"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/trainer_state.json +421 -4
- model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629668403.1334243/events.out.tfevents.1629668403.91f414e9573a.903.231 +3 -0
- model-bin/finetune/base/log/1629669077.8987737/events.out.tfevents.1629669077.91f414e9573a.903.233 +3 -0
- model-bin/finetune/base/log/1629669732.968207/events.out.tfevents.1629669732.91f414e9573a.903.235 +3 -0
- model-bin/finetune/base/log/1629670383.11149/events.out.tfevents.1629670383.91f414e9573a.903.237 +3 -0
- model-bin/finetune/base/log/1629671054.9956064/events.out.tfevents.1629671054.91f414e9573a.903.239 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629668403.91f414e9573a.903.230 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629669077.91f414e9573a.903.232 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629669732.91f414e9573a.903.234 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629670383.91f414e9573a.903.236 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629671054.91f414e9573a.903.238 +3 -0
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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:abd7e349665ee986ad6a1304afd9c973b72ddd083d53d36cfa76421ae9baeecd
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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:a36faae405be235591efa7f95b5ea6b8d13c3e0fec4d72f7fedb49fbc5dea309
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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:a988d2b0d1cb9211dfb121c3edded969769aa61f30d7f7449989a9910b173e2b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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:e27f67a7c467d5c3dc88360c3edd7627a0f7707bcda979a34251e7c0774f15c3
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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:b1c444f1f2aa8f4c6c362da78c8eca9c2f3fc07fdc96b3367ad5299db3980bd5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/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,
|
|
@@ -147129,11 +147129,428 @@
|
|
| 147129 |
"eval_steps_per_second": 0.666,
|
| 147130 |
"eval_wer": 0.21227696600337764,
|
| 147131 |
"step": 36957
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147132 |
}
|
| 147133 |
],
|
| 147134 |
-
"max_steps":
|
| 147135 |
"num_train_epochs": 5000,
|
| 147136 |
-
"total_flos": 1.
|
| 147137 |
"trial_name": null,
|
| 147138 |
"trial_params": null
|
| 147139 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 302.99598393574297,
|
| 5 |
+
"global_step": 37579,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 147129 |
"eval_steps_per_second": 0.666,
|
| 147130 |
"eval_wer": 0.21227696600337764,
|
| 147131 |
"step": 36957
|
| 147132 |
+
},
|
| 147133 |
+
{
|
| 147134 |
+
"epoch": 295.02,
|
| 147135 |
+
"learning_rate": 9.419273021001615e-06,
|
| 147136 |
+
"loss": 0.411,
|
| 147137 |
+
"step": 36960
|
| 147138 |
+
},
|
| 147139 |
+
{
|
| 147140 |
+
"epoch": 295.1,
|
| 147141 |
+
"learning_rate": 9.419111470113085e-06,
|
| 147142 |
