"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629732801.0318525/events.out.tfevents.1629732801.74272264b15c.932.163 +3 -0
- model-bin/finetune/base/log/1629733448.706626/events.out.tfevents.1629733448.74272264b15c.932.165 +3 -0
- model-bin/finetune/base/log/1629734099.0176368/events.out.tfevents.1629734099.74272264b15c.932.167 +3 -0
- model-bin/finetune/base/log/1629734741.4134967/events.out.tfevents.1629734741.74272264b15c.932.169 +3 -0
- model-bin/finetune/base/log/1629735490.0998623/events.out.tfevents.1629735490.74272264b15c.932.171 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629732801.74272264b15c.932.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629733448.74272264b15c.932.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629734099.74272264b15c.932.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629734741.74272264b15c.932.168 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629735490.74272264b15c.932.170 +3 -0
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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 sha256:4a4799d7b86d3f74339b95190899ecd7765f28e93ec2786a22480441d21c0365
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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 sha256:4546597dc171347c940c45d3cd9b0bebdf98ae9ee1d7ba9468bcb1a835529f16
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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 sha256:6f3752d2edd5a1b4cd8998d0d13d3d416d503f384292755ef1b3149033869fb4
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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:c136379ee38426d55f1ef93b66be7d010c1fbe32dea9080246c25e9f7623bbcd
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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:b56c3442822ad5583d94aeac49a7c6ed4a7cefe2c5fbc85fa346a5e1693afa9b
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/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,
|
|
@@ -160800,11 +160800,800 @@
|
|
| 160800 |
"eval_steps_per_second": 0.638,
|
| 160801 |
"eval_wer": 0.21344501408043903,
|
| 160802 |
"step": 48288
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160803 |
}
|
| 160804 |
],
|
| 160805 |
-
"max_steps":
|
| 160806 |
"num_train_epochs": 5000,
|
| 160807 |
-
"total_flos": 1.
|
| 160808 |
"trial_name": null,
|
| 160809 |
"trial_params": null
|
| 160810 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 393.99598393574297,
|
| 5 |
+
"global_step": 48909,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 160800 |
"eval_steps_per_second": 0.638,
|
| 160801 |
"eval_wer": 0.21344501408043903,
|
| 160802 |
"step": 48288
|
| 160803 |
+
},
|
| 160804 |
+
{
|
| 160805 |
+
