"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629977881.4113429/events.out.tfevents.1629977881.8e89bd551565.924.171 +3 -0
- model-bin/finetune/base/log/1629978315.2297533/events.out.tfevents.1629978315.8e89bd551565.924.173 +3 -0
- model-bin/finetune/base/log/1629978748.357351/events.out.tfevents.1629978748.8e89bd551565.924.175 +3 -0
- model-bin/finetune/base/log/1629979170.7944605/events.out.tfevents.1629979170.8e89bd551565.924.177 +3 -0
- model-bin/finetune/base/log/1629979593.1009088/events.out.tfevents.1629979593.8e89bd551565.924.179 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629977880.8e89bd551565.924.170 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629978315.8e89bd551565.924.172 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629978748.8e89bd551565.924.174 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629979170.8e89bd551565.924.176 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629979593.8e89bd551565.924.178 +3 -0
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2f5597ab892dc7bbd063f5a173c1dd124665bf7d5e017c2f7fb8b03707277384
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/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:ebc5f970b96c488c19c31e1cb0e4f66aaa8ff3d2a45e1e91dc593de1a8d70888
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/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 sha256:193ad6dc9913f8bd9eee397c4a62b1b7b066bdf26b0dc1694e324b4569f39718
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/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 sha256:4166c0b981defd24252fdcfa267326c23bb9a67917e23b9f16f5ee97f24c853d
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/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:bdfc4589d4b1984faf3a1eeeac0648257c835994e7c2e69d5928ff2a41e4c405
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -225249,11 +225249,800 @@
|
|
| 225249 |
"eval_steps_per_second": 0.715,
|
| 225250 |
"eval_wer": 0.18095796963250935,
|
| 225251 |
"step": 98940
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 225252 |
}
|
| 225253 |
],
|
| 225254 |
"max_steps": 620000,
|
| 225255 |
"num_train_epochs": 5000,
|
| 225256 |
-
"total_flos": 2.
|
| 225257 |
"trial_name": null,
|
| 225258 |
"trial_params": null
|
| 225259 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 801.995983935743,
|
| 5 |
+
"global_step": 99560,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 225249 |
"eval_steps_per_second": 0.715,
|
| 225250 |
"eval_wer": 0.18095796963250935,
|
| 225251 |
"step": 98940
|
| 225252 |
+
},
|
| 225253 |
+
{
|
| 225254 |
+
"epoch": 797.04,
|
| 225255 |
+
"learning_rate": 8.431009615384617e-06,
|
| 225256 |
+
"loss": 0.2915,
|
| 225257 |
+
"step": 98945
|
| 225258 |
+
},
|
| 225259 |
+
{
|
| 225260 |
+
"epoch": 797.08,
|
| 225261 |
+
"learning_rate": 8.430929487179487e-06,
|
| 225262 |
+
"loss": 0.4093,
|
| 225263 |
+
"step": 98950
|
| 225264 |
+
},
|
| 225265 |
+
{
|
| 225266 |
+
"epoch": 797.12,
|
| 225267 |
+
"learning_rate": 8.43084935897436e-06,
|
| 225268 |
+
"loss": 0.3225,
|
| 225269 |
+
"step": 98955
|
| 225270 |
+
},
|
| 225271 |
+
{
|
| 225272 |
+
"epoch": 797.16,
|
| 225273 |
+
"learning_rate": 8.430769230769231e-06,
|
| 225274 |
+
"loss": 0.4467,
|
| 225275 |
+
"step": 98960
|
| 225276 |
+
},
|
| 225277 |
+
{
|
| 225278 |
+
"epoch": 797.2,
|
| 225279 |
+
"learning_rate": 8.430689102564103e-06,
|
| 225280 |
+
"loss": 1.2335,
|
| 225281 |
+
"step": 98965
|
| 225282 |
+
},
|
| 225283 |
+
{
|
| 225284 |
+
"epoch": 797.24,
|
| 225285 |
+
"learning_rate": 8.430608974358974e-06,
|
| 225286 |
+
"loss": 0.3039,
|
| 225287 |
+
"step": 98970
|
| 225288 |
+
},
|
| 225289 |
