"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629729537.2079122/events.out.tfevents.1629729537.74272264b15c.932.153 +3 -0
- model-bin/finetune/base/log/1629730178.309412/events.out.tfevents.1629730178.74272264b15c.932.155 +3 -0
- model-bin/finetune/base/log/1629730824.9113455/events.out.tfevents.1629730824.74272264b15c.932.157 +3 -0
- model-bin/finetune/base/log/1629731477.0083396/events.out.tfevents.1629731477.74272264b15c.932.159 +3 -0
- model-bin/finetune/base/log/1629732120.6295552/events.out.tfevents.1629732120.74272264b15c.932.161 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629729537.74272264b15c.932.152 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629730178.74272264b15c.932.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629730824.74272264b15c.932.156 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629731477.74272264b15c.932.158 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629732120.74272264b15c.932.160 +3 -0
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/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 sha256:3b6ae36f1470151fcec1ad3152a33a1a67095e63b48c5439b20bcf248749fd62
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/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 sha256:26445c09a29688f4d145954a54e6e0b49beda1836a932a18751c83fe7ab4143d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/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 sha256:8e4fe6d0d49281984fdb840043efd745377fb7293293dd4c999cf5b86fcb8cc9
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/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 sha256:3ff79ea670b032ae76eb17263d0356be4bded002f7ffdf7b390d21305eda812a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/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:a9f8dfa58cc10ed162b1196ed5f37a00019f0b2ee7112b758926599597261135
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -160005,11 +160005,806 @@
|
|
| 160005 |
"eval_steps_per_second": 0.655,
|
| 160006 |
"eval_wer": 0.20148385219668316,
|
| 160007 |
"step": 47663
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160008 |
}
|
| 160009 |
],
|
| 160010 |
"max_steps": 625000,
|
| 160011 |
"num_train_epochs": 5000,
|
| 160012 |
-
"total_flos": 1.
|
| 160013 |
"trial_name": null,
|
| 160014 |
"trial_params": null
|
| 160015 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 386.0,
|
| 5 |
+
"global_step": 48288,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 160005 |
"eval_steps_per_second": 0.655,
|
| 160006 |
"eval_wer": 0.20148385219668316,
|
| 160007 |
"step": 47663
|
| 160008 |
+
},
|
| 160009 |
+
{
|
| 160010 |
+
"epoch": 381.02,
|
| 160011 |
+
"learning_rate": 9.252451923076923e-06,
|
| 160012 |
+
"loss": 0.3971,
|
| 160013 |
+
"step": 47665
|
| 160014 |
+
},
|
| 160015 |
+
{
|
| 160016 |
+
"epoch": 381.06,
|
| 160017 |
+
"learning_rate": 9.252371794871796e-06,
|
| 160018 |
+
"loss": 0.3389,
|
| 160019 |
+
"step": 47670
|
| 160020 |
+
},
|
| 160021 |
+
{
|
| 160022 |
+
"epoch": 381.1,
|
| 160023 |
+
"learning_rate": 9.252291666666668e-06,
|
| 160024 |
+
"loss": 0.323,
|
| 160025 |
+
"step": 47675
|
| 160026 |
+
},
|
| 160027 |
+
{
|
| 160028 |
+
"epoch": 381.14,
|
| 160029 |
+
"learning_rate": 9.25221153846154e-06,
|
| 160030 |
+
"loss": 0.4617,
|
| 160031 |
+
"step": 47680
|
| 160032 |
+
},
|
| 160033 |
+
{
|
| 160034 |
+
"epoch": 381.18,
|
| 160035 |
+
"learning_rate": 9.25213141025641e-06,
|
| 160036 |
+
"loss": 0.7376,
|
| 160037 |
+
"step": 47685
|
| 160038 |
+
},
|
| 160039 |
+
{
|
| 160040 |
+
"epoch": 381.22,
|
| 160041 |
+
"learning_rate": 9.252051282051284e-06,
|
| 160042 |
+
"loss": 1.0757,
|
| 160043 |
+
"step": 47690