+
"loss": 0.4316,
|
| 147143 |
+
"step": 36970
|
| 147144 |
+
},
|
| 147145 |
+
{
|
| 147146 |
+
"epoch": 295.18,
|
| 147147 |
+
"learning_rate": 9.418949919224557e-06,
|
| 147148 |
+
"loss": 0.6597,
|
| 147149 |
+
"step": 36980
|
| 147150 |
+
},
|
| 147151 |
+
{
|
| 147152 |
+
"epoch": 295.26,
|
| 147153 |
+
"learning_rate": 9.418788368336027e-06,
|
| 147154 |
+
"loss": 0.7,
|
| 147155 |
+
"step": 36990
|
| 147156 |
+
},
|
| 147157 |
+
{
|
| 147158 |
+
"epoch": 295.34,
|
| 147159 |
+
"learning_rate": 9.418626817447497e-06,
|
| 147160 |
+
"loss": 0.4122,
|
| 147161 |
+
"step": 37000
|
| 147162 |
+
},
|
| 147163 |
+
{
|
| 147164 |
+
"epoch": 295.42,
|
| 147165 |
+
"learning_rate": 9.418465266558967e-06,
|
| 147166 |
+
"loss": 1.0047,
|
| 147167 |
+
"step": 37010
|
| 147168 |
+
},
|
| 147169 |
+
{
|
| 147170 |
+
"epoch": 295.5,
|
| 147171 |
+
"learning_rate": 9.418303715670437e-06,
|
| 147172 |
+
"loss": 0.3317,
|
| 147173 |
+
"step": 37020
|
| 147174 |
+
},
|
| 147175 |
+
{
|
| 147176 |
+
"epoch": 295.58,
|
| 147177 |
+
"learning_rate": 9.418142164781907e-06,
|
| 147178 |
+
"loss": 0.6214,
|
| 147179 |
+
"step": 37030
|
| 147180 |
+
},
|
| 147181 |
+
{
|
| 147182 |
+
"epoch": 295.66,
|
| 147183 |
+
"learning_rate": 9.417980613893377e-06,
|
| 147184 |
+
"loss": 0.6993,
|
| 147185 |
+
"step": 37040
|
| 147186 |
+
},
|
| 147187 |
+
{
|
| 147188 |
+
"epoch": 295.74,
|
| 147189 |
+
"learning_rate": 9.417819063004847e-06,
|
| 147190 |
+
"loss": 0.4377,
|
| 147191 |
+
"step": 37050
|
| 147192 |
+
},
|
| 147193 |
+
{
|
| 147194 |
+
"epoch": 295.82,
|
| 147195 |
+
"learning_rate": 9.417657512116317e-06,
|
| 147196 |
+
"loss": 1.1317,
|
| 147197 |
+
"step": 37060
|
| 147198 |
+
},
|
| 147199 |
+
{
|
| 147200 |
+
"epoch": 295.9,
|
| 147201 |
+
"learning_rate": 9.417495961227787e-06,
|
| 147202 |
+
"loss": 0.4032,
|
| 147203 |
+
"step": 37070
|
| 147204 |
+
},
|
| 147205 |
+
{
|
| 147206 |
+
"epoch": 295.98,
|
| 147207 |
+
"learning_rate": 9.417334410339257e-06,
|
| 147208 |
+
"loss": 0.6623,
|
| 147209 |
+
"step": 37080
|
| 147210 |
+
},
|
| 147211 |
+
{
|
| 147212 |
+
"epoch": 296.0,
|
| 147213 |
+
"eval_loss": 0.4281972050666809,
|
| 147214 |
+
"eval_runtime": 40.0924,
|
| 147215 |
+
"eval_samples_per_second": 21.051,
|
| 147216 |
+
"eval_steps_per_second": 0.673,
|
| 147217 |
+
"eval_wer": 0.21002162941600577,
|
| 147218 |
+
"step": 37082
|
| 147219 |
+
},
|
| 147220 |
+
{
|
| 147221 |
+
"epoch": 299.06,
|
| 147222 |
+
"learning_rate": 9.417172859450727e-06,
|
| 147223 |
+
"loss": 0.384,
|
| 147224 |
+
"step": 37090
|
| 147225 |
+
},
|
| 147226 |
+
{
|
| 147227 |
+
"epoch": 299.14,
|
| 147228 |
+
"learning_rate": 9.417011308562198e-06,
|
| 147229 |
+
"loss": 0.4483,
|
| 147230 |
+
"step": 37100
|
| 147231 |
+
},
|
| 147232 |
+
{
|
| 147233 |
+
"epoch": 299.22,
|
| 147234 |
+
"learning_rate": 9.416849757673668e-06,
|
| 147235 |
+
"loss": 0.9751,
|
| 147236 |
+
"step": 37110
|
| 147237 |
+
},
|
| 147238 |
+
{
|
| 147239 |
+
"epoch": 299.31,
|
| 147240 |
+