+
{
|
| 225290 |
+
"epoch": 797.28,
|
| 225291 |
+
"learning_rate": 8.430528846153847e-06,
|
| 225292 |
+
"loss": 0.2726,
|
| 225293 |
+
"step": 98975
|
| 225294 |
+
},
|
| 225295 |
+
{
|
| 225296 |
+
"epoch": 797.32,
|
| 225297 |
+
"learning_rate": 8.430448717948719e-06,
|
| 225298 |
+
"loss": 0.3405,
|
| 225299 |
+
"step": 98980
|
| 225300 |
+
},
|
| 225301 |
+
{
|
| 225302 |
+
"epoch": 797.36,
|
| 225303 |
+
"learning_rate": 8.43036858974359e-06,
|
| 225304 |
+
"loss": 0.4582,
|
| 225305 |
+
"step": 98985
|
| 225306 |
+
},
|
| 225307 |
+
{
|
| 225308 |
+
"epoch": 797.4,
|
| 225309 |
+
"learning_rate": 8.430288461538461e-06,
|
| 225310 |
+
"loss": 1.2883,
|
| 225311 |
+
"step": 98990
|
| 225312 |
+
},
|
| 225313 |
+
{
|
| 225314 |
+
"epoch": 797.44,
|
| 225315 |
+
"learning_rate": 8.430208333333334e-06,
|
| 225316 |
+
"loss": 0.3498,
|
| 225317 |
+
"step": 98995
|
| 225318 |
+
},
|
| 225319 |
+
{
|
| 225320 |
+
"epoch": 797.48,
|
| 225321 |
+
"learning_rate": 8.430128205128206e-06,
|
| 225322 |
+
"loss": 0.2927,
|
| 225323 |
+
"step": 99000
|
| 225324 |
+
},
|
| 225325 |
+
{
|
| 225326 |
+
"epoch": 797.52,
|
| 225327 |
+
"learning_rate": 8.430048076923077e-06,
|
| 225328 |
+
"loss": 0.287,
|
| 225329 |
+
"step": 99005
|
| 225330 |
+
},
|
| 225331 |
+
{
|
| 225332 |
+
"epoch": 797.56,
|
| 225333 |
+
"learning_rate": 8.42996794871795e-06,
|
| 225334 |
+
"loss": 0.4129,
|
| 225335 |
+
"step": 99010
|
| 225336 |
+
},
|
| 225337 |
+
{
|
| 225338 |
+
"epoch": 797.6,
|
| 225339 |
+
"learning_rate": 8.429887820512822e-06,
|
| 225340 |
+
"loss": 1.2854,
|
| 225341 |
+
"step": 99015
|
| 225342 |
+
},
|
| 225343 |
+
{
|
| 225344 |
+
"epoch": 797.64,
|
| 225345 |
+
"learning_rate": 8.429807692307693e-06,
|
| 225346 |
+
"loss": 0.3396,
|
| 225347 |
+
"step": 99020
|
| 225348 |
+
},
|
| 225349 |
+
{
|
| 225350 |
+
"epoch": 797.68,
|
| 225351 |
+
"learning_rate": 8.429727564102564e-06,
|
| 225352 |
+
"loss": 0.2914,
|
| 225353 |
+
"step": 99025
|
| 225354 |
+
},
|
| 225355 |
+
{
|
| 225356 |
+
"epoch": 797.72,
|
| 225357 |
+
"learning_rate": 8.429647435897437e-06,
|
| 225358 |
+
"loss": 0.3242,
|
| 225359 |
+
"step": 99030
|
| 225360 |
+
},
|
| 225361 |
+
{
|
| 225362 |
+
"epoch": 797.76,
|
| 225363 |
+
"learning_rate": 8.429567307692309e-06,
|
| 225364 |
+
"loss": 0.4406,
|
| 225365 |
+
"step": 99035
|
| 225366 |
+
},
|
| 225367 |
+
{
|
| 225368 |
+
"epoch": 797.8,
|
| 225369 |
+
"learning_rate": 8.42948717948718e-06,
|
| 225370 |
+
"loss": 1.202,
|
| 225371 |
+
"step": 99040
|
| 225372 |
+
},
|
| 225373 |
+
{
|
| 225374 |
+
"epoch": 797.84,
|
| 225375 |
+
"learning_rate": 8.429407051282053e-06,
|
| 225376 |
+
"loss": 0.2829,
|
| 225377 |
+
"step": 99045
|
| 225378 |
+
},
|
| 225379 |
+
{
|
| 225380 |
+
"epoch": 797.88,
|
| 225381 |
+
"learning_rate": 8.429326923076924e-06,
|
| 225382 |
+
"loss": 0.2988,
|
| 225383 |
+
"step": 99050
|
| 225384 |
+
},
|
| 225385 |
+
{
|
| 225386 |
+
"epoch": 797.92,
|
| 225387 |
+
"learning_rate": 8.429246794871796e-06,
|
| 225388 |
+
"loss": 0.3442,
|
| 225389 |
+
"step": 99055
|
| 225390 |
+
},
|
| 225391 |
+
{
|
| 225392 |
+
"epoch": 797.96,
|
| 225393 |
+
"learning_rate": 8.429166666666667e-06,
|
| 225394 |
+
"loss": 0.5094,
|
| 225395 |
+
"step": 99060
|
| 225396 |
+
},
|
| 225397 |
+
{
|
| 225398 |
+
"epoch": 798.0,
|
| 225399 |
+
"eval_loss": 0.35809361934661865,
|
| 225400 |
+
"eval_runtime": 40.194,
|
| 225401 |
+
"eval_samples_per_second": 20.724,
|
| 225402 |
+
"eval_steps_per_second": 0.672,
|
| 225403 |
+
"eval_wer": 0.1815831058755926,
|
| 225404 |
+
"step": 99064
|
| 225405 |
+
},
|
| 225406 |
+
{
|
| 225407 |
+
"epoch": 798.01,
|
| 225408 |
+
