"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630239529.1803105/events.out.tfevents.1630239529.cc93b136ebf5.1086.159 +3 -0
- model-bin/finetune/base/log/1630239966.3963904/events.out.tfevents.1630239966.cc93b136ebf5.1086.161 +3 -0
- model-bin/finetune/base/log/1630240411.0957363/events.out.tfevents.1630240411.cc93b136ebf5.1086.163 +3 -0
- model-bin/finetune/base/log/1630240854.716574/events.out.tfevents.1630240854.cc93b136ebf5.1086.165 +3 -0
- model-bin/finetune/base/log/1630241290.4871612/events.out.tfevents.1630241290.cc93b136ebf5.1086.167 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630239529.cc93b136ebf5.1086.158 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630239966.cc93b136ebf5.1086.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630240411.cc93b136ebf5.1086.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630240854.cc93b136ebf5.1086.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630241290.cc93b136ebf5.1086.166 +3 -0
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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 sha256:207277e446e0ca60dd83abf90aab7a2af3700e64df25f7d5a5d03195c838cc31
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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 sha256:04af34e30722560d517d6ae868955c6953393b4b21776c8eb7ccd0753bf1f2ea
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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 sha256:953f51e4a6b744d9cf78d43abb72507dea35603620d29a905184d67154569c48
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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 sha256:48a37d54355b773f8b00614fabd924720d9d3f3bbbb3a062ffefc6c699f6469c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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 sha256:0dc231f0815a529224131eede510f2dcb3545c8b449267f729dc8e7d82e5f194
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/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,
|
|
@@ -288891,11 +288891,800 @@
|
|
| 288891 |
"eval_steps_per_second": 0.645,
|
| 288892 |
"eval_wer": 0.17429476108232586,
|
| 288893 |
"step": 148960
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 288894 |
}
|
| 288895 |
],
|
| 288896 |
"max_steps": 625000,
|
| 288897 |
"num_train_epochs": 5000,
|
| 288898 |
-
"total_flos": 4.
|
| 288899 |
"trial_name": null,
|
| 288900 |
"trial_params": null
|
| 288901 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
+
"epoch": 1196.0,
|
| 5 |
+
"global_step": 149584,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 288891 |
"eval_steps_per_second": 0.645,
|
| 288892 |
"eval_wer": 0.17429476108232586,
|
| 288893 |
"step": 148960
|
| 288894 |
+
},
|
| 288895 |
+
{
|
| 288896 |
+
"epoch": 1201.04,
|
| 288897 |
+
"learning_rate": 7.610565428109855e-06,
|
| 288898 |
+
"loss": 0.3312,
|
| 288899 |
+
"step": 148965
|
| 288900 |
+
},
|
| 288901 |
+
{
|
| 288902 |
+
"epoch": 1201.08,
|
| 288903 |
+
"learning_rate": 7.610484652665591e-06,
|
| 288904 |
+
"loss": 0.2707,
|
| 288905 |
+
"step": 148970
|
| 288906 |
+
},
|
| 288907 |
+
{
|
| 288908 |
+
"epoch": 1201.12,
|
| 288909 |
+
"learning_rate": 7.610403877221325e-06,
|
| 288910 |
+
"loss": 0.2662,
|
| 288911 |
+
"step": 148975
|
| 288912 |
+
},
|
| 288913 |
+
{
|
| 288914 |
+
"epoch": 1201.16,
|
| 288915 |
+
"learning_rate": 7.610323101777061e-06,
|
| 288916 |
+
"loss": 0.485,
|
| 288917 |
+
"step": 148980
|
| 288918 |
+
},
|
| 288919 |
+
{
|
| 288920 |
+
"epoch": 1201.2,
|
| 288921 |
+
"learning_rate": 7.610242326332795e-06,
|
| 288922 |
+
"loss": 1.0488,
|
| 288923 |
+
"step": 148985
|
| 288924 |
+
},
|
| 288925 |
+
{
|
| 288926 |
+
"epoch": 1201.24,
|
| 288927 |
+
"learning_rate": 7.610161550888531e-06,
