"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629811388.0786695/events.out.tfevents.1629811388.c435e1c5ee04.920.141 +3 -0
- model-bin/finetune/base/log/1629812021.3954506/events.out.tfevents.1629812021.c435e1c5ee04.920.143 +3 -0
- model-bin/finetune/base/log/1629812663.6763268/events.out.tfevents.1629812663.c435e1c5ee04.920.145 +3 -0
- model-bin/finetune/base/log/1629813319.9052355/events.out.tfevents.1629813319.c435e1c5ee04.920.147 +3 -0
- model-bin/finetune/base/log/1629813976.4157386/events.out.tfevents.1629813976.c435e1c5ee04.920.149 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629811388.c435e1c5ee04.920.140 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629812021.c435e1c5ee04.920.142 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629812663.c435e1c5ee04.920.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629813319.c435e1c5ee04.920.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629813976.c435e1c5ee04.920.148 +3 -0
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5ed31d665ca056cfff755fe3f18ed48e5a01d3602d164496c1af7ac8fcc8ba54
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/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 sha256:0365bb561b98cbc60fefb1ba606c6f7f93b32d0a19ebae812f3252fe5c646f85
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/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 sha256:7d235db26337f9932a1024726950ac14182c0e5561ce0dd571368c9738da8869
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/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:ca871593ba41366731da66f366fb0722e2e31e792097384b0184e23d00ca763e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/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 sha256:7db59327c50f766d3f36f138d8cebc0d39dc04985d598bf70e4f3014370adc34
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -179001,11 +179001,806 @@
|
|
| 179001 |
"eval_steps_per_second": 0.68,
|
| 179002 |
"eval_wer": 0.2003999111308598,
|
| 179003 |
"step": 62593
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 179004 |
}
|
| 179005 |
],
|
| 179006 |
-
"max_steps":
|
| 179007 |
"num_train_epochs": 5000,
|
| 179008 |
-
"total_flos": 1.
|
| 179009 |
"trial_name": null,
|
| 179010 |
"trial_params": null
|
| 179011 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 505.0,
|
| 5 |
+
"global_step": 63215,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 179001 |
"eval_steps_per_second": 0.68,
|
| 179002 |
"eval_wer": 0.2003999111308598,
|
| 179003 |
"step": 62593
|
| 179004 |
+
},
|
| 179005 |
+
{
|
| 179006 |
+
"epoch": 504.02,
|
| 179007 |
+
"learning_rate": 9.013269230769231e-06,
|
| 179008 |
+
"loss": 0.4336,
|
| 179009 |
+
"step": 62595
|
| 179010 |
+
},
|
| 179011 |
+
{
|
| 179012 |
+
"epoch": 504.06,
|
| 179013 |
+
"learning_rate": 9.013189102564104e-06,
|
| 179014 |
+
"loss": 0.3401,
|
| 179015 |
+
"step": 62600
|
| 179016 |
+
},
|
| 179017 |
+
{
|
| 179018 |
+
"epoch": 504.1,
|
| 179019 |
+
"learning_rate": 9.013108974358976e-06,
|
| 179020 |
+
"loss": 0.3122,
|
| 179021 |
+
"step": 62605
|
| 179022 |
+
},
|
| 179023 |
+
{
|
| 179024 |
+
"epoch": 504.14,
|
| 179025 |
+
"learning_rate": 9.013028846153847e-06,
|
| 179026 |
+
"loss": 0.37,
|
| 179027 |
+
"step": 62610
|
| 179028 |
+
},
|
| 179029 |
+
{
|
| 179030 |
+
"epoch": 504.18,
|
| 179031 |
+
"learning_rate": 9.012948717948719e-06,
|
| 179032 |
+
"loss": 0.6859,
|
| 179033 |
+
"step": 62615
|
| 179034 |
+
},
|
| 179035 |
+
{
|
| 179036 |
+
"epoch": 504.22,
|
| 179037 |
+
"learning_rate": 9.012868589743592e-06,
|
| 179038 |
+
"loss": 1.1365,
|
| 179039 |
+
"step": 62620
|
| 179040 |
+
},
|
| 179041 |
+
{
|
| 179042 |
+
"epoch": 504.26,
