"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629726271.057744/events.out.tfevents.1629726271.74272264b15c.932.143 +3 -0
- model-bin/finetune/base/log/1629726936.3025346/events.out.tfevents.1629726936.74272264b15c.932.145 +3 -0
- model-bin/finetune/base/log/1629727569.1115878/events.out.tfevents.1629727569.74272264b15c.932.147 +3 -0
- model-bin/finetune/base/log/1629728226.273403/events.out.tfevents.1629728226.74272264b15c.932.149 +3 -0
- model-bin/finetune/base/log/1629728868.313585/events.out.tfevents.1629728868.74272264b15c.932.151 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629726271.74272264b15c.932.142 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629726936.74272264b15c.932.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629727569.74272264b15c.932.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629728226.74272264b15c.932.148 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629728868.74272264b15c.932.150 +3 -0
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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:c6f1253e46cde7d178f39e6b17a8972a3aad4111872e5bf27d7b6dca137c72c2
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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:e40873fa8da493e92e672079326593cb7f57a26d199c8587c88979165d3b405d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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 sha256:8655ea7d9fd7f1ddbb7f3c0285d61ea693bed9b3fcbbe176e4c78f2515296b7f
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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:f26c7b4da20f1cc9a4ffd34476828968f627bc9309919fb52d579c167c07ee6a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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 sha256:31131468d789f262e6e19dc1f7e277df2e3b8527cff32367b76129cd61996678
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/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,
|
|
@@ -159210,11 +159210,806 @@
|
|
| 159210 |
"eval_steps_per_second": 0.658,
|
| 159211 |
"eval_wer": 0.20545609548167093,
|
| 159212 |
"step": 47039
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159213 |
}
|
| 159214 |
],
|
| 159215 |
-
"max_steps":
|
| 159216 |
"num_train_epochs": 5000,
|
| 159217 |
-
"total_flos": 1.
|
| 159218 |
"trial_name": null,
|
| 159219 |
"trial_params": null
|
| 159220 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 381.0,
|
| 5 |
+
"global_step": 47663,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 159210 |
"eval_steps_per_second": 0.658,
|
| 159211 |
"eval_wer": 0.20545609548167093,
|
| 159212 |
"step": 47039
|
| 159213 |
+
},
|
| 159214 |
+
{
|
| 159215 |
+
"epoch": 376.01,
|
| 159216 |
+
"learning_rate": 9.262451923076923e-06,
|
| 159217 |
+
"loss": 0.4411,
|
| 159218 |
+
"step": 47040
|
| 159219 |
+
},
|
| 159220 |
+
{
|
| 159221 |
+
"epoch": 376.05,
|
| 159222 |
+
"learning_rate": 9.262371794871796e-06,
|
| 159223 |
+
"loss": 0.3758,
|
| 159224 |
+
"step": 47045
|
| 159225 |
+
},
|
| 159226 |
+
{
|
| 159227 |
+
"epoch": 376.09,
|
| 159228 |
+
"learning_rate": 9.262291666666668e-06,
|
| 159229 |
+
"loss": 0.3242,
|
| 159230 |
+
"step": 47050
|
| 159231 |
+
},
|
| 159232 |
+
{
|
| 159233 |
+
"epoch": 376.13,
|
| 159234 |
+
"learning_rate": 9.262211538461539e-06,
|
| 159235 |
+
"loss": 0.3592,
|
| 159236 |
+
"step": 47055
|
| 159237 |
+
},
|
| 159238 |
+
{
|
| 159239 |
+
"epoch": 376.17,
|
| 159240 |
+
"learning_rate": 9.26213141025641e-06,
|
| 159241 |
+
"loss": 0.5519,
|
| 159242 |
+
"step": 47060
|
| 159243 |
+
},
|
| 159244 |
+
{
|
| 159245 |
+
"epoch": 376.21,
|
| 159246 |
+
"learning_rate": 9.262051282051284e-06,
|
| 159247 |
+
"loss": 1.1892,
