"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630257516.869323/events.out.tfevents.1630257516.cc93b136ebf5.1086.239 +3 -0
- model-bin/finetune/base/log/1630257947.7075908/events.out.tfevents.1630257947.cc93b136ebf5.1086.241 +3 -0
- model-bin/finetune/base/log/1630258385.8405457/events.out.tfevents.1630258385.cc93b136ebf5.1086.243 +3 -0
- model-bin/finetune/base/log/1630258817.5863304/events.out.tfevents.1630258817.cc93b136ebf5.1086.245 +3 -0
- model-bin/finetune/base/log/1630259247.5041316/events.out.tfevents.1630259247.cc93b136ebf5.1086.247 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630257516.cc93b136ebf5.1086.238 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630257947.cc93b136ebf5.1086.240 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630258385.cc93b136ebf5.1086.242 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630258817.cc93b136ebf5.1086.244 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630259247.cc93b136ebf5.1086.246 +3 -0
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/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 sha256:74ba9b2dfdf6dee25d004cb565b958e98dcf9c4fbd59c57c5467560d1f4d0fa0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/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:bf7e26fb4cb3b91846741a8ca38d407f10f0f1e4bc6d6d1c322f960abd8cf940
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71bad233cfa0a1d396b5ebc324d1525576abd564039e669d5d2f749724b873a3
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/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 sha256:53047867f5dbec0185f704d1730a6309e0150cb39a35badc8cf7348be089b5bb
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/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:e6d2ae467a50189b89589e279f888de8f1fcdfcd87203d6852aac1cd1477f181
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -295215,11 +295215,806 @@
|
|
| 295215 |
"eval_steps_per_second": 0.642,
|
| 295216 |
"eval_wer": 0.17948529411764705,
|
| 295217 |
"step": 153934
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 295218 |
}
|
| 295219 |
],
|
| 295220 |
-
"max_steps":
|
| 295221 |
"num_train_epochs": 5000,
|
| 295222 |
-
"total_flos": 4.
|
| 295223 |
"trial_name": null,
|
| 295224 |
"trial_params": null
|
| 295225 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
+
"epoch": 1246.0,
|
| 5 |
+
"global_step": 154557,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 295215 |
"eval_steps_per_second": 0.642,
|
| 295216 |
"eval_wer": 0.17948529411764705,
|
| 295217 |
"step": 153934
|
| 295218 |
+
},
|
| 295219 |
+
{
|
| 295220 |
+
"epoch": 1231.01,
|
| 295221 |
+
"learning_rate": 7.530306946688208e-06,
|
| 295222 |
+
"loss": 0.3618,
|
| 295223 |
+
"step": 153935
|
| 295224 |
+
},
|
| 295225 |
+
{
|
| 295226 |
+
"epoch": 1231.05,
|
| 295227 |
+
"learning_rate": 7.530226171243943e-06,
|
| 295228 |
+
"loss": 0.2959,
|
| 295229 |
+
"step": 153940
|
| 295230 |
+
},
|
| 295231 |
+
{
|
| 295232 |
+
"epoch": 1231.09,
|
| 295233 |
+
"learning_rate": 7.530145395799678e-06,
|
| 295234 |
+
"loss": 0.3115,
|
| 295235 |
+
"step": 153945
|
| 295236 |
+
},
|
| 295237 |
+
{
|
| 295238 |
+
"epoch": 1231.13,
|
| 295239 |
+
"learning_rate": 7.530064620355413e-06,
|
| 295240 |
+
"loss": 0.3225,
|
| 295241 |
+
"step": 153950
|
| 295242 |
+
},
|
| 295243 |
+
{
|
| 295244 |
+
"epoch": 1231.17,
|
| 295245 |
+
"learning_rate": 7.529983844911148e-06,
|
| 295246 |
+
"loss": 0.5211,
|
| 295247 |
+
"step": 153955
|
| 295248 |
+
},
|
| 295249 |
+
{
|
| 295250 |
+
"epoch": 1231.21,
|
| 295251 |
+
"learning_rate": 7.5299030694668826e-06,
|
| 295252 |
+
"loss": 1.0981,
|
| 295253 |
+
"step": 153960
|
| 295254 |
+
},
|
| 295255 |
+
{
|
| 295256 |
+
"epoch": 1231.25,
