"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629742661.920259/events.out.tfevents.1629742661.74272264b15c.932.193 +3 -0
- model-bin/finetune/base/log/1629743288.4072373/events.out.tfevents.1629743288.74272264b15c.932.195 +3 -0
- model-bin/finetune/base/log/1629743914.6346037/events.out.tfevents.1629743914.74272264b15c.932.197 +3 -0
- model-bin/finetune/base/log/1629744549.981468/events.out.tfevents.1629744549.74272264b15c.932.199 +3 -0
- model-bin/finetune/base/log/1629745189.2385828/events.out.tfevents.1629745189.74272264b15c.932.201 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629742661.74272264b15c.932.192 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629743288.74272264b15c.932.194 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629743914.74272264b15c.932.196 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629744549.74272264b15c.932.198 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629745189.74272264b15c.932.200 +3 -0
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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:e611ef6eb0932bcbb9e9a1621990693ca8da184a79c2ceaccd5440bf996a183a
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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 sha256:6be8d1f12ce4632815334b9bd0823a8f3b7273af50451b38613eeb06553f2a2b
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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 sha256:6998aba75249b05da7a1b0ab6f24007a6001e564e5598114a39b94a232ecbd24
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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:d269dbaa15c8c996885095e3ec368886bb63539377a7a983489490a1f5f6ebf0
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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:a7bba09a74be081dab7e53048831dda35de58e4ecf91d71c9b69e4547c5bb938
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/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,
|
|
@@ -163173,11 +163173,800 @@
|
|
| 163173 |
"eval_steps_per_second": 0.676,
|
| 163174 |
"eval_wer": 0.2024080463989428,
|
| 163175 |
"step": 50151
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163176 |
}
|
| 163177 |
],
|
| 163178 |
"max_steps": 620000,
|
| 163179 |
"num_train_epochs": 5000,
|
| 163180 |
-
"total_flos": 1.
|
| 163181 |
"trial_name": null,
|
| 163182 |
"trial_params": null
|
| 163183 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 409.0,
|
| 5 |
+
"global_step": 50773,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 163173 |
"eval_steps_per_second": 0.676,
|
| 163174 |
"eval_wer": 0.2024080463989428,
|
| 163175 |
"step": 50151
|
| 163176 |
+
},
|
| 163177 |
+
{
|
| 163178 |
+
"epoch": 401.03,
|
| 163179 |
+
"learning_rate": 9.212564102564103e-06,
|
| 163180 |
+
"loss": 0.4142,
|
| 163181 |
+
"step": 50155
|
| 163182 |
+
},
|
| 163183 |
+
{
|
| 163184 |
+
"epoch": 401.07,
|
| 163185 |
+
"learning_rate": 9.212483974358976e-06,
|
| 163186 |
+
"loss": 0.3919,
|
| 163187 |
+
"step": 50160
|
| 163188 |
+
},
|
| 163189 |
+
{
|
| 163190 |
+
"epoch": 401.11,
|
| 163191 |
+
"learning_rate": 9.212403846153847e-06,
|
| 163192 |
+
"loss": 0.4063,
|
| 163193 |
+
"step": 50165
|
| 163194 |
+
},
|
| 163195 |
+
{
|
| 163196 |
+
"epoch": 401.15,
|
| 163197 |
+
"learning_rate": 9.212323717948719e-06,
|
| 163198 |
+
"loss": 0.4506,
|
| 163199 |
+
"step": 50170
|
| 163200 |
+
},
|
| 163201 |
+
{
|
| 163202 |
+
"epoch": 401.19,
|
| 163203 |
+
"learning_rate": 9.21224358974359e-06,
|
| 163204 |
+
"loss": 1.0087,
|
| 163205 |
+
"step": 50175
|
| 163206 |
+
},
|
| 163207 |
+
{
|
| 163208 |
+
"epoch": 401.23,
|
| 163209 |
+