|
| 159248 |
+
"step": 47065
|
| 159249 |
+
},
|
| 159250 |
+
{
|
| 159251 |
+
"epoch": 376.25,
|
| 159252 |
+
"learning_rate": 9.261971153846155e-06,
|
| 159253 |
+
"loss": 0.3381,
|
| 159254 |
+
"step": 47070
|
| 159255 |
+
},
|
| 159256 |
+
{
|
| 159257 |
+
"epoch": 376.29,
|
| 159258 |
+
"learning_rate": 9.261891025641026e-06,
|
| 159259 |
+
"loss": 0.3679,
|
| 159260 |
+
"step": 47075
|
| 159261 |
+
},
|
| 159262 |
+
{
|
| 159263 |
+
"epoch": 376.33,
|
| 159264 |
+
"learning_rate": 9.261810897435898e-06,
|
| 159265 |
+
"loss": 0.3559,
|
| 159266 |
+
"step": 47080
|
| 159267 |
+
},
|
| 159268 |
+
{
|
| 159269 |
+
"epoch": 376.37,
|
| 159270 |
+
"learning_rate": 9.26173076923077e-06,
|
| 159271 |
+
"loss": 0.768,
|
| 159272 |
+
"step": 47085
|
| 159273 |
+
},
|
| 159274 |
+
{
|
| 159275 |
+
"epoch": 376.41,
|
| 159276 |
+
"learning_rate": 9.261650641025642e-06,
|
| 159277 |
+
"loss": 1.2828,
|
| 159278 |
+
"step": 47090
|
| 159279 |
+
},
|
| 159280 |
+
{
|
| 159281 |
+
"epoch": 376.45,
|
| 159282 |
+
"learning_rate": 9.261570512820513e-06,
|
| 159283 |
+
"loss": 0.3981,
|
| 159284 |
+
"step": 47095
|
| 159285 |
+
},
|
| 159286 |
+
{
|
| 159287 |
+
"epoch": 376.49,
|
| 159288 |
+
"learning_rate": 9.261490384615386e-06,
|
| 159289 |
+
"loss": 0.3617,
|
| 159290 |
+
"step": 47100
|
| 159291 |
+
},
|
| 159292 |
+
{
|
| 159293 |
+
"epoch": 376.53,
|
| 159294 |
+
"learning_rate": 9.261410256410258e-06,
|
| 159295 |
+
"loss": 0.3778,
|
| 159296 |
+
"step": 47105
|
| 159297 |
+
},
|
| 159298 |
+
{
|
| 159299 |
+
"epoch": 376.57,
|
| 159300 |
+
"learning_rate": 9.261330128205129e-06,
|
| 159301 |
+
"loss": 0.5657,
|
| 159302 |
+
"step": 47110
|
| 159303 |
+
},
|
| 159304 |
+
{
|
| 159305 |
+
"epoch": 376.61,
|
| 159306 |
+
"learning_rate": 9.26125e-06,
|
| 159307 |
+
"loss": 1.2655,
|
| 159308 |
+
"step": 47115
|
| 159309 |
+
},
|
| 159310 |
+
{
|
| 159311 |
+
"epoch": 376.65,
|
| 159312 |
+
"learning_rate": 9.261169871794874e-06,
|
| 159313 |
+
"loss": 0.3939,
|
| 159314 |
+
"step": 47120
|
| 159315 |
+
},
|
| 159316 |
+
{
|
| 159317 |
+
"epoch": 376.69,
|
| 159318 |
+
"learning_rate": 9.261089743589743e-06,
|
| 159319 |
+
"loss": 0.3814,
|
| 159320 |
+
"step": 47125
|
| 159321 |
+
},
|
| 159322 |
+
{
|
| 159323 |
+
"epoch": 376.73,
|
| 159324 |
+
"learning_rate": 9.261009615384616e-06,
|
| 159325 |
+
"loss": 0.4081,
|
| 159326 |
+
"step": 47130
|
| 159327 |
+
},
|
| 159328 |
+
{
|
| 159329 |
+
"epoch": 376.77,
|
| 159330 |
+
"learning_rate": 9.260929487179488e-06,
|
| 159331 |
+
"loss": 0.6094,
|
| 159332 |
+
"step": 47135
|
| 159333 |
+
},
|
| 159334 |
+
{
|
| 159335 |
+
"epoch": 376.81,
|
| 159336 |
+
"learning_rate": 9.260849358974359e-06,
|
| 159337 |
+
"loss": 1.2367,
|
| 159338 |
+
"step": 47140
|
| 159339 |
+
},
|
| 159340 |
+
{
|
| 159341 |
+
"epoch": 376.85,
|
| 159342 |
+
"learning_rate": 9.260769230769232e-06,
|
| 159343 |
+
"loss": 0.3702,
|
| 159344 |
+
"step": 47145
|
| 159345 |
+
},
|
| 159346 |
+
{
|
| 159347 |
+
"epoch": 376.89,
|
| 159348 |
+
"learning_rate": 9.260689102564103e-06,
|
| 159349 |
+
"loss": 0.3565,
|
| 159350 |
+
"step": 47150
|
| 159351 |
+
},
|
| 159352 |
+
{
|
| 159353 |
+
"epoch": 376.93,
|
| 159354 |
+
"learning_rate": 9.260608974358975e-06,
|
| 159355 |
+
"loss": 0.3359,
|
| 159356 |
+
"step": 47155
|
| 159357 |
+
},
|
| 159358 |
+
{
|
| 159359 |
+
"epoch": 376.97,
|
| 159360 |
+
"learning_rate": 9.260528846153846e-06,
|
| 159361 |
+
"loss": 0.6193,
|
| 159362 |
+
"step": 47160
|
| 159363 |
+
},
|
| 159364 |
+
{
|
| 159365 |
+
"epoch": 377.0,
|
| 159366 |
+
"eval_loss": 0.482885479927063,
|
| 159367 |
+
"eval_runtime": 40.1976,
|
| 159368 |
+
"eval_samples_per_second": 20.797,