"epoch": 389.02,
|
| 160806 |
+
"learning_rate": 9.242435897435899e-06,
|
| 160807 |
+
"loss": 0.43,
|
| 160808 |
+
"step": 48290
|
| 160809 |
+
},
|
| 160810 |
+
{
|
| 160811 |
+
"epoch": 389.06,
|
| 160812 |
+
"learning_rate": 9.24235576923077e-06,
|
| 160813 |
+
"loss": 0.3888,
|
| 160814 |
+
"step": 48295
|
| 160815 |
+
},
|
| 160816 |
+
{
|
| 160817 |
+
"epoch": 389.1,
|
| 160818 |
+
"learning_rate": 9.242275641025642e-06,
|
| 160819 |
+
"loss": 0.3824,
|
| 160820 |
+
"step": 48300
|
| 160821 |
+
},
|
| 160822 |
+
{
|
| 160823 |
+
"epoch": 389.14,
|
| 160824 |
+
"learning_rate": 9.242195512820513e-06,
|
| 160825 |
+
"loss": 0.3744,
|
| 160826 |
+
"step": 48305
|
| 160827 |
+
},
|
| 160828 |
+
{
|
| 160829 |
+
"epoch": 389.18,
|
| 160830 |
+
"learning_rate": 9.242115384615386e-06,
|
| 160831 |
+
"loss": 0.6971,
|
| 160832 |
+
"step": 48310
|
| 160833 |
+
},
|
| 160834 |
+
{
|
| 160835 |
+
"epoch": 389.22,
|
| 160836 |
+
"learning_rate": 9.242035256410256e-06,
|
| 160837 |
+
"loss": 1.2726,
|
| 160838 |
+
"step": 48315
|
| 160839 |
+
},
|
| 160840 |
+
{
|
| 160841 |
+
"epoch": 389.26,
|
| 160842 |
+
"learning_rate": 9.241955128205129e-06,
|
| 160843 |
+
"loss": 0.3767,
|
| 160844 |
+
"step": 48320
|
| 160845 |
+
},
|
| 160846 |
+
{
|
| 160847 |
+
"epoch": 389.3,
|
| 160848 |
+
"learning_rate": 9.241875000000002e-06,
|
| 160849 |
+
"loss": 1.1581,
|
| 160850 |
+
"step": 48325
|
| 160851 |
+
},
|
| 160852 |
+
{
|
| 160853 |
+
"epoch": 389.34,
|
| 160854 |
+
"learning_rate": 9.241794871794872e-06,
|
| 160855 |
+
"loss": 0.4466,
|
| 160856 |
+
"step": 48330
|
| 160857 |
+
},
|
| 160858 |
+
{
|
| 160859 |
+
"epoch": 389.38,
|
| 160860 |
+
"learning_rate": 9.241714743589745e-06,
|
| 160861 |
+
"loss": 0.6444,
|
| 160862 |
+
"step": 48335
|
| 160863 |
+
},
|
| 160864 |
+
{
|
| 160865 |
+
"epoch": 389.42,
|
| 160866 |
+
"learning_rate": 9.241634615384616e-06,
|
| 160867 |
+
"loss": 1.0077,
|
| 160868 |
+
"step": 48340
|
| 160869 |
+
},
|
| 160870 |
+
{
|
| 160871 |
+
"epoch": 389.46,
|
| 160872 |
+
"learning_rate": 9.241554487179487e-06,
|
| 160873 |
+
"loss": 0.3924,
|
| 160874 |
+
"step": 48345
|
| 160875 |
+
},
|
| 160876 |
+
{
|
| 160877 |
+
"epoch": 389.5,
|
| 160878 |
+
"learning_rate": 9.241474358974359e-06,
|
| 160879 |
+
"loss": 0.3339,
|
| 160880 |
+
"step": 48350
|
| 160881 |
+
},
|
| 160882 |
+
{
|
| 160883 |
+
"epoch": 389.54,
|
| 160884 |
+
"learning_rate": 9.241394230769232e-06,
|
| 160885 |
+
"loss": 0.4085,
|
| 160886 |
+
"step": 48355
|
| 160887 |
+
},
|
| 160888 |
+
{
|
| 160889 |
+
"epoch": 389.58,
|
| 160890 |
+
"learning_rate": 9.241314102564103e-06,
|
| 160891 |
+
"loss": 0.6992,
|
| 160892 |
+
"step": 48360
|
| 160893 |
+
},
|
| 160894 |