"learning_rate": 9.212163461538463e-06,
|
| 163210 |
+
"loss": 0.7284,
|
| 163211 |
+
"step": 50180
|
| 163212 |
+
},
|
| 163213 |
+
{
|
| 163214 |
+
"epoch": 401.27,
|
| 163215 |
+
"learning_rate": 9.212083333333334e-06,
|
| 163216 |
+
"loss": 0.382,
|
| 163217 |
+
"step": 50185
|
| 163218 |
+
},
|
| 163219 |
+
{
|
| 163220 |
+
"epoch": 401.31,
|
| 163221 |
+
"learning_rate": 9.212003205128206e-06,
|
| 163222 |
+
"loss": 0.3614,
|
| 163223 |
+
"step": 50190
|
| 163224 |
+
},
|
| 163225 |
+
{
|
| 163226 |
+
"epoch": 401.35,
|
| 163227 |
+
"learning_rate": 9.211923076923077e-06,
|
| 163228 |
+
"loss": 0.4452,
|
| 163229 |
+
"step": 50195
|
| 163230 |
+
},
|
| 163231 |
+
{
|
| 163232 |
+
"epoch": 401.39,
|
| 163233 |
+
"learning_rate": 9.21184294871795e-06,
|
| 163234 |
+
"loss": 1.0302,
|
| 163235 |
+
"step": 50200
|
| 163236 |
+
},
|
| 163237 |
+
{
|
| 163238 |
+
"epoch": 401.43,
|
| 163239 |
+
"learning_rate": 9.211762820512822e-06,
|
| 163240 |
+
"loss": 0.7231,
|
| 163241 |
+
"step": 50205
|
| 163242 |
+
},
|
| 163243 |
+
{
|
| 163244 |
+
"epoch": 401.47,
|
| 163245 |
+
"learning_rate": 9.211682692307693e-06,
|
| 163246 |
+
"loss": 0.2642,
|
| 163247 |
+
"step": 50210
|
| 163248 |
+
},
|
| 163249 |
+
{
|
| 163250 |
+
"epoch": 401.51,
|
| 163251 |
+
"learning_rate": 9.211602564102566e-06,
|
| 163252 |
+
"loss": 0.3184,
|
| 163253 |
+
"step": 50215
|
| 163254 |
+
},
|
| 163255 |
+
{
|
| 163256 |
+
"epoch": 401.55,
|
| 163257 |
+
"learning_rate": 9.211522435897436e-06,
|
| 163258 |
+
"loss": 0.4984,
|
| 163259 |
+
"step": 50220
|
| 163260 |
+
},
|
| 163261 |
+
{
|
| 163262 |
+
"epoch": 401.59,
|
| 163263 |
+
"learning_rate": 9.211442307692309e-06,
|
| 163264 |
+
"loss": 1.1354,
|
| 163265 |
+
"step": 50225
|
| 163266 |
+
},
|
| 163267 |
+
{
|
| 163268 |
+
"epoch": 401.63,
|
| 163269 |
+
"learning_rate": 9.21136217948718e-06,
|
| 163270 |
+
"loss": 0.6877,
|
| 163271 |
+
"step": 50230
|
| 163272 |
+
},
|
| 163273 |
+
{
|
| 163274 |
+
"epoch": 401.67,
|
| 163275 |
+
"learning_rate": 9.211282051282051e-06,
|
| 163276 |
+
"loss": 0.3542,
|
| 163277 |
+
"step": 50235
|
| 163278 |
+
},
|
| 163279 |
+
{
|
| 163280 |
+
"epoch": 401.71,
|
| 163281 |
+
"learning_rate": 9.211201923076923e-06,
|
| 163282 |
+
"loss": 0.42,
|
| 163283 |
+
"step": 50240
|
| 163284 |
+
},
|
| 163285 |
+
{
|
| 163286 |
+
"epoch": 401.75,
|
| 163287 |
+
"learning_rate": 9.211121794871796e-06,
|
| 163288 |
+
"loss": 0.5262,
|
| 163289 |
+
"step": 50245
|
| 163290 |
+
},
|
| 163291 |
+
{
|
| 163292 |
+
"epoch": 401.79,
|
| 163293 |
+
"learning_rate": 9.211041666666667e-06,
|
| 163294 |
+
"loss": 1.0461,
|
| 163295 |
+
"step": 50250
|
| 163296 |
+
},
|
| 163297 |
+
{
|
| 163298 |
+
"epoch": 401.83,
|
| 163299 |
+
"learning_rate": 9.210961538461539e-06,
|
| 163300 |
+
"loss": 0.7594,
|
| 163301 |
+
"step": 50255
|
| 163302 |
+
},
|
| 163303 |
+
{
|
| 163304 |
+
"epoch": 401.87,
|
| 163305 |
+
"learning_rate": 9.210881410256412e-06,
|
| 163306 |
+
"loss": 0.3436,
|
| 163307 |
+
"step": 50260
|
| 163308 |
+
},
|
| 163309 |
+
{
|
| 163310 |
+
"epoch": 401.91,
|
| 163311 |
+
"learning_rate": 9.210801282051283e-06,
|
| 163312 |
+
"loss": 0.3786,
|
| 163313 |
+
"step": 50265
|
| 163314 |
+
},
|
| 163315 |
+
{
|
| 163316 |
+
"epoch": 401.95,
|
| 163317 |
+
"learning_rate": 9.210721153846154e-06,
|
| 163318 |
+
"loss": 0.439,
|
| 163319 |
+
"step": 50270
|
| 163320 |
+
},
|
| 163321 |
+
{
|
| 163322 |
+
"epoch": 401.99,
|
| 163323 |
+
"learning_rate": 9.210641025641026e-06,
|
| 163324 |
+
"loss": 1.0057,
|
| 163325 |
+
"step": 50275
|
| 163326 |
+
},
|
| 163327 |
+
{
|
| 163328 |
+
"epoch": 402.0,
|
| 163329 |
+
"eval_loss": 0.42510727047920227,