|
| 288928 |
+
"loss": 0.2695,
|
| 288929 |
+
"step": 148990
|
| 288930 |
+
},
|
| 288931 |
+
{
|
| 288932 |
+
"epoch": 1201.28,
|
| 288933 |
+
"learning_rate": 7.610080775444265e-06,
|
| 288934 |
+
"loss": 0.2995,
|
| 288935 |
+
"step": 148995
|
| 288936 |
+
},
|
| 288937 |
+
{
|
| 288938 |
+
"epoch": 1201.32,
|
| 288939 |
+
"learning_rate": 7.610000000000001e-06,
|
| 288940 |
+
"loss": 0.3051,
|
| 288941 |
+
"step": 149000
|
| 288942 |
+
},
|
| 288943 |
+
{
|
| 288944 |
+
"epoch": 1201.36,
|
| 288945 |
+
"learning_rate": 7.609919224555735e-06,
|
| 288946 |
+
"loss": 0.4148,
|
| 288947 |
+
"step": 149005
|
| 288948 |
+
},
|
| 288949 |
+
{
|
| 288950 |
+
"epoch": 1201.4,
|
| 288951 |
+
"learning_rate": 7.609838449111471e-06,
|
| 288952 |
+
"loss": 1.0493,
|
| 288953 |
+
"step": 149010
|
| 288954 |
+
},
|
| 288955 |
+
{
|
| 288956 |
+
"epoch": 1201.44,
|
| 288957 |
+
"learning_rate": 7.609757673667207e-06,
|
| 288958 |
+
"loss": 0.2863,
|
| 288959 |
+
"step": 149015
|
| 288960 |
+
},
|
| 288961 |
+
{
|
| 288962 |
+
"epoch": 1201.48,
|
| 288963 |
+
"learning_rate": 7.609676898222941e-06,
|
| 288964 |
+
"loss": 0.2597,
|
| 288965 |
+
"step": 149020
|
| 288966 |
+
},
|
| 288967 |
+
{
|
| 288968 |
+
"epoch": 1201.52,
|
| 288969 |
+
"learning_rate": 7.609596122778677e-06,
|
| 288970 |
+
"loss": 0.3128,
|
| 288971 |
+
"step": 149025
|
| 288972 |
+
},
|
| 288973 |
+
{
|
| 288974 |
+
"epoch": 1201.56,
|
| 288975 |
+
"learning_rate": 7.609515347334411e-06,
|
| 288976 |
+
"loss": 0.4076,
|
| 288977 |
+
"step": 149030
|
| 288978 |
+
},
|
| 288979 |
+
{
|
| 288980 |
+
"epoch": 1201.6,
|
| 288981 |
+
"learning_rate": 7.609434571890147e-06,
|
| 288982 |
+
"loss": 1.1858,
|
| 288983 |
+
"step": 149035
|
| 288984 |
+
},
|
| 288985 |
+
{
|
| 288986 |
+
"epoch": 1201.65,
|
| 288987 |
+
"learning_rate": 7.609353796445881e-06,
|
| 288988 |
+
"loss": 0.3028,
|
| 288989 |
+
"step": 149040
|
| 288990 |
+
},
|
| 288991 |
+
{
|
| 288992 |
+
"epoch": 1201.69,
|
| 288993 |
+
"learning_rate": 7.6092730210016166e-06,
|
| 288994 |
+
"loss": 0.256,
|
| 288995 |
+
"step": 149045
|
| 288996 |
+
},
|
| 288997 |
+
{
|
| 288998 |
+
"epoch": 1201.73,
|
| 288999 |
+
"learning_rate": 7.609192245557351e-06,
|
| 289000 |
+
"loss": 0.2901,
|
| 289001 |
+
"step": 149050
|
| 289002 |
+
},
|
| 289003 |
+
{
|
| 289004 |
+
"epoch": 1201.77,
|
| 289005 |
+
"learning_rate": 7.6091114701130865e-06,
|
| 289006 |
+
"loss": 0.4104,
|
| 289007 |
+
"step": 149055
|
| 289008 |
+
},
|
| 289009 |
+
{
|
| 289010 |
+
"epoch": 1201.81,
|
| 289011 |
+
"learning_rate": 7.609030694668821e-06,
|
| 289012 |
+
"loss": 1.1365,
|
| 289013 |
+
"step": 149060
|
| 289014 |
+
},
|
| 289015 |
+
{
|
| 289016 |
+
"epoch": 1201.85,
|
| 289017 |
+
"learning_rate": 7.6089499192245565e-06,
|
| 289018 |
+
"loss": 0.3528,
|
| 289019 |
+
"step": 149065
|
| 289020 |
+
},
|
| 289021 |
+
{
|
| 289022 |
+
"epoch": 1201.89,
|
| 289023 |
+
"learning_rate": 7.608869143780292e-06,
|
| 289024 |
+
"loss": 0.2461,
|
| 289025 |
+
"step": 149070
|
| 289026 |
+
},
|
| 289027 |
+
{
|
| 289028 |
+
"epoch": 1201.93,
|
| 289029 |
+
"learning_rate": 7.6087883683360265e-06,
|
| 289030 |
+
"loss": 0.3093,
|
| 289031 |
+
"step": 149075
|
| 289032 |
+
},
|
| 289033 |
+
{
|
| 289034 |
+
"epoch": 1201.97,
|
| 289035 |
+
"learning_rate": 7.608707592891762e-06,
|
| 289036 |
+
"loss": 0.5117,
|
| 289037 |
+
"step": 149080
|
| 289038 |
+
},
|
| 289039 |
+
{
|
| 289040 |
+
"epoch": 1202.0,
|
| 289041 |
+
"eval_loss": 0.4121530055999756,
|
| 289042 |
+
"eval_runtime": 42.1832,
|
| 289043 |
+
"eval_samples_per_second": 19.937,
|
| 289044 |
+
"eval_steps_per_second": 0.64,
|
| 289045 |
+
"eval_wer": 0.17716044858169025,
|
| 289046 |
+
"step": 149084
