"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/trainer_state.json +427 -4
- model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629645072.528472/events.out.tfevents.1629645072.91f414e9573a.903.161 +3 -0
- model-bin/finetune/base/log/1629645744.269193/events.out.tfevents.1629645744.91f414e9573a.903.163 +3 -0
- model-bin/finetune/base/log/1629646405.8412092/events.out.tfevents.1629646405.91f414e9573a.903.165 +3 -0
- model-bin/finetune/base/log/1629647059.354017/events.out.tfevents.1629647059.91f414e9573a.903.167 +3 -0
- model-bin/finetune/base/log/1629647710.367303/events.out.tfevents.1629647710.91f414e9573a.903.169 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629645072.91f414e9573a.903.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629645744.91f414e9573a.903.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629646405.91f414e9573a.903.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629647059.91f414e9573a.903.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629647710.91f414e9573a.903.168 +3 -0
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/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:c2978ec2464aaaebbad9da87d83c047a93e54d4876cbc35b9b00ecea5c80a821
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/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 sha256:35e76857d6f644e475610db33b38d3dc5eabd10226e92bd4717b857894b4837b
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/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 sha256:8e8b29c0187c23518a391935e13f5a53845a3875a64139d8bb7e4f06bc6925c2
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/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:d68831e3f282d8d7e8ff16af2b787b1301c3b0d8d0c1bf86c9a2f591631e2fe7
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/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 sha256:346ac72d455ab09ee7028d006c22ce13b20852c71928982947029223751b91ec
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -144198,11 +144198,434 @@
|
|
| 144198 |
"eval_steps_per_second": 0.64,
|
| 144199 |
"eval_wer": 0.217536623524392,
|
| 144200 |
"step": 32599
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 144201 |
}
|
| 144202 |
],
|
| 144203 |
-
"max_steps":
|
| 144204 |
"num_train_epochs": 5000,
|
| 144205 |
-
"total_flos": 9.
|
| 144206 |
"trial_name": null,
|
| 144207 |
"trial_params": null
|
| 144208 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 266.99598393574297,
|
| 5 |
+
"global_step": 33220,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 144198 |
"eval_steps_per_second": 0.64,
|
| 144199 |
"eval_wer": 0.217536623524392,
|
| 144200 |
"step": 32599
|
| 144201 |
+
},
|
| 144202 |
+
{
|
| 144203 |
+
"epoch": 260.01,
|
| 144204 |
+
"learning_rate": 9.489709208400647e-06,
|
| 144205 |
+
"loss": 0.513,
|
| 144206 |
+
"step": 32600
|
| 144207 |
+
},
|
| 144208 |
+
{
|
| 144209 |
+
"epoch": 260.09,
|
| 144210 |
+
"learning_rate": 9.489547657512117e-06,
|
| 144211 |
+
"loss": 0.4065,
|
| 144212 |
+
"step": 32610
|
| 144213 |
+
},
|