|
| 160044 |
+
},
|
| 160045 |
+
{
|
| 160046 |
+
"epoch": 381.26,
|
| 160047 |
+
"learning_rate": 9.251971153846155e-06,
|
| 160048 |
+
"loss": 0.2988,
|
| 160049 |
+
"step": 47695
|
| 160050 |
+
},
|
| 160051 |
+
{
|
| 160052 |
+
"epoch": 381.3,
|
| 160053 |
+
"learning_rate": 9.251891025641026e-06,
|
| 160054 |
+
"loss": 0.3414,
|
| 160055 |
+
"step": 47700
|
| 160056 |
+
},
|
| 160057 |
+
{
|
| 160058 |
+
"epoch": 381.34,
|
| 160059 |
+
"learning_rate": 9.251810897435898e-06,
|
| 160060 |
+
"loss": 0.3587,
|
| 160061 |
+
"step": 47705
|
| 160062 |
+
},
|
| 160063 |
+
{
|
| 160064 |
+
"epoch": 381.38,
|
| 160065 |
+
"learning_rate": 9.25173076923077e-06,
|
| 160066 |
+
"loss": 0.6515,
|
| 160067 |
+
"step": 47710
|
| 160068 |
+
},
|
| 160069 |
+
{
|
| 160070 |
+
"epoch": 381.42,
|
| 160071 |
+
"learning_rate": 9.251650641025642e-06,
|
| 160072 |
+
"loss": 1.1233,
|
| 160073 |
+
"step": 47715
|
| 160074 |
+
},
|
| 160075 |
+
{
|
| 160076 |
+
"epoch": 381.46,
|
| 160077 |
+
"learning_rate": 9.251570512820513e-06,
|
| 160078 |
+
"loss": 0.4138,
|
| 160079 |
+
"step": 47720
|
| 160080 |
+
},
|
| 160081 |
+
{
|
| 160082 |
+
"epoch": 381.5,
|
| 160083 |
+
"learning_rate": 9.251490384615386e-06,
|
| 160084 |
+
"loss": 0.3441,
|
| 160085 |
+
"step": 47725
|
| 160086 |
+
},
|
| 160087 |
+
{
|
| 160088 |
+
"epoch": 381.54,
|
| 160089 |
+
"learning_rate": 9.251410256410258e-06,
|
| 160090 |
+
"loss": 0.4364,
|
| 160091 |
+
"step": 47730
|
| 160092 |
+
},
|
| 160093 |
+
{
|
| 160094 |
+
"epoch": 381.58,
|
| 160095 |
+
"learning_rate": 9.25133012820513e-06,
|
| 160096 |
+
"loss": 0.7155,
|
| 160097 |
+
"step": 47735
|
| 160098 |
+
},
|
| 160099 |
+
{
|
| 160100 |
+
"epoch": 381.62,
|
| 160101 |
+
"learning_rate": 9.25125e-06,
|
| 160102 |
+
"loss": 1.246,
|
| 160103 |
+
"step": 47740
|
| 160104 |
+
},
|
| 160105 |
+
{
|
| 160106 |
+
"epoch": 381.66,
|
| 160107 |
+
"learning_rate": 9.251169871794874e-06,
|
| 160108 |
+
"loss": 0.3348,
|
| 160109 |
+
"step": 47745
|
| 160110 |
+
},
|
| 160111 |
+
{
|
| 160112 |
+
"epoch": 381.7,
|
| 160113 |
+
"learning_rate": 9.251089743589743e-06,
|
| 160114 |
+
"loss": 0.3715,
|
| 160115 |
+
"step": 47750
|
| 160116 |
+
},
|
| 160117 |
+
{
|
| 160118 |
+
"epoch": 381.74,
|
| 160119 |
+
"learning_rate": 9.251009615384616e-06,
|
| 160120 |
+
"loss": 0.37,
|
| 160121 |
+
"step": 47755
|
| 160122 |
+
},
|
| 160123 |
+
{
|
| 160124 |
+
"epoch": 381.78,
|
| 160125 |
+
"learning_rate": 9.250929487179488e-06,
|
| 160126 |
+
"loss": 0.6945,
|
| 160127 |
+
"step": 47760
|
| 160128 |
+
},
|
| 160129 |
+
{
|
| 160130 |
+
"epoch": 381.82,
|
| 160131 |
+
"learning_rate": 9.250849358974359e-06,
|
| 160132 |
+
"loss": 1.1434,
|
| 160133 |
+
"step": 47765
|
| 160134 |
+
},
|
| 160135 |
+
{
|
| 160136 |
+
"epoch": 381.86,
|
| 160137 |
+
"learning_rate": 9.250769230769232e-06,
|
| 160138 |
+
"loss": 0.3241,
|
| 160139 |
+
"step": 47770
|
| 160140 |
+
},
|
| 160141 |
+
{
|
| 160142 |
+
"epoch": 381.9,
|
| 160143 |
+
"learning_rate": 9.250689102564103e-06,
|
| 160144 |
+
"loss": 0.3328,
|
| 160145 |
+
"step": 47775
|
| 160146 |
+
},
|
| 160147 |
+
{
|
| 160148 |
+
"epoch": 381.94,
|
| 160149 |
+
"learning_rate": 9.250608974358975e-06,
|
| 160150 |
+
"loss": 0.4408,
|
| 160151 |
+
"step": 47780
|
| 160152 |
+
},
|
| 160153 |
+
{
|
| 160154 |
+
"epoch": 381.98,
|
| 160155 |
+
"learning_rate": 9.250528846153846e-06,
|
| 160156 |
+
"loss": 0.7996,
|
| 160157 |
+
"step": 47785
|
| 160158 |
+
},
|
| 160159 |
+
{
|
| 160160 |
+
"epoch": 382.0,
|
| 160161 |
+
"eval_loss": 0.4474487900733948,
|
| 160162 |
+
"eval_runtime": 39.2412,
|
| 160163 |
+
"eval_samples_per_second": 21.279,
|
| 160164 |
+