"learning_rate": 9.416688206785138e-06,
|
| 147241 |
+
"loss": 0.4218,
|
| 147242 |
+
"step": 37120
|
| 147243 |
+
},
|
| 147244 |
+
{
|
| 147245 |
+
"epoch": 299.39,
|
| 147246 |
+
"learning_rate": 9.416526655896608e-06,
|
| 147247 |
+
"loss": 0.6011,
|
| 147248 |
+
"step": 37130
|
| 147249 |
+
},
|
| 147250 |
+
{
|
| 147251 |
+
"epoch": 299.47,
|
| 147252 |
+
"learning_rate": 9.416365105008078e-06,
|
| 147253 |
+
"loss": 0.6866,
|
| 147254 |
+
"step": 37140
|
| 147255 |
+
},
|
| 147256 |
+
{
|
| 147257 |
+
"epoch": 299.55,
|
| 147258 |
+
"learning_rate": 9.416203554119548e-06,
|
| 147259 |
+
"loss": 0.412,
|
| 147260 |
+
"step": 37150
|
| 147261 |
+
},
|
| 147262 |
+
{
|
| 147263 |
+
"epoch": 299.63,
|
| 147264 |
+
"learning_rate": 9.416042003231018e-06,
|
| 147265 |
+
"loss": 0.9878,
|
| 147266 |
+
"step": 37160
|
| 147267 |
+
},
|
| 147268 |
+
{
|
| 147269 |
+
"epoch": 299.71,
|
| 147270 |
+
"learning_rate": 9.415880452342488e-06,
|
| 147271 |
+
"loss": 0.391,
|
| 147272 |
+
"step": 37170
|
| 147273 |
+
},
|
| 147274 |
+
{
|
| 147275 |
+
"epoch": 299.79,
|
| 147276 |
+
"learning_rate": 9.415718901453958e-06,
|
| 147277 |
+
"loss": 0.6808,
|
| 147278 |
+
"step": 37180
|
| 147279 |
+
},
|
| 147280 |
+
{
|
| 147281 |
+
"epoch": 299.87,
|
| 147282 |
+
"learning_rate": 9.415557350565428e-06,
|
| 147283 |
+
"loss": 0.7299,
|
| 147284 |
+
"step": 37190
|
| 147285 |
+
},
|
| 147286 |
+
{
|
| 147287 |
+
"epoch": 299.95,
|
| 147288 |
+
"learning_rate": 9.415395799676898e-06,
|
| 147289 |
+
"loss": 0.4137,
|
| 147290 |
+
"step": 37200
|
| 147291 |
+
},
|
| 147292 |
+
{
|
| 147293 |
+
"epoch": 300.0,
|
| 147294 |
+
"eval_loss": 0.43289971351623535,
|
| 147295 |
+
"eval_runtime": 40.9688,
|
| 147296 |
+
"eval_samples_per_second": 20.601,
|
| 147297 |
+
"eval_steps_per_second": 0.659,
|
| 147298 |
+
"eval_wer": 0.21350112751873135,
|
| 147299 |
+
"step": 37206
|
| 147300 |
+
},
|
| 147301 |
+
{
|
| 147302 |
+
"epoch": 300.03,
|
| 147303 |
+
"learning_rate": 9.415234248788368e-06,
|
| 147304 |
+
"loss": 0.5024,
|
| 147305 |
+
"step": 37210
|
| 147306 |
+
},
|
| 147307 |
+
{
|
| 147308 |
+
"epoch": 300.11,
|
| 147309 |
+
"learning_rate": 9.41507269789984e-06,
|
| 147310 |
+
"loss": 0.3504,
|
| 147311 |
+
"step": 37220
|
| 147312 |
+
},
|
| 147313 |
+
{
|
| 147314 |
+
"epoch": 300.19,
|
| 147315 |
+
"learning_rate": 9.41491114701131e-06,
|
| 147316 |
+
"loss": 0.8002,
|
| 147317 |
+
"step": 37230
|
| 147318 |
+
},
|
| 147319 |
+
{
|
| 147320 |
+
"epoch": 300.27,
|
| 147321 |
+
"learning_rate": 9.41474959612278e-06,
|
| 147322 |
+
"loss": 0.5897,
|
| 147323 |
+
"step": 37240
|
| 147324 |
+
},
|
| 147325 |
+
{
|
| 147326 |
+
"epoch": 300.35,
|
| 147327 |
+
"learning_rate": 9.41458804523425e-06,
|
| 147328 |
+
"loss": 0.4217,
|
| 147329 |
+
"step": 37250
|
| 147330 |
+
},
|
| 147331 |
+
{
|
| 147332 |
+
"epoch": 300.44,
|
| 147333 |
+
"learning_rate": 9.41442649434572e-06,
|
| 147334 |
+
"loss": 0.9469,
|
| 147335 |
+
"step": 37260
|
| 147336 |
+
},
|
| 147337 |
+
{
|
| 147338 |
+
"epoch": 300.52,
|