| 144214 |
+
{
|
| 144215 |
+
"epoch": 260.17,
|
| 144216 |
+
"learning_rate": 9.489386106623587e-06,
|
| 144217 |
+
"loss": 0.726,
|
| 144218 |
+
"step": 32620
|
| 144219 |
+
},
|
| 144220 |
+
{
|
| 144221 |
+
"epoch": 260.25,
|
| 144222 |
+
"learning_rate": 9.489224555735057e-06,
|
| 144223 |
+
"loss": 0.8806,
|
| 144224 |
+
"step": 32630
|
| 144225 |
+
},
|
| 144226 |
+
{
|
| 144227 |
+
"epoch": 260.33,
|
| 144228 |
+
"learning_rate": 9.489063004846527e-06,
|
| 144229 |
+
"loss": 0.4425,
|
| 144230 |
+
"step": 32640
|
| 144231 |
+
},
|
| 144232 |
+
{
|
| 144233 |
+
"epoch": 260.41,
|
| 144234 |
+
"learning_rate": 9.488901453957997e-06,
|
| 144235 |
+
"loss": 0.9212,
|
| 144236 |
+
"step": 32650
|
| 144237 |
+
},
|
| 144238 |
+
{
|
| 144239 |
+
"epoch": 260.49,
|
| 144240 |
+
"learning_rate": 9.488739903069467e-06,
|
| 144241 |
+
"loss": 0.4024,
|
| 144242 |
+
"step": 32660
|
| 144243 |
+
},
|
| 144244 |
+
{
|
| 144245 |
+
"epoch": 260.57,
|
| 144246 |
+
"learning_rate": 9.488578352180938e-06,
|
| 144247 |
+
"loss": 0.5007,
|
| 144248 |
+
"step": 32670
|
| 144249 |
+
},
|
| 144250 |
+
{
|
| 144251 |
+
"epoch": 260.65,
|
| 144252 |
+
"learning_rate": 9.488416801292408e-06,
|
| 144253 |
+
"loss": 0.961,
|
| 144254 |
+
"step": 32680
|
| 144255 |
+
},
|
| 144256 |
+
{
|
| 144257 |
+
"epoch": 260.73,
|
| 144258 |
+
"learning_rate": 9.488255250403878e-06,
|
| 144259 |
+
"loss": 0.4242,
|
| 144260 |
+
"step": 32690
|
| 144261 |
+
},
|
| 144262 |
+
{
|
| 144263 |
+
"epoch": 260.81,
|
| 144264 |
+
"learning_rate": 9.488093699515348e-06,
|
| 144265 |
+
"loss": 1.0445,
|
| 144266 |
+
"step": 32700
|
| 144267 |
+
},
|
| 144268 |
+
{
|
| 144269 |
+
"epoch": 260.89,
|
| 144270 |
+
"learning_rate": 9.487932148626818e-06,
|
| 144271 |
+
"loss": 0.3941,
|
| 144272 |
+
"step": 32710
|
| 144273 |
+
},
|
| 144274 |
+
{
|
| 144275 |
+
"epoch": 260.97,
|
| 144276 |
+
"learning_rate": 9.487770597738288e-06,
|
| 144277 |
+
"loss": 0.5526,
|
| 144278 |
+
"step": 32720
|
| 144279 |
+
},
|
| 144280 |
+
{
|
| 144281 |
+
"epoch": 261.0,
|
| 144282 |
+
"eval_loss": 0.5151455998420715,
|
| 144283 |
+
"eval_runtime": 42.6295,
|
| 144284 |
+
"eval_samples_per_second": 19.705,
|
| 144285 |
+
"eval_steps_per_second": 0.633,
|
| 144286 |
+
"eval_wer": 0.22311237149946828,
|
| 144287 |
+
"step": 32724
|
| 144288 |
+
},
|
| 144289 |
+
{
|
| 144290 |
+
"epoch": 263.05,
|
| 144291 |
+
"learning_rate": 9.487609046849758e-06,
|
| 144292 |
+
"loss": 0.4533,
|
| 144293 |
+
"step": 32730
|
| 144294 |
+
},
|
| 144295 |
+
{
|
| 144296 |
+
"epoch": 263.13,
|
| 144297 |
+
"learning_rate": 9.487447495961228e-06,
|
| 144298 |
+
"loss": 0.5431,
|
| 144299 |
+
"step": 32740
|
| 144300 |
+
},
|
| 144301 |
+
{
|
| 144302 |
+
"epoch": 263.21,
|
| 144303 |
+
"learning_rate": 9.487285945072698e-06,
|
| 144304 |
+
"loss": 1.0422,
|
| 144305 |
+
"step": 32750
|
| 144306 |
+
},
|
| 144307 |
+
{
|
| 144308 |
+
"epoch": 263.29,
|
| 144309 |
+
"learning_rate": 9.487124394184168e-06,
|
| 144310 |
+
"loss": 0.3902,
|
| 144311 |
+
"step": 32760
|