|
| 295257 |
+
"learning_rate": 7.5298222940226176e-06,
|
| 295258 |
+
"loss": 0.2782,
|
| 295259 |
+
"step": 153965
|
| 295260 |
+
},
|
| 295261 |
+
{
|
| 295262 |
+
"epoch": 1231.29,
|
| 295263 |
+
"learning_rate": 7.5297415185783525e-06,
|
| 295264 |
+
"loss": 0.289,
|
| 295265 |
+
"step": 153970
|
| 295266 |
+
},
|
| 295267 |
+
{
|
| 295268 |
+
"epoch": 1231.33,
|
| 295269 |
+
"learning_rate": 7.5296607431340875e-06,
|
| 295270 |
+
"loss": 0.3406,
|
| 295271 |
+
"step": 153975
|
| 295272 |
+
},
|
| 295273 |
+
{
|
| 295274 |
+
"epoch": 1231.37,
|
| 295275 |
+
"learning_rate": 7.5295799676898225e-06,
|
| 295276 |
+
"loss": 0.4844,
|
| 295277 |
+
"step": 153980
|
| 295278 |
+
},
|
| 295279 |
+
{
|
| 295280 |
+
"epoch": 1231.41,
|
| 295281 |
+
"learning_rate": 7.529499192245558e-06,
|
| 295282 |
+
"loss": 1.0779,
|
| 295283 |
+
"step": 153985
|
| 295284 |
+
},
|
| 295285 |
+
{
|
| 295286 |
+
"epoch": 1231.45,
|
| 295287 |
+
"learning_rate": 7.529418416801293e-06,
|
| 295288 |
+
"loss": 0.3267,
|
| 295289 |
+
"step": 153990
|
| 295290 |
+
},
|
| 295291 |
+
{
|
| 295292 |
+
"epoch": 1231.49,
|
| 295293 |
+
"learning_rate": 7.529337641357028e-06,
|
| 295294 |
+
"loss": 0.2785,
|
| 295295 |
+
"step": 153995
|
| 295296 |
+
},
|
| 295297 |
+
{
|
| 295298 |
+
"epoch": 1231.53,
|
| 295299 |
+
"learning_rate": 7.529256865912763e-06,
|
| 295300 |
+
"loss": 0.2886,
|
| 295301 |
+
"step": 154000
|
| 295302 |
+
},
|
| 295303 |
+
{
|
| 295304 |
+
"epoch": 1231.57,
|
| 295305 |
+
"learning_rate": 7.529176090468498e-06,
|
| 295306 |
+
"loss": 0.4818,
|
| 295307 |
+
"step": 154005
|
| 295308 |
+
},
|
| 295309 |
+
{
|
| 295310 |
+
"epoch": 1231.61,
|
| 295311 |
+
"learning_rate": 7.529095315024233e-06,
|
| 295312 |
+
"loss": 1.0733,
|
| 295313 |
+
"step": 154010
|
| 295314 |
+
},
|
| 295315 |
+
{
|
| 295316 |
+
"epoch": 1231.65,
|
| 295317 |
+
"learning_rate": 7.529014539579968e-06,
|
| 295318 |
+
"loss": 0.2831,
|
| 295319 |
+
"step": 154015
|
| 295320 |
+
},
|
| 295321 |
+
{
|
| 295322 |
+
"epoch": 1231.69,
|
| 295323 |
+
"learning_rate": 7.528933764135703e-06,
|
| 295324 |
+
"loss": 0.3198,
|
| 295325 |
+
"step": 154020
|
| 295326 |
+
},
|
| 295327 |
+
{
|
| 295328 |
+
"epoch": 1231.73,
|
| 295329 |
+
"learning_rate": 7.528852988691438e-06,
|
| 295330 |
+
"loss": 0.3014,
|
| 295331 |
+
"step": 154025
|
| 295332 |
+
},
|
| 295333 |
+
{
|
| 295334 |
+
"epoch": 1231.77,
|
| 295335 |
+
"learning_rate": 7.528772213247173e-06,
|
| 295336 |
+
"loss": 0.5665,
|
| 295337 |
+
"step": 154030
|
| 295338 |
+
},
|
| 295339 |
+
{
|
| 295340 |
+
"epoch": 1231.81,
|
| 295341 |
+
"learning_rate": 7.528691437802908e-06,
|
| 295342 |
+
"loss": 1.0155,
|
| 295343 |
+
"step": 154035
|
| 295344 |
+
},
|
| 295345 |
+
{
|
| 295346 |
+
"epoch": 1231.85,
|
| 295347 |
+
"learning_rate": 7.528610662358644e-06,
|
| 295348 |
+
"loss": 0.3132,
|
| 295349 |
+
"step": 154040
|
| 295350 |
+
},
|
| 295351 |
+
{
|
| 295352 |
+
"epoch": 1231.89,
|
| 295353 |
+
"learning_rate": 7.528529886914378e-06,
|
| 295354 |
+
"loss": 0.2465,
|
| 295355 |
+
"step": 154045
|
| 295356 |
+
},
|
| 295357 |
+
{
|
| 295358 |
+
"epoch": 1231.93,
|
| 295359 |
+
"learning_rate": 7.528449111470114e-06,
|
| 295360 |
+
"loss": 0.3236,
|
| 295361 |
+
"step": 154050
|
| 295362 |
+
},
|
| 295363 |
+
{
|
| 295364 |
+
"epoch": 1231.97,
|
| 295365 |
+
"learning_rate": 7.528368336025849e-06,
|
| 295366 |
+
"loss": 0.4988,
|
| 295367 |
+
"step": 154055
|
| 295368 |
+
},
|
| 295369 |
+
{
|
| 295370 |
+
"epoch": 1232.0,
|
| 295371 |
+
"eval_loss": 0.31069421768188477,
|
| 295372 |
+
"eval_runtime": 41.1007,
|
| 295373 |
+
"eval_samples_per_second": 20.559,
|
| 295374 |
+
"eval_steps_per_second": 0.657,
|
| 295375 |
+