|
| 159369 |
+
"eval_steps_per_second": 0.672,
|
| 159370 |
+
"eval_wer": 0.20530342419965833,
|
| 159371 |
+
"step": 47164
|
| 159372 |
+
},
|
| 159373 |
+
{
|
| 159374 |
+
"epoch": 377.01,
|
| 159375 |
+
"learning_rate": 9.260448717948719e-06,
|
| 159376 |
+
"loss": 0.4196,
|
| 159377 |
+
"step": 47165
|
| 159378 |
+
},
|
| 159379 |
+
{
|
| 159380 |
+
"epoch": 377.05,
|
| 159381 |
+
"learning_rate": 9.26036858974359e-06,
|
| 159382 |
+
"loss": 0.3283,
|
| 159383 |
+
"step": 47170
|
| 159384 |
+
},
|
| 159385 |
+
{
|
| 159386 |
+
"epoch": 377.09,
|
| 159387 |
+
"learning_rate": 9.260288461538462e-06,
|
| 159388 |
+
"loss": 0.3014,
|
| 159389 |
+
"step": 47175
|
| 159390 |
+
},
|
| 159391 |
+
{
|
| 159392 |
+
"epoch": 377.13,
|
| 159393 |
+
"learning_rate": 9.260208333333333e-06,
|
| 159394 |
+
"loss": 0.4168,
|
| 159395 |
+
"step": 47180
|
| 159396 |
+
},
|
| 159397 |
+
{
|
| 159398 |
+
"epoch": 377.17,
|
| 159399 |
+
"learning_rate": 9.260128205128206e-06,
|
| 159400 |
+
"loss": 0.5062,
|
| 159401 |
+
"step": 47185
|
| 159402 |
+
},
|
| 159403 |
+
{
|
| 159404 |
+
"epoch": 377.21,
|
| 159405 |
+
"learning_rate": 9.260048076923078e-06,
|
| 159406 |
+
"loss": 1.4582,
|
| 159407 |
+
"step": 47190
|
| 159408 |
+
},
|
| 159409 |
+
{
|
| 159410 |
+
"epoch": 377.25,
|
| 159411 |
+
"learning_rate": 9.259967948717949e-06,
|
| 159412 |
+
"loss": 0.3753,
|
| 159413 |
+
"step": 47195
|
| 159414 |
+
},
|
| 159415 |
+
{
|
| 159416 |
+
"epoch": 377.29,
|
| 159417 |
+
"learning_rate": 9.259887820512822e-06,
|
| 159418 |
+
"loss": 0.3808,
|
| 159419 |
+
"step": 47200
|
| 159420 |
+
},
|
| 159421 |
+
{
|
| 159422 |
+
"epoch": 377.33,
|
| 159423 |
+
"learning_rate": 9.259807692307693e-06,
|
| 159424 |
+
"loss": 0.5026,
|
| 159425 |
+
"step": 47205
|
| 159426 |
+
},
|
| 159427 |
+
{
|
| 159428 |
+
"epoch": 377.37,
|
| 159429 |
+
"learning_rate": 9.259727564102565e-06,
|
| 159430 |
+
"loss": 0.6164,
|
| 159431 |
+
"step": 47210
|
| 159432 |
+
},
|
| 159433 |
+
{
|
| 159434 |
+
"epoch": 377.41,
|
| 159435 |
+
"learning_rate": 9.259647435897436e-06,
|
| 159436 |
+
"loss": 1.2348,
|
| 159437 |
+
"step": 47215
|
| 159438 |
+
},
|
| 159439 |
+
{
|
| 159440 |
+
"epoch": 377.45,
|
| 159441 |
+
"learning_rate": 9.25956730769231e-06,
|
| 159442 |
+
"loss": 0.3612,
|
| 159443 |
+
"step": 47220
|
| 159444 |
+
},
|
| 159445 |
+
{
|
| 159446 |
+
"epoch": 377.49,
|
| 159447 |
+
"learning_rate": 9.25948717948718e-06,
|
| 159448 |
+
"loss": 0.3119,
|
| 159449 |
+
"step": 47225
|
| 159450 |
+
},
|
| 159451 |
+
{
|
| 159452 |
+
"epoch": 377.53,
|
| 159453 |
+
"learning_rate": 9.259407051282052e-06,
|
| 159454 |
+
"loss": 0.3822,
|
| 159455 |
+
"step": 47230
|
| 159456 |
+
},
|
| 159457 |
+
{
|
| 159458 |
+
"epoch": 377.57,
|
| 159459 |
+
"learning_rate": 9.259326923076923e-06,
|
| 159460 |
+
"loss": 0.5769,
|
| 159461 |
+
"step": 47235
|
| 159462 |
+
},
|
| 159463 |
+
{
|
| 159464 |
+
"epoch": 377.61,
|
| 159465 |
+
"learning_rate": 9.259246794871796e-06,
|
| 159466 |
+
"loss": 1.225,
|
| 159467 |
+
"step": 47240
|
| 159468 |
+
},
|
| 159469 |
+
{
|
| 159470 |
+
"epoch": 377.65,
|
| 159471 |
+
"learning_rate": 9.259166666666668e-06,
|
| 159472 |
+
"loss": 0.3264,
|
| 159473 |
+
"step": 47245
|
| 159474 |
+
},
|
| 159475 |
+
{
|
| 159476 |
+
"epoch": 377.69,
|
| 159477 |
+
"learning_rate": 9.259086538461539e-06,
|
| 159478 |
+
"loss": 0.401,
|
| 159479 |
+
"step": 47250
|
| 159480 |
+
},
|
| 159481 |
+
{
|
| 159482 |
+
"epoch": 377.73,
|
| 159483 |
+
"learning_rate": 9.259006410256412e-06,
|
| 159484 |
+
"loss": 0.35,
|
| 159485 |
+
"step": 47255
|
| 159486 |
+
},
|
| 159487 |
+
{
|
| 159488 |
+
"epoch": 377.77,
|
| 159489 |
+