"eval_steps_per_second": 0.688,
|
| 160165 |
+
"eval_wer": 0.20729635920537626,
|
| 160166 |
+
"step": 47788
|
| 160167 |
+
},
|
| 160168 |
+
{
|
| 160169 |
+
"epoch": 382.02,
|
| 160170 |
+
"learning_rate": 9.25044871794872e-06,
|
| 160171 |
+
"loss": 0.4093,
|
| 160172 |
+
"step": 47790
|
| 160173 |
+
},
|
| 160174 |
+
{
|
| 160175 |
+
"epoch": 382.06,
|
| 160176 |
+
"learning_rate": 9.25036858974359e-06,
|
| 160177 |
+
"loss": 0.3645,
|
| 160178 |
+
"step": 47795
|
| 160179 |
+
},
|
| 160180 |
+
{
|
| 160181 |
+
"epoch": 382.1,
|
| 160182 |
+
"learning_rate": 9.250288461538462e-06,
|
| 160183 |
+
"loss": 0.3569,
|
| 160184 |
+
"step": 47800
|
| 160185 |
+
},
|
| 160186 |
+
{
|
| 160187 |
+
"epoch": 382.14,
|
| 160188 |
+
"learning_rate": 9.250208333333333e-06,
|
| 160189 |
+
"loss": 0.361,
|
| 160190 |
+
"step": 47805
|
| 160191 |
+
},
|
| 160192 |
+
{
|
| 160193 |
+
"epoch": 382.18,
|
| 160194 |
+
"learning_rate": 9.250128205128206e-06,
|
| 160195 |
+
"loss": 0.606,
|
| 160196 |
+
"step": 47810
|
| 160197 |
+
},
|
| 160198 |
+
{
|
| 160199 |
+
"epoch": 382.22,
|
| 160200 |
+
"learning_rate": 9.250048076923078e-06,
|
| 160201 |
+
"loss": 1.1471,
|
| 160202 |
+
"step": 47815
|
| 160203 |
+
},
|
| 160204 |
+
{
|
| 160205 |
+
"epoch": 382.26,
|
| 160206 |
+
"learning_rate": 9.249967948717949e-06,
|
| 160207 |
+
"loss": 0.3549,
|
| 160208 |
+
"step": 47820
|
| 160209 |
+
},
|
| 160210 |
+
{
|
| 160211 |
+
"epoch": 382.3,
|
| 160212 |
+
"learning_rate": 9.249887820512822e-06,
|
| 160213 |
+
"loss": 0.2992,
|
| 160214 |
+
"step": 47825
|
| 160215 |
+
},
|
| 160216 |
+
{
|
| 160217 |
+
"epoch": 382.34,
|
| 160218 |
+
"learning_rate": 9.249807692307693e-06,
|
| 160219 |
+
"loss": 0.4853,
|
| 160220 |
+
"step": 47830
|
| 160221 |
+
},
|
| 160222 |
+
{
|
| 160223 |
+
"epoch": 382.38,
|
| 160224 |
+
"learning_rate": 9.249727564102565e-06,
|
| 160225 |
+
"loss": 0.7708,
|
| 160226 |
+
"step": 47835
|
| 160227 |
+
},
|
| 160228 |
+
{
|
| 160229 |
+
"epoch": 382.42,
|
| 160230 |
+
"learning_rate": 9.249647435897436e-06,
|
| 160231 |
+
"loss": 1.1267,
|
| 160232 |
+
"step": 47840
|
| 160233 |
+
},
|
| 160234 |
+
{
|
| 160235 |
+
"epoch": 382.46,
|
| 160236 |
+
"learning_rate": 9.24956730769231e-06,
|
| 160237 |
+
"loss": 0.3346,
|
| 160238 |
+
"step": 47845
|
| 160239 |
+
},
|
| 160240 |
+
{
|
| 160241 |
+
"epoch": 382.5,
|
| 160242 |
+
"learning_rate": 9.24948717948718e-06,
|
| 160243 |
+
"loss": 0.3632,
|
| 160244 |
+
"step": 47850
|
| 160245 |
+
},
|
| 160246 |
+
{
|
| 160247 |
+
"epoch": 382.54,
|
| 160248 |
+
"learning_rate": 9.249407051282052e-06,
|
| 160249 |
+
"loss": 0.3969,
|
| 160250 |
+
"step": 47855
|
| 160251 |
+
},
|
| 160252 |
+
{
|
| 160253 |
+
"epoch": 382.58,
|
| 160254 |
+
"learning_rate": 9.249326923076923e-06,
|
| 160255 |
+
"loss": 0.7573,
|
| 160256 |
+
"step": 47860
|
| 160257 |
+
},
|
| 160258 |
+
{
|
| 160259 |
+
"epoch": 382.62,
|
| 160260 |
+
"learning_rate": 9.249246794871796e-06,
|
| 160261 |
+
"loss": 1.1858,
|
| 160262 |
+
"step": 47865
|
| 160263 |
+
},
|
| 160264 |
+
{
|
| 160265 |
+
"epoch": 382.66,
|
| 160266 |
+
"learning_rate": 9.249166666666668e-06,
|
| 160267 |
+
"loss": 0.3861,
|
| 160268 |
+
"step": 47870
|
| 160269 |
+
},
|
| 160270 |
+
{
|
| 160271 |
+
"epoch": 382.7,
|
| 160272 |
+
"learning_rate": 9.249086538461539e-06,
|
| 160273 |
+
"loss": 0.3233,
|
| 160274 |
+
"step": 47875
|
| 160275 |
+
},
|
| 160276 |
+
{
|
| 160277 |
+
"epoch": 382.74,
|
| 160278 |
+
"learning_rate": 9.249006410256412e-06,
|
| 160279 |
+
"loss": 0.4369,
|
| 160280 |
+
"step": 47880
|
| 160281 |
+
},
|
| 160282 |
+
{
|
| 160283 |
+
"epoch": 382.78,
|
| 160284 |
+
"learning_rate": 9.248926282051282e-06,
|