"eval_wer": 0.17686326291079812,
|
| 295376 |
+
"step": 154059
|
| 295377 |
+
},
|
| 295378 |
+
{
|
| 295379 |
+
"epoch": 1232.01,
|
| 295380 |
+
"learning_rate": 7.528287560581584e-06,
|
| 295381 |
+
"loss": 0.4309,
|
| 295382 |
+
"step": 154060
|
| 295383 |
+
},
|
| 295384 |
+
{
|
| 295385 |
+
"epoch": 1232.05,
|
| 295386 |
+
"learning_rate": 7.528206785137319e-06,
|
| 295387 |
+
"loss": 0.2709,
|
| 295388 |
+
"step": 154065
|
| 295389 |
+
},
|
| 295390 |
+
{
|
| 295391 |
+
"epoch": 1232.09,
|
| 295392 |
+
"learning_rate": 7.528126009693054e-06,
|
| 295393 |
+
"loss": 0.3108,
|
| 295394 |
+
"step": 154070
|
| 295395 |
+
},
|
| 295396 |
+
{
|
| 295397 |
+
"epoch": 1232.13,
|
| 295398 |
+
"learning_rate": 7.528045234248789e-06,
|
| 295399 |
+
"loss": 0.3349,
|
| 295400 |
+
"step": 154075
|
| 295401 |
+
},
|
| 295402 |
+
{
|
| 295403 |
+
"epoch": 1232.17,
|
| 295404 |
+
"learning_rate": 7.527964458804524e-06,
|
| 295405 |
+
"loss": 0.3979,
|
| 295406 |
+
"step": 154080
|
| 295407 |
+
},
|
| 295408 |
+
{
|
| 295409 |
+
"epoch": 1232.21,
|
| 295410 |
+
"learning_rate": 7.527883683360259e-06,
|
| 295411 |
+
"loss": 1.162,
|
| 295412 |
+
"step": 154085
|
| 295413 |
+
},
|
| 295414 |
+
{
|
| 295415 |
+
"epoch": 1232.25,
|
| 295416 |
+
"learning_rate": 7.527802907915994e-06,
|
| 295417 |
+
"loss": 0.2879,
|
| 295418 |
+
"step": 154090
|
| 295419 |
+
},
|
| 295420 |
+
{
|
| 295421 |
+
"epoch": 1232.29,
|
| 295422 |
+
"learning_rate": 7.527722132471729e-06,
|
| 295423 |
+
"loss": 0.2757,
|
| 295424 |
+
"step": 154095
|
| 295425 |
+
},
|
| 295426 |
+
{
|
| 295427 |
+
"epoch": 1232.33,
|
| 295428 |
+
"learning_rate": 7.527641357027464e-06,
|
| 295429 |
+
"loss": 0.2955,
|
| 295430 |
+
"step": 154100
|
| 295431 |
+
},
|
| 295432 |
+
{
|
| 295433 |
+
"epoch": 1232.37,
|
| 295434 |
+
"learning_rate": 7.5275605815832e-06,
|
| 295435 |
+
"loss": 0.6014,
|
| 295436 |
+
"step": 154105
|
| 295437 |
+
},
|
| 295438 |
+
{
|
| 295439 |
+
"epoch": 1232.41,
|
| 295440 |
+
"learning_rate": 7.527479806138934e-06,
|
| 295441 |
+
"loss": 1.1433,
|
| 295442 |
+
"step": 154110
|
| 295443 |
+
},
|
| 295444 |
+
{
|
| 295445 |
+
"epoch": 1232.45,
|
| 295446 |
+
"learning_rate": 7.52739903069467e-06,
|
| 295447 |
+
"loss": 0.3034,
|
| 295448 |
+
"step": 154115
|
| 295449 |
+
},
|
| 295450 |
+
{
|
| 295451 |
+
"epoch": 1232.49,
|
| 295452 |
+
"learning_rate": 7.527318255250404e-06,
|
| 295453 |
+
"loss": 0.271,
|
| 295454 |
+
"step": 154120
|
| 295455 |
+
},
|
| 295456 |
+
{
|
| 295457 |
+
"epoch": 1232.53,
|
| 295458 |
+
"learning_rate": 7.52723747980614e-06,
|
| 295459 |
+
"loss": 0.2804,
|
| 295460 |
+
"step": 154125
|
| 295461 |
+
},
|
| 295462 |
+
{
|
| 295463 |
+
"epoch": 1232.57,
|
| 295464 |
+
"learning_rate": 7.527156704361875e-06,
|
| 295465 |
+
"loss": 0.4664,
|
| 295466 |
+
"step": 154130
|
| 295467 |
+
},
|
| 295468 |
+
{
|
| 295469 |
+
"epoch": 1232.61,
|
| 295470 |
+
"learning_rate": 7.52707592891761e-06,
|
| 295471 |
+
"loss": 1.0826,
|
| 295472 |
+
"step": 154135
|
| 295473 |
+
},
|
| 295474 |
+
{
|
| 295475 |
+
"epoch": 1232.65,
|
| 295476 |
+
"learning_rate": 7.526995153473345e-06,
|
| 295477 |
+
"loss": 0.2752,
|
| 295478 |
+
"step": 154140
|
| 295479 |
+
},
|
| 295480 |
+
{
|
| 295481 |
+
"epoch": 1232.69,
|
| 295482 |
+
"learning_rate": 7.52691437802908e-06,
|
| 295483 |
+
"loss": 0.2435,
|
| 295484 |
+
"step": 154145
|
| 295485 |
+
},
|
| 295486 |
+
{
|
| 295487 |
+
"epoch": 1232.73,
|
| 295488 |
+
"learning_rate": 7.526833602584815e-06,
|
| 295489 |
+
"loss": 0.3137,
|
| 295490 |
+
"step": 154150
|
| 295491 |
+
},
|
| 295492 |
+
{
|
| 295493 |
+
"epoch": 1232.77,
|
| 295494 |
+
"learning_rate": 7.52675282714055e-06,
|
| 295495 |
+
"loss": 0.4692,
|
| 295496 |
+
"step": 154155