| 144312 |
+
},
|
| 144313 |
+
{
|
| 144314 |
+
"epoch": 263.37,
|
| 144315 |
+
"learning_rate": 9.486962843295638e-06,
|
| 144316 |
+
"loss": 0.9624,
|
| 144317 |
+
"step": 32770
|
| 144318 |
+
},
|
| 144319 |
+
{
|
| 144320 |
+
"epoch": 263.45,
|
| 144321 |
+
"learning_rate": 9.48680129240711e-06,
|
| 144322 |
+
"loss": 0.8451,
|
| 144323 |
+
"step": 32780
|
| 144324 |
+
},
|
| 144325 |
+
{
|
| 144326 |
+
"epoch": 263.53,
|
| 144327 |
+
"learning_rate": 9.48663974151858e-06,
|
| 144328 |
+
"loss": 0.4238,
|
| 144329 |
+
"step": 32790
|
| 144330 |
+
},
|
| 144331 |
+
{
|
| 144332 |
+
"epoch": 263.61,
|
| 144333 |
+
"learning_rate": 9.48647819063005e-06,
|
| 144334 |
+
"loss": 1.0257,
|
| 144335 |
+
"step": 32800
|
| 144336 |
+
},
|
| 144337 |
+
{
|
| 144338 |
+
"epoch": 263.69,
|
| 144339 |
+
"learning_rate": 9.48631663974152e-06,
|
| 144340 |
+
"loss": 0.3474,
|
| 144341 |
+
"step": 32810
|
| 144342 |
+
},
|
| 144343 |
+
{
|
| 144344 |
+
"epoch": 263.77,
|
| 144345 |
+
"learning_rate": 9.48615508885299e-06,
|
| 144346 |
+
"loss": 0.5286,
|
| 144347 |
+
"step": 32820
|
| 144348 |
+
},
|
| 144349 |
+
{
|
| 144350 |
+
"epoch": 263.85,
|
| 144351 |
+
"learning_rate": 9.48599353796446e-06,
|
| 144352 |
+
"loss": 0.8974,
|
| 144353 |
+
"step": 32830
|
| 144354 |
+
},
|
| 144355 |
+
{
|
| 144356 |
+
"epoch": 263.93,
|
| 144357 |
+
"learning_rate": 9.48583198707593e-06,
|
| 144358 |
+
"loss": 0.4847,
|
| 144359 |
+
"step": 32840
|
| 144360 |
+
},
|
| 144361 |
+
{
|
| 144362 |
+
"epoch": 264.0,
|
| 144363 |
+
"eval_loss": 0.46656250953674316,
|
| 144364 |
+
"eval_runtime": 41.1189,
|
| 144365 |
+
"eval_samples_per_second": 20.404,
|
| 144366 |
+
"eval_steps_per_second": 0.657,
|
| 144367 |
+
"eval_wer": 0.21536111717215797,
|
| 144368 |
+
"step": 32848
|
| 144369 |
+
},
|
| 144370 |
+
{
|
| 144371 |
+
"epoch": 264.02,
|
| 144372 |
+
"learning_rate": 9.4856704361874e-06,
|
| 144373 |
+
"loss": 0.4346,
|
| 144374 |
+
"step": 32850
|
| 144375 |
+
},
|
| 144376 |
+
{
|
| 144377 |
+
"epoch": 264.1,
|
| 144378 |
+
"learning_rate": 9.48550888529887e-06,
|
| 144379 |
+
"loss": 0.3585,
|
| 144380 |
+
"step": 32860
|
| 144381 |
+
},
|
| 144382 |
+
{
|
| 144383 |
+
"epoch": 264.18,
|
| 144384 |
+
"learning_rate": 9.48534733441034e-06,
|
| 144385 |
+
"loss": 0.5452,
|
| 144386 |
+
"step": 32870
|
| 144387 |
+
},
|
| 144388 |
+
{
|
| 144389 |
+
"epoch": 264.26,
|
| 144390 |
+
"learning_rate": 9.48518578352181e-06,
|
| 144391 |
+
"loss": 0.7754,
|
| 144392 |
+
"step": 32880
|
| 144393 |
+
},
|
| 144394 |
+
{
|
| 144395 |
+
"epoch": 264.34,
|
| 144396 |
+
"learning_rate": 9.48502423263328e-06,
|
| 144397 |
+
"loss": 0.4467,
|
| 144398 |
+
"step": 32890
|
| 144399 |
+
},
|
| 144400 |
+
{
|
| 144401 |
+
"epoch": 264.42,
|
| 144402 |
+
"learning_rate": 9.484862681744751e-06,
|
| 144403 |
+
"loss": 0.909,
|
| 144404 |
+
"step": 32900
|
| 144405 |
+
},
|
| 144406 |
+
{
|
| 144407 |
+
"epoch": 264.5,
|
| 144408 |
+
"learning_rate": 9.484701130856221e-06,
|
| 144409 |
+
"loss": 0.4008,
|
| 144410 |
+
"step": 32910