"learning_rate": 8.42908653846154e-06,
|
| 225409 |
+
"loss": 0.3762,
|
| 225410 |
+
"step": 99065
|
| 225411 |
+
},
|
| 225412 |
+
{
|
| 225413 |
+
"epoch": 798.05,
|
| 225414 |
+
"learning_rate": 8.42900641025641e-06,
|
| 225415 |
+
"loss": 0.3051,
|
| 225416 |
+
"step": 99070
|
| 225417 |
+
},
|
| 225418 |
+
{
|
| 225419 |
+
"epoch": 798.09,
|
| 225420 |
+
"learning_rate": 8.428926282051283e-06,
|
| 225421 |
+
"loss": 0.2992,
|
| 225422 |
+
"step": 99075
|
| 225423 |
+
},
|
| 225424 |
+
{
|
| 225425 |
+
"epoch": 798.13,
|
| 225426 |
+
"learning_rate": 8.428846153846154e-06,
|
| 225427 |
+
"loss": 0.3087,
|
| 225428 |
+
"step": 99080
|
| 225429 |
+
},
|
| 225430 |
+
{
|
| 225431 |
+
"epoch": 798.17,
|
| 225432 |
+
"learning_rate": 8.428766025641026e-06,
|
| 225433 |
+
"loss": 0.5676,
|
| 225434 |
+
"step": 99085
|
| 225435 |
+
},
|
| 225436 |
+
{
|
| 225437 |
+
"epoch": 798.21,
|
| 225438 |
+
"learning_rate": 8.428685897435897e-06,
|
| 225439 |
+
"loss": 1.0171,
|
| 225440 |
+
"step": 99090
|
| 225441 |
+
},
|
| 225442 |
+
{
|
| 225443 |
+
"epoch": 798.25,
|
| 225444 |
+
"learning_rate": 8.42860576923077e-06,
|
| 225445 |
+
"loss": 0.2775,
|
| 225446 |
+
"step": 99095
|
| 225447 |
+
},
|
| 225448 |
+
{
|
| 225449 |
+
"epoch": 798.29,
|
| 225450 |
+
"learning_rate": 8.428525641025641e-06,
|
| 225451 |
+
"loss": 0.3158,
|
| 225452 |
+
"step": 99100
|
| 225453 |
+
},
|
| 225454 |
+
{
|
| 225455 |
+
"epoch": 798.33,
|
| 225456 |
+
"learning_rate": 8.428445512820513e-06,
|
| 225457 |
+
"loss": 0.3832,
|
| 225458 |
+
"step": 99105
|
| 225459 |
+
},
|
| 225460 |
+
{
|
| 225461 |
+
"epoch": 798.37,
|
| 225462 |
+
"learning_rate": 8.428365384615386e-06,
|
| 225463 |
+
"loss": 0.5536,
|
| 225464 |
+
"step": 99110
|
| 225465 |
+
},
|
| 225466 |
+
{
|
| 225467 |
+
"epoch": 798.41,
|
| 225468 |
+
"learning_rate": 8.428285256410257e-06,
|
| 225469 |
+
"loss": 1.2106,
|
| 225470 |
+
"step": 99115
|
| 225471 |
+
},
|
| 225472 |
+
{
|
| 225473 |
+
"epoch": 798.45,
|
| 225474 |
+
"learning_rate": 8.428205128205129e-06,
|
| 225475 |
+
"loss": 0.3579,
|
| 225476 |
+
"step": 99120
|
| 225477 |
+
},
|
| 225478 |
+
{
|
| 225479 |
+
"epoch": 798.49,
|
| 225480 |
+
"learning_rate": 8.428125e-06,
|
| 225481 |
+
"loss": 0.2752,
|
| 225482 |
+
"step": 99125
|
| 225483 |
+
},
|
| 225484 |
+
{
|
| 225485 |
+
"epoch": 798.53,
|
| 225486 |
+
"learning_rate": 8.428044871794873e-06,
|
| 225487 |
+
"loss": 0.3139,
|
| 225488 |
+
"step": 99130
|
| 225489 |
+
},
|
| 225490 |
+
{
|
| 225491 |
+
"epoch": 798.57,
|
| 225492 |
+
"learning_rate": 8.427964743589744e-06,
|
| 225493 |
+
"loss": 0.6959,
|
| 225494 |
+
"step": 99135
|
| 225495 |
+
},
|
| 225496 |
+
{
|
| 225497 |
+
"epoch": 798.61,
|
| 225498 |
+
"learning_rate": 8.427884615384616e-06,
|
| 225499 |
+
"loss": 1.197,
|
| 225500 |
+
"step": 99140
|
| 225501 |
+
},
|
| 225502 |
+
{
|
| 225503 |
+
"epoch": 798.65,
|
| 225504 |
+
"learning_rate": 8.427804487179489e-06,
|
| 225505 |
+
"loss": 0.2704,
|
| 225506 |
+
"step": 99145
|
| 225507 |
+
},
|
| 225508 |
+
{
|
| 225509 |
+
"epoch": 798.69,
|
| 225510 |
+
"learning_rate": 8.42772435897436e-06,
|
| 225511 |
+
"loss": 0.2523,
|
| 225512 |
+
"step": 99150
|
| 225513 |
+
},
|
| 225514 |
+
{
|
| 225515 |
+
"epoch": 798.73,
|
| 225516 |
+
"learning_rate": 8.427644230769231e-06,
|
| 225517 |
+
"loss": 0.3845,
|
| 225518 |
+
"step": 99155
|
| 225519 |
+
},
|
| 225520 |
+
{
|
| 225521 |
+
"epoch": 798.77,
|
| 225522 |
+
"learning_rate": 8.427564102564103e-06,
|
| 225523 |
+
"loss": 0.4935,
|
| 225524 |
+
"step": 99160
|
| 225525 |
+
},
|
| 225526 |
+
{
|
| 225527 |
+
"epoch": 798.81,
|
| 225528 |
+
"learning_rate": 8.427483974358976e-06,