|
| 295497 |
+
},
|
| 295498 |
+
{
|
| 295499 |
+
"epoch": 1232.81,
|
| 295500 |
+
"learning_rate": 7.5266720516962855e-06,
|
| 295501 |
+
"loss": 1.035,
|
| 295502 |
+
"step": 154160
|
| 295503 |
+
},
|
| 295504 |
+
{
|
| 295505 |
+
"epoch": 1232.85,
|
| 295506 |
+
"learning_rate": 7.52659127625202e-06,
|
| 295507 |
+
"loss": 0.3189,
|
| 295508 |
+
"step": 154165
|
| 295509 |
+
},
|
| 295510 |
+
{
|
| 295511 |
+
"epoch": 1232.89,
|
| 295512 |
+
"learning_rate": 7.5265105008077554e-06,
|
| 295513 |
+
"loss": 0.2449,
|
| 295514 |
+
"step": 154170
|
| 295515 |
+
},
|
| 295516 |
+
{
|
| 295517 |
+
"epoch": 1232.93,
|
| 295518 |
+
"learning_rate": 7.5264297253634896e-06,
|
| 295519 |
+
"loss": 0.3198,
|
| 295520 |
+
"step": 154175
|
| 295521 |
+
},
|
| 295522 |
+
{
|
| 295523 |
+
"epoch": 1232.97,
|
| 295524 |
+
"learning_rate": 7.526348949919225e-06,
|
| 295525 |
+
"loss": 0.5268,
|
| 295526 |
+
"step": 154180
|
| 295527 |
+
},
|
| 295528 |
+
{
|
| 295529 |
+
"epoch": 1233.0,
|
| 295530 |
+
"eval_loss": 0.5121481418609619,
|
| 295531 |
+
"eval_runtime": 41.2486,
|
| 295532 |
+
"eval_samples_per_second": 20.486,
|
| 295533 |
+
"eval_steps_per_second": 0.655,
|
| 295534 |
+
"eval_wer": 0.18070791282726342,
|
| 295535 |
+
"step": 154184
|
| 295536 |
+
},
|
| 295537 |
+
{
|
| 295538 |
+
"epoch": 1233.01,
|
| 295539 |
+
"learning_rate": 7.5262681744749595e-06,
|
| 295540 |
+
"loss": 0.4611,
|
| 295541 |
+
"step": 154185
|
| 295542 |
+
},
|
| 295543 |
+
{
|
| 295544 |
+
"epoch": 1233.05,
|
| 295545 |
+
"learning_rate": 7.526187399030695e-06,
|
| 295546 |
+
"loss": 0.247,
|
| 295547 |
+
"step": 154190
|
| 295548 |
+
},
|
| 295549 |
+
{
|
| 295550 |
+
"epoch": 1233.09,
|
| 295551 |
+
"learning_rate": 7.52610662358643e-06,
|
| 295552 |
+
"loss": 0.3827,
|
| 295553 |
+
"step": 154195
|
| 295554 |
+
},
|
| 295555 |
+
{
|
| 295556 |
+
"epoch": 1233.13,
|
| 295557 |
+
"learning_rate": 7.526025848142165e-06,
|
| 295558 |
+
"loss": 0.3205,
|
| 295559 |
+
"step": 154200
|
| 295560 |
+
},
|
| 295561 |
+
{
|
| 295562 |
+
"epoch": 1233.17,
|
| 295563 |
+
"learning_rate": 7.5259450726979e-06,
|
| 295564 |
+
"loss": 0.5641,
|
| 295565 |
+
"step": 154205
|
| 295566 |
+
},
|
| 295567 |
+
{
|
| 295568 |
+
"epoch": 1233.21,
|
| 295569 |
+
"learning_rate": 7.525864297253635e-06,
|
| 295570 |
+
"loss": 0.9883,
|
| 295571 |
+
"step": 154210
|
| 295572 |
+
},
|
| 295573 |
+
{
|
| 295574 |
+
"epoch": 1233.25,
|
| 295575 |
+
"learning_rate": 7.525783521809371e-06,
|
| 295576 |
+
"loss": 0.2971,
|
| 295577 |
+
"step": 154215
|
| 295578 |
+
},
|
| 295579 |
+
{
|
| 295580 |
+
"epoch": 1233.29,
|
| 295581 |
+
"learning_rate": 7.525702746365105e-06,
|
| 295582 |
+
"loss": 0.2764,
|
| 295583 |
+
"step": 154220
|
| 295584 |
+
},
|
| 295585 |
+
{
|
| 295586 |
+
"epoch": 1233.33,
|
| 295587 |
+
"learning_rate": 7.525621970920841e-06,
|
| 295588 |
+
"loss": 0.2942,
|
| 295589 |
+
"step": 154225
|
| 295590 |
+
},
|
| 295591 |
+
{
|
| 295592 |
+
"epoch": 1233.37,
|
| 295593 |
+
"learning_rate": 7.525541195476575e-06,
|
| 295594 |
+
"loss": 0.6157,
|
| 295595 |
+
"step": 154230
|
| 295596 |
+
},
|
| 295597 |
+
{
|
| 295598 |
+
"epoch": 1233.41,
|
| 295599 |
+
"learning_rate": 7.525460420032311e-06,
|
| 295600 |
+
"loss": 1.1009,
|
| 295601 |
+
"step": 154235
|
| 295602 |
+
},
|
| 295603 |
+
{
|
| 295604 |
+
"epoch": 1233.45,
|
| 295605 |
+
"learning_rate": 7.525379644588045e-06,
|
| 295606 |
+
"loss": 0.2929,
|
| 295607 |
+
"step": 154240
|
| 295608 |
+
},
|
| 295609 |
+
{
|
| 295610 |
+
"epoch": 1233.49,
|
| 295611 |
+
"learning_rate": 7.525298869143781e-06,
|
| 295612 |
+
"loss": 0.36,
|
| 295613 |
+
"step": 154245
|
| 295614 |
+
},
|
| 295615 |
+
{
|
| 295616 |
+
"epoch": 1233.53,
|