|
| 289047 |
+
},
|
| 289048 |
+
{
|
| 289049 |
+
"epoch": 1192.01,
|
| 289050 |
+
"learning_rate": 7.6086268174474965e-06,
|
| 289051 |
+
"loss": 0.3826,
|
| 289052 |
+
"step": 149085
|
| 289053 |
+
},
|
| 289054 |
+
{
|
| 289055 |
+
"epoch": 1192.05,
|
| 289056 |
+
"learning_rate": 7.608546042003232e-06,
|
| 289057 |
+
"loss": 0.3115,
|
| 289058 |
+
"step": 149090
|
| 289059 |
+
},
|
| 289060 |
+
{
|
| 289061 |
+
"epoch": 1192.09,
|
| 289062 |
+
"learning_rate": 7.6084652665589664e-06,
|
| 289063 |
+
"loss": 0.2778,
|
| 289064 |
+
"step": 149095
|
| 289065 |
+
},
|
| 289066 |
+
{
|
| 289067 |
+
"epoch": 1192.13,
|
| 289068 |
+
"learning_rate": 7.608384491114702e-06,
|
| 289069 |
+
"loss": 0.4587,
|
| 289070 |
+
"step": 149100
|
| 289071 |
+
},
|
| 289072 |
+
{
|
| 289073 |
+
"epoch": 1192.17,
|
| 289074 |
+
"learning_rate": 7.608303715670436e-06,
|
| 289075 |
+
"loss": 0.699,
|
| 289076 |
+
"step": 149105
|
| 289077 |
+
},
|
| 289078 |
+
{
|
| 289079 |
+
"epoch": 1192.21,
|
| 289080 |
+
"learning_rate": 7.608222940226172e-06,
|
| 289081 |
+
"loss": 1.0862,
|
| 289082 |
+
"step": 149110
|
| 289083 |
+
},
|
| 289084 |
+
{
|
| 289085 |
+
"epoch": 1192.25,
|
| 289086 |
+
"learning_rate": 7.608142164781906e-06,
|
| 289087 |
+
"loss": 0.3359,
|
| 289088 |
+
"step": 149115
|
| 289089 |
+
},
|
| 289090 |
+
{
|
| 289091 |
+
"epoch": 1192.29,
|
| 289092 |
+
"learning_rate": 7.608061389337642e-06,
|
| 289093 |
+
"loss": 0.2895,
|
| 289094 |
+
"step": 149120
|
| 289095 |
+
},
|
| 289096 |
+
{
|
| 289097 |
+
"epoch": 1192.33,
|
| 289098 |
+
"learning_rate": 7.607980613893376e-06,
|
| 289099 |
+
"loss": 0.4508,
|
| 289100 |
+
"step": 149125
|
| 289101 |
+
},
|
| 289102 |
+
{
|
| 289103 |
+
"epoch": 1192.37,
|
| 289104 |
+
"learning_rate": 7.607899838449112e-06,
|
| 289105 |
+
"loss": 0.4802,
|
| 289106 |
+
"step": 149130
|
| 289107 |
+
},
|
| 289108 |
+
{
|
| 289109 |
+
"epoch": 1192.41,
|
| 289110 |
+
"learning_rate": 7.607819063004848e-06,
|
| 289111 |
+
"loss": 0.9843,
|
| 289112 |
+
"step": 149135
|
| 289113 |
+
},
|
| 289114 |
+
{
|
| 289115 |
+
"epoch": 1192.45,
|
| 289116 |
+
"learning_rate": 7.607738287560582e-06,
|
| 289117 |
+
"loss": 0.3298,
|
| 289118 |
+
"step": 149140
|
| 289119 |
+
},
|
| 289120 |
+
{
|
| 289121 |
+
"epoch": 1192.49,
|
| 289122 |
+
"learning_rate": 7.607657512116318e-06,
|
| 289123 |
+
"loss": 0.2377,
|
| 289124 |
+
"step": 149145
|
| 289125 |
+
},
|
| 289126 |
+
{
|
| 289127 |
+
"epoch": 1192.53,
|
| 289128 |
+
"learning_rate": 7.607576736672052e-06,
|
| 289129 |
+
"loss": 0.324,
|
| 289130 |
+
"step": 149150
|
| 289131 |
+
},
|
| 289132 |
+
{
|
| 289133 |
+
"epoch": 1192.57,
|
| 289134 |
+
"learning_rate": 7.607495961227788e-06,
|
| 289135 |
+
"loss": 0.5793,
|
| 289136 |
+
"step": 149155
|
| 289137 |
+
},
|
| 289138 |
+
{
|
| 289139 |
+
"epoch": 1192.61,
|
| 289140 |
+
"learning_rate": 7.607415185783522e-06,
|
| 289141 |
+
"loss": 1.1508,
|
| 289142 |
+
"step": 149160
|
| 289143 |
+
},
|
| 289144 |
+
{
|
| 289145 |
+
"epoch": 1192.65,
|
| 289146 |
+
"learning_rate": 7.607334410339258e-06,
|
| 289147 |
+
"loss": 0.3118,
|
| 289148 |
+
"step": 149165
|
| 289149 |
+
},
|
| 289150 |
+
{
|
| 289151 |
+
"epoch": 1192.69,
|
| 289152 |
+
"learning_rate": 7.607253634894992e-06,
|
| 289153 |
+
"loss": 0.3146,
|
| 289154 |
+
"step": 149170
|
| 289155 |
+
},
|
| 289156 |
+
{
|
| 289157 |
+
"epoch": 1192.73,
|
| 289158 |
+
"learning_rate": 7.607172859450728e-06,
|
| 289159 |
+
"loss": 0.3237,
|
| 289160 |
+
"step": 149175
|
| 289161 |
+
},
|
| 289162 |
+
{
|
| 289163 |
+
"epoch": 1192.77,
|
| 289164 |
+
"learning_rate": 7.607092084006462e-06,
|
| 289165 |
+
"loss": 0.4928,
|
| 289166 |
+
"step": 149180
|
| 289167 |
+
