"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630183506.9783158/events.out.tfevents.1630183506.86bb0ddabf9b.4092.301 +3 -0
- model-bin/finetune/base/log/1630183895.283262/events.out.tfevents.1630183895.86bb0ddabf9b.4092.303 +3 -0
- model-bin/finetune/base/log/1630184282.1676457/events.out.tfevents.1630184282.86bb0ddabf9b.4092.305 +3 -0
- model-bin/finetune/base/log/1630184680.4007196/events.out.tfevents.1630184680.86bb0ddabf9b.4092.307 +3 -0
- model-bin/finetune/base/log/1630185067.7506618/events.out.tfevents.1630185067.86bb0ddabf9b.4092.309 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630183506.86bb0ddabf9b.4092.300 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630183895.86bb0ddabf9b.4092.302 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630184282.86bb0ddabf9b.4092.304 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630184680.86bb0ddabf9b.4092.306 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630185067.86bb0ddabf9b.4092.308 +3 -0
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/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 sha256:76d7a9e2252213272411c461f9164ac3e93d1897f2d008f224c9bed96cad13f9
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/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 sha256:000d7ccbe251fbed150a5f45c57233a29742cf5d5eff85be3ee4e1f6794da44c
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/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 sha256:62621059a4c0e648429ed76eebb1747257125240134b8622f94dd250f5c244fb
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/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:e8695f57df923e22b943b0b0f2b9cc7007008e80b53ccee275b3a35963fe67e9
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/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:a7158013891c1869087253a91158ad1f03385d1ef5853212b8a515594f48c7ec
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -273216,11 +273216,800 @@
|
|
| 273216 |
"eval_steps_per_second": 0.751,
|
| 273217 |
"eval_wer": 0.17527468420659242,
|
| 273218 |
"step": 136640
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 273219 |
}
|
| 273220 |
],
|
| 273221 |
-
"max_steps":
|
| 273222 |
"num_train_epochs": 5000,
|
| 273223 |
-
"total_flos": 3.
|
| 273224 |
"trial_name": null,
|
| 273225 |
"trial_params": null
|
| 273226 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1098.0,
|
| 5 |
+
"global_step": 137263,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 273216 |
"eval_steps_per_second": 0.751,
|
| 273217 |
"eval_wer": 0.17527468420659242,
|
| 273218 |
"step": 136640
|
| 273219 |
+
},
|
| 273220 |
+
{
|
| 273221 |
+
"epoch": 1093.04,
|
| 273222 |
+
"learning_rate": 7.827051282051283e-06,
|
| 273223 |
+
"loss": 0.3093,
|
| 273224 |
+
"step": 136645
|
| 273225 |
+
},
|
| 273226 |
+
{
|
| 273227 |
+
"epoch": 1093.08,
|
| 273228 |
+
"learning_rate": 7.826971153846154e-06,
|
| 273229 |
+
"loss": 0.2836,
|
| 273230 |
+
"step": 136650
|
| 273231 |
+
},
|
| 273232 |
+
{
|
| 273233 |
+
"epoch": 1093.12,
|
| 273234 |
+
"learning_rate": 7.826891025641026e-06,
|
| 273235 |
+
"loss": 0.2721,
|
| 273236 |
+
"step": 136655
|
| 273237 |
+
},
|
| 273238 |
+
{
|
| 273239 |
+
"epoch": 1093.16,
|
| 273240 |
+
"learning_rate": 7.826810897435899e-06,
|
| 273241 |
+
"loss": 0.3567,
|
| 273242 |
+
"step": 136660
|
| 273243 |
+
},
|
| 273244 |
+
{
|
| 273245 |
+
"epoch": 1093.2,
|
| 273246 |
+
"learning_rate": 7.82673076923077e-06,
|
| 273247 |
+
"loss": 1.418,
|
| 273248 |
+
"step": 136665
|
| 273249 |
+
},
|
| 273250 |
+
{
|
| 273251 |
+
"epoch": 1093.24,
|
| 273252 |
+
"learning_rate": 7.826650641025641e-06,
|
| 273253 |
+
"loss": 0.2603,
|
| 273254 |
+
"step": 136670