| 160285 |
+
"loss": 0.6703,
|
| 160286 |
+
"step": 47885
|
| 160287 |
+
},
|
| 160288 |
+
{
|
| 160289 |
+
"epoch": 382.82,
|
| 160290 |
+
"learning_rate": 9.248846153846155e-06,
|
| 160291 |
+
"loss": 1.1758,
|
| 160292 |
+
"step": 47890
|
| 160293 |
+
},
|
| 160294 |
+
{
|
| 160295 |
+
"epoch": 382.86,
|
| 160296 |
+
"learning_rate": 9.248766025641026e-06,
|
| 160297 |
+
"loss": 0.2987,
|
| 160298 |
+
"step": 47895
|
| 160299 |
+
},
|
| 160300 |
+
{
|
| 160301 |
+
"epoch": 382.9,
|
| 160302 |
+
"learning_rate": 9.248685897435898e-06,
|
| 160303 |
+
"loss": 0.3724,
|
| 160304 |
+
"step": 47900
|
| 160305 |
+
},
|
| 160306 |
+
{
|
| 160307 |
+
"epoch": 382.94,
|
| 160308 |
+
"learning_rate": 9.248605769230769e-06,
|
| 160309 |
+
"loss": 0.4235,
|
| 160310 |
+
"step": 47905
|
| 160311 |
+
},
|
| 160312 |
+
{
|
| 160313 |
+
"epoch": 382.98,
|
| 160314 |
+
"learning_rate": 9.248525641025642e-06,
|
| 160315 |
+
"loss": 0.6528,
|
| 160316 |
+
"step": 47910
|
| 160317 |
+
},
|
| 160318 |
+
{
|
| 160319 |
+
"epoch": 383.0,
|
| 160320 |
+
"eval_loss": 0.4338654577732086,
|
| 160321 |
+
"eval_runtime": 39.3369,
|
| 160322 |
+
"eval_samples_per_second": 21.227,
|
| 160323 |
+
"eval_steps_per_second": 0.686,
|
| 160324 |
+
"eval_wer": 0.21281085154483798,
|
| 160325 |
+
"step": 47913
|
| 160326 |
+
},
|
| 160327 |
+
{
|
| 160328 |
+
"epoch": 383.02,
|
| 160329 |
+
"learning_rate": 9.248445512820513e-06,
|
| 160330 |
+
"loss": 0.3667,
|
| 160331 |
+
"step": 47915
|
| 160332 |
+
},
|
| 160333 |
+
{
|
| 160334 |
+
"epoch": 383.06,
|
| 160335 |
+
"learning_rate": 9.248365384615385e-06,
|
| 160336 |
+
"loss": 0.3526,
|
| 160337 |
+
"step": 47920
|
| 160338 |
+
},
|
| 160339 |
+
{
|
| 160340 |
+
"epoch": 383.1,
|
| 160341 |
+
"learning_rate": 9.248285256410258e-06,
|
| 160342 |
+
"loss": 0.3728,
|
| 160343 |
+
"step": 47925
|
| 160344 |
+
},
|
| 160345 |
+
{
|
| 160346 |
+
"epoch": 383.14,
|
| 160347 |
+
"learning_rate": 9.248205128205129e-06,
|
| 160348 |
+
"loss": 0.3497,
|
| 160349 |
+
"step": 47930
|
| 160350 |
+
},
|
| 160351 |
+
{
|
| 160352 |
+
"epoch": 383.18,
|
| 160353 |
+
"learning_rate": 9.248125e-06,
|
| 160354 |
+
"loss": 0.7086,
|
| 160355 |
+
"step": 47935
|
| 160356 |
+
},
|
| 160357 |
+
{
|
| 160358 |
+
"epoch": 383.22,
|
| 160359 |
+
"learning_rate": 9.248044871794872e-06,
|
| 160360 |
+
"loss": 1.2526,
|
| 160361 |
+
"step": 47940
|
| 160362 |
+
},
|
| 160363 |
+
{
|
| 160364 |
+
"epoch": 383.26,
|
| 160365 |
+
"learning_rate": 9.247964743589745e-06,
|
| 160366 |
+
"loss": 0.3158,
|
| 160367 |
+
"step": 47945
|
| 160368 |
+
},
|
| 160369 |
+
{
|
| 160370 |
+
"epoch": 383.3,
|
| 160371 |
+
"learning_rate": 9.247884615384616e-06,
|
| 160372 |
+
"loss": 0.3663,
|
| 160373 |
+
"step": 47950
|
| 160374 |
+
},
|
| 160375 |
+
{
|
| 160376 |
+
"epoch": 383.34,
|
| 160377 |
+
"learning_rate": 9.247804487179488e-06,
|
| 160378 |
+
"loss": 0.4438,
|
| 160379 |
+
"step": 47955
|
| 160380 |
+
},
|
| 160381 |
+
{
|
| 160382 |
+
"epoch": 383.38,
|
| 160383 |
+
"learning_rate": 9.247724358974359e-06,
|
| 160384 |
+
"loss": 0.726,
|
| 160385 |
+
"step": 47960
|
| 160386 |
+
},
|
| 160387 |
+
{
|
| 160388 |
+
"epoch": 383.42,
|
| 160389 |
+
"learning_rate": 9.247644230769232e-06,
|
| 160390 |
+
"loss": 1.2397,
|
| 160391 |
+
"step": 47965
|
| 160392 |
+
},
|
| 160393 |
+
{
|
| 160394 |
+
"epoch": 383.46,
|
| 160395 |
+
"learning_rate": 9.247564102564103e-06,
|
| 160396 |
+
"loss": 0.3575,
|
| 160397 |
+
"step": 47970
|
| 160398 |
+
},
|
| 160399 |
+
{
|
| 160400 |
+
"epoch": 383.5,
|
| 160401 |
+
"learning_rate": 9.247483974358975e-06,
|
| 160402 |
+
"loss": 0.3929,
|
| 160403 |
+
"step": 47975
|
| 160404 |
+
},
|
| 160405 |