},
|
| 289168 |
+
{
|
| 289169 |
+
"epoch": 1192.81,
|
| 289170 |
+
"learning_rate": 7.607011308562198e-06,
|
| 289171 |
+
"loss": 1.108,
|
| 289172 |
+
"step": 149185
|
| 289173 |
+
},
|
| 289174 |
+
{
|
| 289175 |
+
"epoch": 1192.85,
|
| 289176 |
+
"learning_rate": 7.606930533117933e-06,
|
| 289177 |
+
"loss": 0.2488,
|
| 289178 |
+
"step": 149190
|
| 289179 |
+
},
|
| 289180 |
+
{
|
| 289181 |
+
"epoch": 1192.89,
|
| 289182 |
+
"learning_rate": 7.606849757673668e-06,
|
| 289183 |
+
"loss": 0.263,
|
| 289184 |
+
"step": 149195
|
| 289185 |
+
},
|
| 289186 |
+
{
|
| 289187 |
+
"epoch": 1192.93,
|
| 289188 |
+
"learning_rate": 7.606768982229404e-06,
|
| 289189 |
+
"loss": 0.3348,
|
| 289190 |
+
"step": 149200
|
| 289191 |
+
},
|
| 289192 |
+
{
|
| 289193 |
+
"epoch": 1192.97,
|
| 289194 |
+
"learning_rate": 7.606688206785138e-06,
|
| 289195 |
+
"loss": 0.5095,
|
| 289196 |
+
"step": 149205
|
| 289197 |
+
},
|
| 289198 |
+
{
|
| 289199 |
+
"epoch": 1193.0,
|
| 289200 |
+
"eval_loss": 0.32593458890914917,
|
| 289201 |
+
"eval_runtime": 42.4315,
|
| 289202 |
+
"eval_samples_per_second": 19.82,
|
| 289203 |
+
"eval_steps_per_second": 0.636,
|
| 289204 |
+
"eval_wer": 0.1726039323804687,
|
| 289205 |
+
"step": 149209
|
| 289206 |
+
},
|
| 289207 |
+
{
|
| 289208 |
+
"epoch": 1193.01,
|
| 289209 |
+
"learning_rate": 7.606607431340874e-06,
|
| 289210 |
+
"loss": 0.9795,
|
| 289211 |
+
"step": 149210
|
| 289212 |
+
},
|
| 289213 |
+
{
|
| 289214 |
+
"epoch": 1193.05,
|
| 289215 |
+
"learning_rate": 7.606526655896608e-06,
|
| 289216 |
+
"loss": 0.2454,
|
| 289217 |
+
"step": 149215
|
| 289218 |
+
},
|
| 289219 |
+
{
|
| 289220 |
+
"epoch": 1193.09,
|
| 289221 |
+
"learning_rate": 7.606445880452344e-06,
|
| 289222 |
+
"loss": 0.2858,
|
| 289223 |
+
"step": 149220
|
| 289224 |
+
},
|
| 289225 |
+
{
|
| 289226 |
+
"epoch": 1193.13,
|
| 289227 |
+
"learning_rate": 7.606365105008078e-06,
|
| 289228 |
+
"loss": 0.3203,
|
| 289229 |
+
"step": 149225
|
| 289230 |
+
},
|
| 289231 |
+
{
|
| 289232 |
+
"epoch": 1193.17,
|
| 289233 |
+
"learning_rate": 7.606284329563814e-06,
|
| 289234 |
+
"loss": 0.5705,
|
| 289235 |
+
"step": 149230
|
| 289236 |
+
},
|
| 289237 |
+
{
|
| 289238 |
+
"epoch": 1193.21,
|
| 289239 |
+
"learning_rate": 7.606203554119548e-06,
|
| 289240 |
+
"loss": 0.9506,
|
| 289241 |
+
"step": 149235
|
| 289242 |
+
},
|
| 289243 |
+
{
|
| 289244 |
+
"epoch": 1193.25,
|
| 289245 |
+
"learning_rate": 7.606122778675284e-06,
|
| 289246 |
+
"loss": 0.2817,
|
| 289247 |
+
"step": 149240
|
| 289248 |
+
},
|
| 289249 |
+
{
|
| 289250 |
+
"epoch": 1193.29,
|
| 289251 |
+
"learning_rate": 7.606042003231019e-06,
|
| 289252 |
+
"loss": 0.303,
|
| 289253 |
+
"step": 149245
|
| 289254 |
+
},
|
| 289255 |
+
{
|
| 289256 |
+
"epoch": 1193.33,
|
| 289257 |
+
"learning_rate": 7.605961227786754e-06,
|
| 289258 |
+
"loss": 0.9205,
|
| 289259 |
+
"step": 149250
|
| 289260 |
+
},
|
| 289261 |
+
{
|
| 289262 |
+
"epoch": 1193.37,
|
| 289263 |
+
"learning_rate": 7.605880452342489e-06,
|
| 289264 |
+
"loss": 0.4905,
|
| 289265 |
+
"step": 149255
|
| 289266 |
+
},
|
| 289267 |
+
{
|
| 289268 |
+
"epoch": 1193.41,
|
| 289269 |
+
"learning_rate": 7.6057996768982236e-06,
|
| 289270 |
+
"loss": 1.1657,
|
| 289271 |
+
"step": 149260
|
| 289272 |
+
},
|
| 289273 |
+
{
|
| 289274 |
+
"epoch": 1193.45,
|
| 289275 |
+
"learning_rate": 7.6057189014539586e-06,
|
| 289276 |
+
"loss": 0.2844,
|
| 289277 |
+
"step": 149265
|
| 289278 |
+
},
|
| 289279 |
+
{
|
| 289280 |
+
"epoch": 1193.49,
|
| 289281 |
+
"learning_rate": 7.6056381260096935e-06,
|
| 289282 |
+
"loss": 0.3248,
|
| 289283 |
+
"step": 149270
|
| 289284 |
+
},
|
| 289285 |
+
{
|
| 289286 |
+
"epoch": 1193.53,
|
| 289287 |