|
| 225529 |
+
"loss": 1.1547,
|
| 225530 |
+
"step": 99165
|
| 225531 |
+
},
|
| 225532 |
+
{
|
| 225533 |
+
"epoch": 798.85,
|
| 225534 |
+
"learning_rate": 8.427403846153847e-06,
|
| 225535 |
+
"loss": 0.2951,
|
| 225536 |
+
"step": 99170
|
| 225537 |
+
},
|
| 225538 |
+
{
|
| 225539 |
+
"epoch": 798.89,
|
| 225540 |
+
"learning_rate": 8.427323717948719e-06,
|
| 225541 |
+
"loss": 0.2957,
|
| 225542 |
+
"step": 99175
|
| 225543 |
+
},
|
| 225544 |
+
{
|
| 225545 |
+
"epoch": 798.93,
|
| 225546 |
+
"learning_rate": 8.42724358974359e-06,
|
| 225547 |
+
"loss": 0.305,
|
| 225548 |
+
"step": 99180
|
| 225549 |
+
},
|
| 225550 |
+
{
|
| 225551 |
+
"epoch": 798.97,
|
| 225552 |
+
"learning_rate": 8.427163461538463e-06,
|
| 225553 |
+
"loss": 0.7074,
|
| 225554 |
+
"step": 99185
|
| 225555 |
+
},
|
| 225556 |
+
{
|
| 225557 |
+
"epoch": 799.0,
|
| 225558 |
+
"eval_loss": 0.44648560881614685,
|
| 225559 |
+
"eval_runtime": 38.6206,
|
| 225560 |
+
"eval_samples_per_second": 21.569,
|
| 225561 |
+
"eval_steps_per_second": 0.699,
|
| 225562 |
+
"eval_wer": 0.18572896784613127,
|
| 225563 |
+
"step": 99188
|
| 225564 |
+
},
|
| 225565 |
+
{
|
| 225566 |
+
"epoch": 799.02,
|
| 225567 |
+
"learning_rate": 8.427083333333334e-06,
|
| 225568 |
+
"loss": 0.5228,
|
| 225569 |
+
"step": 99190
|
| 225570 |
+
},
|
| 225571 |
+
{
|
| 225572 |
+
"epoch": 799.06,
|
| 225573 |
+
"learning_rate": 8.427003205128206e-06,
|
| 225574 |
+
"loss": 0.3038,
|
| 225575 |
+
"step": 99195
|
| 225576 |
+
},
|
| 225577 |
+
{
|
| 225578 |
+
"epoch": 799.1,
|
| 225579 |
+
"learning_rate": 8.426923076923079e-06,
|
| 225580 |
+
"loss": 0.395,
|
| 225581 |
+
"step": 99200
|
| 225582 |
+
},
|
| 225583 |
+
{
|
| 225584 |
+
"epoch": 799.14,
|
| 225585 |
+
"learning_rate": 8.42684294871795e-06,
|
| 225586 |
+
"loss": 0.3469,
|
| 225587 |
+
"step": 99205
|
| 225588 |
+
},
|
| 225589 |
+
{
|
| 225590 |
+
"epoch": 799.18,
|
| 225591 |
+
"learning_rate": 8.426762820512821e-06,
|
| 225592 |
+
"loss": 0.7684,
|
| 225593 |
+
"step": 99210
|
| 225594 |
+
},
|
| 225595 |
+
{
|
| 225596 |
+
"epoch": 799.22,
|
| 225597 |
+
"learning_rate": 8.426682692307693e-06,
|
| 225598 |
+
"loss": 1.0676,
|
| 225599 |
+
"step": 99215
|
| 225600 |
+
},
|
| 225601 |
+
{
|
| 225602 |
+
"epoch": 799.26,
|
| 225603 |
+
"learning_rate": 8.426602564102566e-06,
|
| 225604 |
+
"loss": 0.2841,
|
| 225605 |
+
"step": 99220
|
| 225606 |
+
},
|
| 225607 |
+
{
|
| 225608 |
+
"epoch": 799.3,
|
| 225609 |
+
"learning_rate": 8.426522435897436e-06,
|
| 225610 |
+
"loss": 0.2621,
|
| 225611 |
+
"step": 99225
|
| 225612 |
+
},
|
| 225613 |
+
{
|
| 225614 |
+
"epoch": 799.34,
|
| 225615 |
+
"learning_rate": 8.426442307692309e-06,
|
| 225616 |
+
"loss": 0.3179,
|
| 225617 |
+
"step": 99230
|
| 225618 |
+
},
|
| 225619 |
+
{
|
| 225620 |
+
"epoch": 799.38,
|
| 225621 |
+
"learning_rate": 8.42636217948718e-06,
|
| 225622 |
+
"loss": 0.5719,
|
| 225623 |
+
"step": 99235
|
| 225624 |
+
},
|
| 225625 |
+
{
|
| 225626 |
+
"epoch": 799.42,
|
| 225627 |
+
"learning_rate": 8.426282051282051e-06,
|
| 225628 |
+
"loss": 1.0278,
|
| 225629 |
+
"step": 99240
|
| 225630 |
+
},
|
| 225631 |
+
{
|
| 225632 |
+
"epoch": 799.46,
|
| 225633 |
+
"learning_rate": 8.426201923076924e-06,
|
| 225634 |
+
"loss": 0.2901,
|
| 225635 |
+
"step": 99245
|
| 225636 |
+
},
|
| 225637 |
+
{
|
| 225638 |
+
"epoch": 799.5,
|
| 225639 |
+
"learning_rate": 8.426121794871796e-06,
|
| 225640 |
+
"loss": 0.3263,
|
| 225641 |
+
"step": 99250
|
| 225642 |
+
},
|
| 225643 |
+
{
|
| 225644 |
+
"epoch": 799.54,
|
| 225645 |
+
"learning_rate": 8.426041666666667e-06,
|
| 225646 |
+
"loss": 0.3654,
|
| 225647 |