| 295617 |
+
"learning_rate": 7.525218093699515e-06,
|
| 295618 |
+
"loss": 0.631,
|
| 295619 |
+
"step": 154250
|
| 295620 |
+
},
|
| 295621 |
+
{
|
| 295622 |
+
"epoch": 1233.57,
|
| 295623 |
+
"learning_rate": 7.525137318255251e-06,
|
| 295624 |
+
"loss": 0.4253,
|
| 295625 |
+
"step": 154255
|
| 295626 |
+
},
|
| 295627 |
+
{
|
| 295628 |
+
"epoch": 1233.61,
|
| 295629 |
+
"learning_rate": 7.525056542810986e-06,
|
| 295630 |
+
"loss": 0.9945,
|
| 295631 |
+
"step": 154260
|
| 295632 |
+
},
|
| 295633 |
+
{
|
| 295634 |
+
"epoch": 1233.65,
|
| 295635 |
+
"learning_rate": 7.524975767366721e-06,
|
| 295636 |
+
"loss": 0.2619,
|
| 295637 |
+
"step": 154265
|
| 295638 |
+
},
|
| 295639 |
+
{
|
| 295640 |
+
"epoch": 1233.69,
|
| 295641 |
+
"learning_rate": 7.524894991922456e-06,
|
| 295642 |
+
"loss": 0.3313,
|
| 295643 |
+
"step": 154270
|
| 295644 |
+
},
|
| 295645 |
+
{
|
| 295646 |
+
"epoch": 1233.73,
|
| 295647 |
+
"learning_rate": 7.524814216478191e-06,
|
| 295648 |
+
"loss": 0.3285,
|
| 295649 |
+
"step": 154275
|
| 295650 |
+
},
|
| 295651 |
+
{
|
| 295652 |
+
"epoch": 1233.76,
|
| 295653 |
+
"learning_rate": 7.524733441033927e-06,
|
| 295654 |
+
"loss": 0.544,
|
| 295655 |
+
"step": 154280
|
| 295656 |
+
},
|
| 295657 |
+
{
|
| 295658 |
+
"epoch": 1233.8,
|
| 295659 |
+
"learning_rate": 7.524652665589661e-06,
|
| 295660 |
+
"loss": 0.8979,
|
| 295661 |
+
"step": 154285
|
| 295662 |
+
},
|
| 295663 |
+
{
|
| 295664 |
+
"epoch": 1233.84,
|
| 295665 |
+
"learning_rate": 7.524571890145397e-06,
|
| 295666 |
+
"loss": 0.2899,
|
| 295667 |
+
"step": 154290
|
| 295668 |
+
},
|
| 295669 |
+
{
|
| 295670 |
+
"epoch": 1233.88,
|
| 295671 |
+
"learning_rate": 7.524491114701131e-06,
|
| 295672 |
+
"loss": 0.3065,
|
| 295673 |
+
"step": 154295
|
| 295674 |
+
},
|
| 295675 |
+
{
|
| 295676 |
+
"epoch": 1233.92,
|
| 295677 |
+
"learning_rate": 7.524410339256867e-06,
|
| 295678 |
+
"loss": 0.3261,
|
| 295679 |
+
"step": 154300
|
| 295680 |
+
},
|
| 295681 |
+
{
|
| 295682 |
+
"epoch": 1233.96,
|
| 295683 |
+
"learning_rate": 7.524329563812601e-06,
|
| 295684 |
+
"loss": 0.5531,
|
| 295685 |
+
"step": 154305
|
| 295686 |
+
},
|
| 295687 |
+
{
|
| 295688 |
+
"epoch": 1234.0,
|
| 295689 |
+
"eval_loss": 0.3117029666900635,
|
| 295690 |
+
"eval_runtime": 41.4162,
|
| 295691 |
+
"eval_samples_per_second": 20.403,
|
| 295692 |
+
"eval_steps_per_second": 0.652,
|
| 295693 |
+
"eval_wer": 0.17106142761643023,
|
| 295694 |
+
"step": 154309
|
| 295695 |
+
},
|
| 295696 |
+
{
|
| 295697 |
+
"epoch": 1244.01,
|
| 295698 |
+
"learning_rate": 7.524248788368337e-06,
|
| 295699 |
+
"loss": 0.6507,
|
| 295700 |
+
"step": 154310
|
| 295701 |
+
},
|
| 295702 |
+
{
|
| 295703 |
+
"epoch": 1244.05,
|
| 295704 |
+
"learning_rate": 7.524168012924071e-06,
|
| 295705 |
+
"loss": 0.2557,
|
| 295706 |
+
"step": 154315
|
| 295707 |
+
},
|
| 295708 |
+
{
|
| 295709 |
+
"epoch": 1244.09,
|
| 295710 |
+
"learning_rate": 7.524087237479807e-06,
|
| 295711 |
+
"loss": 0.2584,
|
| 295712 |
+
"step": 154320
|
| 295713 |
+
},
|
| 295714 |
+
{
|
| 295715 |
+
"epoch": 1244.13,
|
| 295716 |
+
"learning_rate": 7.524006462035541e-06,
|
| 295717 |
+
"loss": 0.3013,
|
| 295718 |
+
"step": 154325
|
| 295719 |
+
},
|
| 295720 |
+
{
|
| 295721 |
+
"epoch": 1244.17,
|
| 295722 |
+
"learning_rate": 7.523925686591277e-06,
|
| 295723 |
+
"loss": 0.4878,
|
| 295724 |
+
"step": 154330
|
| 295725 |
+
},
|
| 295726 |
+
{
|
| 295727 |
+
"epoch": 1244.21,
|
| 295728 |
+
"learning_rate": 7.5238449111470126e-06,
|
| 295729 |
+
"loss": 1.1685,
|
| 295730 |
+
"step": 154335
|
| 295731 |
+
},
|
| 295732 |
+
{
|
| 295733 |
+
"epoch": 1244.25,
|
| 295734 |
+
"learning_rate": 7.523764135702747e-06,
|
| 295735 |
+