+
{
|
| 160895 |
+
"epoch": 389.62,
|
| 160896 |
+
"learning_rate": 9.241233974358975e-06,
|
| 160897 |
+
"loss": 1.134,
|
| 160898 |
+
"step": 48365
|
| 160899 |
+
},
|
| 160900 |
+
{
|
| 160901 |
+
"epoch": 389.66,
|
| 160902 |
+
"learning_rate": 9.241153846153846e-06,
|
| 160903 |
+
"loss": 0.346,
|
| 160904 |
+
"step": 48370
|
| 160905 |
+
},
|
| 160906 |
+
{
|
| 160907 |
+
"epoch": 389.7,
|
| 160908 |
+
"learning_rate": 9.241073717948719e-06,
|
| 160909 |
+
"loss": 0.3594,
|
| 160910 |
+
"step": 48375
|
| 160911 |
+
},
|
| 160912 |
+
{
|
| 160913 |
+
"epoch": 389.74,
|
| 160914 |
+
"learning_rate": 9.24099358974359e-06,
|
| 160915 |
+
"loss": 0.3704,
|
| 160916 |
+
"step": 48380
|
| 160917 |
+
},
|
| 160918 |
+
{
|
| 160919 |
+
"epoch": 389.78,
|
| 160920 |
+
"learning_rate": 9.240913461538462e-06,
|
| 160921 |
+
"loss": 0.6779,
|
| 160922 |
+
"step": 48385
|
| 160923 |
+
},
|
| 160924 |
+
{
|
| 160925 |
+
"epoch": 389.82,
|
| 160926 |
+
"learning_rate": 9.240833333333335e-06,
|
| 160927 |
+
"loss": 1.166,
|
| 160928 |
+
"step": 48390
|
| 160929 |
+
},
|
| 160930 |
+
{
|
| 160931 |
+
"epoch": 389.86,
|
| 160932 |
+
"learning_rate": 9.240753205128206e-06,
|
| 160933 |
+
"loss": 0.3895,
|
| 160934 |
+
"step": 48395
|
| 160935 |
+
},
|
| 160936 |
+
{
|
| 160937 |
+
"epoch": 389.9,
|
| 160938 |
+
"learning_rate": 9.240673076923078e-06,
|
| 160939 |
+
"loss": 0.4289,
|
| 160940 |
+
"step": 48400
|
| 160941 |
+
},
|
| 160942 |
+
{
|
| 160943 |
+
"epoch": 389.94,
|
| 160944 |
+
"learning_rate": 9.240592948717949e-06,
|
| 160945 |
+
"loss": 0.4343,
|
| 160946 |
+
"step": 48405
|
| 160947 |
+
},
|
| 160948 |
+
{
|
| 160949 |
+
"epoch": 389.98,
|
| 160950 |
+
"learning_rate": 9.240512820512822e-06,
|
| 160951 |
+
"loss": 0.91,
|
| 160952 |
+
"step": 48410
|
| 160953 |
+
},
|
| 160954 |
+
{
|
| 160955 |
+
"epoch": 390.0,
|
| 160956 |
+
"eval_loss": 0.42324376106262207,
|
| 160957 |
+
"eval_runtime": 40.5411,
|
| 160958 |
+
"eval_samples_per_second": 20.621,
|
| 160959 |
+
"eval_steps_per_second": 0.666,
|
| 160960 |
+
"eval_wer": 0.2087578582267505,
|
| 160961 |
+
"step": 48412
|
| 160962 |
+
},
|
| 160963 |
+
{
|
| 160964 |
+
"epoch": 390.02,
|
| 160965 |
+
"learning_rate": 9.240432692307693e-06,
|
| 160966 |
+
"loss": 0.3956,
|
| 160967 |
+
"step": 48415
|
| 160968 |
+
},
|
| 160969 |
+
{
|
| 160970 |
+
"epoch": 390.06,
|
| 160971 |
+
"learning_rate": 9.240352564102565e-06,
|
| 160972 |
+
"loss": 0.3495,
|
| 160973 |
+
"step": 48420
|
| 160974 |
+
},
|
| 160975 |
+
{
|
| 160976 |
+
"epoch": 390.1,
|
| 160977 |
+
"learning_rate": 9.240272435897438e-06,
|
| 160978 |
+
"loss": 0.3291,
|
| 160979 |
+
"step": 48425
|
| 160980 |
+
},
|
| 160981 |
+
{
|