"learning_rate": 9.258926282051283e-06,
|
| 159490 |
+
"loss": 0.5652,
|
| 159491 |
+
"step": 47260
|
| 159492 |
+
},
|
| 159493 |
+
{
|
| 159494 |
+
"epoch": 377.81,
|
| 159495 |
+
"learning_rate": 9.258846153846155e-06,
|
| 159496 |
+
"loss": 1.2207,
|
| 159497 |
+
"step": 47265
|
| 159498 |
+
},
|
| 159499 |
+
{
|
| 159500 |
+
"epoch": 377.85,
|
| 159501 |
+
"learning_rate": 9.258766025641026e-06,
|
| 159502 |
+
"loss": 0.3268,
|
| 159503 |
+
"step": 47270
|
| 159504 |
+
},
|
| 159505 |
+
{
|
| 159506 |
+
"epoch": 377.89,
|
| 159507 |
+
"learning_rate": 9.2586858974359e-06,
|
| 159508 |
+
"loss": 0.3757,
|
| 159509 |
+
"step": 47275
|
| 159510 |
+
},
|
| 159511 |
+
{
|
| 159512 |
+
"epoch": 377.93,
|
| 159513 |
+
"learning_rate": 9.258605769230769e-06,
|
| 159514 |
+
"loss": 0.9012,
|
| 159515 |
+
"step": 47280
|
| 159516 |
+
},
|
| 159517 |
+
{
|
| 159518 |
+
"epoch": 377.97,
|
| 159519 |
+
"learning_rate": 9.258525641025642e-06,
|
| 159520 |
+
"loss": 0.7584,
|
| 159521 |
+
"step": 47285
|
| 159522 |
+
},
|
| 159523 |
+
{
|
| 159524 |
+
"epoch": 378.0,
|
| 159525 |
+
"eval_loss": 0.4493987560272217,
|
| 159526 |
+
"eval_runtime": 40.9245,
|
| 159527 |
+
"eval_samples_per_second": 20.428,
|
| 159528 |
+
"eval_steps_per_second": 0.66,
|
| 159529 |
+
"eval_wer": 0.21467253953354032,
|
| 159530 |
+
"step": 47289
|
| 159531 |
+
},
|
| 159532 |
+
{
|
| 159533 |
+
"epoch": 378.01,
|
| 159534 |
+
"learning_rate": 9.25846153846154e-06,
|
| 159535 |
+
"loss": 0.5,
|
| 159536 |
+
"step": 47290
|
| 159537 |
+
},
|
| 159538 |
+
{
|
| 159539 |
+
"epoch": 378.05,
|
| 159540 |
+
"learning_rate": 9.25838141025641e-06,
|
| 159541 |
+
"loss": 0.3277,
|
| 159542 |
+
"step": 47295
|
| 159543 |
+
},
|
| 159544 |
+
{
|
| 159545 |
+
"epoch": 378.09,
|
| 159546 |
+
"learning_rate": 9.258301282051284e-06,
|
| 159547 |
+
"loss": 0.3122,
|
| 159548 |
+
"step": 47300
|
| 159549 |
+
},
|
| 159550 |
+
{
|
| 159551 |
+
"epoch": 378.13,
|
| 159552 |
+
"learning_rate": 9.258221153846153e-06,
|
| 159553 |
+
"loss": 0.4403,
|
| 159554 |
+
"step": 47305
|
| 159555 |
+
},
|
| 159556 |
+
{
|
| 159557 |
+
"epoch": 378.17,
|
| 159558 |
+
"learning_rate": 9.258141025641026e-06,
|
| 159559 |
+
"loss": 0.604,
|
| 159560 |
+
"step": 47310
|
| 159561 |
+
},
|
| 159562 |
+
{
|
| 159563 |
+
"epoch": 378.21,
|
| 159564 |
+
"learning_rate": 9.258060897435898e-06,
|
| 159565 |
+
"loss": 1.1431,
|
| 159566 |
+
"step": 47315
|
| 159567 |
+
},
|
| 159568 |
+
{
|
| 159569 |
+
"epoch": 378.25,
|
| 159570 |
+
"learning_rate": 9.25798076923077e-06,
|
| 159571 |
+
"loss": 0.3546,
|
| 159572 |
+
"step": 47320
|
| 159573 |
+
},
|
| 159574 |
+
{
|
| 159575 |
+
"epoch": 378.29,
|
| 159576 |
+
"learning_rate": 9.257900641025642e-06,
|
| 159577 |
+
"loss": 0.3311,
|
| 159578 |
+
"step": 47325
|
| 159579 |
+
},
|
| 159580 |
+
{
|
| 159581 |
+
"epoch": 378.33,
|
| 159582 |
+
"learning_rate": 9.257820512820514e-06,
|
| 159583 |
+
"loss": 0.3303,
|
| 159584 |
+
"step": 47330
|
| 159585 |
+
},
|
| 159586 |
+
{
|
| 159587 |
+
"epoch": 378.37,
|
| 159588 |
+
"learning_rate": 9.257740384615385e-06,
|
| 159589 |
+
"loss": 0.638,
|
| 159590 |
+
"step": 47335
|
| 159591 |
+
},
|
| 159592 |
+
{
|
| 159593 |
+
"epoch": 378.41,
|
| 159594 |
+
"learning_rate": 9.257660256410256e-06,
|
| 159595 |
+
"loss": 1.1426,
|
| 159596 |
+
"step": 47340
|
| 159597 |
+
},
|
| 159598 |
+
{
|
| 159599 |
+
"epoch": 378.45,
|
| 159600 |
+
"learning_rate": 9.25758012820513e-06,
|
| 159601 |
+
"loss": 0.3966,
|
| 159602 |
+
"step": 47345
|
| 159603 |
+
},
|
| 159604 |
+
{
|
| 159605 |
+
"epoch": 378.49,
|
| 159606 |
+
"learning_rate": 9.2575e-06,
|
| 159607 |
+
"loss": 0.3601,
|
| 159608 |
+
"step": 47350
|
| 159609 |
+
},