|
| 273255 |
+
},
|
| 273256 |
+
{
|
| 273257 |
+
"epoch": 1093.28,
|
| 273258 |
+
"learning_rate": 7.826570512820514e-06,
|
| 273259 |
+
"loss": 0.2787,
|
| 273260 |
+
"step": 136675
|
| 273261 |
+
},
|
| 273262 |
+
{
|
| 273263 |
+
"epoch": 1093.32,
|
| 273264 |
+
"learning_rate": 7.826490384615386e-06,
|
| 273265 |
+
"loss": 0.4668,
|
| 273266 |
+
"step": 136680
|
| 273267 |
+
},
|
| 273268 |
+
{
|
| 273269 |
+
"epoch": 1093.36,
|
| 273270 |
+
"learning_rate": 7.826410256410257e-06,
|
| 273271 |
+
"loss": 0.4332,
|
| 273272 |
+
"step": 136685
|
| 273273 |
+
},
|
| 273274 |
+
{
|
| 273275 |
+
"epoch": 1093.4,
|
| 273276 |
+
"learning_rate": 7.826330128205128e-06,
|
| 273277 |
+
"loss": 1.1931,
|
| 273278 |
+
"step": 136690
|
| 273279 |
+
},
|
| 273280 |
+
{
|
| 273281 |
+
"epoch": 1093.44,
|
| 273282 |
+
"learning_rate": 7.826250000000002e-06,
|
| 273283 |
+
"loss": 0.3168,
|
| 273284 |
+
"step": 136695
|
| 273285 |
+
},
|
| 273286 |
+
{
|
| 273287 |
+
"epoch": 1093.48,
|
| 273288 |
+
"learning_rate": 7.826169871794871e-06,
|
| 273289 |
+
"loss": 0.2578,
|
| 273290 |
+
"step": 136700
|
| 273291 |
+
},
|
| 273292 |
+
{
|
| 273293 |
+
"epoch": 1093.52,
|
| 273294 |
+
"learning_rate": 7.826089743589744e-06,
|
| 273295 |
+
"loss": 0.2864,
|
| 273296 |
+
"step": 136705
|
| 273297 |
+
},
|
| 273298 |
+
{
|
| 273299 |
+
"epoch": 1093.56,
|
| 273300 |
+
"learning_rate": 7.826009615384617e-06,
|
| 273301 |
+
"loss": 0.3803,
|
| 273302 |
+
"step": 136710
|
| 273303 |
+
},
|
| 273304 |
+
{
|
| 273305 |
+
"epoch": 1093.6,
|
| 273306 |
+
"learning_rate": 7.825929487179487e-06,
|
| 273307 |
+
"loss": 1.1531,
|
| 273308 |
+
"step": 136715
|
| 273309 |
+
},
|
| 273310 |
+
{
|
| 273311 |
+
"epoch": 1093.64,
|
| 273312 |
+
"learning_rate": 7.82584935897436e-06,
|
| 273313 |
+
"loss": 0.3286,
|
| 273314 |
+
"step": 136720
|
| 273315 |
+
},
|
| 273316 |
+
{
|
| 273317 |
+
"epoch": 1093.68,
|
| 273318 |
+
"learning_rate": 7.825769230769231e-06,
|
| 273319 |
+
"loss": 0.283,
|
| 273320 |
+
"step": 136725
|
| 273321 |
+
},
|
| 273322 |
+
{
|
| 273323 |
+
"epoch": 1093.72,
|
| 273324 |
+
"learning_rate": 7.825689102564103e-06,
|
| 273325 |
+
"loss": 0.2493,
|
| 273326 |
+
"step": 136730
|
| 273327 |
+
},
|
| 273328 |
+
{
|
| 273329 |
+
"epoch": 1093.76,
|
| 273330 |
+
"learning_rate": 7.825608974358974e-06,
|
| 273331 |
+
"loss": 0.4748,
|
| 273332 |
+
"step": 136735
|
| 273333 |
+
},
|
| 273334 |
+
{
|
| 273335 |
+
"epoch": 1093.8,
|
| 273336 |
+
"learning_rate": 7.825528846153847e-06,
|
| 273337 |
+
"loss": 1.2627,
|
| 273338 |
+
"step": 136740
|
| 273339 |
+
},
|
| 273340 |
+
{
|
| 273341 |
+
"epoch": 1093.84,
|
| 273342 |
+
"learning_rate": 7.825448717948718e-06,
|
| 273343 |
+
"loss": 0.3215,
|
| 273344 |
+
"step": 136745
|
| 273345 |
+
},
|
| 273346 |
+
{
|
| 273347 |
+
"epoch": 1093.88,
|
| 273348 |
+
"learning_rate": 7.82536858974359e-06,
|
| 273349 |
+
"loss": 0.2965,
|
| 273350 |
+
"step": 136750
|
| 273351 |
+
},
|
| 273352 |
+
{
|
| 273353 |
+
"epoch": 1093.92,
|
| 273354 |
+
"learning_rate": 7.825288461538461e-06,
|
| 273355 |
+
"loss": 0.2845,
|
| 273356 |
+
"step": 136755
|
| 273357 |
+
},
|
| 273358 |
+
{
|
| 273359 |
+
"epoch": 1093.96,
|
| 273360 |
+
"learning_rate": 7.825208333333334e-06,
|
| 273361 |
+
"loss": 0.4017,
|
| 273362 |
+
"step": 136760
|
| 273363 |
+
},
|
| 273364 |
+
{
|
| 273365 |
+
"epoch": 1094.0,
|
| 273366 |
+
"learning_rate": 7.825128205128206e-06,
|
| 273367 |
+
"loss": 1.0981,
|
| 273368 |
+
"step": 136765
|
| 273369 |
+
},
|
| 273370 |
+
{
|
| 273371 |
+
"epoch": 1094.0,
|
| 273372 |
+
"eval_loss": 0.5501209497451782,
|
| 273373 |
+
"eval_runtime": 36.8982,
|
| 273374 |
+
"eval_samples_per_second": 22.847,