+
"learning_rate": 7.605557350565429e-06,
|
| 289288 |
+
"loss": 0.3391,
|
| 289289 |
+
"step": 149275
|
| 289290 |
+
},
|
| 289291 |
+
{
|
| 289292 |
+
"epoch": 1193.57,
|
| 289293 |
+
"learning_rate": 7.6054765751211635e-06,
|
| 289294 |
+
"loss": 0.5131,
|
| 289295 |
+
"step": 149280
|
| 289296 |
+
},
|
| 289297 |
+
{
|
| 289298 |
+
"epoch": 1193.61,
|
| 289299 |
+
"learning_rate": 7.605395799676899e-06,
|
| 289300 |
+
"loss": 1.119,
|
| 289301 |
+
"step": 149285
|
| 289302 |
+
},
|
| 289303 |
+
{
|
| 289304 |
+
"epoch": 1193.65,
|
| 289305 |
+
"learning_rate": 7.6053150242326335e-06,
|
| 289306 |
+
"loss": 0.2677,
|
| 289307 |
+
"step": 149290
|
| 289308 |
+
},
|
| 289309 |
+
{
|
| 289310 |
+
"epoch": 1193.69,
|
| 289311 |
+
"learning_rate": 7.605234248788369e-06,
|
| 289312 |
+
"loss": 0.2582,
|
| 289313 |
+
"step": 149295
|
| 289314 |
+
},
|
| 289315 |
+
{
|
| 289316 |
+
"epoch": 1193.73,
|
| 289317 |
+
"learning_rate": 7.6051534733441035e-06,
|
| 289318 |
+
"loss": 0.3005,
|
| 289319 |
+
"step": 149300
|
| 289320 |
+
},
|
| 289321 |
+
{
|
| 289322 |
+
"epoch": 1193.77,
|
| 289323 |
+
"learning_rate": 7.605072697899839e-06,
|
| 289324 |
+
"loss": 0.4774,
|
| 289325 |
+
"step": 149305
|
| 289326 |
+
},
|
| 289327 |
+
{
|
| 289328 |
+
"epoch": 1193.81,
|
| 289329 |
+
"learning_rate": 7.604991922455574e-06,
|
| 289330 |
+
"loss": 1.1529,
|
| 289331 |
+
"step": 149310
|
| 289332 |
+
},
|
| 289333 |
+
{
|
| 289334 |
+
"epoch": 1193.85,
|
| 289335 |
+
"learning_rate": 7.604911147011309e-06,
|
| 289336 |
+
"loss": 0.274,
|
| 289337 |
+
"step": 149315
|
| 289338 |
+
},
|
| 289339 |
+
{
|
| 289340 |
+
"epoch": 1193.89,
|
| 289341 |
+
"learning_rate": 7.604830371567044e-06,
|
| 289342 |
+
"loss": 0.3157,
|
| 289343 |
+
"step": 149320
|
| 289344 |
+
},
|
| 289345 |
+
{
|
| 289346 |
+
"epoch": 1193.93,
|
| 289347 |
+
"learning_rate": 7.604749596122779e-06,
|
| 289348 |
+
"loss": 0.2767,
|
| 289349 |
+
"step": 149325
|
| 289350 |
+
},
|
| 289351 |
+
{
|
| 289352 |
+
"epoch": 1193.97,
|
| 289353 |
+
"learning_rate": 7.604668820678514e-06,
|
| 289354 |
+
"loss": 0.492,
|
| 289355 |
+
"step": 149330
|
| 289356 |
+
},
|
| 289357 |
+
{
|
| 289358 |
+
"epoch": 1194.0,
|
| 289359 |
+
"eval_loss": 0.3644717037677765,
|
| 289360 |
+
"eval_runtime": 43.3033,
|
| 289361 |
+
"eval_samples_per_second": 19.421,
|
| 289362 |
+
"eval_steps_per_second": 0.624,
|
| 289363 |
+
"eval_wer": 0.17845911949685533,
|
| 289364 |
+
"step": 149334
|
| 289365 |
+
},
|
| 289366 |
+
{
|
| 289367 |
+
"epoch": 1194.01,
|
| 289368 |
+
"learning_rate": 7.604588045234249e-06,
|
| 289369 |
+
"loss": 0.3145,
|
| 289370 |
+
"step": 149335
|
| 289371 |
+
},
|
| 289372 |
+
{
|
| 289373 |
+
"epoch": 1194.05,
|
| 289374 |
+
"learning_rate": 7.604507269789985e-06,
|
| 289375 |
+
"loss": 0.2788,
|
| 289376 |
+
"step": 149340
|
| 289377 |
+
},
|
| 289378 |
+
{
|
| 289379 |
+
"epoch": 1194.09,
|
| 289380 |
+
"learning_rate": 7.604426494345719e-06,
|
| 289381 |
+
"loss": 0.2532,
|
| 289382 |
+
"step": 149345
|
| 289383 |
+
},
|
| 289384 |
+
{
|
| 289385 |
+
"epoch": 1194.13,
|
| 289386 |
+
"learning_rate": 7.604345718901455e-06,
|
| 289387 |
+
"loss": 0.3359,
|
| 289388 |
+
"step": 149350
|
| 289389 |
+
},
|
| 289390 |
+
{
|
| 289391 |
+
"epoch": 1194.17,
|
| 289392 |
+
"learning_rate": 7.604264943457189e-06,
|
| 289393 |
+
"loss": 0.4988,
|
| 289394 |
+
"step": 149355
|
| 289395 |
+
},
|
| 289396 |
+
{
|
| 289397 |
+
"epoch": 1194.21,
|
| 289398 |
+
"learning_rate": 7.604184168012925e-06,
|
| 289399 |
+
"loss": 1.105,
|
| 289400 |
+
"step": 149360
|
| 289401 |
+
},
|
| 289402 |
+
{
|
| 289403 |
+
"epoch": 1194.25,
|
| 289404 |
+
"learning_rate": 7.60410339256866e-06,
|
| 289405 |
+