|
| 159610 |
+
{
|
| 159611 |
+
"epoch": 378.53,
|
| 159612 |
+
"learning_rate": 9.257419871794872e-06,
|
| 159613 |
+
"loss": 0.3695,
|
| 159614 |
+
"step": 47355
|
| 159615 |
+
},
|
| 159616 |
+
{
|
| 159617 |
+
"epoch": 378.57,
|
| 159618 |
+
"learning_rate": 9.257339743589743e-06,
|
| 159619 |
+
"loss": 0.7932,
|
| 159620 |
+
"step": 47360
|
| 159621 |
+
},
|
| 159622 |
+
{
|
| 159623 |
+
"epoch": 378.61,
|
| 159624 |
+
"learning_rate": 9.257259615384616e-06,
|
| 159625 |
+
"loss": 1.4832,
|
| 159626 |
+
"step": 47365
|
| 159627 |
+
},
|
| 159628 |
+
{
|
| 159629 |
+
"epoch": 378.65,
|
| 159630 |
+
"learning_rate": 9.257179487179488e-06,
|
| 159631 |
+
"loss": 0.3544,
|
| 159632 |
+
"step": 47370
|
| 159633 |
+
},
|
| 159634 |
+
{
|
| 159635 |
+
"epoch": 378.69,
|
| 159636 |
+
"learning_rate": 9.25709935897436e-06,
|
| 159637 |
+
"loss": 0.3521,
|
| 159638 |
+
"step": 47375
|
| 159639 |
+
},
|
| 159640 |
+
{
|
| 159641 |
+
"epoch": 378.73,
|
| 159642 |
+
"learning_rate": 9.257019230769232e-06,
|
| 159643 |
+
"loss": 0.3812,
|
| 159644 |
+
"step": 47380
|
| 159645 |
+
},
|
| 159646 |
+
{
|
| 159647 |
+
"epoch": 378.77,
|
| 159648 |
+
"learning_rate": 9.256939102564104e-06,
|
| 159649 |
+
"loss": 0.5812,
|
| 159650 |
+
"step": 47385
|
| 159651 |
+
},
|
| 159652 |
+
{
|
| 159653 |
+
"epoch": 378.81,
|
| 159654 |
+
"learning_rate": 9.256858974358975e-06,
|
| 159655 |
+
"loss": 1.0635,
|
| 159656 |
+
"step": 47390
|
| 159657 |
+
},
|
| 159658 |
+
{
|
| 159659 |
+
"epoch": 378.85,
|
| 159660 |
+
"learning_rate": 9.256778846153846e-06,
|
| 159661 |
+
"loss": 0.3511,
|
| 159662 |
+
"step": 47395
|
| 159663 |
+
},
|
| 159664 |
+
{
|
| 159665 |
+
"epoch": 378.89,
|
| 159666 |
+
"learning_rate": 9.25669871794872e-06,
|
| 159667 |
+
"loss": 0.4201,
|
| 159668 |
+
"step": 47400
|
| 159669 |
+
},
|
| 159670 |
+
{
|
| 159671 |
+
"epoch": 378.93,
|
| 159672 |
+
"learning_rate": 9.25661858974359e-06,
|
| 159673 |
+
"loss": 0.3906,
|
| 159674 |
+
"step": 47405
|
| 159675 |
+
},
|
| 159676 |
+
{
|
| 159677 |
+
"epoch": 378.97,
|
| 159678 |
+
"learning_rate": 9.256538461538462e-06,
|
| 159679 |
+
"loss": 0.5379,
|
| 159680 |
+
"step": 47410
|
| 159681 |
+
},
|
| 159682 |
+
{
|
| 159683 |
+
"epoch": 379.0,
|
| 159684 |
+
"eval_loss": 0.41098281741142273,
|
| 159685 |
+
"eval_runtime": 41.7422,
|
| 159686 |
+
"eval_samples_per_second": 20.028,
|
| 159687 |
+
"eval_steps_per_second": 0.647,
|
| 159688 |
+
"eval_wer": 0.20962497290266638,
|
| 159689 |
+
"step": 47414
|
| 159690 |
+
},
|
| 159691 |
+
{
|
| 159692 |
+
"epoch": 382.01,
|
| 159693 |
+
"learning_rate": 9.256458333333333e-06,
|
| 159694 |
+
"loss": 0.4079,
|
| 159695 |
+
"step": 47415
|
| 159696 |
+
},
|
| 159697 |
+
{
|
| 159698 |
+
"epoch": 382.05,
|
| 159699 |
+
"learning_rate": 9.256378205128207e-06,
|
| 159700 |
+
"loss": 0.4051,
|
| 159701 |
+
"step": 47420
|
| 159702 |
+
},
|
| 159703 |
+
{
|
| 159704 |
+
"epoch": 382.09,
|
| 159705 |
+
"learning_rate": 9.256298076923078e-06,
|
| 159706 |
+
"loss": 0.3244,
|
| 159707 |
+
"step": 47425
|
| 159708 |
+
},
|
| 159709 |
+
{
|
| 159710 |
+
"epoch": 382.13,
|
| 159711 |
+
"learning_rate": 9.25621794871795e-06,
|
| 159712 |
+
"loss": 0.3739,
|
| 159713 |
+
"step": 47430
|
| 159714 |
+
},
|
| 159715 |
+
{
|
| 159716 |
+
"epoch": 382.17,
|
| 159717 |
+
"learning_rate": 9.256137820512822e-06,
|
| 159718 |
+
"loss": 0.5677,
|
| 159719 |
+
"step": 47435
|
| 159720 |
+
},
|
| 159721 |
+
{
|
| 159722 |
+
"epoch": 382.21,
|
| 159723 |
+
"learning_rate": 9.256057692307692e-06,
|
| 159724 |
+
"loss": 1.2913,
|
| 159725 |
+
"step": 47440
|
| 159726 |
+
},
|
| 159727 |
+
{
|
| 159728 |
+
"epoch": 382.25,
|
| 159729 |
+