| 147339 |
+
"learning_rate": 9.41426494345719e-06,
|
| 147340 |
+
"loss": 0.4014,
|
| 147341 |
+
"step": 37270
|
| 147342 |
+
},
|
| 147343 |
+
{
|
| 147344 |
+
"epoch": 300.6,
|
| 147345 |
+
"learning_rate": 9.41410339256866e-06,
|
| 147346 |
+
"loss": 0.7525,
|
| 147347 |
+
"step": 37280
|
| 147348 |
+
},
|
| 147349 |
+
{
|
| 147350 |
+
"epoch": 300.68,
|
| 147351 |
+
"learning_rate": 9.41394184168013e-06,
|
| 147352 |
+
"loss": 0.5929,
|
| 147353 |
+
"step": 37290
|
| 147354 |
+
},
|
| 147355 |
+
{
|
| 147356 |
+
"epoch": 300.76,
|
| 147357 |
+
"learning_rate": 9.4137802907916e-06,
|
| 147358 |
+
"loss": 0.4489,
|
| 147359 |
+
"step": 37300
|
| 147360 |
+
},
|
| 147361 |
+
{
|
| 147362 |
+
"epoch": 300.84,
|
| 147363 |
+
"learning_rate": 9.41361873990307e-06,
|
| 147364 |
+
"loss": 0.9527,
|
| 147365 |
+
"step": 37310
|
| 147366 |
+
},
|
| 147367 |
+
{
|
| 147368 |
+
"epoch": 300.92,
|
| 147369 |
+
"learning_rate": 9.41345718901454e-06,
|
| 147370 |
+
"loss": 0.3861,
|
| 147371 |
+
"step": 37320
|
| 147372 |
+
},
|
| 147373 |
+
{
|
| 147374 |
+
"epoch": 301.0,
|
| 147375 |
+
"learning_rate": 9.413295638126011e-06,
|
| 147376 |
+
"loss": 0.9791,
|
| 147377 |
+
"step": 37330
|
| 147378 |
+
},
|
| 147379 |
+
{
|
| 147380 |
+
"epoch": 301.0,
|
| 147381 |
+
"eval_loss": 0.4430272877216339,
|
| 147382 |
+
"eval_runtime": 40.5642,
|
| 147383 |
+
"eval_samples_per_second": 20.807,
|
| 147384 |
+
"eval_steps_per_second": 0.666,
|
| 147385 |
+
"eval_wer": 0.2157237325495959,
|
| 147386 |
+
"step": 37330
|
| 147387 |
+
},
|
| 147388 |
+
{
|
| 147389 |
+
"epoch": 298.08,
|
| 147390 |
+
"learning_rate": 9.413134087237481e-06,
|
| 147391 |
+
"loss": 0.3867,
|
| 147392 |
+
"step": 37340
|
| 147393 |
+
},
|
| 147394 |
+
{
|
| 147395 |
+
"epoch": 298.16,
|
| 147396 |
+
"learning_rate": 9.412972536348951e-06,
|
| 147397 |
+
"loss": 0.4678,
|
| 147398 |
+
"step": 37350
|
| 147399 |
+
},
|
| 147400 |
+
{
|
| 147401 |
+
"epoch": 298.24,
|
| 147402 |
+
"learning_rate": 9.412810985460421e-06,
|
| 147403 |
+
"loss": 0.9091,
|
| 147404 |
+
"step": 37360
|
| 147405 |
+
},
|
| 147406 |
+
{
|
| 147407 |
+
"epoch": 298.32,
|
| 147408 |
+
"learning_rate": 9.412649434571891e-06,
|
| 147409 |
+
"loss": 0.3784,
|
| 147410 |
+
"step": 37370
|
| 147411 |
+
},
|
| 147412 |
+
{
|
| 147413 |
+
"epoch": 298.4,
|
| 147414 |
+
"learning_rate": 9.412487883683361e-06,
|
| 147415 |
+
"loss": 0.9351,
|
| 147416 |
+
"step": 37380
|
| 147417 |
+
},
|
| 147418 |
+
{
|
| 147419 |
+
"epoch": 298.48,
|
| 147420 |
+
"learning_rate": 9.412326332794831e-06,
|
| 147421 |
+
"loss": 0.3865,
|
| 147422 |
+
"step": 37390
|
| 147423 |
+
},
|
| 147424 |
+
{
|
| 147425 |
+
"epoch": 298.56,
|
| 147426 |
+
"learning_rate": 9.412164781906301e-06,
|
| 147427 |
+
"loss": 0.4235,
|
| 147428 |
+
"step": 37400
|
| 147429 |
+
},
|
| 147430 |
+
{
|
| 147431 |
+
"epoch": 298.64,
|
| 147432 |
+
"learning_rate": 9.412003231017771e-06,
|
| 147433 |
+
"loss": 0.9001,
|
| 147434 |
+
"step": 37410
|
| 147435 |
+
},
|
| 147436 |
+
{
|
| 147437 |
+