|
| 163330 |
+
"eval_runtime": 38.7383,
|
| 163331 |
+
"eval_samples_per_second": 21.684,
|
| 163332 |
+
"eval_steps_per_second": 0.697,
|
| 163333 |
+
"eval_wer": 0.20118041387350627,
|
| 163334 |
+
"step": 50276
|
| 163335 |
+
},
|
| 163336 |
+
{
|
| 163337 |
+
"epoch": 405.03,
|
| 163338 |
+
"learning_rate": 9.210560897435899e-06,
|
| 163339 |
+
"loss": 0.3261,
|
| 163340 |
+
"step": 50280
|
| 163341 |
+
},
|
| 163342 |
+
{
|
| 163343 |
+
"epoch": 405.07,
|
| 163344 |
+
"learning_rate": 9.21048076923077e-06,
|
| 163345 |
+
"loss": 0.3447,
|
| 163346 |
+
"step": 50285
|
| 163347 |
+
},
|
| 163348 |
+
{
|
| 163349 |
+
"epoch": 405.11,
|
| 163350 |
+
"learning_rate": 9.210400641025641e-06,
|
| 163351 |
+
"loss": 0.4499,
|
| 163352 |
+
"step": 50290
|
| 163353 |
+
},
|
| 163354 |
+
{
|
| 163355 |
+
"epoch": 405.15,
|
| 163356 |
+
"learning_rate": 9.210320512820513e-06,
|
| 163357 |
+
"loss": 0.4307,
|
| 163358 |
+
"step": 50295
|
| 163359 |
+
},
|
| 163360 |
+
{
|
| 163361 |
+
"epoch": 405.19,
|
| 163362 |
+
"learning_rate": 9.210240384615386e-06,
|
| 163363 |
+
"loss": 0.9117,
|
| 163364 |
+
"step": 50300
|
| 163365 |
+
},
|
| 163366 |
+
{
|
| 163367 |
+
"epoch": 405.23,
|
| 163368 |
+
"learning_rate": 9.210160256410257e-06,
|
| 163369 |
+
"loss": 0.8096,
|
| 163370 |
+
"step": 50305
|
| 163371 |
+
},
|
| 163372 |
+
{
|
| 163373 |
+
"epoch": 405.27,
|
| 163374 |
+
"learning_rate": 9.210080128205129e-06,
|
| 163375 |
+
"loss": 0.3745,
|
| 163376 |
+
"step": 50310
|
| 163377 |
+
},
|
| 163378 |
+
{
|
| 163379 |
+
"epoch": 405.31,
|
| 163380 |
+
"learning_rate": 9.210000000000002e-06,
|
| 163381 |
+
"loss": 0.3634,
|
| 163382 |
+
"step": 50315
|
| 163383 |
+
},
|
| 163384 |
+
{
|
| 163385 |
+
"epoch": 405.35,
|
| 163386 |
+
"learning_rate": 9.209919871794873e-06,
|
| 163387 |
+
"loss": 0.447,
|
| 163388 |
+
"step": 50320
|
| 163389 |
+
},
|
| 163390 |
+
{
|
| 163391 |
+
"epoch": 405.4,
|
| 163392 |
+
"learning_rate": 9.209839743589744e-06,
|
| 163393 |
+
"loss": 1.0058,
|
| 163394 |
+
"step": 50325
|
| 163395 |
+
},
|
| 163396 |
+
{
|
| 163397 |
+
"epoch": 405.44,
|
| 163398 |
+
"learning_rate": 9.209759615384616e-06,
|
| 163399 |
+
"loss": 0.6426,
|
| 163400 |
+
"step": 50330
|
| 163401 |
+
},
|
| 163402 |
+
{
|
| 163403 |
+
"epoch": 405.48,
|
| 163404 |
+
"learning_rate": 9.209679487179489e-06,
|
| 163405 |
+
"loss": 0.3408,
|
| 163406 |
+
"step": 50335
|
| 163407 |
+
},
|
| 163408 |
+
{
|
| 163409 |
+
"epoch": 405.52,
|
| 163410 |
+
"learning_rate": 9.20959935897436e-06,
|
| 163411 |
+
"loss": 0.3103,
|
| 163412 |
+
"step": 50340
|
| 163413 |
+
},
|
| 163414 |
+
{
|
| 163415 |
+
"epoch": 405.56,
|
| 163416 |
+
"learning_rate": 9.209519230769231e-06,
|
| 163417 |
+
"loss": 0.5689,
|
| 163418 |
+
"step": 50345
|
| 163419 |
+
},
|
| 163420 |
+
{
|
| 163421 |
+
"epoch": 405.6,
|
| 163422 |
+
"learning_rate": 9.209439102564103e-06,
|
| 163423 |
+
"loss": 0.9575,
|
| 163424 |
+
"step": 50350
|
| 163425 |
+
},
|
| 163426 |
+
{
|
| 163427 |
+
"epoch": 405.64,
|
| 163428 |
+
"learning_rate": 9.209358974358976e-06,
|
| 163429 |
+
"loss": 0.6906,
|
| 163430 |
+
"step": 50355
|
| 163431 |
+
},
|
| 163432 |
+
{
|
| 163433 |
+
"epoch": 405.68,
|
| 163434 |
+
"learning_rate": 9.209278846153847e-06,
|
| 163435 |
+
"loss": 0.4068,
|
| 163436 |
+
"step": 50360
|
| 163437 |
+
},
|
| 163438 |
+
{
|
| 163439 |
+
"epoch": 405.72,
|
| 163440 |
+
"learning_rate": 9.209198717948719e-06,
|
| 163441 |
+
"loss": 0.357,
|
| 163442 |
+
"step": 50365
|
| 163443 |
+
},
|
| 163444 |
+
{
|
| 163445 |
+
"epoch": 405.76,
|
| 163446 |
+
"learning_rate": 9.209118589743592e-06,
|
| 163447 |
+
"loss": 0.4558,