|
| 273375 |
+
"eval_steps_per_second": 0.732,
|
| 273376 |
+
"eval_wer": 0.17850520610469262,
|
| 273377 |
+
"step": 136765
|
| 273378 |
+
},
|
| 273379 |
+
{
|
| 273380 |
+
"epoch": 1102.04,
|
| 273381 |
+
"learning_rate": 7.825048076923077e-06,
|
| 273382 |
+
"loss": 0.3279,
|
| 273383 |
+
"step": 136770
|
| 273384 |
+
},
|
| 273385 |
+
{
|
| 273386 |
+
"epoch": 1102.08,
|
| 273387 |
+
"learning_rate": 7.82496794871795e-06,
|
| 273388 |
+
"loss": 0.2974,
|
| 273389 |
+
"step": 136775
|
| 273390 |
+
},
|
| 273391 |
+
{
|
| 273392 |
+
"epoch": 1102.12,
|
| 273393 |
+
"learning_rate": 7.824887820512821e-06,
|
| 273394 |
+
"loss": 0.3185,
|
| 273395 |
+
"step": 136780
|
| 273396 |
+
},
|
| 273397 |
+
{
|
| 273398 |
+
"epoch": 1102.16,
|
| 273399 |
+
"learning_rate": 7.824807692307693e-06,
|
| 273400 |
+
"loss": 0.3698,
|
| 273401 |
+
"step": 136785
|
| 273402 |
+
},
|
| 273403 |
+
{
|
| 273404 |
+
"epoch": 1102.2,
|
| 273405 |
+
"learning_rate": 7.824727564102564e-06,
|
| 273406 |
+
"loss": 1.0618,
|
| 273407 |
+
"step": 136790
|
| 273408 |
+
},
|
| 273409 |
+
{
|
| 273410 |
+
"epoch": 1102.24,
|
| 273411 |
+
"learning_rate": 7.824647435897437e-06,
|
| 273412 |
+
"loss": 0.3564,
|
| 273413 |
+
"step": 136795
|
| 273414 |
+
},
|
| 273415 |
+
{
|
| 273416 |
+
"epoch": 1102.28,
|
| 273417 |
+
"learning_rate": 7.824567307692309e-06,
|
| 273418 |
+
"loss": 0.2904,
|
| 273419 |
+
"step": 136800
|
| 273420 |
+
},
|
| 273421 |
+
{
|
| 273422 |
+
"epoch": 1102.32,
|
| 273423 |
+
"learning_rate": 7.82448717948718e-06,
|
| 273424 |
+
"loss": 0.2934,
|
| 273425 |
+
"step": 136805
|
| 273426 |
+
},
|
| 273427 |
+
{
|
| 273428 |
+
"epoch": 1102.36,
|
| 273429 |
+
"learning_rate": 7.824407051282053e-06,
|
| 273430 |
+
"loss": 0.484,
|
| 273431 |
+
"step": 136810
|
| 273432 |
+
},
|
| 273433 |
+
{
|
| 273434 |
+
"epoch": 1102.4,
|
| 273435 |
+
"learning_rate": 7.824326923076924e-06,
|
| 273436 |
+
"loss": 1.0554,
|
| 273437 |
+
"step": 136815
|
| 273438 |
+
},
|
| 273439 |
+
{
|
| 273440 |
+
"epoch": 1102.44,
|
| 273441 |
+
"learning_rate": 7.824246794871796e-06,
|
| 273442 |
+
"loss": 0.3041,
|
| 273443 |
+
"step": 136820
|
| 273444 |
+
},
|
| 273445 |
+
{
|
| 273446 |
+
"epoch": 1102.48,
|
| 273447 |
+
"learning_rate": 7.824166666666667e-06,
|
| 273448 |
+
"loss": 0.3005,
|
| 273449 |
+
"step": 136825
|
| 273450 |
+
},
|
| 273451 |
+
{
|
| 273452 |
+
"epoch": 1102.52,
|
| 273453 |
+
"learning_rate": 7.82408653846154e-06,
|
| 273454 |
+
"loss": 0.3459,
|
| 273455 |
+
"step": 136830
|
| 273456 |
+
},
|
| 273457 |
+
{
|
| 273458 |
+
"epoch": 1102.56,
|
| 273459 |
+
"learning_rate": 7.824006410256411e-06,
|
| 273460 |
+
"loss": 0.4353,
|
| 273461 |
+
"step": 136835
|
| 273462 |
+
},
|
| 273463 |
+
{
|
| 273464 |
+
"epoch": 1102.6,
|
| 273465 |
+
"learning_rate": 7.823926282051283e-06,
|
| 273466 |
+
"loss": 1.3489,
|
| 273467 |
+
"step": 136840
|
| 273468 |
+
},
|
| 273469 |
+
{
|
| 273470 |
+
"epoch": 1102.65,
|
| 273471 |
+
"learning_rate": 7.823846153846154e-06,
|
| 273472 |
+
"loss": 0.3342,
|
| 273473 |
+
"step": 136845
|
| 273474 |
+
},
|
| 273475 |
+
{
|
| 273476 |
+
"epoch": 1102.69,
|
| 273477 |
+
"learning_rate": 7.823766025641027e-06,
|
| 273478 |
+
"loss": 0.293,
|
| 273479 |
+
"step": 136850
|
| 273480 |
+
},
|
| 273481 |
+
{
|
| 273482 |
+
"epoch": 1102.73,
|
| 273483 |
+
"learning_rate": 7.823685897435897e-06,
|
| 273484 |
+
"loss": 0.3329,
|
| 273485 |
+
"step": 136855
|
| 273486 |
+
},
|
| 273487 |
+
{
|
| 273488 |
+
"epoch": 1102.77,
|
| 273489 |
+
"learning_rate": 7.82360576923077e-06,
|
| 273490 |
+
"loss": 0.5113,
|
| 273491 |
+
"step": 136860
|
| 273492 |
+
},
|
| 273493 |
+
{
|
| 273494 |
+
"epoch": 1102.81,
|
| 273495 |