|
| 179043 |
+
"learning_rate": 9.012788461538461e-06,
|
| 179044 |
+
"loss": 0.379,
|
| 179045 |
+
"step": 62625
|
| 179046 |
+
},
|
| 179047 |
+
{
|
| 179048 |
+
"epoch": 504.3,
|
| 179049 |
+
"learning_rate": 9.012708333333334e-06,
|
| 179050 |
+
"loss": 0.3206,
|
| 179051 |
+
"step": 62630
|
| 179052 |
+
},
|
| 179053 |
+
{
|
| 179054 |
+
"epoch": 504.34,
|
| 179055 |
+
"learning_rate": 9.012628205128206e-06,
|
| 179056 |
+
"loss": 0.3921,
|
| 179057 |
+
"step": 62635
|
| 179058 |
+
},
|
| 179059 |
+
{
|
| 179060 |
+
"epoch": 504.38,
|
| 179061 |
+
"learning_rate": 9.012548076923077e-06,
|
| 179062 |
+
"loss": 0.7003,
|
| 179063 |
+
"step": 62640
|
| 179064 |
+
},
|
| 179065 |
+
{
|
| 179066 |
+
"epoch": 504.42,
|
| 179067 |
+
"learning_rate": 9.012467948717948e-06,
|
| 179068 |
+
"loss": 1.0953,
|
| 179069 |
+
"step": 62645
|
| 179070 |
+
},
|
| 179071 |
+
{
|
| 179072 |
+
"epoch": 504.46,
|
| 179073 |
+
"learning_rate": 9.012387820512821e-06,
|
| 179074 |
+
"loss": 0.3642,
|
| 179075 |
+
"step": 62650
|
| 179076 |
+
},
|
| 179077 |
+
{
|
| 179078 |
+
"epoch": 504.5,
|
| 179079 |
+
"learning_rate": 9.012307692307693e-06,
|
| 179080 |
+
"loss": 0.2942,
|
| 179081 |
+
"step": 62655
|
| 179082 |
+
},
|
| 179083 |
+
{
|
| 179084 |
+
"epoch": 504.54,
|
| 179085 |
+
"learning_rate": 9.012227564102564e-06,
|
| 179086 |
+
"loss": 0.3649,
|
| 179087 |
+
"step": 62660
|
| 179088 |
+
},
|
| 179089 |
+
{
|
| 179090 |
+
"epoch": 504.58,
|
| 179091 |
+
"learning_rate": 9.012147435897437e-06,
|
| 179092 |
+
"loss": 0.7285,
|
| 179093 |
+
"step": 62665
|
| 179094 |
+
},
|
| 179095 |
+
{
|
| 179096 |
+
"epoch": 504.62,
|
| 179097 |
+
"learning_rate": 9.012067307692309e-06,
|
| 179098 |
+
"loss": 1.0353,
|
| 179099 |
+
"step": 62670
|
| 179100 |
+
},
|
| 179101 |
+
{
|
| 179102 |
+
"epoch": 504.66,
|
| 179103 |
+
"learning_rate": 9.01198717948718e-06,
|
| 179104 |
+
"loss": 0.3504,
|
| 179105 |
+
"step": 62675
|
| 179106 |
+
},
|
| 179107 |
+
{
|
| 179108 |
+
"epoch": 504.7,
|
| 179109 |
+
"learning_rate": 9.011907051282051e-06,
|
| 179110 |
+
"loss": 0.3285,
|
| 179111 |
+
"step": 62680
|
| 179112 |
+
},
|
| 179113 |
+
{
|
| 179114 |
+
"epoch": 504.74,
|
| 179115 |
+
"learning_rate": 9.011826923076924e-06,
|
| 179116 |
+
"loss": 0.3982,
|
| 179117 |
+
"step": 62685
|
| 179118 |
+
},
|
| 179119 |
+
{
|
| 179120 |
+
"epoch": 504.78,
|
| 179121 |
+
"learning_rate": 9.011746794871796e-06,
|
| 179122 |
+
"loss": 0.6853,
|
| 179123 |
+
"step": 62690
|
| 179124 |
+
},
|
| 179125 |
+
{
|
| 179126 |
+
"epoch": 504.82,
|
| 179127 |
+
"learning_rate": 9.011666666666667e-06,
|
| 179128 |
+
"loss": 1.1931,
|
| 179129 |
+
"step": 62695
|
| 179130 |
+
},
|
| 179131 |
+
{
|
| 179132 |
+
"epoch": 504.86,
|
| 179133 |
+
"learning_rate": 9.01158653846154e-06,
|
| 179134 |
+
"loss": 0.3007,
|
| 179135 |
+
"step": 62700
|
| 179136 |
+
},
|
| 179137 |
+
{
|
| 179138 |
+
"epoch": 504.9,
|
| 179139 |
+
"learning_rate": 9.011506410256411e-06,
|
| 179140 |
+
"loss": 0.3688,
|
| 179141 |
+
"step": 62705
|
| 179142 |
+
},
|
| 179143 |
+
{
|
| 179144 |
+
"epoch": 504.94,
|
| 179145 |
+
"learning_rate": 9.011426282051283e-06,
|
| 179146 |
+
"loss": 0.4134,
|
| 179147 |
+
"step": 62710
|
| 179148 |
+
},
|
| 179149 |
+
{
|
| 179150 |
+
"epoch": 504.98,
|
| 179151 |
+
"learning_rate": 9.011346153846154e-06,
|
| 179152 |
+
"loss": 0.6534,
|
| 179153 |
+
"step": 62715
|
| 179154 |
+
},
|
| 179155 |
+
{
|
| 179156 |
+
"epoch": 505.0,
|
| 179157 |
+
"eval_loss": 0.37348735332489014,
|
| 179158 |
+
"eval_runtime": 40.0793,
|
| 179159 |
+
"eval_samples_per_second": 21.058,
|
| 179160 |
+
"eval_steps_per_second": 0.674,
|
| 179161 |
+
"eval_wer": 0.1937600233270156,