+
"step": 99255
|
| 225648 |
+
},
|
| 225649 |
+
{
|
| 225650 |
+
"epoch": 799.58,
|
| 225651 |
+
"learning_rate": 8.425961538461538e-06,
|
| 225652 |
+
"loss": 0.6779,
|
| 225653 |
+
"step": 99260
|
| 225654 |
+
},
|
| 225655 |
+
{
|
| 225656 |
+
"epoch": 799.62,
|
| 225657 |
+
"learning_rate": 8.425881410256411e-06,
|
| 225658 |
+
"loss": 1.0469,
|
| 225659 |
+
"step": 99265
|
| 225660 |
+
},
|
| 225661 |
+
{
|
| 225662 |
+
"epoch": 799.66,
|
| 225663 |
+
"learning_rate": 8.425801282051283e-06,
|
| 225664 |
+
"loss": 0.3481,
|
| 225665 |
+
"step": 99270
|
| 225666 |
+
},
|
| 225667 |
+
{
|
| 225668 |
+
"epoch": 799.7,
|
| 225669 |
+
"learning_rate": 8.425721153846154e-06,
|
| 225670 |
+
"loss": 0.3132,
|
| 225671 |
+
"step": 99275
|
| 225672 |
+
},
|
| 225673 |
+
{
|
| 225674 |
+
"epoch": 799.74,
|
| 225675 |
+
"learning_rate": 8.425641025641026e-06,
|
| 225676 |
+
"loss": 0.3516,
|
| 225677 |
+
"step": 99280
|
| 225678 |
+
},
|
| 225679 |
+
{
|
| 225680 |
+
"epoch": 799.78,
|
| 225681 |
+
"learning_rate": 8.425560897435899e-06,
|
| 225682 |
+
"loss": 0.6171,
|
| 225683 |
+
"step": 99285
|
| 225684 |
+
},
|
| 225685 |
+
{
|
| 225686 |
+
"epoch": 799.82,
|
| 225687 |
+
"learning_rate": 8.42548076923077e-06,
|
| 225688 |
+
"loss": 0.9037,
|
| 225689 |
+
"step": 99290
|
| 225690 |
+
},
|
| 225691 |
+
{
|
| 225692 |
+
"epoch": 799.86,
|
| 225693 |
+
"learning_rate": 8.425400641025641e-06,
|
| 225694 |
+
"loss": 0.2661,
|
| 225695 |
+
"step": 99295
|
| 225696 |
+
},
|
| 225697 |
+
{
|
| 225698 |
+
"epoch": 799.9,
|
| 225699 |
+
"learning_rate": 8.425320512820514e-06,
|
| 225700 |
+
"loss": 0.3241,
|
| 225701 |
+
"step": 99300
|
| 225702 |
+
},
|
| 225703 |
+
{
|
| 225704 |
+
"epoch": 799.94,
|
| 225705 |
+
"learning_rate": 8.425240384615386e-06,
|
| 225706 |
+
"loss": 0.3294,
|
| 225707 |
+
"step": 99305
|
| 225708 |
+
},
|
| 225709 |
+
{
|
| 225710 |
+
"epoch": 799.98,
|
| 225711 |
+
"learning_rate": 8.425160256410257e-06,
|
| 225712 |
+
"loss": 0.7607,
|
| 225713 |
+
"step": 99310
|
| 225714 |
+
},
|
| 225715 |
+
{
|
| 225716 |
+
"epoch": 800.0,
|
| 225717 |
+
"eval_loss": 0.3901221454143524,
|
| 225718 |
+
"eval_runtime": 38.0646,
|
| 225719 |
+
"eval_samples_per_second": 21.884,
|
| 225720 |
+
"eval_steps_per_second": 0.709,
|
| 225721 |
+
"eval_wer": 0.1879355400696864,
|
| 225722 |
+
"step": 99312
|
| 225723 |
+
},
|
| 225724 |
+
{
|
| 225725 |
+
"epoch": 800.02,
|
| 225726 |
+
"learning_rate": 8.425080128205128e-06,
|
| 225727 |
+
"loss": 0.3674,
|
| 225728 |
+
"step": 99315
|
| 225729 |
+
},
|
| 225730 |
+
{
|
| 225731 |
+
"epoch": 800.06,
|
| 225732 |
+
"learning_rate": 8.425000000000001e-06,
|
| 225733 |
+
"loss": 0.2658,
|
| 225734 |
+
"step": 99320
|
| 225735 |
+
},
|
| 225736 |
+
{
|
| 225737 |
+
"epoch": 800.1,
|
| 225738 |
+
"learning_rate": 8.424919871794873e-06,
|
| 225739 |
+
"loss": 0.2953,
|
| 225740 |
+
"step": 99325
|
| 225741 |
+
},
|
| 225742 |
+
{
|
| 225743 |
+
"epoch": 800.15,
|
| 225744 |
+
"learning_rate": 8.424839743589744e-06,
|
| 225745 |
+
"loss": 0.3954,
|
| 225746 |
+
"step": 99330
|
| 225747 |
+
},
|
| 225748 |
+
{
|
| 225749 |
+
"epoch": 800.19,
|
| 225750 |
+
"learning_rate": 8.424759615384616e-06,
|
| 225751 |
+
"loss": 0.7494,
|
| 225752 |
+
"step": 99335
|
| 225753 |
+
},
|
| 225754 |
+
{
|
| 225755 |
+
"epoch": 800.23,
|
| 225756 |
+
"learning_rate": 8.424679487179489e-06,
|
| 225757 |
+
"loss": 1.0121,
|
| 225758 |
+
"step": 99340
|
| 225759 |
+
},
|
| 225760 |
+
{
|
| 225761 |
+
"epoch": 800.27,
|
| 225762 |
+
"learning_rate": 8.42459935897436e-06,
|
| 225763 |
+
"loss": 0.2721,
|
| 225764 |
+
"step": 99345
|
| 225765 |
+
},
|
| 225766 |
+