| 160982 |
+
"epoch": 390.14,
|
| 160983 |
+
"learning_rate": 9.240192307692309e-06,
|
| 160984 |
+
"loss": 0.4074,
|
| 160985 |
+
"step": 48430
|
| 160986 |
+
},
|
| 160987 |
+
{
|
| 160988 |
+
"epoch": 390.18,
|
| 160989 |
+
"learning_rate": 9.24011217948718e-06,
|
| 160990 |
+
"loss": 0.8909,
|
| 160991 |
+
"step": 48435
|
| 160992 |
+
},
|
| 160993 |
+
{
|
| 160994 |
+
"epoch": 390.22,
|
| 160995 |
+
"learning_rate": 9.240032051282052e-06,
|
| 160996 |
+
"loss": 0.9895,
|
| 160997 |
+
"step": 48440
|
| 160998 |
+
},
|
| 160999 |
+
{
|
| 161000 |
+
"epoch": 390.27,
|
| 161001 |
+
"learning_rate": 9.239951923076925e-06,
|
| 161002 |
+
"loss": 0.2848,
|
| 161003 |
+
"step": 48445
|
| 161004 |
+
},
|
| 161005 |
+
{
|
| 161006 |
+
"epoch": 390.31,
|
| 161007 |
+
"learning_rate": 9.239871794871794e-06,
|
| 161008 |
+
"loss": 0.3077,
|
| 161009 |
+
"step": 48450
|
| 161010 |
+
},
|
| 161011 |
+
{
|
| 161012 |
+
"epoch": 390.35,
|
| 161013 |
+
"learning_rate": 9.239791666666668e-06,
|
| 161014 |
+
"loss": 0.4015,
|
| 161015 |
+
"step": 48455
|
| 161016 |
+
},
|
| 161017 |
+
{
|
| 161018 |
+
"epoch": 390.39,
|
| 161019 |
+
"learning_rate": 9.239711538461539e-06,
|
| 161020 |
+
"loss": 0.7797,
|
| 161021 |
+
"step": 48460
|
| 161022 |
+
},
|
| 161023 |
+
{
|
| 161024 |
+
"epoch": 390.43,
|
| 161025 |
+
"learning_rate": 9.23963141025641e-06,
|
| 161026 |
+
"loss": 1.0206,
|
| 161027 |
+
"step": 48465
|
| 161028 |
+
},
|
| 161029 |
+
{
|
| 161030 |
+
"epoch": 390.47,
|
| 161031 |
+
"learning_rate": 9.239551282051282e-06,
|
| 161032 |
+
"loss": 0.351,
|
| 161033 |
+
"step": 48470
|
| 161034 |
+
},
|
| 161035 |
+
{
|
| 161036 |
+
"epoch": 390.51,
|
| 161037 |
+
"learning_rate": 9.239471153846155e-06,
|
| 161038 |
+
"loss": 0.3726,
|
| 161039 |
+
"step": 48475
|
| 161040 |
+
},
|
| 161041 |
+
{
|
| 161042 |
+
"epoch": 390.55,
|
| 161043 |
+
"learning_rate": 9.239391025641026e-06,
|
| 161044 |
+
"loss": 0.3983,
|
| 161045 |
+
"step": 48480
|
| 161046 |
+
},
|
| 161047 |
+
{
|
| 161048 |
+
"epoch": 390.59,
|
| 161049 |
+
"learning_rate": 9.239310897435897e-06,
|
| 161050 |
+
"loss": 0.9361,
|
| 161051 |
+
"step": 48485
|
| 161052 |
+
},
|
| 161053 |
+
{
|
| 161054 |
+
"epoch": 390.63,
|
| 161055 |
+
"learning_rate": 9.23923076923077e-06,
|
| 161056 |
+
"loss": 1.0041,
|
| 161057 |
+
"step": 48490
|
| 161058 |
+
},
|
| 161059 |
+
{
|
| 161060 |
+
"epoch": 390.67,
|
| 161061 |
+
"learning_rate": 9.239150641025642e-06,
|
| 161062 |
+
"loss": 0.3347,
|
| 161063 |
+
"step": 48495
|
| 161064 |
+
},
|
| 161065 |
+
{
|
| 161066 |
+
"epoch": 390.71,
|
| 161067 |
+
"learning_rate": 9.239070512820513e-06,
|
| 161068 |
+
"loss": 0.3517,
|
| 161069 |
+
"step": 48500
|
| 161070 |