|
| 179162 |
+
"step": 62717
|
| 179163 |
+
},
|
| 179164 |
+
{
|
| 179165 |
+
"epoch": 505.02,
|
| 179166 |
+
"learning_rate": 9.011266025641027e-06,
|
| 179167 |
+
"loss": 0.3351,
|
| 179168 |
+
"step": 62720
|
| 179169 |
+
},
|
| 179170 |
+
{
|
| 179171 |
+
"epoch": 505.06,
|
| 179172 |
+
"learning_rate": 9.011185897435899e-06,
|
| 179173 |
+
"loss": 0.372,
|
| 179174 |
+
"step": 62725
|
| 179175 |
+
},
|
| 179176 |
+
{
|
| 179177 |
+
"epoch": 505.1,
|
| 179178 |
+
"learning_rate": 9.01110576923077e-06,
|
| 179179 |
+
"loss": 0.2816,
|
| 179180 |
+
"step": 62730
|
| 179181 |
+
},
|
| 179182 |
+
{
|
| 179183 |
+
"epoch": 505.15,
|
| 179184 |
+
"learning_rate": 9.011025641025641e-06,
|
| 179185 |
+
"loss": 0.3572,
|
| 179186 |
+
"step": 62735
|
| 179187 |
+
},
|
| 179188 |
+
{
|
| 179189 |
+
"epoch": 505.19,
|
| 179190 |
+
"learning_rate": 9.010945512820514e-06,
|
| 179191 |
+
"loss": 0.7106,
|
| 179192 |
+
"step": 62740
|
| 179193 |
+
},
|
| 179194 |
+
{
|
| 179195 |
+
"epoch": 505.23,
|
| 179196 |
+
"learning_rate": 9.010865384615384e-06,
|
| 179197 |
+
"loss": 0.9575,
|
| 179198 |
+
"step": 62745
|
| 179199 |
+
},
|
| 179200 |
+
{
|
| 179201 |
+
"epoch": 505.27,
|
| 179202 |
+
"learning_rate": 9.010785256410257e-06,
|
| 179203 |
+
"loss": 0.3145,
|
| 179204 |
+
"step": 62750
|
| 179205 |
+
},
|
| 179206 |
+
{
|
| 179207 |
+
"epoch": 505.31,
|
| 179208 |
+
"learning_rate": 9.01070512820513e-06,
|
| 179209 |
+
"loss": 0.3618,
|
| 179210 |
+
"step": 62755
|
| 179211 |
+
},
|
| 179212 |
+
{
|
| 179213 |
+
"epoch": 505.35,
|
| 179214 |
+
"learning_rate": 9.010625e-06,
|
| 179215 |
+
"loss": 0.3789,
|
| 179216 |
+
"step": 62760
|
| 179217 |
+
},
|
| 179218 |
+
{
|
| 179219 |
+
"epoch": 505.39,
|
| 179220 |
+
"learning_rate": 9.010544871794873e-06,
|
| 179221 |
+
"loss": 0.8679,
|
| 179222 |
+
"step": 62765
|
| 179223 |
+
},
|
| 179224 |
+
{
|
| 179225 |
+
"epoch": 505.43,
|
| 179226 |
+
"learning_rate": 9.010464743589744e-06,
|
| 179227 |
+
"loss": 0.9309,
|
| 179228 |
+
"step": 62770
|
| 179229 |
+
},
|
| 179230 |
+
{
|
| 179231 |
+
"epoch": 505.47,
|
| 179232 |
+
"learning_rate": 9.010384615384616e-06,
|
| 179233 |
+
"loss": 0.3612,
|
| 179234 |
+
"step": 62775
|
| 179235 |
+
},
|
| 179236 |
+
{
|
| 179237 |
+
"epoch": 505.51,
|
| 179238 |
+
"learning_rate": 9.010304487179487e-06,
|
| 179239 |
+
"loss": 0.3099,
|
| 179240 |
+
"step": 62780
|
| 179241 |
+
},
|
| 179242 |
+
{
|
| 179243 |
+
"epoch": 505.55,
|
| 179244 |
+
"learning_rate": 9.01022435897436e-06,
|
| 179245 |
+
"loss": 0.361,
|
| 179246 |
+
"step": 62785
|
| 179247 |
+
},
|
| 179248 |
+
{
|
| 179249 |
+
"epoch": 505.59,
|
| 179250 |
+
"learning_rate": 9.010144230769231e-06,
|
| 179251 |
+
"loss": 0.6729,
|
| 179252 |
+
"step": 62790
|
| 179253 |
+
},
|
| 179254 |
+
{
|
| 179255 |
+
"epoch": 505.63,
|
| 179256 |
+
"learning_rate": 9.010064102564103e-06,
|
| 179257 |
+
"loss": 0.8956,
|
| 179258 |
+
"step": 62795
|
| 179259 |
+
},
|
| 179260 |
+
{
|
| 179261 |
+
"epoch": 505.67,
|
| 179262 |
+
"learning_rate": 9.009983974358976e-06,
|
| 179263 |
+
"loss": 0.3684,
|
| 179264 |
+
"step": 62800
|
| 179265 |
+
},
|
| 179266 |
+
{
|
| 179267 |
+
"epoch": 505.71,
|
| 179268 |
+
"learning_rate": 9.009903846153847e-06,
|
| 179269 |
+
"loss": 0.3877,
|
| 179270 |
+
"step": 62805
|
| 179271 |
+
},
|
| 179272 |
+
{
|
| 179273 |
+
"epoch": 505.75,
|
| 179274 |
+
"learning_rate": 9.009823717948718e-06,
|
| 179275 |
+
"loss": 0.4419,
|
| 179276 |
+
"step": 62810
|
| 179277 |
+
},
|
| 179278 |
+
{
|
| 179279 |
+
"epoch": 505.79,
|
| 179280 |
+
"learning_rate": 9.00974358974359e-06,
|
| 179281 |
+
"loss": 0.8221,
|
| 179282 |
+
"step": 62815