|
| 144411 |
+
},
|
| 144412 |
+
{
|
| 144413 |
+
"epoch": 264.58,
|
| 144414 |
+
"learning_rate": 9.484539579967691e-06,
|
| 144415 |
+
"loss": 0.5642,
|
| 144416 |
+
"step": 32920
|
| 144417 |
+
},
|
| 144418 |
+
{
|
| 144419 |
+
"epoch": 264.66,
|
| 144420 |
+
"learning_rate": 9.484378029079161e-06,
|
| 144421 |
+
"loss": 0.7662,
|
| 144422 |
+
"step": 32930
|
| 144423 |
+
},
|
| 144424 |
+
{
|
| 144425 |
+
"epoch": 264.74,
|
| 144426 |
+
"learning_rate": 9.484216478190631e-06,
|
| 144427 |
+
"loss": 0.4462,
|
| 144428 |
+
"step": 32940
|
| 144429 |
+
},
|
| 144430 |
+
{
|
| 144431 |
+
"epoch": 264.82,
|
| 144432 |
+
"learning_rate": 9.484054927302101e-06,
|
| 144433 |
+
"loss": 0.8801,
|
| 144434 |
+
"step": 32950
|
| 144435 |
+
},
|
| 144436 |
+
{
|
| 144437 |
+
"epoch": 264.9,
|
| 144438 |
+
"learning_rate": 9.483893376413571e-06,
|
| 144439 |
+
"loss": 0.4076,
|
| 144440 |
+
"step": 32960
|
| 144441 |
+
},
|
| 144442 |
+
{
|
| 144443 |
+
"epoch": 264.98,
|
| 144444 |
+
"learning_rate": 9.483731825525041e-06,
|
| 144445 |
+
"loss": 0.6219,
|
| 144446 |
+
"step": 32970
|
| 144447 |
+
},
|
| 144448 |
+
{
|
| 144449 |
+
"epoch": 265.0,
|
| 144450 |
+
"eval_loss": 0.45929187536239624,
|
| 144451 |
+
"eval_runtime": 41.8207,
|
| 144452 |
+
"eval_samples_per_second": 20.062,
|
| 144453 |
+
"eval_steps_per_second": 0.646,
|
| 144454 |
+
"eval_wer": 0.2201124662236179,
|
| 144455 |
+
"step": 32972
|
| 144456 |
+
},
|
| 144457 |
+
{
|
| 144458 |
+
"epoch": 265.06,
|
| 144459 |
+
"learning_rate": 9.483570274636511e-06,
|
| 144460 |
+
"loss": 0.4056,
|
| 144461 |
+
"step": 32980
|
| 144462 |
+
},
|
| 144463 |
+
{
|
| 144464 |
+
"epoch": 265.14,
|
| 144465 |
+
"learning_rate": 9.483408723747981e-06,
|
| 144466 |
+
"loss": 0.4099,
|
| 144467 |
+
"step": 32990
|
| 144468 |
+
},
|
| 144469 |
+
{
|
| 144470 |
+
"epoch": 265.22,
|
| 144471 |
+
"learning_rate": 9.483247172859451e-06,
|
| 144472 |
+
"loss": 0.9526,
|
| 144473 |
+
"step": 33000
|
| 144474 |
+
},
|
| 144475 |
+
{
|
| 144476 |
+
"epoch": 265.31,
|
| 144477 |
+
"learning_rate": 9.483085621970921e-06,
|
| 144478 |
+
"loss": 0.3707,
|
| 144479 |
+
"step": 33010
|
| 144480 |
+
},
|
| 144481 |
+
{
|
| 144482 |
+
"epoch": 265.39,
|
| 144483 |
+
"learning_rate": 9.482924071082393e-06,
|
| 144484 |
+
"loss": 0.6603,
|
| 144485 |
+
"step": 33020
|
| 144486 |
+
},
|
| 144487 |
+
{
|
| 144488 |
+
"epoch": 265.47,
|
| 144489 |
+
"learning_rate": 9.482762520193863e-06,
|
| 144490 |
+
"loss": 0.7363,
|
| 144491 |
+
"step": 33030
|
| 144492 |
+
},
|
| 144493 |
+
{
|
| 144494 |
+
"epoch": 265.55,
|
| 144495 |
+
"learning_rate": 9.482600969305333e-06,
|
| 144496 |
+
"loss": 0.4319,
|
| 144497 |
+
"step": 33040
|
| 144498 |
+
},
|
| 144499 |
+
{
|
| 144500 |
+
"epoch": 265.63,
|
| 144501 |
+
"learning_rate": 9.482439418416803e-06,
|
| 144502 |
+
"loss": 0.986,
|
| 144503 |
+
"step": 33050
|
| 144504 |
+
},
|
| 144505 |
+
{
|
| 144506 |
+
"epoch": 265.71,
|
| 144507 |
+
"learning_rate": 9.482277867528273e-06,
|
| 144508 |
+
"loss": 0.3868,
|
| 144509 |