|
| 163448 |
+
"step": 50370
|
| 163449 |
+
},
|
| 163450 |
+
{
|
| 163451 |
+
"epoch": 405.8,
|
| 163452 |
+
"learning_rate": 9.209038461538461e-06,
|
| 163453 |
+
"loss": 0.9309,
|
| 163454 |
+
"step": 50375
|
| 163455 |
+
},
|
| 163456 |
+
{
|
| 163457 |
+
"epoch": 405.84,
|
| 163458 |
+
"learning_rate": 9.208958333333334e-06,
|
| 163459 |
+
"loss": 0.8142,
|
| 163460 |
+
"step": 50380
|
| 163461 |
+
},
|
| 163462 |
+
{
|
| 163463 |
+
"epoch": 405.88,
|
| 163464 |
+
"learning_rate": 9.208878205128206e-06,
|
| 163465 |
+
"loss": 0.2887,
|
| 163466 |
+
"step": 50385
|
| 163467 |
+
},
|
| 163468 |
+
{
|
| 163469 |
+
"epoch": 405.92,
|
| 163470 |
+
"learning_rate": 9.208798076923077e-06,
|
| 163471 |
+
"loss": 0.3882,
|
| 163472 |
+
"step": 50390
|
| 163473 |
+
},
|
| 163474 |
+
{
|
| 163475 |
+
"epoch": 405.96,
|
| 163476 |
+
"learning_rate": 9.208717948717948e-06,
|
| 163477 |
+
"loss": 0.4638,
|
| 163478 |
+
"step": 50395
|
| 163479 |
+
},
|
| 163480 |
+
{
|
| 163481 |
+
"epoch": 406.0,
|
| 163482 |
+
"learning_rate": 9.208637820512821e-06,
|
| 163483 |
+
"loss": 1.2828,
|
| 163484 |
+
"step": 50400
|
| 163485 |
+
},
|
| 163486 |
+
{
|
| 163487 |
+
"epoch": 406.0,
|
| 163488 |
+
"eval_loss": 0.3657532036304474,
|
| 163489 |
+
"eval_runtime": 39.292,
|
| 163490 |
+
"eval_samples_per_second": 21.378,
|
| 163491 |
+
"eval_steps_per_second": 0.687,
|
| 163492 |
+
"eval_wer": 0.20113438045375218,
|
| 163493 |
+
"step": 50400
|
| 163494 |
+
},
|
| 163495 |
+
{
|
| 163496 |
+
"epoch": 403.04,
|
| 163497 |
+
"learning_rate": 9.208557692307693e-06,
|
| 163498 |
+
"loss": 0.3397,
|
| 163499 |
+
"step": 50405
|
| 163500 |
+
},
|
| 163501 |
+
{
|
| 163502 |
+
"epoch": 403.08,
|
| 163503 |
+
"learning_rate": 9.208477564102564e-06,
|
| 163504 |
+
"loss": 0.3581,
|
| 163505 |
+
"step": 50410
|
| 163506 |
+
},
|
| 163507 |
+
{
|
| 163508 |
+
"epoch": 403.12,
|
| 163509 |
+
"learning_rate": 9.208397435897437e-06,
|
| 163510 |
+
"loss": 0.324,
|
| 163511 |
+
"step": 50415
|
| 163512 |
+
},
|
| 163513 |
+
{
|
| 163514 |
+
"epoch": 403.16,
|
| 163515 |
+
"learning_rate": 9.208317307692309e-06,
|
| 163516 |
+
"loss": 0.4713,
|
| 163517 |
+
"step": 50420
|
| 163518 |
+
},
|
| 163519 |
+
{
|
| 163520 |
+
"epoch": 403.2,
|
| 163521 |
+
"learning_rate": 9.20823717948718e-06,
|
| 163522 |
+
"loss": 1.3346,
|
| 163523 |
+
"step": 50425
|
| 163524 |
+
},
|
| 163525 |
+
{
|
| 163526 |
+
"epoch": 403.24,
|
| 163527 |
+
"learning_rate": 9.208157051282051e-06,
|
| 163528 |
+
"loss": 0.4149,
|
| 163529 |
+
"step": 50430
|
| 163530 |
+
},
|
| 163531 |
+
{
|
| 163532 |
+
"epoch": 403.28,
|
| 163533 |
+
"learning_rate": 9.208076923076924e-06,
|
| 163534 |
+
"loss": 0.2844,
|
| 163535 |
+
"step": 50435
|
| 163536 |
+
},
|
| 163537 |
+
{
|
| 163538 |
+
"epoch": 403.32,
|
| 163539 |
+
"learning_rate": 9.207996794871796e-06,
|
| 163540 |
+
"loss": 0.3906,
|
| 163541 |
+
"step": 50440
|
| 163542 |
+
},
|
| 163543 |
+
{
|
| 163544 |
+
"epoch": 403.36,
|
| 163545 |
+
"learning_rate": 9.207916666666667e-06,
|
| 163546 |
+
"loss": 0.4264,
|
| 163547 |
+
"step": 50445
|
| 163548 |
+
},
|
| 163549 |
+
{
|
| 163550 |
+
"epoch": 403.4,
|
| 163551 |
+
"learning_rate": 9.207836538461538e-06,
|
| 163552 |
+
"loss": 1.1413,
|
| 163553 |
+
"step": 50450
|
| 163554 |
+
},
|
| 163555 |
+
{
|
| 163556 |
+
"epoch": 403.44,
|
| 163557 |
+
"learning_rate": 9.207756410256412e-06,
|
| 163558 |
+
"loss": 0.3735,
|
| 163559 |
+
"step": 50455
|
| 163560 |
+
},
|
| 163561 |
+
{
|
| 163562 |
+
"epoch": 403.48,
|
| 163563 |
+
"learning_rate": 9.207676282051283e-06,
|
| 163564 |
+
"loss": 0.3853,
|
| 163565 |
+
"step": 50460
|