|
| 179283 |
+
},
|
| 179284 |
+
{
|
| 179285 |
+
"epoch": 505.83,
|
| 179286 |
+
"learning_rate": 9.009663461538463e-06,
|
| 179287 |
+
"loss": 0.9572,
|
| 179288 |
+
"step": 62820
|
| 179289 |
+
},
|
| 179290 |
+
{
|
| 179291 |
+
"epoch": 505.87,
|
| 179292 |
+
"learning_rate": 9.009583333333334e-06,
|
| 179293 |
+
"loss": 0.3106,
|
| 179294 |
+
"step": 62825
|
| 179295 |
+
},
|
| 179296 |
+
{
|
| 179297 |
+
"epoch": 505.91,
|
| 179298 |
+
"learning_rate": 9.009503205128206e-06,
|
| 179299 |
+
"loss": 0.3388,
|
| 179300 |
+
"step": 62830
|
| 179301 |
+
},
|
| 179302 |
+
{
|
| 179303 |
+
"epoch": 505.95,
|
| 179304 |
+
"learning_rate": 9.009423076923077e-06,
|
| 179305 |
+
"loss": 0.3729,
|
| 179306 |
+
"step": 62835
|
| 179307 |
+
},
|
| 179308 |
+
{
|
| 179309 |
+
"epoch": 505.99,
|
| 179310 |
+
"learning_rate": 9.00934294871795e-06,
|
| 179311 |
+
"loss": 0.8673,
|
| 179312 |
+
"step": 62840
|
| 179313 |
+
},
|
| 179314 |
+
{
|
| 179315 |
+
"epoch": 506.0,
|
| 179316 |
+
"eval_loss": 0.4431955814361572,
|
| 179317 |
+
"eval_runtime": 39.2969,
|
| 179318 |
+
"eval_samples_per_second": 21.478,
|
| 179319 |
+
"eval_steps_per_second": 0.687,
|
| 179320 |
+
"eval_wer": 0.21121718377088305,
|
| 179321 |
+
"step": 62841
|
| 179322 |
+
},
|
| 179323 |
+
{
|
| 179324 |
+
"epoch": 506.03,
|
| 179325 |
+
"learning_rate": 9.009262820512821e-06,
|
| 179326 |
+
"loss": 0.3755,
|
| 179327 |
+
"step": 62845
|
| 179328 |
+
},
|
| 179329 |
+
{
|
| 179330 |
+
"epoch": 506.07,
|
| 179331 |
+
"learning_rate": 9.009182692307693e-06,
|
| 179332 |
+
"loss": 0.296,
|
| 179333 |
+
"step": 62850
|
| 179334 |
+
},
|
| 179335 |
+
{
|
| 179336 |
+
"epoch": 506.11,
|
| 179337 |
+
"learning_rate": 9.009102564102566e-06,
|
| 179338 |
+
"loss": 0.3318,
|
| 179339 |
+
"step": 62855
|
| 179340 |
+
},
|
| 179341 |
+
{
|
| 179342 |
+
"epoch": 506.15,
|
| 179343 |
+
"learning_rate": 9.009022435897437e-06,
|
| 179344 |
+
"loss": 0.4693,
|
| 179345 |
+
"step": 62860
|
| 179346 |
+
},
|
| 179347 |
+
{
|
| 179348 |
+
"epoch": 506.19,
|
| 179349 |
+
"learning_rate": 9.008942307692309e-06,
|
| 179350 |
+
"loss": 0.8523,
|
| 179351 |
+
"step": 62865
|
| 179352 |
+
},
|
| 179353 |
+
{
|
| 179354 |
+
"epoch": 506.23,
|
| 179355 |
+
"learning_rate": 9.00886217948718e-06,
|
| 179356 |
+
"loss": 0.7526,
|
| 179357 |
+
"step": 62870
|
| 179358 |
+
},
|
| 179359 |
+
{
|
| 179360 |
+
"epoch": 506.27,
|
| 179361 |
+
"learning_rate": 9.008782051282053e-06,
|
| 179362 |
+
"loss": 0.3106,
|
| 179363 |
+
"step": 62875
|
| 179364 |
+
},
|
| 179365 |
+
{
|
| 179366 |
+
"epoch": 506.31,
|
| 179367 |
+
"learning_rate": 9.008701923076924e-06,
|
| 179368 |
+
"loss": 0.3951,
|
| 179369 |
+
"step": 62880
|
| 179370 |
+
},
|
| 179371 |
+
{
|
| 179372 |
+
"epoch": 506.35,
|
| 179373 |
+
"learning_rate": 9.008621794871796e-06,
|
| 179374 |
+
"loss": 0.5123,
|
| 179375 |
+
"step": 62885
|
| 179376 |
+
},
|
| 179377 |
+
{
|
| 179378 |
+
"epoch": 506.4,
|
| 179379 |
+
"learning_rate": 9.008541666666667e-06,
|
| 179380 |
+
"loss": 1.0387,
|
| 179381 |
+
"step": 62890
|
| 179382 |
+
},
|
| 179383 |
+
{
|
| 179384 |
+
"epoch": 506.44,
|
| 179385 |
+
"learning_rate": 9.00846153846154e-06,
|
| 179386 |
+
"loss": 0.7079,
|
| 179387 |
+
"step": 62895
|
| 179388 |
+
},
|
| 179389 |
+
{
|
| 179390 |
+
"epoch": 506.48,
|
| 179391 |
+
"learning_rate": 9.00838141025641e-06,
|
| 179392 |
+
"loss": 0.293,
|
| 179393 |
+
"step": 62900
|
| 179394 |
+
},
|
| 179395 |
+
{
|
| 179396 |
+
"epoch": 506.52,
|
| 179397 |
+
"learning_rate": 9.008301282051283e-06,
|
| 179398 |
+
"loss": 0.3958,
|
| 179399 |
+
"step": 62905
|
| 179400 |
+
},
|
| 179401 |
+
{
|
| 179402 |
+
"epoch": 506.56,
|