+
"step": 33060
|
| 144510 |
+
},
|
| 144511 |
+
{
|
| 144512 |
+
"epoch": 265.79,
|
| 144513 |
+
"learning_rate": 9.482116316639743e-06,
|
| 144514 |
+
"loss": 0.6805,
|
| 144515 |
+
"step": 33070
|
| 144516 |
+
},
|
| 144517 |
+
{
|
| 144518 |
+
"epoch": 265.87,
|
| 144519 |
+
"learning_rate": 9.481954765751212e-06,
|
| 144520 |
+
"loss": 0.6665,
|
| 144521 |
+
"step": 33080
|
| 144522 |
+
},
|
| 144523 |
+
{
|
| 144524 |
+
"epoch": 265.95,
|
| 144525 |
+
"learning_rate": 9.481793214862682e-06,
|
| 144526 |
+
"loss": 0.4761,
|
| 144527 |
+
"step": 33090
|
| 144528 |
+
},
|
| 144529 |
+
{
|
| 144530 |
+
"epoch": 266.0,
|
| 144531 |
+
"eval_loss": 0.47401806712150574,
|
| 144532 |
+
"eval_runtime": 40.3343,
|
| 144533 |
+
"eval_samples_per_second": 20.826,
|
| 144534 |
+
"eval_steps_per_second": 0.669,
|
| 144535 |
+
"eval_wer": 0.21947170900692842,
|
| 144536 |
+
"step": 33096
|
| 144537 |
+
},
|
| 144538 |
+
{
|
| 144539 |
+
"epoch": 266.03,
|
| 144540 |
+
"learning_rate": 9.481631663974152e-06,
|
| 144541 |
+
"loss": 0.4343,
|
| 144542 |
+
"step": 33100
|
| 144543 |
+
},
|
| 144544 |
+
{
|
| 144545 |
+
"epoch": 266.11,
|
| 144546 |
+
"learning_rate": 9.481470113085622e-06,
|
| 144547 |
+
"loss": 0.3692,
|
| 144548 |
+
"step": 33110
|
| 144549 |
+
},
|
| 144550 |
+
{
|
| 144551 |
+
"epoch": 266.19,
|
| 144552 |
+
"learning_rate": 9.481308562197092e-06,
|
| 144553 |
+
"loss": 0.8184,
|
| 144554 |
+
"step": 33120
|
| 144555 |
+
},
|
| 144556 |
+
{
|
| 144557 |
+
"epoch": 266.27,
|
| 144558 |
+
"learning_rate": 9.481147011308564e-06,
|
| 144559 |
+
"loss": 0.5983,
|
| 144560 |
+
"step": 33130
|
| 144561 |
+
},
|
| 144562 |
+
{
|
| 144563 |
+
"epoch": 266.35,
|
| 144564 |
+
"learning_rate": 9.480985460420034e-06,
|
| 144565 |
+
"loss": 0.4754,
|
| 144566 |
+
"step": 33140
|
| 144567 |
+
},
|
| 144568 |
+
{
|
| 144569 |
+
"epoch": 266.43,
|
| 144570 |
+
"learning_rate": 9.480823909531504e-06,
|
| 144571 |
+
"loss": 0.9689,
|
| 144572 |
+
"step": 33150
|
| 144573 |
+
},
|
| 144574 |
+
{
|
| 144575 |
+
"epoch": 266.51,
|
| 144576 |
+
"learning_rate": 9.480662358642974e-06,
|
| 144577 |
+
"loss": 0.3819,
|
| 144578 |
+
"step": 33160
|
| 144579 |
+
},
|
| 144580 |
+
{
|
| 144581 |
+
"epoch": 266.59,
|
| 144582 |
+
"learning_rate": 9.480500807754444e-06,
|
| 144583 |
+
"loss": 0.9042,
|
| 144584 |
+
"step": 33170
|
| 144585 |
+
},
|
| 144586 |
+
{
|
| 144587 |
+
"epoch": 266.67,
|
| 144588 |
+
"learning_rate": 9.480339256865914e-06,
|
| 144589 |
+
"loss": 0.5742,
|
| 144590 |
+
"step": 33180
|
| 144591 |
+
},
|
| 144592 |
+
{
|
| 144593 |
+
"epoch": 266.76,
|
| 144594 |
+
"learning_rate": 9.480177705977384e-06,
|
| 144595 |
+
"loss": 0.4939,
|
| 144596 |
+
"step": 33190
|
| 144597 |
+
},
|
| 144598 |
+
{
|
| 144599 |
+
"epoch": 266.84,
|
| 144600 |
+
"learning_rate": 9.480016155088854e-06,
|
| 144601 |
+
"loss": 1.0403,
|
| 144602 |
+
"step": 33200
|
| 144603 |
+
},
|
| 144604 |
+
{
|
| 144605 |
+
"epoch": 266.92,
|
| 144606 |
+
"learning_rate": 9.479854604200324e-06,
|
| 144607 |
+