| 163566 |
+
},
|
| 163567 |
+
{
|
| 163568 |
+
"epoch": 403.52,
|
| 163569 |
+
"learning_rate": 9.207596153846154e-06,
|
| 163570 |
+
"loss": 0.3136,
|
| 163571 |
+
"step": 50465
|
| 163572 |
+
},
|
| 163573 |
+
{
|
| 163574 |
+
"epoch": 403.56,
|
| 163575 |
+
"learning_rate": 9.207516025641027e-06,
|
| 163576 |
+
"loss": 0.5076,
|
| 163577 |
+
"step": 50470
|
| 163578 |
+
},
|
| 163579 |
+
{
|
| 163580 |
+
"epoch": 403.6,
|
| 163581 |
+
"learning_rate": 9.207435897435899e-06,
|
| 163582 |
+
"loss": 1.4854,
|
| 163583 |
+
"step": 50475
|
| 163584 |
+
},
|
| 163585 |
+
{
|
| 163586 |
+
"epoch": 403.64,
|
| 163587 |
+
"learning_rate": 9.20735576923077e-06,
|
| 163588 |
+
"loss": 0.3648,
|
| 163589 |
+
"step": 50480
|
| 163590 |
+
},
|
| 163591 |
+
{
|
| 163592 |
+
"epoch": 403.68,
|
| 163593 |
+
"learning_rate": 9.207275641025641e-06,
|
| 163594 |
+
"loss": 0.3605,
|
| 163595 |
+
"step": 50485
|
| 163596 |
+
},
|
| 163597 |
+
{
|
| 163598 |
+
"epoch": 403.72,
|
| 163599 |
+
"learning_rate": 9.207195512820514e-06,
|
| 163600 |
+
"loss": 0.4836,
|
| 163601 |
+
"step": 50490
|
| 163602 |
+
},
|
| 163603 |
+
{
|
| 163604 |
+
"epoch": 403.76,
|
| 163605 |
+
"learning_rate": 9.207115384615384e-06,
|
| 163606 |
+
"loss": 0.5073,
|
| 163607 |
+
"step": 50495
|
| 163608 |
+
},
|
| 163609 |
+
{
|
| 163610 |
+
"epoch": 403.8,
|
| 163611 |
+
"learning_rate": 9.207035256410257e-06,
|
| 163612 |
+
"loss": 1.3218,
|
| 163613 |
+
"step": 50500
|
| 163614 |
+
},
|
| 163615 |
+
{
|
| 163616 |
+
"epoch": 403.84,
|
| 163617 |
+
"learning_rate": 9.20695512820513e-06,
|
| 163618 |
+
"loss": 0.3526,
|
| 163619 |
+
"step": 50505
|
| 163620 |
+
},
|
| 163621 |
+
{
|
| 163622 |
+
"epoch": 403.88,
|
| 163623 |
+
"learning_rate": 9.206875e-06,
|
| 163624 |
+
"loss": 0.2995,
|
| 163625 |
+
"step": 50510
|
| 163626 |
+
},
|
| 163627 |
+
{
|
| 163628 |
+
"epoch": 403.92,
|
| 163629 |
+
"learning_rate": 9.206794871794873e-06,
|
| 163630 |
+
"loss": 0.4108,
|
| 163631 |
+
"step": 50515
|
| 163632 |
+
},
|
| 163633 |
+
{
|
| 163634 |
+
"epoch": 403.96,
|
| 163635 |
+
"learning_rate": 9.206714743589744e-06,
|
| 163636 |
+
"loss": 0.5077,
|
| 163637 |
+
"step": 50520
|
| 163638 |
+
},
|
| 163639 |
+
{
|
| 163640 |
+
"epoch": 404.0,
|
| 163641 |
+
"learning_rate": 9.206634615384616e-06,
|
| 163642 |
+
"loss": 1.3173,
|
| 163643 |
+
"step": 50525
|
| 163644 |
+
},
|
| 163645 |
+
{
|
| 163646 |
+
"epoch": 404.0,
|
| 163647 |
+
"eval_loss": 0.4317208528518677,
|
| 163648 |
+
"eval_runtime": 39.1425,
|
| 163649 |
+
"eval_samples_per_second": 21.46,
|
| 163650 |
+
"eval_steps_per_second": 0.69,
|
| 163651 |
+
"eval_wer": 0.19772340897331658,
|
| 163652 |
+
"step": 50525
|
| 163653 |
+
},
|
| 163654 |
+
{
|
| 163655 |
+
"epoch": 407.04,
|
| 163656 |
+
"learning_rate": 9.206554487179487e-06,
|
| 163657 |
+
"loss": 0.2909,
|
| 163658 |
+
"step": 50530
|
| 163659 |
+
},
|
| 163660 |
+
{
|
| 163661 |
+
"epoch": 407.08,
|
| 163662 |
+
"learning_rate": 9.20647435897436e-06,
|
| 163663 |
+
"loss": 0.3695,
|
| 163664 |
+
"step": 50535
|
| 163665 |
+
},
|
| 163666 |
+
{
|
| 163667 |
+
"epoch": 407.12,
|
| 163668 |
+
"learning_rate": 9.206394230769231e-06,
|
| 163669 |
+
"loss": 0.3823,
|
| 163670 |
+
"step": 50540
|
| 163671 |
+
},
|
| 163672 |
+
{
|
| 163673 |
+
"epoch": 407.16,
|
| 163674 |
+
"learning_rate": 9.206314102564103e-06,
|
| 163675 |
+
"loss": 0.5045,
|
| 163676 |
+
"step": 50545
|
| 163677 |
+
},
|
| 163678 |
+
{
|
| 163679 |
+
"epoch": 407.2,
|
| 163680 |
+
"learning_rate": 9.206233974358974e-06,
|
| 163681 |
+
"loss": 1.2308,
|
| 163682 |
+
"step": 50550
|
| 163683 |
+
},
|
| 163684 |
+
{
|
| 163685 |
+
"epoch": 407.24,