"loss": 0.2893,
|
| 295736 |
+
"step": 154340
|
| 295737 |
+
},
|
| 295738 |
+
{
|
| 295739 |
+
"epoch": 1244.29,
|
| 295740 |
+
"learning_rate": 7.5236833602584825e-06,
|
| 295741 |
+
"loss": 0.2858,
|
| 295742 |
+
"step": 154345
|
| 295743 |
+
},
|
| 295744 |
+
{
|
| 295745 |
+
"epoch": 1244.33,
|
| 295746 |
+
"learning_rate": 7.523602584814217e-06,
|
| 295747 |
+
"loss": 0.36,
|
| 295748 |
+
"step": 154350
|
| 295749 |
+
},
|
| 295750 |
+
{
|
| 295751 |
+
"epoch": 1244.37,
|
| 295752 |
+
"learning_rate": 7.5235218093699525e-06,
|
| 295753 |
+
"loss": 0.492,
|
| 295754 |
+
"step": 154355
|
| 295755 |
+
},
|
| 295756 |
+
{
|
| 295757 |
+
"epoch": 1244.41,
|
| 295758 |
+
"learning_rate": 7.523441033925687e-06,
|
| 295759 |
+
"loss": 1.1067,
|
| 295760 |
+
"step": 154360
|
| 295761 |
+
},
|
| 295762 |
+
{
|
| 295763 |
+
"epoch": 1244.45,
|
| 295764 |
+
"learning_rate": 7.5233602584814225e-06,
|
| 295765 |
+
"loss": 0.2341,
|
| 295766 |
+
"step": 154365
|
| 295767 |
+
},
|
| 295768 |
+
{
|
| 295769 |
+
"epoch": 1244.49,
|
| 295770 |
+
"learning_rate": 7.523279483037157e-06,
|
| 295771 |
+
"loss": 0.2546,
|
| 295772 |
+
"step": 154370
|
| 295773 |
+
},
|
| 295774 |
+
{
|
| 295775 |
+
"epoch": 1244.53,
|
| 295776 |
+
"learning_rate": 7.5231987075928925e-06,
|
| 295777 |
+
"loss": 0.2621,
|
| 295778 |
+
"step": 154375
|
| 295779 |
+
},
|
| 295780 |
+
{
|
| 295781 |
+
"epoch": 1244.57,
|
| 295782 |
+
"learning_rate": 7.523117932148627e-06,
|
| 295783 |
+
"loss": 0.4435,
|
| 295784 |
+
"step": 154380
|
| 295785 |
+
},
|
| 295786 |
+
{
|
| 295787 |
+
"epoch": 1244.61,
|
| 295788 |
+
"learning_rate": 7.5230371567043624e-06,
|
| 295789 |
+
"loss": 0.966,
|
| 295790 |
+
"step": 154385
|
| 295791 |
+
},
|
| 295792 |
+
{
|
| 295793 |
+
"epoch": 1244.65,
|
| 295794 |
+
"learning_rate": 7.522956381260098e-06,
|
| 295795 |
+
"loss": 0.277,
|
| 295796 |
+
"step": 154390
|
| 295797 |
+
},
|
| 295798 |
+
{
|
| 295799 |
+
"epoch": 1244.69,
|
| 295800 |
+
"learning_rate": 7.522875605815832e-06,
|
| 295801 |
+
"loss": 0.2706,
|
| 295802 |
+
"step": 154395
|
| 295803 |
+
},
|
| 295804 |
+
{
|
| 295805 |
+
"epoch": 1244.73,
|
| 295806 |
+
"learning_rate": 7.522794830371568e-06,
|
| 295807 |
+
"loss": 0.2751,
|
| 295808 |
+
"step": 154400
|
| 295809 |
+
},
|
| 295810 |
+
{
|
| 295811 |
+
"epoch": 1244.77,
|
| 295812 |
+
"learning_rate": 7.522714054927302e-06,
|
| 295813 |
+
"loss": 0.4626,
|
| 295814 |
+
"step": 154405
|
| 295815 |
+
},
|
| 295816 |
+
{
|
| 295817 |
+
"epoch": 1244.81,
|
| 295818 |
+
"learning_rate": 7.522633279483038e-06,
|
| 295819 |
+
"loss": 1.1171,
|
| 295820 |
+
"step": 154410
|
| 295821 |
+
},
|
| 295822 |
+
{
|
| 295823 |
+
"epoch": 1244.85,
|
| 295824 |
+
"learning_rate": 7.522552504038772e-06,
|
| 295825 |
+
"loss": 0.2921,
|
| 295826 |
+
"step": 154415
|
| 295827 |
+
},
|
| 295828 |
+
{
|
| 295829 |
+
"epoch": 1244.89,
|
| 295830 |
+
"learning_rate": 7.522471728594508e-06,
|
| 295831 |
+
"loss": 0.2548,
|
| 295832 |
+
"step": 154420
|
| 295833 |
+
},
|
| 295834 |
+
{
|
| 295835 |
+
"epoch": 1244.93,
|
| 295836 |
+
"learning_rate": 7.522390953150242e-06,
|
| 295837 |
+
"loss": 0.3238,
|
| 295838 |
+
"step": 154425
|
| 295839 |
+
},
|
| 295840 |
+
{
|
| 295841 |
+
"epoch": 1244.97,
|
| 295842 |
+
"learning_rate": 7.522310177705978e-06,
|
| 295843 |
+
"loss": 0.5491,
|
| 295844 |
+
"step": 154430
|
| 295845 |
+
},
|
| 295846 |
+
{
|
| 295847 |
+
"epoch": 1245.0,
|
| 295848 |
+
"eval_loss": 0.3593205511569977,
|
| 295849 |
+
"eval_runtime": 41.3281,
|
| 295850 |
+
"eval_samples_per_second": 20.446,
|
| 295851 |
+
"eval_steps_per_second": 0.653,
|
| 295852 |
+
"eval_wer": 0.17205071269966335,
|
| 295853 |
+
"step": 154433
|
| 295854 |
+
},
|
| 295855 |
+