+
},
|
| 161071 |
+
{
|
| 161072 |
+
"epoch": 390.75,
|
| 161073 |
+
"learning_rate": 9.238990384615385e-06,
|
| 161074 |
+
"loss": 0.4371,
|
| 161075 |
+
"step": 48505
|
| 161076 |
+
},
|
| 161077 |
+
{
|
| 161078 |
+
"epoch": 390.79,
|
| 161079 |
+
"learning_rate": 9.238910256410258e-06,
|
| 161080 |
+
"loss": 0.8136,
|
| 161081 |
+
"step": 48510
|
| 161082 |
+
},
|
| 161083 |
+
{
|
| 161084 |
+
"epoch": 390.83,
|
| 161085 |
+
"learning_rate": 9.238830128205129e-06,
|
| 161086 |
+
"loss": 1.0253,
|
| 161087 |
+
"step": 48515
|
| 161088 |
+
},
|
| 161089 |
+
{
|
| 161090 |
+
"epoch": 390.87,
|
| 161091 |
+
"learning_rate": 9.23875e-06,
|
| 161092 |
+
"loss": 0.3152,
|
| 161093 |
+
"step": 48520
|
| 161094 |
+
},
|
| 161095 |
+
{
|
| 161096 |
+
"epoch": 390.91,
|
| 161097 |
+
"learning_rate": 9.238669871794873e-06,
|
| 161098 |
+
"loss": 0.3272,
|
| 161099 |
+
"step": 48525
|
| 161100 |
+
},
|
| 161101 |
+
{
|
| 161102 |
+
"epoch": 390.95,
|
| 161103 |
+
"learning_rate": 9.238589743589745e-06,
|
| 161104 |
+
"loss": 0.6504,
|
| 161105 |
+
"step": 48530
|
| 161106 |
+
},
|
| 161107 |
+
{
|
| 161108 |
+
"epoch": 390.99,
|
| 161109 |
+
"learning_rate": 9.238509615384616e-06,
|
| 161110 |
+
"loss": 0.9365,
|
| 161111 |
+
"step": 48535
|
| 161112 |
+
},
|
| 161113 |
+
{
|
| 161114 |
+
"epoch": 391.0,
|
| 161115 |
+
"eval_loss": 0.43879130482673645,
|
| 161116 |
+
"eval_runtime": 40.2761,
|
| 161117 |
+
"eval_samples_per_second": 20.757,
|
| 161118 |
+
"eval_steps_per_second": 0.67,
|
| 161119 |
+
"eval_wer": 0.20352941176470588,
|
| 161120 |
+
"step": 48536
|
| 161121 |
+
},
|
| 161122 |
+
{
|
| 161123 |
+
"epoch": 388.03,
|
| 161124 |
+
"learning_rate": 9.238429487179487e-06,
|
| 161125 |
+
"loss": 0.3571,
|
| 161126 |
+
"step": 48540
|
| 161127 |
+
},
|
| 161128 |
+
{
|
| 161129 |
+
"epoch": 388.07,
|
| 161130 |
+
"learning_rate": 9.23834935897436e-06,
|
| 161131 |
+
"loss": 0.3744,
|
| 161132 |
+
"step": 48545
|
| 161133 |
+
},
|
| 161134 |
+
{
|
| 161135 |
+
"epoch": 388.11,
|
| 161136 |
+
"learning_rate": 9.238269230769232e-06,
|
| 161137 |
+
"loss": 0.3287,
|
| 161138 |
+
"step": 48550
|
| 161139 |
+
},
|
| 161140 |
+
{
|
| 161141 |
+
"epoch": 388.15,
|
| 161142 |
+
"learning_rate": 9.238189102564103e-06,
|
| 161143 |
+
"loss": 0.446,
|
| 161144 |
+
"step": 48555
|
| 161145 |
+
},
|
| 161146 |
+
{
|
| 161147 |
+
"epoch": 388.19,
|
| 161148 |
+
"learning_rate": 9.238108974358975e-06,
|
| 161149 |
+
"loss": 1.5071,
|
| 161150 |
+
"step": 48560
|
| 161151 |
+
},
|
| 161152 |
+
{
|
| 161153 |
+
"epoch": 388.23,
|
| 161154 |
+
"learning_rate": 9.238028846153848e-06,
|
| 161155 |
+
"loss": 0.6705,
|
| 161156 |
+
"step": 48565
|
| 161157 |
+
},
|
| 161158 |