+
"learning_rate": 7.823525641025643e-06,
|
| 273496 |
+
"loss": 1.1694,
|
| 273497 |
+
"step": 136865
|
| 273498 |
+
},
|
| 273499 |
+
{
|
| 273500 |
+
"epoch": 1102.85,
|
| 273501 |
+
"learning_rate": 7.823445512820513e-06,
|
| 273502 |
+
"loss": 0.2982,
|
| 273503 |
+
"step": 136870
|
| 273504 |
+
},
|
| 273505 |
+
{
|
| 273506 |
+
"epoch": 1102.89,
|
| 273507 |
+
"learning_rate": 7.823365384615386e-06,
|
| 273508 |
+
"loss": 0.297,
|
| 273509 |
+
"step": 136875
|
| 273510 |
+
},
|
| 273511 |
+
{
|
| 273512 |
+
"epoch": 1102.93,
|
| 273513 |
+
"learning_rate": 7.823285256410257e-06,
|
| 273514 |
+
"loss": 0.3415,
|
| 273515 |
+
"step": 136880
|
| 273516 |
+
},
|
| 273517 |
+
{
|
| 273518 |
+
"epoch": 1102.97,
|
| 273519 |
+
"learning_rate": 7.823205128205128e-06,
|
| 273520 |
+
"loss": 0.4936,
|
| 273521 |
+
"step": 136885
|
| 273522 |
+
},
|
| 273523 |
+
{
|
| 273524 |
+
"epoch": 1103.0,
|
| 273525 |
+
"eval_loss": 0.371574729681015,
|
| 273526 |
+
"eval_runtime": 35.6428,
|
| 273527 |
+
"eval_samples_per_second": 23.651,
|
| 273528 |
+
"eval_steps_per_second": 0.758,
|
| 273529 |
+
"eval_wer": 0.19113115370161593,
|
| 273530 |
+
"step": 136889
|
| 273531 |
+
},
|
| 273532 |
+
{
|
| 273533 |
+
"epoch": 1095.01,
|
| 273534 |
+
"learning_rate": 7.823125e-06,
|
| 273535 |
+
"loss": 0.3023,
|
| 273536 |
+
"step": 136890
|
| 273537 |
+
},
|
| 273538 |
+
{
|
| 273539 |
+
"epoch": 1095.05,
|
| 273540 |
+
"learning_rate": 7.823044871794873e-06,
|
| 273541 |
+
"loss": 0.2835,
|
| 273542 |
+
"step": 136895
|
| 273543 |
+
},
|
| 273544 |
+
{
|
| 273545 |
+
"epoch": 1095.09,
|
| 273546 |
+
"learning_rate": 7.822964743589744e-06,
|
| 273547 |
+
"loss": 0.2638,
|
| 273548 |
+
"step": 136900
|
| 273549 |
+
},
|
| 273550 |
+
{
|
| 273551 |
+
"epoch": 1095.13,
|
| 273552 |
+
"learning_rate": 7.822884615384616e-06,
|
| 273553 |
+
"loss": 0.267,
|
| 273554 |
+
"step": 136905
|
| 273555 |
+
},
|
| 273556 |
+
{
|
| 273557 |
+
"epoch": 1095.17,
|
| 273558 |
+
"learning_rate": 7.822804487179489e-06,
|
| 273559 |
+
"loss": 0.5787,
|
| 273560 |
+
"step": 136910
|
| 273561 |
+
},
|
| 273562 |
+
{
|
| 273563 |
+
"epoch": 1095.21,
|
| 273564 |
+
"learning_rate": 7.82272435897436e-06,
|
| 273565 |
+
"loss": 1.0743,
|
| 273566 |
+
"step": 136915
|
| 273567 |
+
},
|
| 273568 |
+
{
|
| 273569 |
+
"epoch": 1095.25,
|
| 273570 |
+
"learning_rate": 7.822644230769231e-06,
|
| 273571 |
+
"loss": 0.2557,
|
| 273572 |
+
"step": 136920
|
| 273573 |
+
},
|
| 273574 |
+
{
|
| 273575 |
+
"epoch": 1095.29,
|
| 273576 |
+
"learning_rate": 7.822564102564103e-06,
|
| 273577 |
+
"loss": 0.3078,
|
| 273578 |
+
"step": 136925
|
| 273579 |
+
},
|
| 273580 |
+
{
|
| 273581 |
+
"epoch": 1095.33,
|
| 273582 |
+
"learning_rate": 7.822483974358976e-06,
|
| 273583 |
+
"loss": 0.3404,
|
| 273584 |
+
"step": 136930
|
| 273585 |
+
},
|
| 273586 |
+
{
|
| 273587 |
+
"epoch": 1095.37,
|
| 273588 |
+
"learning_rate": 7.822403846153847e-06,
|
| 273589 |
+
"loss": 0.5916,
|
| 273590 |
+
"step": 136935
|
| 273591 |
+
},
|
| 273592 |
+
{
|
| 273593 |
+
"epoch": 1095.41,
|
| 273594 |
+
"learning_rate": 7.822323717948718e-06,
|
| 273595 |
+
"loss": 0.9323,
|
| 273596 |
+
"step": 136940
|
| 273597 |
+
},
|
| 273598 |
+
{
|
| 273599 |
+
"epoch": 1095.45,
|
| 273600 |
+
"learning_rate": 7.82224358974359e-06,
|
| 273601 |
+
"loss": 0.3015,
|
| 273602 |
+
"step": 136945
|
| 273603 |
+
},
|
| 273604 |
+
{
|
| 273605 |
+
"epoch": 1095.49,
|
| 273606 |
+
"learning_rate": 7.822163461538463e-06,
|
| 273607 |
+
"loss": 0.3338,
|
| 273608 |
+
"step": 136950
|
| 273609 |
+
},
|
| 273610 |
+
{
|
| 273611 |
+
"epoch": 1095.53,
|
| 273612 |
+
"learning_rate": 7.822083333333334e-06,
|
| 273613 |
+
"loss": 0.3108,