{
|
| 225767 |
+
"epoch": 800.31,
|
| 225768 |
+
"learning_rate": 8.424519230769231e-06,
|
| 225769 |
+
"loss": 0.36,
|
| 225770 |
+
"step": 99350
|
| 225771 |
+
},
|
| 225772 |
+
{
|
| 225773 |
+
"epoch": 800.35,
|
| 225774 |
+
"learning_rate": 8.424439102564104e-06,
|
| 225775 |
+
"loss": 0.3509,
|
| 225776 |
+
"step": 99355
|
| 225777 |
+
},
|
| 225778 |
+
{
|
| 225779 |
+
"epoch": 800.39,
|
| 225780 |
+
"learning_rate": 8.424358974358974e-06,
|
| 225781 |
+
"loss": 0.6597,
|
| 225782 |
+
"step": 99360
|
| 225783 |
+
},
|
| 225784 |
+
{
|
| 225785 |
+
"epoch": 800.43,
|
| 225786 |
+
"learning_rate": 8.424278846153847e-06,
|
| 225787 |
+
"loss": 0.9188,
|
| 225788 |
+
"step": 99365
|
| 225789 |
+
},
|
| 225790 |
+
{
|
| 225791 |
+
"epoch": 800.47,
|
| 225792 |
+
"learning_rate": 8.424198717948718e-06,
|
| 225793 |
+
"loss": 0.2928,
|
| 225794 |
+
"step": 99370
|
| 225795 |
+
},
|
| 225796 |
+
{
|
| 225797 |
+
"epoch": 800.51,
|
| 225798 |
+
"learning_rate": 8.42411858974359e-06,
|
| 225799 |
+
"loss": 0.3367,
|
| 225800 |
+
"step": 99375
|
| 225801 |
+
},
|
| 225802 |
+
{
|
| 225803 |
+
"epoch": 800.55,
|
| 225804 |
+
"learning_rate": 8.424038461538461e-06,
|
| 225805 |
+
"loss": 0.4375,
|
| 225806 |
+
"step": 99380
|
| 225807 |
+
},
|
| 225808 |
+
{
|
| 225809 |
+
"epoch": 800.59,
|
| 225810 |
+
"learning_rate": 8.423958333333334e-06,
|
| 225811 |
+
"loss": 0.7852,
|
| 225812 |
+
"step": 99385
|
| 225813 |
+
},
|
| 225814 |
+
{
|
| 225815 |
+
"epoch": 800.63,
|
| 225816 |
+
"learning_rate": 8.423878205128206e-06,
|
| 225817 |
+
"loss": 0.9107,
|
| 225818 |
+
"step": 99390
|
| 225819 |
+
},
|
| 225820 |
+
{
|
| 225821 |
+
"epoch": 800.67,
|
| 225822 |
+
"learning_rate": 8.423798076923077e-06,
|
| 225823 |
+
"loss": 0.2874,
|
| 225824 |
+
"step": 99395
|
| 225825 |
+
},
|
| 225826 |
+
{
|
| 225827 |
+
"epoch": 800.71,
|
| 225828 |
+
"learning_rate": 8.42371794871795e-06,
|
| 225829 |
+
"loss": 0.3251,
|
| 225830 |
+
"step": 99400
|
| 225831 |
+
},
|
| 225832 |
+
{
|
| 225833 |
+
"epoch": 800.75,
|
| 225834 |
+
"learning_rate": 8.423637820512821e-06,
|
| 225835 |
+
"loss": 0.5027,
|
| 225836 |
+
"step": 99405
|
| 225837 |
+
},
|
| 225838 |
+
{
|
| 225839 |
+
"epoch": 800.79,
|
| 225840 |
+
"learning_rate": 8.423557692307693e-06,
|
| 225841 |
+
"loss": 0.7723,
|
| 225842 |
+
"step": 99410
|
| 225843 |
+
},
|
| 225844 |
+
{
|
| 225845 |
+
"epoch": 800.83,
|
| 225846 |
+
"learning_rate": 8.423477564102564e-06,
|
| 225847 |
+
"loss": 0.9051,
|
| 225848 |
+
"step": 99415
|
| 225849 |
+
},
|
| 225850 |
+
{
|
| 225851 |
+
"epoch": 800.87,
|
| 225852 |
+
"learning_rate": 8.423397435897437e-06,
|
| 225853 |
+
"loss": 0.3167,
|
| 225854 |
+
"step": 99420
|
| 225855 |
+
},
|
| 225856 |
+
{
|
| 225857 |
+
"epoch": 800.91,
|
| 225858 |
+
"learning_rate": 8.423317307692308e-06,
|
| 225859 |
+
"loss": 0.3365,
|
| 225860 |
+
"step": 99425
|
| 225861 |
+
},
|
| 225862 |
+
{
|
| 225863 |
+
"epoch": 800.95,
|
| 225864 |
+
"learning_rate": 8.42323717948718e-06,
|
| 225865 |
+
"loss": 0.4145,
|
| 225866 |
+
"step": 99430
|
| 225867 |
+
},
|
| 225868 |
+
{
|
| 225869 |
+
"epoch": 800.99,
|
| 225870 |
+
"learning_rate": 8.423157051282051e-06,
|
| 225871 |
+
"loss": 1.0965,
|
| 225872 |
+
"step": 99435
|
| 225873 |
+
},
|
| 225874 |
+
{
|
| 225875 |
+
"epoch": 801.0,
|
| 225876 |
+
"eval_loss": 0.33939823508262634,
|
| 225877 |
+
"eval_runtime": 38.5691,
|
| 225878 |
+
"eval_samples_per_second": 21.598,
|
| 225879 |
+
"eval_steps_per_second": 0.7,
|
| 225880 |
+
"eval_wer": 0.19310196472106697,
|
| 225881 |
+
"step": 99436
|
| 225882 |
+
},
|
| 225883 |
+
{
|
| 225884 |
+
"epoch": 801.03,
|
| 225885 |
+