+
{
|
| 160406 |
+
"epoch": 383.54,
|
| 160407 |
+
"learning_rate": 9.247403846153848e-06,
|
| 160408 |
+
"loss": 0.3886,
|
| 160409 |
+
"step": 47980
|
| 160410 |
+
},
|
| 160411 |
+
{
|
| 160412 |
+
"epoch": 383.58,
|
| 160413 |
+
"learning_rate": 9.24732371794872e-06,
|
| 160414 |
+
"loss": 0.6418,
|
| 160415 |
+
"step": 47985
|
| 160416 |
+
},
|
| 160417 |
+
{
|
| 160418 |
+
"epoch": 383.62,
|
| 160419 |
+
"learning_rate": 9.24724358974359e-06,
|
| 160420 |
+
"loss": 1.1219,
|
| 160421 |
+
"step": 47990
|
| 160422 |
+
},
|
| 160423 |
+
{
|
| 160424 |
+
"epoch": 383.66,
|
| 160425 |
+
"learning_rate": 9.247163461538462e-06,
|
| 160426 |
+
"loss": 0.3736,
|
| 160427 |
+
"step": 47995
|
| 160428 |
+
},
|
| 160429 |
+
{
|
| 160430 |
+
"epoch": 383.7,
|
| 160431 |
+
"learning_rate": 9.247083333333335e-06,
|
| 160432 |
+
"loss": 0.3584,
|
| 160433 |
+
"step": 48000
|
| 160434 |
+
},
|
| 160435 |
+
{
|
| 160436 |
+
"epoch": 383.74,
|
| 160437 |
+
"learning_rate": 9.247003205128206e-06,
|
| 160438 |
+
"loss": 0.4055,
|
| 160439 |
+
"step": 48005
|
| 160440 |
+
},
|
| 160441 |
+
{
|
| 160442 |
+
"epoch": 383.78,
|
| 160443 |
+
"learning_rate": 9.246923076923078e-06,
|
| 160444 |
+
"loss": 0.7403,
|
| 160445 |
+
"step": 48010
|
| 160446 |
+
},
|
| 160447 |
+
{
|
| 160448 |
+
"epoch": 383.82,
|
| 160449 |
+
"learning_rate": 9.246842948717949e-06,
|
| 160450 |
+
"loss": 1.0718,
|
| 160451 |
+
"step": 48015
|
| 160452 |
+
},
|
| 160453 |
+
{
|
| 160454 |
+
"epoch": 383.86,
|
| 160455 |
+
"learning_rate": 9.246762820512822e-06,
|
| 160456 |
+
"loss": 0.3165,
|
| 160457 |
+
"step": 48020
|
| 160458 |
+
},
|
| 160459 |
+
{
|
| 160460 |
+
"epoch": 383.9,
|
| 160461 |
+
"learning_rate": 9.246682692307693e-06,
|
| 160462 |
+
"loss": 0.3457,
|
| 160463 |
+
"step": 48025
|
| 160464 |
+
},
|
| 160465 |
+
{
|
| 160466 |
+
"epoch": 383.94,
|
| 160467 |
+
"learning_rate": 9.246602564102565e-06,
|
| 160468 |
+
"loss": 0.3643,
|
| 160469 |
+
"step": 48030
|
| 160470 |
+
},
|
| 160471 |
+
{
|
| 160472 |
+
"epoch": 383.98,
|
| 160473 |
+
"learning_rate": 9.246522435897438e-06,
|
| 160474 |
+
"loss": 0.6838,
|
| 160475 |
+
"step": 48035
|
| 160476 |
+
},
|
| 160477 |
+
{
|
| 160478 |
+
"epoch": 384.0,
|
| 160479 |
+
"eval_loss": 0.4506882131099701,
|
| 160480 |
+
"eval_runtime": 40.789,
|
| 160481 |
+
"eval_samples_per_second": 20.471,
|
| 160482 |
+
"eval_steps_per_second": 0.662,
|
| 160483 |
+
"eval_wer": 0.20609548167092925,
|
| 160484 |
+
"step": 48038
|
| 160485 |
+
},
|
| 160486 |
+
{
|
| 160487 |
+
"epoch": 384.02,
|
| 160488 |
+
"learning_rate": 9.246442307692307e-06,
|
| 160489 |
+
"loss": 0.4569,
|
| 160490 |
+
"step": 48040
|
| 160491 |
+
},
|
| 160492 |
+
{
|
| 160493 |
+
"epoch": 384.06,
|
| 160494 |
+
"learning_rate": 9.24636217948718e-06,
|
| 160495 |
+
"loss": 0.3582,
|
| 160496 |
+
"step": 48045
|
| 160497 |
+
},
|
| 160498 |
+
{
|
| 160499 |
+
"epoch": 384.1,
|
| 160500 |
+
"learning_rate": 9.246282051282052e-06,
|
| 160501 |
+
"loss": 0.334,
|
| 160502 |
+
"step": 48050
|
| 160503 |
+
},
|
| 160504 |
+
{
|
| 160505 |
+
"epoch": 384.14,
|
| 160506 |
+
"learning_rate": 9.246201923076923e-06,
|
| 160507 |
+
"loss": 0.4399,
|
| 160508 |
+
"step": 48055
|
| 160509 |
+
},
|
| 160510 |
+
{
|
| 160511 |
+
"epoch": 384.18,
|
| 160512 |
+
"learning_rate": 9.246121794871795e-06,
|
| 160513 |
+
"loss": 0.6171,
|
| 160514 |
+
"step": 48060
|
| 160515 |
+
},
|
| 160516 |
+
{
|
| 160517 |
+
"epoch": 384.22,
|
| 160518 |
+
"learning_rate": 9.246041666666668e-06,
|
| 160519 |
+
"loss": 1.0357,
|
| 160520 |
+
"step": 48065
|
| 160521 |
+
},
|
| 160522 |
+
{
|
| 160523 |
+
"epoch": 384.26,
|
| 160524 |
+
"learning_rate": 9.245961538461539e-06,