| 179403 |
+
"learning_rate": 9.008221153846156e-06,
|
| 179404 |
+
"loss": 0.4341,
|
| 179405 |
+
"step": 62910
|
| 179406 |
+
},
|
| 179407 |
+
{
|
| 179408 |
+
"epoch": 506.6,
|
| 179409 |
+
"learning_rate": 9.008141025641025e-06,
|
| 179410 |
+
"loss": 1.0019,
|
| 179411 |
+
"step": 62915
|
| 179412 |
+
},
|
| 179413 |
+
{
|
| 179414 |
+
"epoch": 506.64,
|
| 179415 |
+
"learning_rate": 9.008060897435899e-06,
|
| 179416 |
+
"loss": 0.7332,
|
| 179417 |
+
"step": 62920
|
| 179418 |
+
},
|
| 179419 |
+
{
|
| 179420 |
+
"epoch": 506.68,
|
| 179421 |
+
"learning_rate": 9.00798076923077e-06,
|
| 179422 |
+
"loss": 0.3169,
|
| 179423 |
+
"step": 62925
|
| 179424 |
+
},
|
| 179425 |
+
{
|
| 179426 |
+
"epoch": 506.72,
|
| 179427 |
+
"learning_rate": 9.007900641025641e-06,
|
| 179428 |
+
"loss": 0.3726,
|
| 179429 |
+
"step": 62930
|
| 179430 |
+
},
|
| 179431 |
+
{
|
| 179432 |
+
"epoch": 506.76,
|
| 179433 |
+
"learning_rate": 9.007820512820513e-06,
|
| 179434 |
+
"loss": 0.4221,
|
| 179435 |
+
"step": 62935
|
| 179436 |
+
},
|
| 179437 |
+
{
|
| 179438 |
+
"epoch": 506.8,
|
| 179439 |
+
"learning_rate": 9.007740384615386e-06,
|
| 179440 |
+
"loss": 1.0042,
|
| 179441 |
+
"step": 62940
|
| 179442 |
+
},
|
| 179443 |
+
{
|
| 179444 |
+
"epoch": 506.84,
|
| 179445 |
+
"learning_rate": 9.007660256410257e-06,
|
| 179446 |
+
"loss": 0.6921,
|
| 179447 |
+
"step": 62945
|
| 179448 |
+
},
|
| 179449 |
+
{
|
| 179450 |
+
"epoch": 506.88,
|
| 179451 |
+
"learning_rate": 9.007580128205128e-06,
|
| 179452 |
+
"loss": 0.2953,
|
| 179453 |
+
"step": 62950
|
| 179454 |
+
},
|
| 179455 |
+
{
|
| 179456 |
+
"epoch": 506.92,
|
| 179457 |
+
"learning_rate": 9.007500000000001e-06,
|
| 179458 |
+
"loss": 0.3853,
|
| 179459 |
+
"step": 62955
|
| 179460 |
+
},
|
| 179461 |
+
{
|
| 179462 |
+
"epoch": 506.96,
|
| 179463 |
+
"learning_rate": 9.007419871794873e-06,
|
| 179464 |
+
"loss": 0.7195,
|
| 179465 |
+
"step": 62960
|
| 179466 |
+
},
|
| 179467 |
+
{
|
| 179468 |
+
"epoch": 507.0,
|
| 179469 |
+
"learning_rate": 9.007339743589744e-06,
|
| 179470 |
+
"loss": 1.3909,
|
| 179471 |
+
"step": 62965
|
| 179472 |
+
},
|
| 179473 |
+
{
|
| 179474 |
+
"epoch": 507.0,
|
| 179475 |
+
"eval_loss": 0.40501317381858826,
|
| 179476 |
+
"eval_runtime": 39.635,
|
| 179477 |
+
"eval_samples_per_second": 21.294,
|
| 179478 |
+
"eval_steps_per_second": 0.681,
|
| 179479 |
+
"eval_wer": 0.19239262055701972,
|
| 179480 |
+
"step": 62965
|
| 179481 |
+
},
|
| 179482 |
+
{
|
| 179483 |
+
"epoch": 503.04,
|
| 179484 |
+
"learning_rate": 9.007259615384616e-06,
|
| 179485 |
+
"loss": 0.3549,
|
| 179486 |
+
"step": 62970
|
| 179487 |
+
},
|
| 179488 |
+
{
|
| 179489 |
+
"epoch": 503.08,
|
| 179490 |
+
"learning_rate": 9.007179487179489e-06,
|
| 179491 |
+
"loss": 0.3541,
|
| 179492 |
+
"step": 62975
|
| 179493 |
+
},
|
| 179494 |
+
{
|
| 179495 |
+
"epoch": 503.12,
|
| 179496 |
+
"learning_rate": 9.00709935897436e-06,
|
| 179497 |
+
"loss": 0.3363,
|
| 179498 |
+
"step": 62980
|
| 179499 |
+
},
|
| 179500 |
+
{
|
| 179501 |
+
"epoch": 503.16,
|
| 179502 |
+
"learning_rate": 9.007019230769231e-06,
|
| 179503 |
+
"loss": 0.5528,
|
| 179504 |
+
"step": 62985
|
| 179505 |
+
},
|
| 179506 |
+
{
|
| 179507 |
+
"epoch": 503.2,
|
| 179508 |
+
"learning_rate": 9.006939102564103e-06,
|
| 179509 |
+
"loss": 1.1822,
|
| 179510 |
+
"step": 62990
|
| 179511 |
+
},
|
| 179512 |
+
{
|
| 179513 |
+
"epoch": 503.24,
|
| 179514 |
+
"learning_rate": 9.006858974358976e-06,
|
| 179515 |
+
"loss": 0.3426,
|
| 179516 |
+
"step": 62995
|
| 179517 |
+
},
|
| 179518 |
+
{
|
| 179519 |
+
"epoch": 503.28,
|
| 179520 |
+
"learning_rate": 9.006778846153847e-06,
|
| 179521 |
+