"learning_rate": 8.423076923076924e-06,
|
| 225886 |
+
"loss": 0.3326,
|
| 225887 |
+
"step": 99440
|
| 225888 |
+
},
|
| 225889 |
+
{
|
| 225890 |
+
"epoch": 801.07,
|
| 225891 |
+
"learning_rate": 8.422996794871796e-06,
|
| 225892 |
+
"loss": 0.3037,
|
| 225893 |
+
"step": 99445
|
| 225894 |
+
},
|
| 225895 |
+
{
|
| 225896 |
+
"epoch": 801.11,
|
| 225897 |
+
"learning_rate": 8.422916666666667e-06,
|
| 225898 |
+
"loss": 0.3343,
|
| 225899 |
+
"step": 99450
|
| 225900 |
+
},
|
| 225901 |
+
{
|
| 225902 |
+
"epoch": 801.15,
|
| 225903 |
+
"learning_rate": 8.42283653846154e-06,
|
| 225904 |
+
"loss": 0.4206,
|
| 225905 |
+
"step": 99455
|
| 225906 |
+
},
|
| 225907 |
+
{
|
| 225908 |
+
"epoch": 801.19,
|
| 225909 |
+
"learning_rate": 8.422756410256411e-06,
|
| 225910 |
+
"loss": 0.9343,
|
| 225911 |
+
"step": 99460
|
| 225912 |
+
},
|
| 225913 |
+
{
|
| 225914 |
+
"epoch": 801.23,
|
| 225915 |
+
"learning_rate": 8.422676282051283e-06,
|
| 225916 |
+
"loss": 0.6053,
|
| 225917 |
+
"step": 99465
|
| 225918 |
+
},
|
| 225919 |
+
{
|
| 225920 |
+
"epoch": 801.27,
|
| 225921 |
+
"learning_rate": 8.422596153846154e-06,
|
| 225922 |
+
"loss": 0.3313,
|
| 225923 |
+
"step": 99470
|
| 225924 |
+
},
|
| 225925 |
+
{
|
| 225926 |
+
"epoch": 801.31,
|
| 225927 |
+
"learning_rate": 8.422516025641027e-06,
|
| 225928 |
+
"loss": 0.316,
|
| 225929 |
+
"step": 99475
|
| 225930 |
+
},
|
| 225931 |
+
{
|
| 225932 |
+
"epoch": 801.35,
|
| 225933 |
+
"learning_rate": 8.422435897435899e-06,
|
| 225934 |
+
"loss": 0.4048,
|
| 225935 |
+
"step": 99480
|
| 225936 |
+
},
|
| 225937 |
+
{
|
| 225938 |
+
"epoch": 801.39,
|
| 225939 |
+
"learning_rate": 8.42235576923077e-06,
|
| 225940 |
+
"loss": 0.905,
|
| 225941 |
+
"step": 99485
|
| 225942 |
+
},
|
| 225943 |
+
{
|
| 225944 |
+
"epoch": 801.43,
|
| 225945 |
+
"learning_rate": 8.422275641025641e-06,
|
| 225946 |
+
"loss": 0.6495,
|
| 225947 |
+
"step": 99490
|
| 225948 |
+
},
|
| 225949 |
+
{
|
| 225950 |
+
"epoch": 801.47,
|
| 225951 |
+
"learning_rate": 8.422195512820514e-06,
|
| 225952 |
+
"loss": 0.3253,
|
| 225953 |
+
"step": 99495
|
| 225954 |
+
},
|
| 225955 |
+
{
|
| 225956 |
+
"epoch": 801.51,
|
| 225957 |
+
"learning_rate": 8.422115384615386e-06,
|
| 225958 |
+
"loss": 0.2806,
|
| 225959 |
+
"step": 99500
|
| 225960 |
+
},
|
| 225961 |
+
{
|
| 225962 |
+
"epoch": 801.55,
|
| 225963 |
+
"learning_rate": 8.422035256410257e-06,
|
| 225964 |
+
"loss": 0.4059,
|
| 225965 |
+
"step": 99505
|
| 225966 |
+
},
|
| 225967 |
+
{
|
| 225968 |
+
"epoch": 801.59,
|
| 225969 |
+
"learning_rate": 8.42195512820513e-06,
|
| 225970 |
+
"loss": 0.8931,
|
| 225971 |
+
"step": 99510
|
| 225972 |
+
},
|
| 225973 |
+
{
|
| 225974 |
+
"epoch": 801.63,
|
| 225975 |
+
"learning_rate": 8.421875e-06,
|
| 225976 |
+
"loss": 0.6791,
|
| 225977 |
+
"step": 99515
|
| 225978 |
+
},
|
| 225979 |
+
{
|
| 225980 |
+
"epoch": 801.67,
|
| 225981 |
+
"learning_rate": 8.421794871794873e-06,
|
| 225982 |
+
"loss": 0.3056,
|
| 225983 |
+
"step": 99520
|
| 225984 |
+
},
|
| 225985 |
+
{
|
| 225986 |
+
"epoch": 801.71,
|
| 225987 |
+
"learning_rate": 8.421714743589744e-06,
|
| 225988 |
+
"loss": 0.2884,
|
| 225989 |
+
"step": 99525
|
| 225990 |
+
},
|
| 225991 |
+
{
|
| 225992 |
+
"epoch": 801.76,
|
| 225993 |
+
"learning_rate": 8.421634615384615e-06,
|
| 225994 |
+
"loss": 0.378,
|
| 225995 |
+
"step": 99530
|
| 225996 |
+
},
|
| 225997 |
+
{
|
| 225998 |
+
"epoch": 801.8,
|
| 225999 |
+
"learning_rate": 8.421554487179487e-06,
|
| 226000 |
+
"loss": 0.8309,
|
| 226001 |
+
"step": 99535
|
| 226002 |
+
},
|
| 226003 |
+
{
|
| 226004 |
+
"epoch": 801.84,
|
| 226005 |
+
"learning_rate": 8.42147435897436e-06,