|
| 160525 |
+
"loss": 0.3912,
|
| 160526 |
+
"step": 48070
|
| 160527 |
+
},
|
| 160528 |
+
{
|
| 160529 |
+
"epoch": 384.3,
|
| 160530 |
+
"learning_rate": 9.24588141025641e-06,
|
| 160531 |
+
"loss": 0.438,
|
| 160532 |
+
"step": 48075
|
| 160533 |
+
},
|
| 160534 |
+
{
|
| 160535 |
+
"epoch": 384.34,
|
| 160536 |
+
"learning_rate": 9.245801282051283e-06,
|
| 160537 |
+
"loss": 0.402,
|
| 160538 |
+
"step": 48080
|
| 160539 |
+
},
|
| 160540 |
+
{
|
| 160541 |
+
"epoch": 384.38,
|
| 160542 |
+
"learning_rate": 9.245721153846155e-06,
|
| 160543 |
+
"loss": 0.6041,
|
| 160544 |
+
"step": 48085
|
| 160545 |
+
},
|
| 160546 |
+
{
|
| 160547 |
+
"epoch": 384.42,
|
| 160548 |
+
"learning_rate": 9.245641025641026e-06,
|
| 160549 |
+
"loss": 1.202,
|
| 160550 |
+
"step": 48090
|
| 160551 |
+
},
|
| 160552 |
+
{
|
| 160553 |
+
"epoch": 384.46,
|
| 160554 |
+
"learning_rate": 9.245560897435898e-06,
|
| 160555 |
+
"loss": 0.3859,
|
| 160556 |
+
"step": 48095
|
| 160557 |
+
},
|
| 160558 |
+
{
|
| 160559 |
+
"epoch": 384.5,
|
| 160560 |
+
"learning_rate": 9.24548076923077e-06,
|
| 160561 |
+
"loss": 0.3481,
|
| 160562 |
+
"step": 48100
|
| 160563 |
+
},
|
| 160564 |
+
{
|
| 160565 |
+
"epoch": 384.54,
|
| 160566 |
+
"learning_rate": 9.245400641025642e-06,
|
| 160567 |
+
"loss": 0.4098,
|
| 160568 |
+
"step": 48105
|
| 160569 |
+
},
|
| 160570 |
+
{
|
| 160571 |
+
"epoch": 384.58,
|
| 160572 |
+
"learning_rate": 9.245320512820513e-06,
|
| 160573 |
+
"loss": 0.7351,
|
| 160574 |
+
"step": 48110
|
| 160575 |
+
},
|
| 160576 |
+
{
|
| 160577 |
+
"epoch": 384.62,
|
| 160578 |
+
"learning_rate": 9.245240384615385e-06,
|
| 160579 |
+
"loss": 1.0795,
|
| 160580 |
+
"step": 48115
|
| 160581 |
+
},
|
| 160582 |
+
{
|
| 160583 |
+
"epoch": 384.66,
|
| 160584 |
+
"learning_rate": 9.245160256410258e-06,
|
| 160585 |
+
"loss": 0.3069,
|
| 160586 |
+
"step": 48120
|
| 160587 |
+
},
|
| 160588 |
+
{
|
| 160589 |
+
"epoch": 384.7,
|
| 160590 |
+
"learning_rate": 9.245080128205129e-06,
|
| 160591 |
+
"loss": 0.3185,
|
| 160592 |
+
"step": 48125
|
| 160593 |
+
},
|
| 160594 |
+
{
|
| 160595 |
+
"epoch": 384.74,
|
| 160596 |
+
"learning_rate": 9.245e-06,
|
| 160597 |
+
"loss": 0.4013,
|
| 160598 |
+
"step": 48130
|
| 160599 |
+
},
|
| 160600 |
+
{
|
| 160601 |
+
"epoch": 384.78,
|
| 160602 |
+
"learning_rate": 9.244919871794873e-06,
|
| 160603 |
+
"loss": 0.6792,
|
| 160604 |
+
"step": 48135
|
| 160605 |
+
},
|
| 160606 |
+
{
|
| 160607 |
+
"epoch": 384.82,
|
| 160608 |
+
"learning_rate": 9.244839743589745e-06,
|
| 160609 |
+
"loss": 1.153,
|
| 160610 |
+
"step": 48140
|
| 160611 |
+
},
|
| 160612 |
+
{
|
| 160613 |
+
"epoch": 384.86,
|
| 160614 |
+
"learning_rate": 9.244759615384616e-06,
|
| 160615 |
+
"loss": 0.2832,
|
| 160616 |
+
"step": 48145
|
| 160617 |
+
},
|
| 160618 |
+
{
|
| 160619 |
+
"epoch": 384.9,
|
| 160620 |
+
"learning_rate": 9.244679487179488e-06,
|
| 160621 |
+
"loss": 0.4042,
|
| 160622 |
+
"step": 48150
|
| 160623 |
+
},
|
| 160624 |
+
{
|
| 160625 |
+
"epoch": 384.94,
|
| 160626 |
+
"learning_rate": 9.24459935897436e-06,
|
| 160627 |
+
"loss": 0.4558,
|
| 160628 |
+
"step": 48155
|
| 160629 |
+
},
|
| 160630 |
+
{
|
| 160631 |
+
"epoch": 384.98,
|
| 160632 |
+
"learning_rate": 9.24451923076923e-06,
|
| 160633 |
+
"loss": 0.7045,
|
| 160634 |
+
"step": 48160
|
| 160635 |
+
},
|
| 160636 |
+
{
|
| 160637 |
+
"epoch": 385.0,
|
| 160638 |
+
"eval_loss": 0.487787127494812,
|
| 160639 |
+
"eval_runtime": 39.5507,
|
| 160640 |
+
"eval_samples_per_second": 21.137,
|
| 160641 |
+
"eval_steps_per_second": 0.683,
|
| 160642 |
+
"eval_wer": 0.2028522617544381,
|
| 160643 |
+
"step": 48163
|
| 160644 |
+