"learning_rate": 9.255977564102565e-06,
|
| 159730 |
+
"loss": 0.3349,
|
| 159731 |
+
"step": 47445
|
| 159732 |
+
},
|
| 159733 |
+
{
|
| 159734 |
+
"epoch": 382.29,
|
| 159735 |
+
"learning_rate": 9.255897435897436e-06,
|
| 159736 |
+
"loss": 0.3091,
|
| 159737 |
+
"step": 47450
|
| 159738 |
+
},
|
| 159739 |
+
{
|
| 159740 |
+
"epoch": 382.33,
|
| 159741 |
+
"learning_rate": 9.255817307692308e-06,
|
| 159742 |
+
"loss": 0.3982,
|
| 159743 |
+
"step": 47455
|
| 159744 |
+
},
|
| 159745 |
+
{
|
| 159746 |
+
"epoch": 382.37,
|
| 159747 |
+
"learning_rate": 9.255737179487179e-06,
|
| 159748 |
+
"loss": 0.7085,
|
| 159749 |
+
"step": 47460
|
| 159750 |
+
},
|
| 159751 |
+
{
|
| 159752 |
+
"epoch": 382.41,
|
| 159753 |
+
"learning_rate": 9.255657051282052e-06,
|
| 159754 |
+
"loss": 1.1537,
|
| 159755 |
+
"step": 47465
|
| 159756 |
+
},
|
| 159757 |
+
{
|
| 159758 |
+
"epoch": 382.45,
|
| 159759 |
+
"learning_rate": 9.255576923076925e-06,
|
| 159760 |
+
"loss": 0.3985,
|
| 159761 |
+
"step": 47470
|
| 159762 |
+
},
|
| 159763 |
+
{
|
| 159764 |
+
"epoch": 382.49,
|
| 159765 |
+
"learning_rate": 9.255496794871795e-06,
|
| 159766 |
+
"loss": 0.3551,
|
| 159767 |
+
"step": 47475
|
| 159768 |
+
},
|
| 159769 |
+
{
|
| 159770 |
+
"epoch": 382.53,
|
| 159771 |
+
"learning_rate": 9.255416666666668e-06,
|
| 159772 |
+
"loss": 0.3884,
|
| 159773 |
+
"step": 47480
|
| 159774 |
+
},
|
| 159775 |
+
{
|
| 159776 |
+
"epoch": 382.57,
|
| 159777 |
+
"learning_rate": 9.25533653846154e-06,
|
| 159778 |
+
"loss": 0.5744,
|
| 159779 |
+
"step": 47485
|
| 159780 |
+
},
|
| 159781 |
+
{
|
| 159782 |
+
"epoch": 382.61,
|
| 159783 |
+
"learning_rate": 9.25525641025641e-06,
|
| 159784 |
+
"loss": 1.2079,
|
| 159785 |
+
"step": 47490
|
| 159786 |
+
},
|
| 159787 |
+
{
|
| 159788 |
+
"epoch": 382.65,
|
| 159789 |
+
"learning_rate": 9.255176282051282e-06,
|
| 159790 |
+
"loss": 0.3573,
|
| 159791 |
+
"step": 47495
|
| 159792 |
+
},
|
| 159793 |
+
{
|
| 159794 |
+
"epoch": 382.69,
|
| 159795 |
+
"learning_rate": 9.255096153846155e-06,
|
| 159796 |
+
"loss": 0.3638,
|
| 159797 |
+
"step": 47500
|
| 159798 |
+
},
|
| 159799 |
+
{
|
| 159800 |
+
"epoch": 382.73,
|
| 159801 |
+
"learning_rate": 9.255016025641026e-06,
|
| 159802 |
+
"loss": 0.373,
|
| 159803 |
+
"step": 47505
|
| 159804 |
+
},
|
| 159805 |
+
{
|
| 159806 |
+
"epoch": 382.77,
|
| 159807 |
+
"learning_rate": 9.254935897435898e-06,
|
| 159808 |
+
"loss": 0.5704,
|
| 159809 |
+
"step": 47510
|
| 159810 |
+
},
|
| 159811 |
+
{
|
| 159812 |
+
"epoch": 382.81,
|
| 159813 |
+
"learning_rate": 9.254855769230769e-06,
|
| 159814 |
+
"loss": 1.3223,
|
| 159815 |
+
"step": 47515
|
| 159816 |
+
},
|
| 159817 |
+
{
|
| 159818 |
+
"epoch": 382.85,
|
| 159819 |
+
"learning_rate": 9.254775641025642e-06,
|
| 159820 |
+
"loss": 0.4135,
|
| 159821 |
+
"step": 47520
|
| 159822 |
+
},
|
| 159823 |
+
{
|
| 159824 |
+
"epoch": 382.89,
|
| 159825 |
+
"learning_rate": 9.254695512820514e-06,
|
| 159826 |
+
"loss": 0.3511,
|
| 159827 |
+
"step": 47525
|
| 159828 |
+
},
|
| 159829 |
+
{
|
| 159830 |
+
"epoch": 382.93,
|
| 159831 |
+
"learning_rate": 9.254615384615385e-06,
|
| 159832 |
+
"loss": 0.443,
|
| 159833 |
+
"step": 47530
|
| 159834 |
+
},
|
| 159835 |
+
{
|
| 159836 |
+
"epoch": 382.97,
|
| 159837 |
+
"learning_rate": 9.254535256410258e-06,
|
| 159838 |
+
"loss": 0.843,
|
| 159839 |
+
"step": 47535
|
| 159840 |
+
},
|
| 159841 |
+
{
|
| 159842 |
+
"epoch": 383.0,
|
| 159843 |
+
"eval_loss": 0.40662431716918945,
|
| 159844 |
+
"eval_runtime": 40.2188,
|
| 159845 |
+
"eval_samples_per_second": 20.786,
|
| 159846 |
+
"eval_steps_per_second": 0.671,
|
| 159847 |
+
"eval_wer": 0.20183823529411765,
|
| 159848 |
+