"epoch": 298.72,
|
| 147438 |
+
"learning_rate": 9.411841680129241e-06,
|
| 147439 |
+
"loss": 0.3678,
|
| 147440 |
+
"step": 37420
|
| 147441 |
+
},
|
| 147442 |
+
{
|
| 147443 |
+
"epoch": 298.8,
|
| 147444 |
+
"learning_rate": 9.411680129240711e-06,
|
| 147445 |
+
"loss": 0.9599,
|
| 147446 |
+
"step": 37430
|
| 147447 |
+
},
|
| 147448 |
+
{
|
| 147449 |
+
"epoch": 298.88,
|
| 147450 |
+
"learning_rate": 9.411518578352181e-06,
|
| 147451 |
+
"loss": 0.4261,
|
| 147452 |
+
"step": 37440
|
| 147453 |
+
},
|
| 147454 |
+
{
|
| 147455 |
+
"epoch": 298.96,
|
| 147456 |
+
"learning_rate": 9.411357027463653e-06,
|
| 147457 |
+
"loss": 0.4676,
|
| 147458 |
+
"step": 37450
|
| 147459 |
+
},
|
| 147460 |
+
{
|
| 147461 |
+
"epoch": 299.0,
|
| 147462 |
+
"eval_loss": 0.4521419405937195,
|
| 147463 |
+
"eval_runtime": 41.3572,
|
| 147464 |
+
"eval_samples_per_second": 20.408,
|
| 147465 |
+
"eval_steps_per_second": 0.653,
|
| 147466 |
+
"eval_wer": 0.21275675675675676,
|
| 147467 |
+
"step": 37455
|
| 147468 |
+
},
|
| 147469 |
+
{
|
| 147470 |
+
"epoch": 302.04,
|
| 147471 |
+
"learning_rate": 9.411195476575123e-06,
|
| 147472 |
+
"loss": 0.4283,
|
| 147473 |
+
"step": 37460
|
| 147474 |
+
},
|
| 147475 |
+
{
|
| 147476 |
+
"epoch": 302.12,
|
| 147477 |
+
"learning_rate": 9.411033925686593e-06,
|
| 147478 |
+
"loss": 0.4122,
|
| 147479 |
+
"step": 37470
|
| 147480 |
+
},
|
| 147481 |
+
{
|
| 147482 |
+
"epoch": 302.2,
|
| 147483 |
+
"learning_rate": 9.410872374798063e-06,
|
| 147484 |
+
"loss": 0.9011,
|
| 147485 |
+
"step": 37480
|
| 147486 |
+
},
|
| 147487 |
+
{
|
| 147488 |
+
"epoch": 302.28,
|
| 147489 |
+
"learning_rate": 9.410710823909532e-06,
|
| 147490 |
+
"loss": 0.4033,
|
| 147491 |
+
"step": 37490
|
| 147492 |
+
},
|
| 147493 |
+
{
|
| 147494 |
+
"epoch": 302.36,
|
| 147495 |
+
"learning_rate": 9.410549273021002e-06,
|
| 147496 |
+
"loss": 0.5024,
|
| 147497 |
+
"step": 37500
|
| 147498 |
+
},
|
| 147499 |
+
{
|
| 147500 |
+
"epoch": 302.44,
|
| 147501 |
+
"learning_rate": 9.410387722132472e-06,
|
| 147502 |
+
"loss": 0.864,
|
| 147503 |
+
"step": 37510
|
| 147504 |
+
},
|
| 147505 |
+
{
|
| 147506 |
+
"epoch": 302.52,
|
| 147507 |
+
"learning_rate": 9.410226171243942e-06,
|
| 147508 |
+
"loss": 0.4052,
|
| 147509 |
+
"step": 37520
|
| 147510 |
+
},
|
| 147511 |
+
{
|
| 147512 |
+
"epoch": 302.6,
|
| 147513 |
+
"learning_rate": 9.410064620355412e-06,
|
| 147514 |
+
"loss": 0.9824,
|
| 147515 |
+
"step": 37530
|
| 147516 |
+
},
|
| 147517 |
+
{
|
| 147518 |
+
"epoch": 302.68,
|
| 147519 |
+
"learning_rate": 9.409903069466882e-06,
|
| 147520 |
+
"loss": 0.3639,
|
| 147521 |
+
"step": 37540
|
| 147522 |
+
},
|
| 147523 |
+
{
|
| 147524 |
+
"epoch": 302.76,
|
| 147525 |
+
"learning_rate": 9.409741518578352e-06,
|
| 147526 |
+
"loss": 0.5923,
|
| 147527 |
+
"step": 37550
|
| 147528 |
+
},
|
| 147529 |
+
{
|
| 147530 |
+
"epoch": 302.84,
|
| 147531 |
+
"learning_rate": 9.409579967689822e-06,
|
| 147532 |
+
"loss": 0.8179,
|
| 147533 |
+
"step": 37560
|
| 147534 |
+
},