},
|
| 160645 |
+
{
|
| 160646 |
+
"epoch": 385.02,
|
| 160647 |
+
"learning_rate": 9.244439102564103e-06,
|
| 160648 |
+
"loss": 0.3803,
|
| 160649 |
+
"step": 48165
|
| 160650 |
+
},
|
| 160651 |
+
{
|
| 160652 |
+
"epoch": 385.06,
|
| 160653 |
+
"learning_rate": 9.244358974358976e-06,
|
| 160654 |
+
"loss": 0.6276,
|
| 160655 |
+
"step": 48170
|
| 160656 |
+
},
|
| 160657 |
+
{
|
| 160658 |
+
"epoch": 385.1,
|
| 160659 |
+
"learning_rate": 9.244278846153846e-06,
|
| 160660 |
+
"loss": 0.3612,
|
| 160661 |
+
"step": 48175
|
| 160662 |
+
},
|
| 160663 |
+
{
|
| 160664 |
+
"epoch": 385.14,
|
| 160665 |
+
"learning_rate": 9.244198717948719e-06,
|
| 160666 |
+
"loss": 0.4303,
|
| 160667 |
+
"step": 48180
|
| 160668 |
+
},
|
| 160669 |
+
{
|
| 160670 |
+
"epoch": 385.18,
|
| 160671 |
+
"learning_rate": 9.24411858974359e-06,
|
| 160672 |
+
"loss": 0.7484,
|
| 160673 |
+
"step": 48185
|
| 160674 |
+
},
|
| 160675 |
+
{
|
| 160676 |
+
"epoch": 385.22,
|
| 160677 |
+
"learning_rate": 9.244038461538462e-06,
|
| 160678 |
+
"loss": 1.0026,
|
| 160679 |
+
"step": 48190
|
| 160680 |
+
},
|
| 160681 |
+
{
|
| 160682 |
+
"epoch": 385.26,
|
| 160683 |
+
"learning_rate": 9.243958333333333e-06,
|
| 160684 |
+
"loss": 0.3777,
|
| 160685 |
+
"step": 48195
|
| 160686 |
+
},
|
| 160687 |
+
{
|
| 160688 |
+
"epoch": 385.3,
|
| 160689 |
+
"learning_rate": 9.243878205128206e-06,
|
| 160690 |
+
"loss": 0.3769,
|
| 160691 |
+
"step": 48200
|
| 160692 |
+
},
|
| 160693 |
+
{
|
| 160694 |
+
"epoch": 385.34,
|
| 160695 |
+
"learning_rate": 9.243798076923078e-06,
|
| 160696 |
+
"loss": 0.3786,
|
| 160697 |
+
"step": 48205
|
| 160698 |
+
},
|
| 160699 |
+
{
|
| 160700 |
+
"epoch": 385.38,
|
| 160701 |
+
"learning_rate": 9.243717948717949e-06,
|
| 160702 |
+
"loss": 0.8006,
|
| 160703 |
+
"step": 48210
|
| 160704 |
+
},
|
| 160705 |
+
{
|
| 160706 |
+
"epoch": 385.42,
|
| 160707 |
+
"learning_rate": 9.24363782051282e-06,
|
| 160708 |
+
"loss": 1.2286,
|
| 160709 |
+
"step": 48215
|
| 160710 |
+
},
|
| 160711 |
+
{
|
| 160712 |
+
"epoch": 385.46,
|
| 160713 |
+
"learning_rate": 9.243557692307693e-06,
|
| 160714 |
+
"loss": 0.3582,
|
| 160715 |
+
"step": 48220
|
| 160716 |
+
},
|
| 160717 |
+
{
|
| 160718 |
+
"epoch": 385.5,
|
| 160719 |
+
"learning_rate": 9.243477564102565e-06,
|
| 160720 |
+
"loss": 0.4509,
|
| 160721 |
+
"step": 48225
|
| 160722 |
+
},
|
| 160723 |
+
{
|
| 160724 |
+
"epoch": 385.54,
|
| 160725 |
+
"learning_rate": 9.243397435897436e-06,
|
| 160726 |
+
"loss": 0.4411,
|
| 160727 |
+
"step": 48230
|
| 160728 |
+
},
|
| 160729 |
+
{
|
| 160730 |
+
"epoch": 385.58,
|
| 160731 |
+
"learning_rate": 9.243317307692309e-06,
|
| 160732 |
+
"loss": 0.9501,
|
| 160733 |
+
"step": 48235
|
| 160734 |
+
},
|
| 160735 |
+
{
|
| 160736 |
+
"epoch": 385.62,
|
| 160737 |
+
"learning_rate": 9.24323717948718e-06,
|
| 160738 |
+
"loss": 1.132,
|
| 160739 |
+
"step": 48240
|
| 160740 |
+
},
|
| 160741 |
+
{
|
| 160742 |
+
"epoch": 385.66,
|
| 160743 |
+
"learning_rate": 9.243157051282052e-06,
|
| 160744 |
+
"loss": 0.3727,
|
| 160745 |
+
"step": 48245
|
| 160746 |
+
},
|
| 160747 |
+
{
|
| 160748 |
+
"epoch": 385.7,
|
| 160749 |
+
"learning_rate": 9.243076923076923e-06,
|
| 160750 |
+
"loss": 0.349,
|
| 160751 |
+
"step": 48250
|
| 160752 |
+
},
|
| 160753 |
+
{
|
| 160754 |
+
"epoch": 385.74,
|
| 160755 |
+
"learning_rate": 9.242996794871796e-06,
|
| 160756 |
+
"loss": 0.3987,
|
| 160757 |
+
"step": 48255
|
| 160758 |
+
},
|
| 160759 |
+
{
|
| 160760 |
+
"epoch": 385.78,
|
| 160761 |
+
"learning_rate": 9.242916666666668e-06,
|
| 160762 |
+
"loss": 0.6941,
|
| 160763 |
+
"step": 48260
|
| 160764 |
+
},
|
| 160765 |
+
{
|
| 160766 |
+