|
| 273614 |
+
"step": 136955
|
| 273615 |
+
},
|
| 273616 |
+
{
|
| 273617 |
+
"epoch": 1095.57,
|
| 273618 |
+
"learning_rate": 7.822003205128206e-06,
|
| 273619 |
+
"loss": 0.4635,
|
| 273620 |
+
"step": 136960
|
| 273621 |
+
},
|
| 273622 |
+
{
|
| 273623 |
+
"epoch": 1095.61,
|
| 273624 |
+
"learning_rate": 7.821923076923079e-06,
|
| 273625 |
+
"loss": 1.0018,
|
| 273626 |
+
"step": 136965
|
| 273627 |
+
},
|
| 273628 |
+
{
|
| 273629 |
+
"epoch": 1095.65,
|
| 273630 |
+
"learning_rate": 7.82184294871795e-06,
|
| 273631 |
+
"loss": 0.3006,
|
| 273632 |
+
"step": 136970
|
| 273633 |
+
},
|
| 273634 |
+
{
|
| 273635 |
+
"epoch": 1095.69,
|
| 273636 |
+
"learning_rate": 7.821762820512821e-06,
|
| 273637 |
+
"loss": 0.3277,
|
| 273638 |
+
"step": 136975
|
| 273639 |
+
},
|
| 273640 |
+
{
|
| 273641 |
+
"epoch": 1095.73,
|
| 273642 |
+
"learning_rate": 7.821682692307693e-06,
|
| 273643 |
+
"loss": 0.2904,
|
| 273644 |
+
"step": 136980
|
| 273645 |
+
},
|
| 273646 |
+
{
|
| 273647 |
+
"epoch": 1095.76,
|
| 273648 |
+
"learning_rate": 7.821602564102566e-06,
|
| 273649 |
+
"loss": 0.5041,
|
| 273650 |
+
"step": 136985
|
| 273651 |
+
},
|
| 273652 |
+
{
|
| 273653 |
+
"epoch": 1095.8,
|
| 273654 |
+
"learning_rate": 7.821522435897435e-06,
|
| 273655 |
+
"loss": 1.0441,
|
| 273656 |
+
"step": 136990
|
| 273657 |
+
},
|
| 273658 |
+
{
|
| 273659 |
+
"epoch": 1095.84,
|
| 273660 |
+
"learning_rate": 7.821442307692308e-06,
|
| 273661 |
+
"loss": 0.3226,
|
| 273662 |
+
"step": 136995
|
| 273663 |
+
},
|
| 273664 |
+
{
|
| 273665 |
+
"epoch": 1095.88,
|
| 273666 |
+
"learning_rate": 7.82136217948718e-06,
|
| 273667 |
+
"loss": 0.3103,
|
| 273668 |
+
"step": 137000
|
| 273669 |
+
},
|
| 273670 |
+
{
|
| 273671 |
+
"epoch": 1095.92,
|
| 273672 |
+
"learning_rate": 7.821282051282051e-06,
|
| 273673 |
+
"loss": 0.2987,
|
| 273674 |
+
"step": 137005
|
| 273675 |
+
},
|
| 273676 |
+
{
|
| 273677 |
+
"epoch": 1095.96,
|
| 273678 |
+
"learning_rate": 7.821201923076923e-06,
|
| 273679 |
+
"loss": 0.4378,
|
| 273680 |
+
"step": 137010
|
| 273681 |
+
},
|
| 273682 |
+
{
|
| 273683 |
+
"epoch": 1096.0,
|
| 273684 |
+
"eval_loss": 0.35713261365890503,
|
| 273685 |
+
"eval_runtime": 36.8635,
|
| 273686 |
+
"eval_samples_per_second": 22.868,
|
| 273687 |
+
"eval_steps_per_second": 0.732,
|
| 273688 |
+
"eval_wer": 0.17508296061174433,
|
| 273689 |
+
"step": 137014
|
| 273690 |
+
},
|
| 273691 |
+
{
|
| 273692 |
+
"epoch": 1104.01,
|
| 273693 |
+
"learning_rate": 7.821121794871796e-06,
|
| 273694 |
+
"loss": 0.4439,
|
| 273695 |
+
"step": 137015
|
| 273696 |
+
},
|
| 273697 |
+
{
|
| 273698 |
+
"epoch": 1104.05,
|
| 273699 |
+
"learning_rate": 7.821041666666667e-06,
|
| 273700 |
+
"loss": 0.3557,
|
| 273701 |
+
"step": 137020
|
| 273702 |
+
},
|
| 273703 |
+
{
|
| 273704 |
+
"epoch": 1104.09,
|
| 273705 |
+
"learning_rate": 7.820961538461538e-06,
|
| 273706 |
+
"loss": 0.2562,
|
| 273707 |
+
"step": 137025
|
| 273708 |
+
},
|
| 273709 |
+
{
|
| 273710 |
+
"epoch": 1104.13,
|
| 273711 |
+
"learning_rate": 7.820881410256411e-06,
|
| 273712 |
+
"loss": 0.3016,
|
| 273713 |
+
"step": 137030
|
| 273714 |
+
},
|
| 273715 |
+
{
|
| 273716 |
+
"epoch": 1104.17,
|
| 273717 |
+
"learning_rate": 7.820801282051283e-06,
|
| 273718 |
+
"loss": 0.5951,
|
| 273719 |
+
"step": 137035
|
| 273720 |
+
},
|
| 273721 |
+
{
|
| 273722 |
+
"epoch": 1104.21,
|
| 273723 |
+
"learning_rate": 7.820721153846154e-06,
|
| 273724 |
+
"loss": 1.1195,
|
| 273725 |
+
"step": 137040
|
| 273726 |
+
},
|
| 273727 |
+
{
|
| 273728 |
+
"epoch": 1104.25,
|
| 273729 |
+
"learning_rate": 7.820641025641025e-06,
|
| 273730 |
+
"loss": 0.3118,
|
| 273731 |
+
"step": 137045
|
| 273732 |
+
},