|
| 163686 |
+
"learning_rate": 9.206153846153847e-06,
|
| 163687 |
+
"loss": 0.3866,
|
| 163688 |
+
"step": 50555
|
| 163689 |
+
},
|
| 163690 |
+
{
|
| 163691 |
+
"epoch": 407.28,
|
| 163692 |
+
"learning_rate": 9.206073717948719e-06,
|
| 163693 |
+
"loss": 0.3491,
|
| 163694 |
+
"step": 50560
|
| 163695 |
+
},
|
| 163696 |
+
{
|
| 163697 |
+
"epoch": 407.32,
|
| 163698 |
+
"learning_rate": 9.20599358974359e-06,
|
| 163699 |
+
"loss": 0.295,
|
| 163700 |
+
"step": 50565
|
| 163701 |
+
},
|
| 163702 |
+
{
|
| 163703 |
+
"epoch": 407.36,
|
| 163704 |
+
"learning_rate": 9.205913461538463e-06,
|
| 163705 |
+
"loss": 0.5069,
|
| 163706 |
+
"step": 50570
|
| 163707 |
+
},
|
| 163708 |
+
{
|
| 163709 |
+
"epoch": 407.4,
|
| 163710 |
+
"learning_rate": 9.205833333333334e-06,
|
| 163711 |
+
"loss": 1.1688,
|
| 163712 |
+
"step": 50575
|
| 163713 |
+
},
|
| 163714 |
+
{
|
| 163715 |
+
"epoch": 407.44,
|
| 163716 |
+
"learning_rate": 9.205753205128206e-06,
|
| 163717 |
+
"loss": 0.346,
|
| 163718 |
+
"step": 50580
|
| 163719 |
+
},
|
| 163720 |
+
{
|
| 163721 |
+
"epoch": 407.48,
|
| 163722 |
+
"learning_rate": 9.205673076923077e-06,
|
| 163723 |
+
"loss": 0.2988,
|
| 163724 |
+
"step": 50585
|
| 163725 |
+
},
|
| 163726 |
+
{
|
| 163727 |
+
"epoch": 407.52,
|
| 163728 |
+
"learning_rate": 9.20559294871795e-06,
|
| 163729 |
+
"loss": 0.3789,
|
| 163730 |
+
"step": 50590
|
| 163731 |
+
},
|
| 163732 |
+
{
|
| 163733 |
+
"epoch": 407.56,
|
| 163734 |
+
"learning_rate": 9.205512820512821e-06,
|
| 163735 |
+
"loss": 0.5441,
|
| 163736 |
+
"step": 50595
|
| 163737 |
+
},
|
| 163738 |
+
{
|
| 163739 |
+
"epoch": 407.6,
|
| 163740 |
+
"learning_rate": 9.205432692307693e-06,
|
| 163741 |
+
"loss": 1.4562,
|
| 163742 |
+
"step": 50600
|
| 163743 |
+
},
|
| 163744 |
+
{
|
| 163745 |
+
"epoch": 407.65,
|
| 163746 |
+
"learning_rate": 9.205352564102566e-06,
|
| 163747 |
+
"loss": 0.3585,
|
| 163748 |
+
"step": 50605
|
| 163749 |
+
},
|
| 163750 |
+
{
|
| 163751 |
+
"epoch": 407.69,
|
| 163752 |
+
"learning_rate": 9.205272435897437e-06,
|
| 163753 |
+
"loss": 0.3477,
|
| 163754 |
+
"step": 50610
|
| 163755 |
+
},
|
| 163756 |
+
{
|
| 163757 |
+
"epoch": 407.73,
|
| 163758 |
+
"learning_rate": 9.205192307692309e-06,
|
| 163759 |
+
"loss": 0.3761,
|
| 163760 |
+
"step": 50615
|
| 163761 |
+
},
|
| 163762 |
+
{
|
| 163763 |
+
"epoch": 407.77,
|
| 163764 |
+
"learning_rate": 9.20511217948718e-06,
|
| 163765 |
+
"loss": 0.5942,
|
| 163766 |
+
"step": 50620
|
| 163767 |
+
},
|
| 163768 |
+
{
|
| 163769 |
+
"epoch": 407.81,
|
| 163770 |
+
"learning_rate": 9.205032051282053e-06,
|
| 163771 |
+
"loss": 1.4338,
|
| 163772 |
+
"step": 50625
|
| 163773 |
+
},
|
| 163774 |
+
{
|
| 163775 |
+
"epoch": 407.85,
|
| 163776 |
+
"learning_rate": 9.204951923076924e-06,
|
| 163777 |
+
"loss": 0.4019,
|
| 163778 |
+
"step": 50630
|
| 163779 |
+
},
|
| 163780 |
+
{
|
| 163781 |
+
"epoch": 407.89,
|
| 163782 |
+
"learning_rate": 9.204871794871796e-06,
|
| 163783 |
+
"loss": 0.3051,
|
| 163784 |
+
"step": 50635
|
| 163785 |
+
},
|
| 163786 |
+
{
|
| 163787 |
+
"epoch": 407.93,
|
| 163788 |
+
"learning_rate": 9.204791666666667e-06,
|
| 163789 |
+
"loss": 0.3793,
|
| 163790 |
+
"step": 50640
|
| 163791 |
+
},
|
| 163792 |
+
{
|
| 163793 |
+
"epoch": 407.97,
|
| 163794 |
+
"learning_rate": 9.20471153846154e-06,
|
| 163795 |
+
"loss": 0.5285,
|
| 163796 |
+
"step": 50645
|
| 163797 |
+
},
|
| 163798 |
+
{
|
| 163799 |
+
"epoch": 408.0,
|
| 163800 |
+
"eval_loss": 0.4244321882724762,
|
| 163801 |
+
"eval_runtime": 39.8011,
|
| 163802 |
+
"eval_samples_per_second": 21.08,
|
| 163803 |
+
"eval_steps_per_second": 0.678,
|