|
| 147535 |
+
{
|
| 147536 |
+
"epoch": 302.92,
|
| 147537 |
+
"learning_rate": 9.409418416801294e-06,
|
| 147538 |
+
"loss": 0.3718,
|
| 147539 |
+
"step": 37570
|
| 147540 |
+
},
|
| 147541 |
+
{
|
| 147542 |
+
"epoch": 303.0,
|
| 147543 |
+
"eval_loss": 0.46862301230430603,
|
| 147544 |
+
"eval_runtime": 41.3914,
|
| 147545 |
+
"eval_samples_per_second": 20.391,
|
| 147546 |
+
"eval_steps_per_second": 0.652,
|
| 147547 |
+
"eval_wer": 0.21095204008589835,
|
| 147548 |
+
"step": 37579
|
| 147549 |
}
|
| 147550 |
],
|
| 147551 |
+
"max_steps": 620000,
|
| 147552 |
"num_train_epochs": 5000,
|
| 147553 |
+
"total_flos": 1.0575630859167164e+20,
|
| 147554 |
"trial_name": null,
|
| 147555 |
"trial_params": null
|
| 147556 |
}
|
model-bin/finetune/base/{checkpoint-36957 → checkpoint-37579}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629668403.1334243/events.out.tfevents.1629668403.91f414e9573a.903.231
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52a5a58693902d84854dff675eab84d90bd245b1a199b34b2fc6f1ed559ade6f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629669077.8987737/events.out.tfevents.1629669077.91f414e9573a.903.233
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aae3f06fe94d5e06ff88e67e1d0b362cacf1a878dde457c23e84de7389dac9c5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629669732.968207/events.out.tfevents.1629669732.91f414e9573a.903.235
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec89027d695396488b1852d2f6bbe9032ebed5c7bd22853bb95a4323ff9e2b4d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629670383.11149/events.out.tfevents.1629670383.91f414e9573a.903.237
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f81c6ac3fc41623d600d30abd63dfb84803c2a43827644da3c16c438abec5c0b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629671054.9956064/events.out.tfevents.1629671054.91f414e9573a.903.239
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cbb339c55c145502c57281651f5fdeebba066f80ec285ea0665fa370b42beec0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629668403.91f414e9573a.903.230
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a137ca361b0c9eb43e77d0c6b461051dd10ee654e631bb02e0adc867795546b9
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629669077.91f414e9573a.903.232
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:43f714afc7e573b49c1795f6bbc93177691a6ede6dae72400c7bd2da3cffdeee
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629669732.91f414e9573a.903.234
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf487872540836fc7d027003bd360a3fcf2c3416501e0eb238deed1b1d94aee2
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629670383.91f414e9573a.903.236
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5336ad0afa69eae6a4d574b506648b48c67e58fde1687308747100116c06ec90
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629671054.91f414e9573a.903.238
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8f6ef1d4aef41fb5cd409ef98fde280649a940f3f14c499ef5273d47a1ecd23
|
| 3 |
+
size 6543
|