|
| 273733 |
+
{
|
| 273734 |
+
"epoch": 1104.29,
|
| 273735 |
+
"learning_rate": 7.820560897435898e-06,
|
| 273736 |
+
"loss": 0.3051,
|
| 273737 |
+
"step": 137050
|
| 273738 |
+
},
|
| 273739 |
+
{
|
| 273740 |
+
"epoch": 1104.33,
|
| 273741 |
+
"learning_rate": 7.82048076923077e-06,
|
| 273742 |
+
"loss": 0.3043,
|
| 273743 |
+
"step": 137055
|
| 273744 |
+
},
|
| 273745 |
+
{
|
| 273746 |
+
"epoch": 1104.37,
|
| 273747 |
+
"learning_rate": 7.820400641025641e-06,
|
| 273748 |
+
"loss": 0.5864,
|
| 273749 |
+
"step": 137060
|
| 273750 |
+
},
|
| 273751 |
+
{
|
| 273752 |
+
"epoch": 1104.41,
|
| 273753 |
+
"learning_rate": 7.820320512820514e-06,
|
| 273754 |
+
"loss": 1.057,
|
| 273755 |
+
"step": 137065
|
| 273756 |
+
},
|
| 273757 |
+
{
|
| 273758 |
+
"epoch": 1104.45,
|
| 273759 |
+
"learning_rate": 7.820240384615386e-06,
|
| 273760 |
+
"loss": 0.2954,
|
| 273761 |
+
"step": 137070
|
| 273762 |
+
},
|
| 273763 |
+
{
|
| 273764 |
+
"epoch": 1104.49,
|
| 273765 |
+
"learning_rate": 7.820160256410257e-06,
|
| 273766 |
+
"loss": 0.2967,
|
| 273767 |
+
"step": 137075
|
| 273768 |
+
},
|
| 273769 |
+
{
|
| 273770 |
+
"epoch": 1104.53,
|
| 273771 |
+
"learning_rate": 7.820080128205128e-06,
|
| 273772 |
+
"loss": 0.2834,
|
| 273773 |
+
"step": 137080
|
| 273774 |
+
},
|
| 273775 |
+
{
|
| 273776 |
+
"epoch": 1104.57,
|
| 273777 |
+
"learning_rate": 7.820000000000001e-06,
|
| 273778 |
+
"loss": 0.5314,
|
| 273779 |
+
"step": 137085
|
| 273780 |
+
},
|
| 273781 |
+
{
|
| 273782 |
+
"epoch": 1104.61,
|
| 273783 |
+
"learning_rate": 7.819919871794873e-06,
|
| 273784 |
+
"loss": 1.1084,
|
| 273785 |
+
"step": 137090
|
| 273786 |
+
},
|
| 273787 |
+
{
|
| 273788 |
+
"epoch": 1104.65,
|
| 273789 |
+
"learning_rate": 7.819839743589744e-06,
|
| 273790 |
+
"loss": 0.3065,
|
| 273791 |
+
"step": 137095
|
| 273792 |
+
},
|
| 273793 |
+
{
|
| 273794 |
+
"epoch": 1104.69,
|
| 273795 |
+
"learning_rate": 7.819759615384615e-06,
|
| 273796 |
+
"loss": 0.2879,
|
| 273797 |
+
"step": 137100
|
| 273798 |
+
},
|
| 273799 |
+
{
|
| 273800 |
+
"epoch": 1104.73,
|
| 273801 |
+
"learning_rate": 7.819679487179488e-06,
|
| 273802 |
+
"loss": 0.323,
|
| 273803 |
+
"step": 137105
|
| 273804 |
+
},
|
| 273805 |
+
{
|
| 273806 |
+
"epoch": 1104.77,
|
| 273807 |
+
"learning_rate": 7.81959935897436e-06,
|
| 273808 |
+
"loss": 0.5632,
|
| 273809 |
+
"step": 137110
|
| 273810 |
+
},
|
| 273811 |
+
{
|
| 273812 |
+
"epoch": 1104.81,
|
| 273813 |
+
"learning_rate": 7.819519230769231e-06,
|
| 273814 |
+
"loss": 1.0946,
|
| 273815 |
+
"step": 137115
|
| 273816 |
+
},
|
| 273817 |
+
{
|
| 273818 |
+
"epoch": 1104.85,
|
| 273819 |
+
"learning_rate": 7.819439102564104e-06,
|
| 273820 |
+
"loss": 0.2465,
|
| 273821 |
+
"step": 137120
|
| 273822 |
+
},
|
| 273823 |
+
{
|
| 273824 |
+
"epoch": 1104.89,
|
| 273825 |
+
"learning_rate": 7.819358974358976e-06,
|
| 273826 |
+
"loss": 0.2458,
|
| 273827 |
+
"step": 137125
|
| 273828 |
+
},
|
| 273829 |
+
{
|
| 273830 |
+
"epoch": 1104.93,
|
| 273831 |
+
"learning_rate": 7.819278846153847e-06,
|
| 273832 |
+
"loss": 0.3166,
|
| 273833 |
+
"step": 137130
|
| 273834 |
+
},
|
| 273835 |
+
{
|
| 273836 |
+
"epoch": 1104.97,
|
| 273837 |
+
"learning_rate": 7.819198717948718e-06,
|
| 273838 |
+
"loss": 0.5424,
|
| 273839 |
+
"step": 137135
|
| 273840 |
+
},
|
| 273841 |
+
{
|
| 273842 |
+
"epoch": 1105.0,
|
| 273843 |
+
"eval_loss": 0.3176847994327545,
|
| 273844 |
+
"eval_runtime": 37.4683,
|
| 273845 |
+
"eval_samples_per_second": 22.499,
|
| 273846 |
+
"eval_steps_per_second": 0.721,
|
| 273847 |
+
"eval_wer": 0.1728868984147295,
|
| 273848 |
+
"step": 137138
|
| 273849 |
+
},
|
| 273850 |
+
{
|
| 273851 |
+
"epoch": 1097.02,
|