| 163804 |
+
"eval_wer": 0.20871475649945076,
|
| 163805 |
+
"step": 50649
|
| 163806 |
+
},
|
| 163807 |
+
{
|
| 163808 |
+
"epoch": 408.01,
|
| 163809 |
+
"learning_rate": 9.20463141025641e-06,
|
| 163810 |
+
"loss": 0.4483,
|
| 163811 |
+
"step": 50650
|
| 163812 |
+
},
|
| 163813 |
+
{
|
| 163814 |
+
"epoch": 408.05,
|
| 163815 |
+
"learning_rate": 9.204551282051283e-06,
|
| 163816 |
+
"loss": 0.3694,
|
| 163817 |
+
"step": 50655
|
| 163818 |
+
},
|
| 163819 |
+
{
|
| 163820 |
+
"epoch": 408.09,
|
| 163821 |
+
"learning_rate": 9.204471153846156e-06,
|
| 163822 |
+
"loss": 0.3167,
|
| 163823 |
+
"step": 50660
|
| 163824 |
+
},
|
| 163825 |
+
{
|
| 163826 |
+
"epoch": 408.13,
|
| 163827 |
+
"learning_rate": 9.204391025641026e-06,
|
| 163828 |
+
"loss": 0.3573,
|
| 163829 |
+
"step": 50665
|
| 163830 |
+
},
|
| 163831 |
+
{
|
| 163832 |
+
"epoch": 408.17,
|
| 163833 |
+
"learning_rate": 9.204310897435899e-06,
|
| 163834 |
+
"loss": 0.6234,
|
| 163835 |
+
"step": 50670
|
| 163836 |
+
},
|
| 163837 |
+
{
|
| 163838 |
+
"epoch": 408.21,
|
| 163839 |
+
"learning_rate": 9.20423076923077e-06,
|
| 163840 |
+
"loss": 1.3824,
|
| 163841 |
+
"step": 50675
|
| 163842 |
+
},
|
| 163843 |
+
{
|
| 163844 |
+
"epoch": 408.25,
|
| 163845 |
+
"learning_rate": 9.204150641025641e-06,
|
| 163846 |
+
"loss": 0.3399,
|
| 163847 |
+
"step": 50680
|
| 163848 |
+
},
|
| 163849 |
+
{
|
| 163850 |
+
"epoch": 408.29,
|
| 163851 |
+
"learning_rate": 9.204070512820513e-06,
|
| 163852 |
+
"loss": 0.3786,
|
| 163853 |
+
"step": 50685
|
| 163854 |
+
},
|
| 163855 |
+
{
|
| 163856 |
+
"epoch": 408.33,
|
| 163857 |
+
"learning_rate": 9.203990384615386e-06,
|
| 163858 |
+
"loss": 0.3745,
|
| 163859 |
+
"step": 50690
|
| 163860 |
+
},
|
| 163861 |
+
{
|
| 163862 |
+
"epoch": 408.37,
|
| 163863 |
+
"learning_rate": 9.203910256410257e-06,
|
| 163864 |
+
"loss": 0.6151,
|
| 163865 |
+
"step": 50695
|
| 163866 |
+
},
|
| 163867 |
+
{
|
| 163868 |
+
"epoch": 408.41,
|
| 163869 |
+
"learning_rate": 9.203830128205128e-06,
|
| 163870 |
+
"loss": 1.2514,
|
| 163871 |
+
"step": 50700
|
| 163872 |
+
},
|
| 163873 |
+
{
|
| 163874 |
+
"epoch": 408.45,
|
| 163875 |
+
"learning_rate": 9.203750000000001e-06,
|
| 163876 |
+
"loss": 0.6378,
|
| 163877 |
+
"step": 50705
|
| 163878 |
+
},
|
| 163879 |
+
{
|
| 163880 |
+
"epoch": 408.49,
|
| 163881 |
+
"learning_rate": 9.203669871794873e-06,
|
| 163882 |
+
"loss": 0.401,
|
| 163883 |
+
"step": 50710
|
| 163884 |
+
},
|
| 163885 |
+
{
|
| 163886 |
+
"epoch": 408.53,
|
| 163887 |
+
"learning_rate": 9.203589743589744e-06,
|
| 163888 |
+
"loss": 0.3908,
|
| 163889 |
+
"step": 50715
|
| 163890 |
+
},
|
| 163891 |
+
{
|
| 163892 |
+
"epoch": 408.57,
|
| 163893 |
+
"learning_rate": 9.203509615384616e-06,
|
| 163894 |
+
"loss": 0.6353,
|
| 163895 |
+
"step": 50720
|
| 163896 |
+
},
|
| 163897 |
+
{
|
| 163898 |
+
"epoch": 408.61,
|
| 163899 |
+
"learning_rate": 9.203429487179489e-06,
|
| 163900 |
+
"loss": 1.1265,
|
| 163901 |
+
"step": 50725
|
| 163902 |
+
},
|
| 163903 |
+
{
|
| 163904 |
+
"epoch": 408.65,
|
| 163905 |
+
"learning_rate": 9.20334935897436e-06,
|
| 163906 |
+
"loss": 0.3404,
|
| 163907 |
+
"step": 50730
|
| 163908 |
+
},
|
| 163909 |
+
{
|
| 163910 |
+
"epoch": 408.69,
|
| 163911 |
+
"learning_rate": 9.203269230769231e-06,
|
| 163912 |
+
"loss": 0.374,
|
| 163913 |
+
"step": 50735
|
| 163914 |
+
},
|
| 163915 |
+
{
|
| 163916 |
+
"epoch": 408.73,
|
| 163917 |
+
"learning_rate": 9.203189102564103e-06,
|
| 163918 |
+
"loss": 0.3945,
|
| 163919 |
+
"step": 50740
|
| 163920 |
+
},
|
| 163921 |
+
{
|
| 163922 |
+
"epoch": 408.77,
|
| 163923 |
+
"learning_rate": 9.203108974358976e-06,