"loss": 0.2978,
|
| 289406 |
+
"step": 149365
|
| 289407 |
+
},
|
| 289408 |
+
{
|
| 289409 |
+
"epoch": 1194.29,
|
| 289410 |
+
"learning_rate": 7.604022617124395e-06,
|
| 289411 |
+
"loss": 0.2537,
|
| 289412 |
+
"step": 149370
|
| 289413 |
+
},
|
| 289414 |
+
{
|
| 289415 |
+
"epoch": 1194.33,
|
| 289416 |
+
"learning_rate": 7.60394184168013e-06,
|
| 289417 |
+
"loss": 0.2694,
|
| 289418 |
+
"step": 149375
|
| 289419 |
+
},
|
| 289420 |
+
{
|
| 289421 |
+
"epoch": 1194.37,
|
| 289422 |
+
"learning_rate": 7.603861066235865e-06,
|
| 289423 |
+
"loss": 0.4095,
|
| 289424 |
+
"step": 149380
|
| 289425 |
+
},
|
| 289426 |
+
{
|
| 289427 |
+
"epoch": 1194.41,
|
| 289428 |
+
"learning_rate": 7.6037802907916e-06,
|
| 289429 |
+
"loss": 1.1433,
|
| 289430 |
+
"step": 149385
|
| 289431 |
+
},
|
| 289432 |
+
{
|
| 289433 |
+
"epoch": 1194.45,
|
| 289434 |
+
"learning_rate": 7.603699515347335e-06,
|
| 289435 |
+
"loss": 0.287,
|
| 289436 |
+
"step": 149390
|
| 289437 |
+
},
|
| 289438 |
+
{
|
| 289439 |
+
"epoch": 1194.49,
|
| 289440 |
+
"learning_rate": 7.60361873990307e-06,
|
| 289441 |
+
"loss": 0.2701,
|
| 289442 |
+
"step": 149395
|
| 289443 |
+
},
|
| 289444 |
+
{
|
| 289445 |
+
"epoch": 1194.53,
|
| 289446 |
+
"learning_rate": 7.603537964458805e-06,
|
| 289447 |
+
"loss": 0.3218,
|
| 289448 |
+
"step": 149400
|
| 289449 |
+
},
|
| 289450 |
+
{
|
| 289451 |
+
"epoch": 1194.57,
|
| 289452 |
+
"learning_rate": 7.603457189014541e-06,
|
| 289453 |
+
"loss": 0.5814,
|
| 289454 |
+
"step": 149405
|
| 289455 |
+
},
|
| 289456 |
+
{
|
| 289457 |
+
"epoch": 1194.61,
|
| 289458 |
+
"learning_rate": 7.603376413570275e-06,
|
| 289459 |
+
"loss": 1.0341,
|
| 289460 |
+
"step": 149410
|
| 289461 |
+
},
|
| 289462 |
+
{
|
| 289463 |
+
"epoch": 1194.65,
|
| 289464 |
+
"learning_rate": 7.603295638126011e-06,
|
| 289465 |
+
"loss": 0.2933,
|
| 289466 |
+
"step": 149415
|
| 289467 |
+
},
|
| 289468 |
+
{
|
| 289469 |
+
"epoch": 1194.69,
|
| 289470 |
+
"learning_rate": 7.603214862681746e-06,
|
| 289471 |
+
"loss": 0.295,
|
| 289472 |
+
"step": 149420
|
| 289473 |
+
},
|
| 289474 |
+
{
|
| 289475 |
+
"epoch": 1194.73,
|
| 289476 |
+
"learning_rate": 7.603134087237481e-06,
|
| 289477 |
+
"loss": 0.3104,
|
| 289478 |
+
"step": 149425
|
| 289479 |
+
},
|
| 289480 |
+
{
|
| 289481 |
+
"epoch": 1194.77,
|
| 289482 |
+
"learning_rate": 7.603053311793216e-06,
|
| 289483 |
+
"loss": 0.4882,
|
| 289484 |
+
"step": 149430
|
| 289485 |
+
},
|
| 289486 |
+
{
|
| 289487 |
+
"epoch": 1194.81,
|
| 289488 |
+
"learning_rate": 7.602972536348951e-06,
|
| 289489 |
+
"loss": 1.1231,
|
| 289490 |
+
"step": 149435
|
| 289491 |
+
},
|
| 289492 |
+
{
|
| 289493 |
+
"epoch": 1194.85,
|
| 289494 |
+
"learning_rate": 7.602891760904686e-06,
|
| 289495 |
+
"loss": 0.2896,
|
| 289496 |
+
"step": 149440
|
| 289497 |
+
},
|
| 289498 |
+
{
|
| 289499 |
+
"epoch": 1194.89,
|
| 289500 |
+
"learning_rate": 7.602810985460421e-06,
|
| 289501 |
+
"loss": 0.2373,
|
| 289502 |
+
"step": 149445
|
| 289503 |
+
},
|
| 289504 |
+
{
|
| 289505 |
+
"epoch": 1194.93,
|
| 289506 |
+
"learning_rate": 7.602730210016156e-06,
|
| 289507 |
+
"loss": 0.3018,
|
| 289508 |
+
"step": 149450
|
| 289509 |
+
},
|
| 289510 |
+
{
|
| 289511 |
+
"epoch": 1194.97,
|
| 289512 |
+
"learning_rate": 7.602649434571891e-06,
|
| 289513 |
+
"loss": 0.4384,
|
| 289514 |
+
"step": 149455
|
| 289515 |
+
},
|
| 289516 |
+
{
|
| 289517 |
+
"epoch": 1195.0,
|
| 289518 |
+
"eval_loss": 0.34299078583717346,
|
| 289519 |
+
"eval_runtime": 43.1632,
|
| 289520 |
+
"eval_samples_per_second": 19.484,
|
| 289521 |
+
"eval_steps_per_second": 0.626,
|
| 289522 |
+
"eval_wer": 0.1696774193548387,
|
| 289523 |
+
"step": 149459
|
| 289524 |
+
},
|