"step": 47538
|
| 159849 |
+
},
|
| 159850 |
+
{
|
| 159851 |
+
"epoch": 380.02,
|
| 159852 |
+
"learning_rate": 9.25445512820513e-06,
|
| 159853 |
+
"loss": 0.3762,
|
| 159854 |
+
"step": 47540
|
| 159855 |
+
},
|
| 159856 |
+
{
|
| 159857 |
+
"epoch": 380.06,
|
| 159858 |
+
"learning_rate": 9.254375e-06,
|
| 159859 |
+
"loss": 0.2878,
|
| 159860 |
+
"step": 47545
|
| 159861 |
+
},
|
| 159862 |
+
{
|
| 159863 |
+
"epoch": 380.1,
|
| 159864 |
+
"learning_rate": 9.254294871794872e-06,
|
| 159865 |
+
"loss": 0.3521,
|
| 159866 |
+
"step": 47550
|
| 159867 |
+
},
|
| 159868 |
+
{
|
| 159869 |
+
"epoch": 380.14,
|
| 159870 |
+
"learning_rate": 9.254214743589745e-06,
|
| 159871 |
+
"loss": 0.4816,
|
| 159872 |
+
"step": 47555
|
| 159873 |
+
},
|
| 159874 |
+
{
|
| 159875 |
+
"epoch": 380.18,
|
| 159876 |
+
"learning_rate": 9.254134615384616e-06,
|
| 159877 |
+
"loss": 0.6926,
|
| 159878 |
+
"step": 47560
|
| 159879 |
+
},
|
| 159880 |
+
{
|
| 159881 |
+
"epoch": 380.22,
|
| 159882 |
+
"learning_rate": 9.254054487179488e-06,
|
| 159883 |
+
"loss": 1.1028,
|
| 159884 |
+
"step": 47565
|
| 159885 |
+
},
|
| 159886 |
+
{
|
| 159887 |
+
"epoch": 380.26,
|
| 159888 |
+
"learning_rate": 9.25397435897436e-06,
|
| 159889 |
+
"loss": 0.333,
|
| 159890 |
+
"step": 47570
|
| 159891 |
+
},
|
| 159892 |
+
{
|
| 159893 |
+
"epoch": 380.3,
|
| 159894 |
+
"learning_rate": 9.253894230769232e-06,
|
| 159895 |
+
"loss": 0.3393,
|
| 159896 |
+
"step": 47575
|
| 159897 |
+
},
|
| 159898 |
+
{
|
| 159899 |
+
"epoch": 380.34,
|
| 159900 |
+
"learning_rate": 9.253814102564104e-06,
|
| 159901 |
+
"loss": 0.3715,
|
| 159902 |
+
"step": 47580
|
| 159903 |
+
},
|
| 159904 |
+
{
|
| 159905 |
+
"epoch": 380.38,
|
| 159906 |
+
"learning_rate": 9.253733974358975e-06,
|
| 159907 |
+
"loss": 0.7846,
|
| 159908 |
+
"step": 47585
|
| 159909 |
+
},
|
| 159910 |
+
{
|
| 159911 |
+
"epoch": 380.42,
|
| 159912 |
+
"learning_rate": 9.253653846153848e-06,
|
| 159913 |
+
"loss": 1.1017,
|
| 159914 |
+
"step": 47590
|
| 159915 |
+
},
|
| 159916 |
+
{
|
| 159917 |
+
"epoch": 380.46,
|
| 159918 |
+
"learning_rate": 9.253573717948718e-06,
|
| 159919 |
+
"loss": 0.3653,
|
| 159920 |
+
"step": 47595
|
| 159921 |
+
},
|
| 159922 |
+
{
|
| 159923 |
+
"epoch": 380.5,
|
| 159924 |
+
"learning_rate": 9.25349358974359e-06,
|
| 159925 |
+
"loss": 0.3604,
|
| 159926 |
+
"step": 47600
|
| 159927 |
+
},
|
| 159928 |
+
{
|
| 159929 |
+
"epoch": 380.54,
|
| 159930 |
+
"learning_rate": 9.253413461538462e-06,
|
| 159931 |
+
"loss": 0.4365,
|
| 159932 |
+
"step": 47605
|
| 159933 |
+
},
|
| 159934 |
+
{
|
| 159935 |
+
"epoch": 380.58,
|
| 159936 |
+
"learning_rate": 9.253333333333333e-06,
|
| 159937 |
+
"loss": 0.6824,
|
| 159938 |
+
"step": 47610
|
| 159939 |
+
},
|
| 159940 |
+
{
|
| 159941 |
+
"epoch": 380.62,
|
| 159942 |
+
"learning_rate": 9.253253205128205e-06,
|
| 159943 |
+
"loss": 1.1141,
|
| 159944 |
+
"step": 47615
|
| 159945 |
+
},
|
| 159946 |
+
{
|
| 159947 |
+
"epoch": 380.66,
|
| 159948 |
+
"learning_rate": 9.253173076923078e-06,
|
| 159949 |
+
"loss": 0.4078,
|
| 159950 |
+
"step": 47620
|
| 159951 |
+
},
|
| 159952 |
+
{
|
| 159953 |
+
"epoch": 380.7,
|
| 159954 |
+
"learning_rate": 9.25309294871795e-06,
|
| 159955 |
+
"loss": 0.3401,
|
| 159956 |
+
"step": 47625
|
| 159957 |
+
},
|
| 159958 |
+
{
|
| 159959 |
+
"epoch": 380.74,
|
| 159960 |
+
"learning_rate": 9.25301282051282e-06,
|
| 159961 |
+
"loss": 0.4202,
|
| 159962 |
+
"step": 47630
|
| 159963 |
+
},
|
| 159964 |
+
{
|
| 159965 |
+
"epoch": 380.78,
|
| 159966 |
+
"learning_rate": 9.252932692307694e-06,
|
| 159967 |
+
"loss": 0.8217,
|
| 159968 |
+
"step": 47635
|
| 159969 |
+
},
|
| 159970 |
+
{
|
| 159971 |