{
|
| 295856 |
+
"epoch": 1245.02,
|
| 295857 |
+
"learning_rate": 7.522229402261712e-06,
|
| 295858 |
+
"loss": 0.4303,
|
| 295859 |
+
"step": 154435
|
| 295860 |
+
},
|
| 295861 |
+
{
|
| 295862 |
+
"epoch": 1245.06,
|
| 295863 |
+
"learning_rate": 7.522148626817448e-06,
|
| 295864 |
+
"loss": 0.291,
|
| 295865 |
+
"step": 154440
|
| 295866 |
+
},
|
| 295867 |
+
{
|
| 295868 |
+
"epoch": 1245.1,
|
| 295869 |
+
"learning_rate": 7.522067851373182e-06,
|
| 295870 |
+
"loss": 0.234,
|
| 295871 |
+
"step": 154445
|
| 295872 |
+
},
|
| 295873 |
+
{
|
| 295874 |
+
"epoch": 1245.14,
|
| 295875 |
+
"learning_rate": 7.521987075928918e-06,
|
| 295876 |
+
"loss": 0.2705,
|
| 295877 |
+
"step": 154450
|
| 295878 |
+
},
|
| 295879 |
+
{
|
| 295880 |
+
"epoch": 1245.18,
|
| 295881 |
+
"learning_rate": 7.521906300484654e-06,
|
| 295882 |
+
"loss": 0.6283,
|
| 295883 |
+
"step": 154455
|
| 295884 |
+
},
|
| 295885 |
+
{
|
| 295886 |
+
"epoch": 1245.22,
|
| 295887 |
+
"learning_rate": 7.521825525040388e-06,
|
| 295888 |
+
"loss": 1.0106,
|
| 295889 |
+
"step": 154460
|
| 295890 |
+
},
|
| 295891 |
+
{
|
| 295892 |
+
"epoch": 1245.26,
|
| 295893 |
+
"learning_rate": 7.521744749596124e-06,
|
| 295894 |
+
"loss": 0.3233,
|
| 295895 |
+
"step": 154465
|
| 295896 |
+
},
|
| 295897 |
+
{
|
| 295898 |
+
"epoch": 1245.3,
|
| 295899 |
+
"learning_rate": 7.521663974151858e-06,
|
| 295900 |
+
"loss": 0.3337,
|
| 295901 |
+
"step": 154470
|
| 295902 |
+
},
|
| 295903 |
+
{
|
| 295904 |
+
"epoch": 1245.34,
|
| 295905 |
+
"learning_rate": 7.521583198707594e-06,
|
| 295906 |
+
"loss": 0.2771,
|
| 295907 |
+
"step": 154475
|
| 295908 |
+
},
|
| 295909 |
+
{
|
| 295910 |
+
"epoch": 1245.38,
|
| 295911 |
+
"learning_rate": 7.521502423263328e-06,
|
| 295912 |
+
"loss": 0.5839,
|
| 295913 |
+
"step": 154480
|
| 295914 |
+
},
|
| 295915 |
+
{
|
| 295916 |
+
"epoch": 1245.42,
|
| 295917 |
+
"learning_rate": 7.521421647819064e-06,
|
| 295918 |
+
"loss": 1.0925,
|
| 295919 |
+
"step": 154485
|
| 295920 |
+
},
|
| 295921 |
+
{
|
| 295922 |
+
"epoch": 1245.46,
|
| 295923 |
+
"learning_rate": 7.521340872374798e-06,
|
| 295924 |
+
"loss": 0.3022,
|
| 295925 |
+
"step": 154490
|
| 295926 |
+
},
|
| 295927 |
+
{
|
| 295928 |
+
"epoch": 1245.5,
|
| 295929 |
+
"learning_rate": 7.521260096930534e-06,
|
| 295930 |
+
"loss": 0.2368,
|
| 295931 |
+
"step": 154495
|
| 295932 |
+
},
|
| 295933 |
+
{
|
| 295934 |
+
"epoch": 1245.54,
|
| 295935 |
+
"learning_rate": 7.521179321486268e-06,
|
| 295936 |
+
"loss": 0.3563,
|
| 295937 |
+
"step": 154500
|
| 295938 |
+
},
|
| 295939 |
+
{
|
| 295940 |
+
"epoch": 1245.58,
|
| 295941 |
+
"learning_rate": 7.521098546042004e-06,
|
| 295942 |
+
"loss": 0.5765,
|
| 295943 |
+
"step": 154505
|
| 295944 |
+
},
|
| 295945 |
+
{
|
| 295946 |
+
"epoch": 1245.62,
|
| 295947 |
+
"learning_rate": 7.52101777059774e-06,
|
| 295948 |
+
"loss": 0.9627,
|
| 295949 |
+
"step": 154510
|
| 295950 |
+
},
|
| 295951 |
+
{
|
| 295952 |
+
"epoch": 1245.66,
|
| 295953 |
+
"learning_rate": 7.520936995153474e-06,
|
| 295954 |
+
"loss": 0.2524,
|
| 295955 |
+
"step": 154515
|
| 295956 |
+
},
|
| 295957 |
+
{
|
| 295958 |
+
"epoch": 1245.7,
|
| 295959 |
+
"learning_rate": 7.52085621970921e-06,
|
| 295960 |
+
"loss": 0.2469,
|
| 295961 |
+
"step": 154520
|
| 295962 |
+
},
|
| 295963 |
+
{
|
| 295964 |
+
"epoch": 1245.74,
|
| 295965 |
+
"learning_rate": 7.520775444264944e-06,
|
| 295966 |
+
"loss": 0.2765,
|
| 295967 |
+
"step": 154525
|
| 295968 |
+
},
|
| 295969 |
+
{
|
| 295970 |
+
"epoch": 1245.78,
|
| 295971 |
+
"learning_rate": 7.52069466882068e-06,
|
| 295972 |
+
"loss": 0.6153,
|
| 295973 |
+
"step": 154530
|
| 295974 |
+
},
|
| 295975 |
+
{
|
| 295976 |
+
"epoch": 1245.82,
|
| 295977 |