| 289525 |
+
{
|
| 289526 |
+
"epoch": 1195.01,
|
| 289527 |
+
"learning_rate": 7.602568659127626e-06,
|
| 289528 |
+
"loss": 0.3513,
|
| 289529 |
+
"step": 149460
|
| 289530 |
+
},
|
| 289531 |
+
{
|
| 289532 |
+
"epoch": 1195.05,
|
| 289533 |
+
"learning_rate": 7.602487883683361e-06,
|
| 289534 |
+
"loss": 0.2511,
|
| 289535 |
+
"step": 149465
|
| 289536 |
+
},
|
| 289537 |
+
{
|
| 289538 |
+
"epoch": 1195.09,
|
| 289539 |
+
"learning_rate": 7.602407108239096e-06,
|
| 289540 |
+
"loss": 0.2822,
|
| 289541 |
+
"step": 149470
|
| 289542 |
+
},
|
| 289543 |
+
{
|
| 289544 |
+
"epoch": 1195.13,
|
| 289545 |
+
"learning_rate": 7.6023263327948306e-06,
|
| 289546 |
+
"loss": 0.3133,
|
| 289547 |
+
"step": 149475
|
| 289548 |
+
},
|
| 289549 |
+
{
|
| 289550 |
+
"epoch": 1195.17,
|
| 289551 |
+
"learning_rate": 7.602245557350566e-06,
|
| 289552 |
+
"loss": 0.4591,
|
| 289553 |
+
"step": 149480
|
| 289554 |
+
},
|
| 289555 |
+
{
|
| 289556 |
+
"epoch": 1195.21,
|
| 289557 |
+
"learning_rate": 7.602164781906301e-06,
|
| 289558 |
+
"loss": 0.9383,
|
| 289559 |
+
"step": 149485
|
| 289560 |
+
},
|
| 289561 |
+
{
|
| 289562 |
+
"epoch": 1195.25,
|
| 289563 |
+
"learning_rate": 7.602084006462036e-06,
|
| 289564 |
+
"loss": 0.2895,
|
| 289565 |
+
"step": 149490
|
| 289566 |
+
},
|
| 289567 |
+
{
|
| 289568 |
+
"epoch": 1195.29,
|
| 289569 |
+
"learning_rate": 7.602003231017771e-06,
|
| 289570 |
+
"loss": 0.2608,
|
| 289571 |
+
"step": 149495
|
| 289572 |
+
},
|
| 289573 |
+
{
|
| 289574 |
+
"epoch": 1195.33,
|
| 289575 |
+
"learning_rate": 7.601922455573506e-06,
|
| 289576 |
+
"loss": 0.3578,
|
| 289577 |
+
"step": 149500
|
| 289578 |
+
},
|
| 289579 |
+
{
|
| 289580 |
+
"epoch": 1195.37,
|
| 289581 |
+
"learning_rate": 7.601841680129241e-06,
|
| 289582 |
+
"loss": 0.5226,
|
| 289583 |
+
"step": 149505
|
| 289584 |
+
},
|
| 289585 |
+
{
|
| 289586 |
+
"epoch": 1195.41,
|
| 289587 |
+
"learning_rate": 7.601760904684976e-06,
|
| 289588 |
+
"loss": 1.1544,
|
| 289589 |
+
"step": 149510
|
| 289590 |
+
},
|
| 289591 |
+
{
|
| 289592 |
+
"epoch": 1195.45,
|
| 289593 |
+
"learning_rate": 7.601680129240711e-06,
|
| 289594 |
+
"loss": 0.2645,
|
| 289595 |
+
"step": 149515
|
| 289596 |
+
},
|
| 289597 |
+
{
|
| 289598 |
+
"epoch": 1195.49,
|
| 289599 |
+
"learning_rate": 7.601599353796446e-06,
|
| 289600 |
+
"loss": 0.2919,
|
| 289601 |
+
"step": 149520
|
| 289602 |
+
},
|
| 289603 |
+
{
|
| 289604 |
+
"epoch": 1195.53,
|
| 289605 |
+
"learning_rate": 7.601518578352181e-06,
|
| 289606 |
+
"loss": 0.3071,
|
| 289607 |
+
"step": 149525
|
| 289608 |
+
},
|
| 289609 |
+
{
|
| 289610 |
+
"epoch": 1195.57,
|
| 289611 |
+
"learning_rate": 7.601437802907916e-06,
|
| 289612 |
+
"loss": 0.4781,
|
| 289613 |
+
"step": 149530
|
| 289614 |
+
},
|
| 289615 |
+
{
|
| 289616 |
+
"epoch": 1195.61,
|
| 289617 |
+
"learning_rate": 7.601357027463651e-06,
|
| 289618 |
+
"loss": 1.0807,
|
| 289619 |
+
"step": 149535
|
| 289620 |
+
},
|
| 289621 |
+
{
|
| 289622 |
+
"epoch": 1195.65,
|
| 289623 |
+
"learning_rate": 7.601276252019387e-06,
|
| 289624 |
+
"loss": 0.2925,
|
| 289625 |
+
"step": 149540
|
| 289626 |
+
},
|
| 289627 |
+
{
|
| 289628 |
+
"epoch": 1195.69,
|
| 289629 |
+
"learning_rate": 7.601195476575122e-06,
|
| 289630 |
+
"loss": 0.2557,
|
| 289631 |
+
"step": 149545
|
| 289632 |
+
},
|
| 289633 |
+
{
|
| 289634 |
+
"epoch": 1195.73,
|
| 289635 |
+
"learning_rate": 7.601114701130857e-06,
|
| 289636 |
+
"loss": 0.3068,
|
| 289637 |
+
"step": 149550
|
| 289638 |
+
},
|
| 289639 |
+
{
|
| 289640 |
+
"epoch": 1195.77,
|
| 289641 |
+
"learning_rate": 7.601033925686592e-06,
|
| 289642 |
+
"loss": 0.5187,
|
| 289643 |
+
"step": 149555
|
| 289644 |
+
},
|
| 289645 |
+
{
|
| 289646 |