+
"epoch": 380.82,
|
| 159972 |
+
"learning_rate": 9.252852564102565e-06,
|
| 159973 |
+
"loss": 1.1662,
|
| 159974 |
+
"step": 47640
|
| 159975 |
+
},
|
| 159976 |
+
{
|
| 159977 |
+
"epoch": 380.86,
|
| 159978 |
+
"learning_rate": 9.252772435897436e-06,
|
| 159979 |
+
"loss": 0.384,
|
| 159980 |
+
"step": 47645
|
| 159981 |
+
},
|
| 159982 |
+
{
|
| 159983 |
+
"epoch": 380.9,
|
| 159984 |
+
"learning_rate": 9.252692307692308e-06,
|
| 159985 |
+
"loss": 0.3267,
|
| 159986 |
+
"step": 47650
|
| 159987 |
+
},
|
| 159988 |
+
{
|
| 159989 |
+
"epoch": 380.94,
|
| 159990 |
+
"learning_rate": 9.25261217948718e-06,
|
| 159991 |
+
"loss": 0.4336,
|
| 159992 |
+
"step": 47655
|
| 159993 |
+
},
|
| 159994 |
+
{
|
| 159995 |
+
"epoch": 380.98,
|
| 159996 |
+
"learning_rate": 9.252532051282052e-06,
|
| 159997 |
+
"loss": 0.8251,
|
| 159998 |
+
"step": 47660
|
| 159999 |
+
},
|
| 160000 |
+
{
|
| 160001 |
+
"epoch": 381.0,
|
| 160002 |
+
"eval_loss": 0.4215746521949768,
|
| 160003 |
+
"eval_runtime": 41.2472,
|
| 160004 |
+
"eval_samples_per_second": 20.268,
|
| 160005 |
+
"eval_steps_per_second": 0.655,
|
| 160006 |
+
"eval_wer": 0.20148385219668316,
|
| 160007 |
+
"step": 47663
|
| 160008 |
}
|
| 160009 |
],
|
| 160010 |
+
"max_steps": 625000,
|
| 160011 |
"num_train_epochs": 5000,
|
| 160012 |
+
"total_flos": 1.3410923324603713e+20,
|
| 160013 |
"trial_name": null,
|
| 160014 |
"trial_params": null
|
| 160015 |
}
|
model-bin/finetune/base/{checkpoint-47039 → checkpoint-47663}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629726271.057744/events.out.tfevents.1629726271.74272264b15c.932.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b9db3e9d3ea351bc26cb8b9983127bb0f702a9106cac7ea3dc41b274977c333a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629726936.3025346/events.out.tfevents.1629726936.74272264b15c.932.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:165e9315c50ee5d69f05a748092d591eedf130eefcb7df44fc318cd85193aede
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629727569.1115878/events.out.tfevents.1629727569.74272264b15c.932.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10f156f766741d9487a186c530690042f45e38b1470071e7c67342fc2f4de57e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629728226.273403/events.out.tfevents.1629728226.74272264b15c.932.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1c9dc7390405e3fe4eccecef2b743e161f9072b9190fa500eadc2be3b6d7a01
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629728868.313585/events.out.tfevents.1629728868.74272264b15c.932.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c64cddb3bc7b984e1ab64dd8d370fdd804d9667df449b32078a4e5821bd879f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629726271.74272264b15c.932.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03b96f74365f459c84063de933519e6e22eed7bd3f11ef3bff26358ce5781c45
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629726936.74272264b15c.932.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f8cad111a8dc44b3e0e2e3199a42f6d4778ea382a280132cd13a1a32431c443
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629727569.74272264b15c.932.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68b30bee97f18985d11b998fbada5fda3910081917a0d4e3297fea859b2dba95
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629728226.74272264b15c.932.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22eefdf23fa9bf2fa42d1587ed28ef33d72d72209393e4b6d6cf5b6b8edc19a5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629728868.74272264b15c.932.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54c523d74d675d81de49b234b5990db7bd9e205ec6bd89e2b2413b51fcdf1e5b
|
| 3 |
+
size 8622
|