| 273852 |
+
"learning_rate": 7.819118589743591e-06,
|
| 273853 |
+
"loss": 0.4155,
|
| 273854 |
+
"step": 137140
|
| 273855 |
+
},
|
| 273856 |
+
{
|
| 273857 |
+
"epoch": 1097.06,
|
| 273858 |
+
"learning_rate": 7.819038461538461e-06,
|
| 273859 |
+
"loss": 0.3272,
|
| 273860 |
+
"step": 137145
|
| 273861 |
+
},
|
| 273862 |
+
{
|
| 273863 |
+
"epoch": 1097.1,
|
| 273864 |
+
"learning_rate": 7.818958333333334e-06,
|
| 273865 |
+
"loss": 0.2476,
|
| 273866 |
+
"step": 137150
|
| 273867 |
+
},
|
| 273868 |
+
{
|
| 273869 |
+
"epoch": 1097.14,
|
| 273870 |
+
"learning_rate": 7.818878205128205e-06,
|
| 273871 |
+
"loss": 0.3344,
|
| 273872 |
+
"step": 137155
|
| 273873 |
+
},
|
| 273874 |
+
{
|
| 273875 |
+
"epoch": 1097.18,
|
| 273876 |
+
"learning_rate": 7.818798076923077e-06,
|
| 273877 |
+
"loss": 0.5918,
|
| 273878 |
+
"step": 137160
|
| 273879 |
+
},
|
| 273880 |
+
{
|
| 273881 |
+
"epoch": 1097.22,
|
| 273882 |
+
"learning_rate": 7.81871794871795e-06,
|
| 273883 |
+
"loss": 0.9623,
|
| 273884 |
+
"step": 137165
|
| 273885 |
+
},
|
| 273886 |
+
{
|
| 273887 |
+
"epoch": 1097.26,
|
| 273888 |
+
"learning_rate": 7.818637820512821e-06,
|
| 273889 |
+
"loss": 0.2835,
|
| 273890 |
+
"step": 137170
|
| 273891 |
+
},
|
| 273892 |
+
{
|
| 273893 |
+
"epoch": 1097.3,
|
| 273894 |
+
"learning_rate": 7.818557692307693e-06,
|
| 273895 |
+
"loss": 0.2989,
|
| 273896 |
+
"step": 137175
|
| 273897 |
+
},
|
| 273898 |
+
{
|
| 273899 |
+
"epoch": 1097.34,
|
| 273900 |
+
"learning_rate": 7.818477564102564e-06,
|
| 273901 |
+
"loss": 0.3434,
|
| 273902 |
+
"step": 137180
|
| 273903 |
+
},
|
| 273904 |
+
{
|
| 273905 |
+
"epoch": 1097.38,
|
| 273906 |
+
"learning_rate": 7.818397435897437e-06,
|
| 273907 |
+
"loss": 0.6796,
|
| 273908 |
+
"step": 137185
|
| 273909 |
+
},
|
| 273910 |
+
{
|
| 273911 |
+
"epoch": 1097.42,
|
| 273912 |
+
"learning_rate": 7.818317307692308e-06,
|
| 273913 |
+
"loss": 1.0891,
|
| 273914 |
+
"step": 137190
|
| 273915 |
+
},
|
| 273916 |
+
{
|
| 273917 |
+
"epoch": 1097.46,
|
| 273918 |
+
"learning_rate": 7.81823717948718e-06,
|
| 273919 |
+
"loss": 0.316,
|
| 273920 |
+
"step": 137195
|
| 273921 |
+
},
|
| 273922 |
+
{
|
| 273923 |
+
"epoch": 1097.5,
|
| 273924 |
+
"learning_rate": 7.818157051282051e-06,
|
| 273925 |
+
"loss": 0.2804,
|
| 273926 |
+
"step": 137200
|
| 273927 |
+
},
|
| 273928 |
+
{
|
| 273929 |
+
"epoch": 1097.54,
|
| 273930 |
+
"learning_rate": 7.818076923076924e-06,
|
| 273931 |
+
"loss": 0.3776,
|
| 273932 |
+
"step": 137205
|
| 273933 |
+
},
|
| 273934 |
+
{
|
| 273935 |
+
"epoch": 1097.58,
|
| 273936 |
+
"learning_rate": 7.817996794871795e-06,
|
| 273937 |
+
"loss": 0.4947,
|
| 273938 |
+
"step": 137210
|
| 273939 |
+
},
|
| 273940 |
+
{
|
| 273941 |
+
"epoch": 1097.62,
|
| 273942 |
+
"learning_rate": 7.817916666666667e-06,
|
| 273943 |
+
"loss": 0.9031,
|
| 273944 |
+
"step": 137215
|
| 273945 |
+
},
|
| 273946 |
+
{
|
| 273947 |
+
"epoch": 1097.66,
|
| 273948 |
+
"learning_rate": 7.81783653846154e-06,
|
| 273949 |
+
"loss": 0.2662,
|
| 273950 |
+
"step": 137220
|
| 273951 |
+
},
|
| 273952 |
+
{
|
| 273953 |
+
"epoch": 1097.7,
|
| 273954 |
+
"learning_rate": 7.817756410256411e-06,
|
| 273955 |
+
"loss": 0.2473,
|
| 273956 |
+
"step": 137225
|
| 273957 |
+
},
|
| 273958 |
+
{
|
| 273959 |
+
"epoch": 1097.74,
|
| 273960 |
+
"learning_rate": 7.817676282051283e-06,
|
| 273961 |
+
"loss": 0.3746,
|
| 273962 |
+
"step": 137230
|
| 273963 |
+
},
|
| 273964 |
+
{
|
| 273965 |
+
"epoch": 1097.78,
|
| 273966 |
+
"learning_rate": 7.817596153846154e-06,
|
| 273967 |
+
"loss": 0.526,
|
| 273968 |
+
"step": 137235
|
| 273969 |
+
},
|
| 273970 |
+
{
|
| 273971 |
+
"epoch": 1097.82,
|
| 273972 |
+