+
{
|
| 161159 |
+
"epoch": 388.27,
|
| 161160 |
+
"learning_rate": 9.237948717948719e-06,
|
| 161161 |
+
"loss": 0.3709,
|
| 161162 |
+
"step": 48570
|
| 161163 |
+
},
|
| 161164 |
+
{
|
| 161165 |
+
"epoch": 388.31,
|
| 161166 |
+
"learning_rate": 9.23786858974359e-06,
|
| 161167 |
+
"loss": 0.3886,
|
| 161168 |
+
"step": 48575
|
| 161169 |
+
},
|
| 161170 |
+
{
|
| 161171 |
+
"epoch": 388.35,
|
| 161172 |
+
"learning_rate": 9.237788461538463e-06,
|
| 161173 |
+
"loss": 0.453,
|
| 161174 |
+
"step": 48580
|
| 161175 |
+
},
|
| 161176 |
+
{
|
| 161177 |
+
"epoch": 388.39,
|
| 161178 |
+
"learning_rate": 9.237708333333335e-06,
|
| 161179 |
+
"loss": 1.04,
|
| 161180 |
+
"step": 48585
|
| 161181 |
+
},
|
| 161182 |
+
{
|
| 161183 |
+
"epoch": 388.43,
|
| 161184 |
+
"learning_rate": 9.237628205128206e-06,
|
| 161185 |
+
"loss": 0.7673,
|
| 161186 |
+
"step": 48590
|
| 161187 |
+
},
|
| 161188 |
+
{
|
| 161189 |
+
"epoch": 388.47,
|
| 161190 |
+
"learning_rate": 9.237548076923077e-06,
|
| 161191 |
+
"loss": 0.3269,
|
| 161192 |
+
"step": 48595
|
| 161193 |
+
},
|
| 161194 |
+
{
|
| 161195 |
+
"epoch": 388.51,
|
| 161196 |
+
"learning_rate": 9.23746794871795e-06,
|
| 161197 |
+
"loss": 0.3925,
|
| 161198 |
+
"step": 48600
|
| 161199 |
+
},
|
| 161200 |
+
{
|
| 161201 |
+
"epoch": 388.55,
|
| 161202 |
+
"learning_rate": 9.23738782051282e-06,
|
| 161203 |
+
"loss": 0.4776,
|
| 161204 |
+
"step": 48605
|
| 161205 |
+
},
|
| 161206 |
+
{
|
| 161207 |
+
"epoch": 388.59,
|
| 161208 |
+
"learning_rate": 9.237307692307693e-06,
|
| 161209 |
+
"loss": 1.0403,
|
| 161210 |
+
"step": 48610
|
| 161211 |
+
},
|
| 161212 |
+
{
|
| 161213 |
+
"epoch": 388.63,
|
| 161214 |
+
"learning_rate": 9.237227564102565e-06,
|
| 161215 |
+
"loss": 0.7185,
|
| 161216 |
+
"step": 48615
|
| 161217 |
+
},
|
| 161218 |
+
{
|
| 161219 |
+
"epoch": 388.67,
|
| 161220 |
+
"learning_rate": 9.237147435897436e-06,
|
| 161221 |
+
"loss": 0.3245,
|
| 161222 |
+
"step": 48620
|
| 161223 |
+
},
|
| 161224 |
+
{
|
| 161225 |
+
"epoch": 388.71,
|
| 161226 |
+
"learning_rate": 9.237067307692309e-06,
|
| 161227 |
+
"loss": 0.3332,
|
| 161228 |
+
"step": 48625
|
| 161229 |
+
},
|
| 161230 |
+
{
|
| 161231 |
+
"epoch": 388.75,
|
| 161232 |
+
"learning_rate": 9.23698717948718e-06,
|
| 161233 |
+
"loss": 0.4575,
|
| 161234 |
+
"step": 48630
|
| 161235 |
+
},
|
| 161236 |
+
{
|
| 161237 |
+
"epoch": 388.79,
|
| 161238 |
+
"learning_rate": 9.236907051282052e-06,
|
| 161239 |
+
"loss": 0.9737,
|
| 161240 |
+
"step": 48635
|
| 161241 |
+
},
|
| 161242 |
+
{
|
| 161243 |
+
"epoch": 388.83,
|
| 161244 |
+
"learning_rate": 9.236826923076923e-06,
|
| 161245 |
+
"loss": 0.745,
|
| 161246 |
+
"step": 48640