"loss": 0.3765,
|
| 144608 |
+
"step": 33210
|
| 144609 |
+
},
|
| 144610 |
+
{
|
| 144611 |
+
"epoch": 267.0,
|
| 144612 |
+
"learning_rate": 9.479693053311794e-06,
|
| 144613 |
+
"loss": 0.859,
|
| 144614 |
+
"step": 33220
|
| 144615 |
+
},
|
| 144616 |
+
{
|
| 144617 |
+
"epoch": 267.0,
|
| 144618 |
+
"eval_loss": 0.4426024556159973,
|
| 144619 |
+
"eval_runtime": 41.113,
|
| 144620 |
+
"eval_samples_per_second": 20.431,
|
| 144621 |
+
"eval_steps_per_second": 0.657,
|
| 144622 |
+
"eval_wer": 0.2105802297504939,
|
| 144623 |
+
"step": 33220
|
| 144624 |
}
|
| 144625 |
],
|
| 144626 |
+
"max_steps": 620000,
|
| 144627 |
"num_train_epochs": 5000,
|
| 144628 |
+
"total_flos": 9.348704046455015e+19,
|
| 144629 |
"trial_name": null,
|
| 144630 |
"trial_params": null
|
| 144631 |
}
|
model-bin/finetune/base/{checkpoint-32599 → checkpoint-33220}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629645072.528472/events.out.tfevents.1629645072.91f414e9573a.903.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:79a8f3c886a657bf3d4cd844d9b21485f572b5b18baf874576b0313a6b865fd9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629645744.269193/events.out.tfevents.1629645744.91f414e9573a.903.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:003ba0751b885dff2462269ca91f46172f142624eddba69ca96c5cdd74e77323
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629646405.8412092/events.out.tfevents.1629646405.91f414e9573a.903.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5872f5cd4f82f868c4f37138c1d4a8c44a6d1c828a291cfcef88a1225d7d9e6d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629647059.354017/events.out.tfevents.1629647059.91f414e9573a.903.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7f074be13e3b1f7c4aa8137a03ba77bcfd15ebddca2673c85defde32125f3dd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629647710.367303/events.out.tfevents.1629647710.91f414e9573a.903.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86c96350856fea8a12a116ef00c10a08afdb61e9d760f51ed6970dc9d1f8440d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629645072.91f414e9573a.903.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9628f937691368671220abecbeacd67c70bebbfed8934ce1512883c0971f4fa7
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629645744.91f414e9573a.903.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c7d4e8d0973ccb90b206db5a416f5f1021aac70bf8428f721e2ec453105fcbf
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629646405.91f414e9573a.903.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec479c169d8615a1742222c732b1b4a2b6896a2f6dd2cb32683c8fa7ae09dcc8
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629647059.91f414e9573a.903.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83c9953772e9968d856f400ecebd15b07b7dc8605da44c9360974cef7195e2db
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629647710.91f414e9573a.903.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d517635b01f28158d8aac716002781f2ed1d05b101f795164ac680015a696f5e
|
| 3 |
+
size 6703
|