"epoch": 385.82,
|
| 160767 |
+
"learning_rate": 9.242836538461539e-06,
|
| 160768 |
+
"loss": 1.2013,
|
| 160769 |
+
"step": 48265
|
| 160770 |
+
},
|
| 160771 |
+
{
|
| 160772 |
+
"epoch": 385.86,
|
| 160773 |
+
"learning_rate": 9.24275641025641e-06,
|
| 160774 |
+
"loss": 0.3445,
|
| 160775 |
+
"step": 48270
|
| 160776 |
+
},
|
| 160777 |
+
{
|
| 160778 |
+
"epoch": 385.9,
|
| 160779 |
+
"learning_rate": 9.242676282051283e-06,
|
| 160780 |
+
"loss": 0.4103,
|
| 160781 |
+
"step": 48275
|
| 160782 |
+
},
|
| 160783 |
+
{
|
| 160784 |
+
"epoch": 385.94,
|
| 160785 |
+
"learning_rate": 9.242596153846155e-06,
|
| 160786 |
+
"loss": 0.3933,
|
| 160787 |
+
"step": 48280
|
| 160788 |
+
},
|
| 160789 |
+
{
|
| 160790 |
+
"epoch": 385.98,
|
| 160791 |
+
"learning_rate": 9.242516025641026e-06,
|
| 160792 |
+
"loss": 0.7593,
|
| 160793 |
+
"step": 48285
|
| 160794 |
+
},
|
| 160795 |
+
{
|
| 160796 |
+
"epoch": 386.0,
|
| 160797 |
+
"eval_loss": 0.43986836075782776,
|
| 160798 |
+
"eval_runtime": 42.306,
|
| 160799 |
+
"eval_samples_per_second": 19.761,
|
| 160800 |
+
"eval_steps_per_second": 0.638,
|
| 160801 |
+
"eval_wer": 0.21344501408043903,
|
| 160802 |
+
"step": 48288
|
| 160803 |
}
|
| 160804 |
],
|
| 160805 |
"max_steps": 625000,
|
| 160806 |
"num_train_epochs": 5000,
|
| 160807 |
+
"total_flos": 1.3586798852174347e+20,
|
| 160808 |
"trial_name": null,
|
| 160809 |
"trial_params": null
|
| 160810 |
}
|
model-bin/finetune/base/{checkpoint-47663 → checkpoint-48288}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629729537.2079122/events.out.tfevents.1629729537.74272264b15c.932.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:657a136ddcc4499ea27567347ebe2644f8af1c6172d51209aa7d9f1afd5234f2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629730178.309412/events.out.tfevents.1629730178.74272264b15c.932.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3636cd11adfc23bec228de24951199d9184dce52516bc83e1def29ee958e610
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629730824.9113455/events.out.tfevents.1629730824.74272264b15c.932.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1fdd45dfe899c2a80fa2732458736b1f6bfd7216dfa6317d1507c410aa37accb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629731477.0083396/events.out.tfevents.1629731477.74272264b15c.932.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:432a4a4291df8feb956101dd81997cb49db02a729ba2dc6661ec43bb2c43872b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629732120.6295552/events.out.tfevents.1629732120.74272264b15c.932.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5a7f4c731a57cc22e079308f9ff1b2ea6e0efcf0d7d49fb5f5784bc07fc60890
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629729537.74272264b15c.932.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7fa03af12aac867931e68772572da097d641158f5792d54f3699b993f372e9c8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629730178.74272264b15c.932.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e2d6c61d8e9be86d165815c4208a5ceb4459fe1a29e8ae544d373b891b096cdc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629730824.74272264b15c.932.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:708d537067aa55f754951c51ce323249dde515b8a154107449dc039375296f39
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629731477.74272264b15c.932.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c69aa0234d4e6b4c46ea7eeb5e5ebb8eea5e78d5c5a08152c02397a5418345bd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629732120.74272264b15c.932.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0ee6172e7980f77301b88fe7e6f9a1c84039a3df2e22026fe0b7a1323ed3bcc
|
| 3 |
+
size 8622
|