_rate": 8.964743589743591e-06,
|
| 182860 |
+
"loss": 0.5314,
|
| 182861 |
+
"step": 65625
|
| 182862 |
+
},
|
| 182863 |
+
{
|
| 182864 |
+
"epoch": 524.4,
|
| 182865 |
+
"learning_rate": 8.964663461538462e-06,
|
| 182866 |
+
"loss": 1.2524,
|
| 182867 |
+
"step": 65630
|
| 182868 |
+
},
|
| 182869 |
+
{
|
| 182870 |
+
"epoch": 524.44,
|
| 182871 |
+
"learning_rate": 8.964583333333334e-06,
|
| 182872 |
+
"loss": 0.3583,
|
| 182873 |
+
"step": 65635
|
| 182874 |
+
},
|
| 182875 |
+
{
|
| 182876 |
+
"epoch": 524.48,
|
| 182877 |
+
"learning_rate": 8.964503205128205e-06,
|
| 182878 |
+
"loss": 0.3134,
|
| 182879 |
+
"step": 65640
|
| 182880 |
+
},
|
| 182881 |
+
{
|
| 182882 |
+
"epoch": 524.52,
|
| 182883 |
+
"learning_rate": 8.964423076923078e-06,
|
| 182884 |
+
"loss": 0.3641,
|
| 182885 |
+
"step": 65645
|
| 182886 |
+
},
|
| 182887 |
+
{
|
| 182888 |
+
"epoch": 524.56,
|
| 182889 |
+
"learning_rate": 8.96434294871795e-06,
|
| 182890 |
+
"loss": 0.5193,
|
| 182891 |
+
"step": 65650
|
| 182892 |
+
},
|
| 182893 |
+
{
|
| 182894 |
+
"epoch": 524.6,
|
| 182895 |
+
"learning_rate": 8.964262820512821e-06,
|
| 182896 |
+
"loss": 1.3569,
|
| 182897 |
+
"step": 65655
|
| 182898 |
+
},
|
| 182899 |
+
{
|
| 182900 |
+
"epoch": 524.64,
|
| 182901 |
+
"learning_rate": 8.964182692307694e-06,
|
| 182902 |
+
"loss": 0.377,
|
| 182903 |
+
"step": 65660
|
| 182904 |
+
},
|
| 182905 |
+
{
|
| 182906 |
+
"epoch": 524.68,
|
| 182907 |
+
"learning_rate": 8.964102564102565e-06,
|
| 182908 |
+
"loss": 0.2879,
|
| 182909 |
+
"step": 65665
|
| 182910 |
+
},
|
| 182911 |
+
{
|
| 182912 |
+
"epoch": 524.72,
|
| 182913 |
+
"learning_rate": 8.964022435897437e-06,
|
| 182914 |
+
"loss": 0.4234,
|
| 182915 |
+
"step": 65670
|
| 182916 |
+
},
|
| 182917 |
+
{
|
| 182918 |
+
"epoch": 524.76,
|
| 182919 |
+
"learning_rate": 8.963942307692308e-06,
|
| 182920 |
+
"loss": 0.4474,
|
| 182921 |
+
"step": 65675
|
| 182922 |
+
},
|
| 182923 |
+
{
|
| 182924 |
+
"epoch": 524.8,
|
| 182925 |
+
"learning_rate": 8.963862179487181e-06,
|
| 182926 |
+
"loss": 1.2905,
|
| 182927 |
+
"step": 65680
|
| 182928 |
+
},
|
| 182929 |
+
{
|
| 182930 |
+
"epoch": 524.84,
|
| 182931 |
+
"learning_rate": 8.963782051282052e-06,
|
| 182932 |
+
"loss": 0.3868,
|
| 182933 |
+
"step": 65685
|
| 182934 |
+
},
|
| 182935 |
+
{
|
| 182936 |
+
"epoch": 524.88,
|
| 182937 |
+
"learning_rate": 8.963701923076924e-06,
|
| 182938 |
+
"loss": 0.2714,
|
| 182939 |
+
"step": 65690
|
| 182940 |
+
},
|
| 182941 |
+
{
|
| 182942 |
+
"epoch": 524.92,
|
| 182943 |
+
"learning_rate": 8.963621794871795e-06,
|
| 182944 |
+
"loss": 0.3472,
|
| 182945 |
+
"step": 65695
|
| 182946 |
+
},
|
| 182947 |
+
{
|
| 182948 |
+
"epoch": 524.96,
|
| 182949 |
+
"learning_rate": 8.963541666666668e-06,
|
| 182950 |
+
"loss": 0.4991,
|
| 182951 |
+
"step": 65700
|
| 182952 |
+
},
|
| 182953 |
+
{
|
| 182954 |
+
"epoch": 525.0,
|
| 182955 |
+
"learning_rate": 8.963461538461538e-06,
|
| 182956 |
+
"loss": 1.4589,
|
| 182957 |
+
"step": 65705
|
| 182958 |
+
},
|
| 182959 |
+
{
|
| 182960 |
+
"epoch": 525.0,
|
| 182961 |
+
"eval_loss": 0.35489851236343384,
|
| 182962 |
+
"eval_runtime": 39.8626,
|
| 182963 |
+
"eval_samples_per_second": 20.972,
|
| 182964 |
+
"eval_steps_per_second": 0.677,
|
| 182965 |
+
"eval_wer": 0.18776816231546797,
|
| 182966 |
+
"step": 65705
|
| 182967 |
}
|
| 182968 |
],
|
| 182969 |
+
"max_steps": 625000,
|
| 182970 |
"num_train_epochs": 5000,
|
| 182971 |
+
"total_flos": 1.8489006770031778e+20,
|
| 182972 |
"trial_name": null,
|
| 182973 |
"trial_params": null
|
| 182974 |
}
|
model-bin/finetune/base/{checkpoint-65082 → checkpoint-65705}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629824464.0486395/events.out.tfevents.1629824464.c435e1c5ee04.920.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afd1781e9b939405fef5e5bf3715a59b9e03a407c6271932f6ca1e74ec36ae6c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629825098.0807111/events.out.tfevents.1629825098.c435e1c5ee04.920.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6741bb880a38aba2e299859da6b5a54f7f82ce30b0e91b245e0bb7229ff198ed
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629825728.7070112/events.out.tfevents.1629825728.c435e1c5ee04.920.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:70e0eca55997e7c9b99d376a65eb8d77e4ed14f61d177f33a5c2fe41dcb8eb95
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629826356.1797123/events.out.tfevents.1629826356.c435e1c5ee04.920.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:21e33bcde35a424c597ccb96bc28ee21766a5080b89734b2ed638c39282cd6a1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629826982.8940318/events.out.tfevents.1629826982.c435e1c5ee04.920.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7affe51f0ef5059d5e791afa7a663a43bb8d95a56c9d1a9f7dee156bcb66249f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629824464.c435e1c5ee04.920.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd819e1f62bb65e4cbb11c44647a5fd43bfe144667dc265872c5a7ad39044c2f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629825098.c435e1c5ee04.920.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5c78c6ecbc72d6d3b0a46152331771e5b191b9c5142f5f3fc7b6882cadd7e6b8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629825728.c435e1c5ee04.920.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fd8add0f7db6fbbf8c56e56eb973d70b4c64a3b571a2851445e36e787011b25
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629826356.c435e1c5ee04.920.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca4d936560402723af9076c529712d77a773c199e938dd1225a0b7f09c4c51e7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629826982.c435e1c5ee04.920.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77712062afa9f26add612c38fa88e050b0e1d03c1aad4f9837ec767820ceda96
|
| 3 |
+
size 8622
|