"learning_rate": 7.817516025641027e-06,
|
| 273973 |
+
"loss": 0.9866,
|
| 273974 |
+
"step": 137240
|
| 273975 |
+
},
|
| 273976 |
+
{
|
| 273977 |
+
"epoch": 1097.86,
|
| 273978 |
+
"learning_rate": 7.817435897435898e-06,
|
| 273979 |
+
"loss": 0.2647,
|
| 273980 |
+
"step": 137245
|
| 273981 |
+
},
|
| 273982 |
+
{
|
| 273983 |
+
"epoch": 1097.9,
|
| 273984 |
+
"learning_rate": 7.81735576923077e-06,
|
| 273985 |
+
"loss": 0.321,
|
| 273986 |
+
"step": 137250
|
| 273987 |
+
},
|
| 273988 |
+
{
|
| 273989 |
+
"epoch": 1097.94,
|
| 273990 |
+
"learning_rate": 7.817275641025641e-06,
|
| 273991 |
+
"loss": 0.3774,
|
| 273992 |
+
"step": 137255
|
| 273993 |
+
},
|
| 273994 |
+
{
|
| 273995 |
+
"epoch": 1097.98,
|
| 273996 |
+
"learning_rate": 7.817195512820514e-06,
|
| 273997 |
+
"loss": 0.7104,
|
| 273998 |
+
"step": 137260
|
| 273999 |
+
},
|
| 274000 |
+
{
|
| 274001 |
+
"epoch": 1098.0,
|
| 274002 |
+
"eval_loss": 0.3407905697822571,
|
| 274003 |
+
"eval_runtime": 36.8965,
|
| 274004 |
+
"eval_samples_per_second": 22.848,
|
| 274005 |
+
"eval_steps_per_second": 0.732,
|
| 274006 |
+
"eval_wer": 0.1787237226004339,
|
| 274007 |
+
"step": 137263
|
| 274008 |
}
|
| 274009 |
],
|
| 274010 |
+
"max_steps": 625000,
|
| 274011 |
"num_train_epochs": 5000,
|
| 274012 |
+
"total_flos": 3.862579605221591e+20,
|
| 274013 |
"trial_name": null,
|
| 274014 |
"trial_params": null
|
| 274015 |
}
|
model-bin/finetune/base/{checkpoint-136640 → checkpoint-137263}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630183506.9783158/events.out.tfevents.1630183506.86bb0ddabf9b.4092.301
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f35090eab11daa3e1aaa1cdd2e3b06161ea437f886eda4d6f87146d1fc9765b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630183895.283262/events.out.tfevents.1630183895.86bb0ddabf9b.4092.303
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bee463dc96f9f228e8a299634990ca855989b8fdbf6f2d084130da97968cc32b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630184282.1676457/events.out.tfevents.1630184282.86bb0ddabf9b.4092.305
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:381d68cdc2224b9fb4e5b358e9e3551d1114af909cc2b5ce11aa395cdc46cdb5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630184680.4007196/events.out.tfevents.1630184680.86bb0ddabf9b.4092.307
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:984f394b3282ff8e326c39c4a8fcc850865e7501501eb0252678960badccc42e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630185067.7506618/events.out.tfevents.1630185067.86bb0ddabf9b.4092.309
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ffbfe384781f3b1ed7c78506f148d1c856da24430bb8499644abfc5c7089c340
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630183506.86bb0ddabf9b.4092.300
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7e5b058d3aa46c5ef537c21fc9b1b7829af9c7fbfbd28b75c1ae4b1383232372
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630183895.86bb0ddabf9b.4092.302
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b81a04f854aa7b0b1825a870593c2bef11f39a8b8a01e05a8be359908737a928
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630184282.86bb0ddabf9b.4092.304
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b974654456ce9dd4a2a376e54ee57e19d07175fcc46044dc8b09dbeac3f94890
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630184680.86bb0ddabf9b.4092.306
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1298af849e411874290babb846a56dc0d824299eecf653480de6f6b512db640
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630185067.86bb0ddabf9b.4092.308
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68a56fddc25ad06b2577262e34c27b878433883fa69890c8aef72e818c81e75d
|
| 3 |
+
size 8622
|