|
| 163924 |
+
"loss": 0.558,
|
| 163925 |
+
"step": 50745
|
| 163926 |
+
},
|
| 163927 |
+
{
|
| 163928 |
+
"epoch": 408.81,
|
| 163929 |
+
"learning_rate": 9.203028846153847e-06,
|
| 163930 |
+
"loss": 1.4867,
|
| 163931 |
+
"step": 50750
|
| 163932 |
+
},
|
| 163933 |
+
{
|
| 163934 |
+
"epoch": 408.85,
|
| 163935 |
+
"learning_rate": 9.202948717948718e-06,
|
| 163936 |
+
"loss": 0.302,
|
| 163937 |
+
"step": 50755
|
| 163938 |
+
},
|
| 163939 |
+
{
|
| 163940 |
+
"epoch": 408.9,
|
| 163941 |
+
"learning_rate": 9.202868589743591e-06,
|
| 163942 |
+
"loss": 0.3243,
|
| 163943 |
+
"step": 50760
|
| 163944 |
+
},
|
| 163945 |
+
{
|
| 163946 |
+
"epoch": 408.94,
|
| 163947 |
+
"learning_rate": 9.202788461538463e-06,
|
| 163948 |
+
"loss": 0.4848,
|
| 163949 |
+
"step": 50765
|
| 163950 |
+
},
|
| 163951 |
+
{
|
| 163952 |
+
"epoch": 408.98,
|
| 163953 |
+
"learning_rate": 9.202708333333334e-06,
|
| 163954 |
+
"loss": 0.6898,
|
| 163955 |
+
"step": 50770
|
| 163956 |
+
},
|
| 163957 |
+
{
|
| 163958 |
+
"epoch": 409.0,
|
| 163959 |
+
"eval_loss": 0.4491475522518158,
|
| 163960 |
+
"eval_runtime": 40.2449,
|
| 163961 |
+
"eval_samples_per_second": 20.847,
|
| 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": 1.4286455484088766e+20,
|
| 163970 |
"trial_name": null,
|
| 163971 |
"trial_params": null
|
| 163972 |
}
|
model-bin/finetune/base/{checkpoint-50151 → checkpoint-50773}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629742661.920259/events.out.tfevents.1629742661.74272264b15c.932.193
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:78ffdbf982f06e32e5b2ab9926afc575a987c1210027c65bf051d595727cf7af
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629743288.4072373/events.out.tfevents.1629743288.74272264b15c.932.195
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41280ecd00ac98cdaec21d419cdf619f53ec21f4c29d9e1ad7721887727e3005
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629743914.6346037/events.out.tfevents.1629743914.74272264b15c.932.197
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d031131a40b440911b1c9a827ed858eb99f64c81a13430cf1cd1301f06d6a1fc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629744549.981468/events.out.tfevents.1629744549.74272264b15c.932.199
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd18e65e70c5c710ee4343c5b41da81255ace763dc5aa236e17b18b8523d3ada
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629745189.2385828/events.out.tfevents.1629745189.74272264b15c.932.201
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:731d7e65ce6025184f6bdd769675693bd0912630411cc8840794067abb4cbde8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629742661.74272264b15c.932.192
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:57fc5cb3c65f6ea6966c770c21c31c529f0d214f65e70d58bd74b761e2ff6722
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629743288.74272264b15c.932.194
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff63995d2556edf0439f7dfcdf13f20cba1ea03c8caecd73c83ce5b74cdd3e2c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629743914.74272264b15c.932.196
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01bd0535bae274e2ad8e893cb115f3480a259d0df52dfe91a96b6f38e880a0b6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629744549.74272264b15c.932.198
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7aae2025eac32670ee8611b5e2b0b0d63975a0591d982c621e22e95248e3a8a3
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629745189.74272264b15c.932.200
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3e9dcf9607982ab33a285375af9a55bfe1fdadc8d834f1da739317dbf955d29
|
| 3 |
+
size 8622
|