|
| 226006 |
+
"loss": 0.8093,
|
| 226007 |
+
"step": 99540
|
| 226008 |
+
},
|
| 226009 |
+
{
|
| 226010 |
+
"epoch": 801.88,
|
| 226011 |
+
"learning_rate": 8.421394230769231e-06,
|
| 226012 |
+
"loss": 0.3122,
|
| 226013 |
+
"step": 99545
|
| 226014 |
+
},
|
| 226015 |
+
{
|
| 226016 |
+
"epoch": 801.92,
|
| 226017 |
+
"learning_rate": 8.421314102564103e-06,
|
| 226018 |
+
"loss": 0.4138,
|
| 226019 |
+
"step": 99550
|
| 226020 |
+
},
|
| 226021 |
+
{
|
| 226022 |
+
"epoch": 801.96,
|
| 226023 |
+
"learning_rate": 8.421233974358976e-06,
|
| 226024 |
+
"loss": 0.4062,
|
| 226025 |
+
"step": 99555
|
| 226026 |
+
},
|
| 226027 |
+
{
|
| 226028 |
+
"epoch": 802.0,
|
| 226029 |
+
"learning_rate": 8.421153846153847e-06,
|
| 226030 |
+
"loss": 0.9545,
|
| 226031 |
+
"step": 99560
|
| 226032 |
+
},
|
| 226033 |
+
{
|
| 226034 |
+
"epoch": 802.0,
|
| 226035 |
+
"eval_loss": 0.3275456130504608,
|
| 226036 |
+
"eval_runtime": 39.2426,
|
| 226037 |
+
"eval_samples_per_second": 21.252,
|
| 226038 |
+
"eval_steps_per_second": 0.688,
|
| 226039 |
+
"eval_wer": 0.17951169289601412,
|
| 226040 |
+
"step": 99560
|
| 226041 |
}
|
| 226042 |
],
|
| 226043 |
"max_steps": 620000,
|
| 226044 |
"num_train_epochs": 5000,
|
| 226045 |
+
"total_flos": 2.801814193016803e+20,
|
| 226046 |
"trial_name": null,
|
| 226047 |
"trial_params": null
|
| 226048 |
}
|
model-bin/finetune/base/{checkpoint-98940 → checkpoint-99560}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629977881.4113429/events.out.tfevents.1629977881.8e89bd551565.924.171
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c3fbc54e2f704a80511df74bc3deb71bbf3c191f5404537c9d49e7a174f77a93
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629978315.2297533/events.out.tfevents.1629978315.8e89bd551565.924.173
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05dbf93a0ed8975d808dde83ebfca2ad72a95e90c57e0946dc3fc05ad878edf6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629978748.357351/events.out.tfevents.1629978748.8e89bd551565.924.175
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3f267470078333c46e88d73a682a9edf34daa50c6c0a2af5f2ba20357ae2b93
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629979170.7944605/events.out.tfevents.1629979170.8e89bd551565.924.177
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a4049ed73f6a66cdd1fd18f17f868c326614105090591df8f33b58d4f7b4fc0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629979593.1009088/events.out.tfevents.1629979593.8e89bd551565.924.179
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3d0358afe7a84a09d8b16dffc8f4b9e8c8b5c2b621ff5792eed5c8c4512f1d3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629977880.8e89bd551565.924.170
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:06bcba4133a4ac9682b222499de39a676d61443adc3bbe13670bb8b3a6a0ac20
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629978315.8e89bd551565.924.172
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:863d92bd975ad330fc0258c948566e3bdfc786a1fbfb7bfefd445f81015c574e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629978748.8e89bd551565.924.174
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9b3cd5dc21c0be7f19c6b139a065aa7cf7accc830dab24ec9e31e646727eee99
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629979170.8e89bd551565.924.176
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d3040723f25d22de07a7ad4942aac8df338fb62ce9f54311843cc16fb12ef95
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629979593.8e89bd551565.924.178
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d174e05e550ebb48f620fd40948279291d41fa4ba6c24ea80a32e011b4b1b228
|
| 3 |
+
size 8622
|