+
"learning_rate": 7.520613893376414e-06,
|
| 295978 |
+
"loss": 1.0515,
|
| 295979 |
+
"step": 154535
|
| 295980 |
+
},
|
| 295981 |
+
{
|
| 295982 |
+
"epoch": 1245.86,
|
| 295983 |
+
"learning_rate": 7.52053311793215e-06,
|
| 295984 |
+
"loss": 0.234,
|
| 295985 |
+
"step": 154540
|
| 295986 |
+
},
|
| 295987 |
+
{
|
| 295988 |
+
"epoch": 1245.9,
|
| 295989 |
+
"learning_rate": 7.520452342487884e-06,
|
| 295990 |
+
"loss": 0.3025,
|
| 295991 |
+
"step": 154545
|
| 295992 |
+
},
|
| 295993 |
+
{
|
| 295994 |
+
"epoch": 1245.94,
|
| 295995 |
+
"learning_rate": 7.5203715670436196e-06,
|
| 295996 |
+
"loss": 0.397,
|
| 295997 |
+
"step": 154550
|
| 295998 |
+
},
|
| 295999 |
+
{
|
| 296000 |
+
"epoch": 1245.98,
|
| 296001 |
+
"learning_rate": 7.520290791599354e-06,
|
| 296002 |
+
"loss": 0.7449,
|
| 296003 |
+
"step": 154555
|
| 296004 |
+
},
|
| 296005 |
+
{
|
| 296006 |
+
"epoch": 1246.0,
|
| 296007 |
+
"eval_loss": 0.3774397671222687,
|
| 296008 |
+
"eval_runtime": 41.2529,
|
| 296009 |
+
"eval_samples_per_second": 20.483,
|
| 296010 |
+
"eval_steps_per_second": 0.654,
|
| 296011 |
+
"eval_wer": 0.17398648648648649,
|
| 296012 |
+
"step": 154557
|
| 296013 |
}
|
| 296014 |
],
|
| 296015 |
+
"max_steps": 620000,
|
| 296016 |
"num_train_epochs": 5000,
|
| 296017 |
+
"total_flos": 4.349352656879201e+20,
|
| 296018 |
"trial_name": null,
|
| 296019 |
"trial_params": null
|
| 296020 |
}
|
model-bin/finetune/base/{checkpoint-153934 → checkpoint-154557}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630257516.869323/events.out.tfevents.1630257516.cc93b136ebf5.1086.239
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:47564b61dbc795fbff8ffb584987ed8e45290a50c8654ff6d6b721f69de0552d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630257947.7075908/events.out.tfevents.1630257947.cc93b136ebf5.1086.241
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5eba5a6fdf49c95649c2e42f6cbd9df77ce1f295c630f4b5a7fb8288d8f10edb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630258385.8405457/events.out.tfevents.1630258385.cc93b136ebf5.1086.243
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9622dfc357fbd2e16d483e3d604f89b527ef37a6c7bf1532e2e88c25cf3308a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630258817.5863304/events.out.tfevents.1630258817.cc93b136ebf5.1086.245
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b5ad18bb6152dbce91ff419fa98b664cad125426c8e4b639dbbd56c574a2b4c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630259247.5041316/events.out.tfevents.1630259247.cc93b136ebf5.1086.247
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:292a1010920050304adef6551b9b5d503b6ea8e379eff8bebc7bbfca9e339431
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630257516.cc93b136ebf5.1086.238
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce8a03899b5fa87bd3d8fc1f6589ec7e1e86ce608c3218a446389363fae9ff14
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630257947.cc93b136ebf5.1086.240
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53aab3424f08d1bb1a16984264879ec0b58616b1ae4a70ae63f52cdbd5587ecf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630258385.cc93b136ebf5.1086.242
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b123bb5fe7fb966dedde9824ff10389dfb4aae067cbf6de8b20f125753846cc8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630258817.cc93b136ebf5.1086.244
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6d4a8b2488c4f860f00627704e43251101cf16498afbb29e86722924531d2062
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630259247.cc93b136ebf5.1086.246
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc711fe15994869da2da61c1357526bf656193c26bd60275c4cbfff1c554852e
|
| 3 |
+
size 8622
|