"loss": 0.3087,
|
| 179522 |
+
"step": 63000
|
| 179523 |
+
},
|
| 179524 |
+
{
|
| 179525 |
+
"epoch": 503.32,
|
| 179526 |
+
"learning_rate": 9.006698717948718e-06,
|
| 179527 |
+
"loss": 0.337,
|
| 179528 |
+
"step": 63005
|
| 179529 |
+
},
|
| 179530 |
+
{
|
| 179531 |
+
"epoch": 503.36,
|
| 179532 |
+
"learning_rate": 9.006618589743591e-06,
|
| 179533 |
+
"loss": 0.4048,
|
| 179534 |
+
"step": 63010
|
| 179535 |
+
},
|
| 179536 |
+
{
|
| 179537 |
+
"epoch": 503.4,
|
| 179538 |
+
"learning_rate": 9.006538461538463e-06,
|
| 179539 |
+
"loss": 1.2429,
|
| 179540 |
+
"step": 63015
|
| 179541 |
+
},
|
| 179542 |
+
{
|
| 179543 |
+
"epoch": 503.44,
|
| 179544 |
+
"learning_rate": 9.006458333333334e-06,
|
| 179545 |
+
"loss": 0.3723,
|
| 179546 |
+
"step": 63020
|
| 179547 |
+
},
|
| 179548 |
+
{
|
| 179549 |
+
"epoch": 503.48,
|
| 179550 |
+
"learning_rate": 9.006378205128206e-06,
|
| 179551 |
+
"loss": 0.3246,
|
| 179552 |
+
"step": 63025
|
| 179553 |
+
},
|
| 179554 |
+
{
|
| 179555 |
+
"epoch": 503.52,
|
| 179556 |
+
"learning_rate": 9.006298076923079e-06,
|
| 179557 |
+
"loss": 0.3242,
|
| 179558 |
+
"step": 63030
|
| 179559 |
+
},
|
| 179560 |
+
{
|
| 179561 |
+
"epoch": 503.56,
|
| 179562 |
+
"learning_rate": 9.006217948717948e-06,
|
| 179563 |
+
"loss": 0.4674,
|
| 179564 |
+
"step": 63035
|
| 179565 |
+
},
|
| 179566 |
+
{
|
| 179567 |
+
"epoch": 503.6,
|
| 179568 |
+
"learning_rate": 9.006137820512821e-06,
|
| 179569 |
+
"loss": 1.3906,
|
| 179570 |
+
"step": 63040
|
| 179571 |
+
},
|
| 179572 |
+
{
|
| 179573 |
+
"epoch": 503.64,
|
| 179574 |
+
"learning_rate": 9.006057692307693e-06,
|
| 179575 |
+
"loss": 0.377,
|
| 179576 |
+
"step": 63045
|
| 179577 |
+
},
|
| 179578 |
+
{
|
| 179579 |
+
"epoch": 503.68,
|
| 179580 |
+
"learning_rate": 9.005977564102564e-06,
|
| 179581 |
+
"loss": 0.335,
|
| 179582 |
+
"step": 63050
|
| 179583 |
+
},
|
| 179584 |
+
{
|
| 179585 |
+
"epoch": 503.72,
|
| 179586 |
+
"learning_rate": 9.005897435897437e-06,
|
| 179587 |
+
"loss": 0.326,
|
| 179588 |
+
"step": 63055
|
| 179589 |
+
},
|
| 179590 |
+
{
|
| 179591 |
+
"epoch": 503.76,
|
| 179592 |
+
"learning_rate": 9.005817307692308e-06,
|
| 179593 |
+
"loss": 0.4857,
|
| 179594 |
+
"step": 63060
|
| 179595 |
+
},
|
| 179596 |
+
{
|
| 179597 |
+
"epoch": 503.8,
|
| 179598 |
+
"learning_rate": 9.00573717948718e-06,
|
| 179599 |
+
"loss": 1.2913,
|
| 179600 |
+
"step": 63065
|
| 179601 |
+
},
|
| 179602 |
+
{
|
| 179603 |
+
"epoch": 503.84,
|
| 179604 |
+
"learning_rate": 9.005657051282051e-06,
|
| 179605 |
+
"loss": 0.3554,
|
| 179606 |
+
"step": 63070
|
| 179607 |
+
},
|
| 179608 |
+
{
|
| 179609 |
+
"epoch": 503.88,
|
| 179610 |
+
"learning_rate": 9.005576923076924e-06,
|
| 179611 |
+
"loss": 0.3489,
|
| 179612 |
+
"step": 63075
|
| 179613 |
+
},
|
| 179614 |
+
{
|
| 179615 |
+
"epoch": 503.92,
|
| 179616 |
+
"learning_rate": 9.005496794871796e-06,
|
| 179617 |
+
"loss": 0.3709,
|
| 179618 |
+
"step": 63080
|
| 179619 |
+
},
|
| 179620 |
+
{
|
| 179621 |
+
"epoch": 503.96,
|
| 179622 |
+
"learning_rate": 9.005416666666667e-06,
|
| 179623 |
+
"loss": 0.4934,
|
| 179624 |
+
"step": 63085
|
| 179625 |
+
},
|
| 179626 |
+
{
|
| 179627 |
+
"epoch": 504.0,
|
| 179628 |
+
"learning_rate": 9.005336538461538e-06,
|
| 179629 |
+
"loss": 1.1345,
|
| 179630 |
+
"step": 63090
|
| 179631 |
+
},
|
| 179632 |
+
{
|
| 179633 |
+
"epoch": 504.0,
|
| 179634 |
+
"eval_loss": 0.4057326912879944,
|
| 179635 |
+
"eval_runtime": 39.8834,
|
| 179636 |
+
"eval_samples_per_second": 21.187,
|
| 179637 |
+
"eval_steps_per_second": 0.677,
|
| 179638 |
+
"eval_wer": 0.1971050882504985,
|
| 179639 |
+
"step": 63090
|
| 179640 |
+
},
|
| 179641 |
+