+
"epoch": 1195.81,
|
| 289647 |
+
"learning_rate": 7.600953150242327e-06,
|
| 289648 |
+
"loss": 1.051,
|
| 289649 |
+
"step": 149560
|
| 289650 |
+
},
|
| 289651 |
+
{
|
| 289652 |
+
"epoch": 1195.85,
|
| 289653 |
+
"learning_rate": 7.600872374798062e-06,
|
| 289654 |
+
"loss": 0.2859,
|
| 289655 |
+
"step": 149565
|
| 289656 |
+
},
|
| 289657 |
+
{
|
| 289658 |
+
"epoch": 1195.89,
|
| 289659 |
+
"learning_rate": 7.600791599353797e-06,
|
| 289660 |
+
"loss": 0.287,
|
| 289661 |
+
"step": 149570
|
| 289662 |
+
},
|
| 289663 |
+
{
|
| 289664 |
+
"epoch": 1195.93,
|
| 289665 |
+
"learning_rate": 7.600710823909532e-06,
|
| 289666 |
+
"loss": 0.2811,
|
| 289667 |
+
"step": 149575
|
| 289668 |
+
},
|
| 289669 |
+
{
|
| 289670 |
+
"epoch": 1195.97,
|
| 289671 |
+
"learning_rate": 7.600630048465267e-06,
|
| 289672 |
+
"loss": 0.6269,
|
| 289673 |
+
"step": 149580
|
| 289674 |
+
},
|
| 289675 |
+
{
|
| 289676 |
+
"epoch": 1196.0,
|
| 289677 |
+
"eval_loss": 0.38900554180145264,
|
| 289678 |
+
"eval_runtime": 41.2898,
|
| 289679 |
+
"eval_samples_per_second": 20.344,
|
| 289680 |
+
"eval_steps_per_second": 0.654,
|
| 289681 |
+
"eval_wer": 0.17671063178605262,
|
| 289682 |
+
"step": 149584
|
| 289683 |
}
|
| 289684 |
],
|
| 289685 |
"max_steps": 625000,
|
| 289686 |
"num_train_epochs": 5000,
|
| 289687 |
+
"total_flos": 4.209383244536534e+20,
|
| 289688 |
"trial_name": null,
|
| 289689 |
"trial_params": null
|
| 289690 |
}
|
model-bin/finetune/base/{checkpoint-148960 → checkpoint-149584}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630239529.1803105/events.out.tfevents.1630239529.cc93b136ebf5.1086.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76a6c607bb1f273f37cfa4e901082ce17f3208dcdef15c5e00ffeec5942c14a5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630239966.3963904/events.out.tfevents.1630239966.cc93b136ebf5.1086.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:465503fdbfc5d14fe84de9165f5655f2f41899d4710673d47929361d53f4da9f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630240411.0957363/events.out.tfevents.1630240411.cc93b136ebf5.1086.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eefc1989d1ee8eaa00563a2fcb553a484a28403c140f2956db55c0815692238f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630240854.716574/events.out.tfevents.1630240854.cc93b136ebf5.1086.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e436a9bd18839678c4870a6f0e8d53b675b0be4ddc7c81d238bab21a45b5ced
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630241290.4871612/events.out.tfevents.1630241290.cc93b136ebf5.1086.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9935ffb6871781f5fde92a2ca9aa1ef0ee7ebe9d4a003bc62d7e63ae5d489e52
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630239529.cc93b136ebf5.1086.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7bd992b4010c4d02a462be43082a58163dfa1264a8431572032b88b2f642b13
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630239966.cc93b136ebf5.1086.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:edeeff26e0625c2de04d71d445030c8743f15f661e060dc708d0eca351945718
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630240411.cc93b136ebf5.1086.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c695844c0087efb4d224e5d10d7a2b51e5f3d7d05abf5a41d9ef95ed6b6d803
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630240854.cc93b136ebf5.1086.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d9bc797344cccc2e0dc808a81839738ed492e23658851f6af1dddd0fad7633fc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630241290.cc93b136ebf5.1086.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc7eac72b2bde57d0a474806abbceee03f3ed33d14024fb4c1146660850929de
|
| 3 |
+
size 8622
|