|
| 161247 |
+
},
|
| 161248 |
+
{
|
| 161249 |
+
"epoch": 388.87,
|
| 161250 |
+
"learning_rate": 9.236746794871796e-06,
|
| 161251 |
+
"loss": 0.2776,
|
| 161252 |
+
"step": 48645
|
| 161253 |
+
},
|
| 161254 |
+
{
|
| 161255 |
+
"epoch": 388.91,
|
| 161256 |
+
"learning_rate": 9.236666666666667e-06,
|
| 161257 |
+
"loss": 0.3983,
|
| 161258 |
+
"step": 48650
|
| 161259 |
+
},
|
| 161260 |
+
{
|
| 161261 |
+
"epoch": 388.95,
|
| 161262 |
+
"learning_rate": 9.236586538461539e-06,
|
| 161263 |
+
"loss": 0.485,
|
| 161264 |
+
"step": 48655
|
| 161265 |
+
},
|
| 161266 |
+
{
|
| 161267 |
+
"epoch": 388.99,
|
| 161268 |
+
"learning_rate": 9.23650641025641e-06,
|
| 161269 |
+
"loss": 1.6334,
|
| 161270 |
+
"step": 48660
|
| 161271 |
+
},
|
| 161272 |
+
{
|
| 161273 |
+
"epoch": 389.0,
|
| 161274 |
+
"eval_loss": 0.42802751064300537,
|
| 161275 |
+
"eval_runtime": 41.6818,
|
| 161276 |
+
"eval_samples_per_second": 20.057,
|
| 161277 |
+
"eval_steps_per_second": 0.648,
|
| 161278 |
+
"eval_wer": 0.19799243526331103,
|
| 161279 |
+
"step": 48661
|
| 161280 |
+
},
|
| 161281 |
+
{
|
| 161282 |
+
"epoch": 392.03,
|
| 161283 |
+
"learning_rate": 9.236426282051283e-06,
|
| 161284 |
+
"loss": 0.4317,
|
| 161285 |
+
"step": 48665
|
| 161286 |
+
},
|
| 161287 |
+
{
|
| 161288 |
+
"epoch": 392.07,
|
| 161289 |
+
"learning_rate": 9.236346153846155e-06,
|
| 161290 |
+
"loss": 1.1302,
|
| 161291 |
+
"step": 48670
|
| 161292 |
+
},
|
| 161293 |
+
{
|
| 161294 |
+
"epoch": 392.11,
|
| 161295 |
+
"learning_rate": 9.236266025641026e-06,
|
| 161296 |
+
"loss": 0.3995,
|
| 161297 |
+
"step": 48675
|
| 161298 |
+
},
|
| 161299 |
+
{
|
| 161300 |
+
"epoch": 392.15,
|
| 161301 |
+
"learning_rate": 9.236185897435899e-06,
|
| 161302 |
+
"loss": 0.5221,
|
| 161303 |
+
"step": 48680
|
| 161304 |
+
},
|
| 161305 |
+
{
|
| 161306 |
+
"epoch": 392.19,
|
| 161307 |
+
"learning_rate": 9.23610576923077e-06,
|
| 161308 |
+
"loss": 1.0194,
|
| 161309 |
+
"step": 48685
|
| 161310 |
+
},
|
| 161311 |
+
{
|
| 161312 |
+
"epoch": 392.23,
|
| 161313 |
+
"learning_rate": 9.236025641025642e-06,
|
| 161314 |
+
"loss": 0.8184,
|
| 161315 |
+
"step": 48690
|
| 161316 |
+
},
|
| 161317 |
+
{
|
| 161318 |
+
"epoch": 392.27,
|
| 161319 |
+
"learning_rate": 9.235945512820513e-06,
|
| 161320 |
+
"loss": 0.3666,
|
| 161321 |
+
"step": 48695
|
| 161322 |
+
},
|
| 161323 |
+
{
|
| 161324 |
+
"epoch": 392.31,
|
| 161325 |
+
"learning_rate": 9.235865384615386e-06,
|
| 161326 |
+
"loss": 0.392,
|
| 161327 |
+
"step": 48700
|
| 161328 |
+
},
|
| 161329 |
+
{
|
| 161330 |
+
"epoch": 392.35,
|
| 161331 |
+
"learning_rate": 9.235785256410257e-06,
|
| 161332 |
+
"loss": 0.4886,
|
| 161333 |
+
"step": 48705
|