{
|
| 179642 |
+
"epoch": 504.04,
|
| 179643 |
+
"learning_rate": 9.005256410256411e-06,
|
| 179644 |
+
"loss": 0.3216,
|
| 179645 |
+
"step": 63095
|
| 179646 |
+
},
|
| 179647 |
+
{
|
| 179648 |
+
"epoch": 504.08,
|
| 179649 |
+
"learning_rate": 9.005176282051283e-06,
|
| 179650 |
+
"loss": 0.3453,
|
| 179651 |
+
"step": 63100
|
| 179652 |
+
},
|
| 179653 |
+
{
|
| 179654 |
+
"epoch": 504.12,
|
| 179655 |
+
"learning_rate": 9.005096153846154e-06,
|
| 179656 |
+
"loss": 0.2979,
|
| 179657 |
+
"step": 63105
|
| 179658 |
+
},
|
| 179659 |
+
{
|
| 179660 |
+
"epoch": 504.16,
|
| 179661 |
+
"learning_rate": 9.005016025641027e-06,
|
| 179662 |
+
"loss": 0.5189,
|
| 179663 |
+
"step": 63110
|
| 179664 |
+
},
|
| 179665 |
+
{
|
| 179666 |
+
"epoch": 504.2,
|
| 179667 |
+
"learning_rate": 9.004935897435898e-06,
|
| 179668 |
+
"loss": 1.2128,
|
| 179669 |
+
"step": 63115
|
| 179670 |
+
},
|
| 179671 |
+
{
|
| 179672 |
+
"epoch": 504.24,
|
| 179673 |
+
"learning_rate": 9.00485576923077e-06,
|
| 179674 |
+
"loss": 0.485,
|
| 179675 |
+
"step": 63120
|
| 179676 |
+
},
|
| 179677 |
+
{
|
| 179678 |
+
"epoch": 504.28,
|
| 179679 |
+
"learning_rate": 9.004775641025641e-06,
|
| 179680 |
+
"loss": 0.314,
|
| 179681 |
+
"step": 63125
|
| 179682 |
+
},
|
| 179683 |
+
{
|
| 179684 |
+
"epoch": 504.32,
|
| 179685 |
+
"learning_rate": 9.004695512820514e-06,
|
| 179686 |
+
"loss": 0.3533,
|
| 179687 |
+
"step": 63130
|
| 179688 |
+
},
|
| 179689 |
+
{
|
| 179690 |
+
"epoch": 504.36,
|
| 179691 |
+
"learning_rate": 9.004615384615386e-06,
|
| 179692 |
+
"loss": 0.5181,
|
| 179693 |
+
"step": 63135
|
| 179694 |
+
},
|
| 179695 |
+
{
|
| 179696 |
+
"epoch": 504.4,
|
| 179697 |
+
"learning_rate": 9.004535256410257e-06,
|
| 179698 |
+
"loss": 1.4735,
|
| 179699 |
+
"step": 63140
|
| 179700 |
+
},
|
| 179701 |
+
{
|
| 179702 |
+
"epoch": 504.44,
|
| 179703 |
+
"learning_rate": 9.004455128205128e-06,
|
| 179704 |
+
"loss": 0.3573,
|
| 179705 |
+
"step": 63145
|
| 179706 |
+
},
|
| 179707 |
+
{
|
| 179708 |
+
"epoch": 504.48,
|
| 179709 |
+
"learning_rate": 9.004375000000001e-06,
|
| 179710 |
+
"loss": 0.3469,
|
| 179711 |
+
"step": 63150
|
| 179712 |
+
},
|
| 179713 |
+
{
|
| 179714 |
+
"epoch": 504.52,
|
| 179715 |
+
"learning_rate": 9.004294871794873e-06,
|
| 179716 |
+
"loss": 0.3977,
|
| 179717 |
+
"step": 63155
|
| 179718 |
+
},
|
| 179719 |
+
{
|
| 179720 |
+
"epoch": 504.56,
|
| 179721 |
+
"learning_rate": 9.004214743589744e-06,
|
| 179722 |
+
"loss": 0.5014,
|
| 179723 |
+
"step": 63160
|
| 179724 |
+
},
|
| 179725 |
+
{
|
| 179726 |
+
"epoch": 504.6,
|
| 179727 |
+
"learning_rate": 9.004134615384617e-06,
|
| 179728 |
+
"loss": 1.2272,
|
| 179729 |
+
"step": 63165
|
| 179730 |
+
},
|
| 179731 |
+
{
|
| 179732 |
+
"epoch": 504.64,
|
| 179733 |
+
"learning_rate": 9.004054487179488e-06,
|
| 179734 |
+
"loss": 0.3564,
|
| 179735 |
+
"step": 63170
|
| 179736 |
+
},
|
| 179737 |
+
{
|
| 179738 |
+
"epoch": 504.68,
|
| 179739 |
+
"learning_rate": 9.00397435897436e-06,
|
| 179740 |
+
"loss": 0.3117,
|
| 179741 |
+
"step": 63175
|
| 179742 |
+
},
|
| 179743 |
+
{
|
| 179744 |
+
"epoch": 504.72,
|
| 179745 |
+
"learning_rate": 9.003894230769231e-06,
|
| 179746 |
+
"loss": 0.3833,
|
| 179747 |
+
"step": 63180
|
| 179748 |
+
},
|
| 179749 |
+
{
|
| 179750 |
+
"epoch": 504.76,
|
| 179751 |
+
"learning_rate": 9.003814102564104e-06,
|
| 179752 |
+
"loss": 0.5139,
|
| 179753 |
+
"step": 63185
|
| 179754 |
+
},
|
| 179755 |
+
{
|
| 179756 |
+
"epoch": 504.8,
|
| 179757 |
+
"learning_rate": 9.003733974358974e-06,
|
| 179758 |
+
"loss": 1.2669,
|
| 179759 |
+
"step": 63190
|
| 179760 |
+
},
|
| 179761 |
+
{
|
| 179762 |
+
"epoch": 504.84,
|
| 179763 |
+
"learning_rate": 9.003653846153847e-06,
|
| 179764 |
+
"loss": 0.3736,
|
| 179765 |
+
"step": 63195
|
| 179766 |
+
},
|
| 179767 |
+
{
|
| 179768 |
+
"epoch": 504.88,
|
| 179769 |
+
"learning_rate": 9.003573717948718e-06,
|
| 179770 |
+
"loss": 0.3299,
|
| 179771 |
+
"step": 63200
|
| 179772 |
+
},
|
| 179773 |
+
{
|
| 179774 |
+
"epoch": 504.92,
|
| 179775 |
+
"learning_rate": 9.00349358974359e-06,
|
| 179776 |
+
"loss": 0.3797,
|
| 179777 |
+
"step": 63205
|
| 179778 |
+
},
|
| 179779 |
+
{
|
| 179780 |
+
"epoch": 504.96,
|
| 179781 |
+
"learning_rate": 9.003413461538463e-06,
|
| 179782 |
+
"loss": 0.5037,
|
| 179783 |
+
"step": 63210
|
| 179784 |
+
},
|
| 179785 |
+
{
|
| 179786 |
+
"epoch": 505.0,
|
| 179787 |
+
"learning_rate": 9.003333333333334e-06,
|
| 179788 |
+
"loss": 1.3314,
|
| 179789 |
+
"step": 63215
|
| 179790 |
+
},
|
| 179791 |
+
{
|
| 179792 |
+
"epoch": 505.0,
|
| 179793 |
+
"eval_loss": 0.3925876021385193,
|
| 179794 |
+
"eval_runtime": 39.7228,
|
| 179795 |
+
"eval_samples_per_second": 21.272,
|
| 179796 |
+
"eval_steps_per_second": 0.68,
|
| 179797 |
+
"eval_wer": 0.19423467905896022,
|
| 179798 |
+
"step": 63215
|
| 179799 |
}
|
| 179800 |
],
|
| 179801 |
+
"max_steps": 625000,
|
| 179802 |
"num_train_epochs": 5000,
|
| 179803 |
+
"total_flos": 1.778898213213315e+20,
|
| 179804 |
"trial_name": null,
|
| 179805 |
"trial_params": null
|
| 179806 |
}
|
model-bin/finetune/base/{checkpoint-62593 → checkpoint-63215}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629811388.0786695/events.out.tfevents.1629811388.c435e1c5ee04.920.141
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5903cc258c3c23cb26dffd7482f598fc20b4dd1e5abf551c370909809e4da56b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629812021.3954506/events.out.tfevents.1629812021.c435e1c5ee04.920.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd1647d4f2c1ce52ed3f2cf26b9e1bf2aeba9768fa851bfae7dc39301dfeb463
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629812663.6763268/events.out.tfevents.1629812663.c435e1c5ee04.920.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:974b3ea2ebc298003a10ea1cb4e65b72ab9a367660c89ca981690d7772105b5d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629813319.9052355/events.out.tfevents.1629813319.c435e1c5ee04.920.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:26c8c4e45ea5ce5f8a1801c3c774d63eacb439c69e56fc104b1998a3b94c6269
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629813976.4157386/events.out.tfevents.1629813976.c435e1c5ee04.920.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:20ac08c9bdd1bf1d1d300e263ecfa1af4a6fe03094c8aa69ddbcbb3d63b7395e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629811388.c435e1c5ee04.920.140
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8c410d9ff7ae0fbd4b53054effc5e059b09f8e5e6aa3edcc0af276201ca1be2
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629812021.c435e1c5ee04.920.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f601d9cfcbc596127dc100f62dc090b835491a2baa1708e4bbd2e384c6ec5104
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629812663.c435e1c5ee04.920.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f29ed24104897c1fa5c4116f89ca5b3e186ba2c4980dbe4c267794cd5c56b7e3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629813319.c435e1c5ee04.920.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd50c791e62f003cd5d0eb4b6870a9a04e1ac5a7a0c5748dd46814d47d56ab7d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629813976.c435e1c5ee04.920.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:534a60851ad323742bbd0c269fb085a022b8cb975c413230415cc9b454634014
|
| 3 |
+
size 8622
|