| 161334 |
+
},
|
| 161335 |
+
{
|
| 161336 |
+
"epoch": 392.39,
|
| 161337 |
+
"learning_rate": 9.235705128205129e-06,
|
| 161338 |
+
"loss": 1.0533,
|
| 161339 |
+
"step": 48710
|
| 161340 |
+
},
|
| 161341 |
+
{
|
| 161342 |
+
"epoch": 392.43,
|
| 161343 |
+
"learning_rate": 9.235625e-06,
|
| 161344 |
+
"loss": 0.7594,
|
| 161345 |
+
"step": 48715
|
| 161346 |
+
},
|
| 161347 |
+
{
|
| 161348 |
+
"epoch": 392.47,
|
| 161349 |
+
"learning_rate": 9.235544871794873e-06,
|
| 161350 |
+
"loss": 0.3777,
|
| 161351 |
+
"step": 48720
|
| 161352 |
+
},
|
| 161353 |
+
{
|
| 161354 |
+
"epoch": 392.51,
|
| 161355 |
+
"learning_rate": 9.235464743589745e-06,
|
| 161356 |
+
"loss": 0.4546,
|
| 161357 |
+
"step": 48725
|
| 161358 |
+
},
|
| 161359 |
+
{
|
| 161360 |
+
"epoch": 392.55,
|
| 161361 |
+
"learning_rate": 9.235384615384616e-06,
|
| 161362 |
+
"loss": 0.4622,
|
| 161363 |
+
"step": 48730
|
| 161364 |
+
},
|
| 161365 |
+
{
|
| 161366 |
+
"epoch": 392.59,
|
| 161367 |
+
"learning_rate": 9.235304487179489e-06,
|
| 161368 |
+
"loss": 1.1097,
|
| 161369 |
+
"step": 48735
|
| 161370 |
+
},
|
| 161371 |
+
{
|
| 161372 |
+
"epoch": 392.63,
|
| 161373 |
+
"learning_rate": 9.23522435897436e-06,
|
| 161374 |
+
"loss": 0.7749,
|
| 161375 |
+
"step": 48740
|
| 161376 |
+
},
|
| 161377 |
+
{
|
| 161378 |
+
"epoch": 392.67,
|
| 161379 |
+
"learning_rate": 9.235144230769232e-06,
|
| 161380 |
+
"loss": 0.3444,
|
| 161381 |
+
"step": 48745
|
| 161382 |
+
},
|
| 161383 |
+
{
|
| 161384 |
+
"epoch": 392.71,
|
| 161385 |
+
"learning_rate": 9.235064102564103e-06,
|
| 161386 |
+
"loss": 0.4619,
|
| 161387 |
+
"step": 48750
|
| 161388 |
+
},
|
| 161389 |
+
{
|
| 161390 |
+
"epoch": 392.76,
|
| 161391 |
+
"learning_rate": 9.234983974358976e-06,
|
| 161392 |
+
"loss": 0.4614,
|
| 161393 |
+
"step": 48755
|
| 161394 |
+
},
|
| 161395 |
+
{
|
| 161396 |
+
"epoch": 392.8,
|
| 161397 |
+
"learning_rate": 9.234903846153846e-06,
|
| 161398 |
+
"loss": 0.8623,
|
| 161399 |
+
"step": 48760
|
| 161400 |
+
},
|
| 161401 |
+
{
|
| 161402 |
+
"epoch": 392.84,
|
| 161403 |
+
"learning_rate": 9.234823717948719e-06,
|
| 161404 |
+
"loss": 0.7069,
|
| 161405 |
+
"step": 48765
|
| 161406 |
+
},
|
| 161407 |
+
{
|
| 161408 |
+
"epoch": 392.88,
|
| 161409 |
+
"learning_rate": 9.23474358974359e-06,
|
| 161410 |
+
"loss": 0.3161,
|
| 161411 |
+
"step": 48770
|
| 161412 |
+
},
|
| 161413 |
+
{
|
| 161414 |
+
"epoch": 392.92,
|
| 161415 |
+
"learning_rate": 9.234663461538462e-06,
|
| 161416 |
+
"loss": 0.3604,
|
| 161417 |
+
"step": 48775
|
| 161418 |
+
},
|
| 161419 |
+
{
|
| 161420 |
+
"epoch": 392.96,
|
| 161421 |
+
"learning_rate": 9.234583333333335e-06,
|
| 161422 |
+
"loss": 0.5099,
|
| 161423 |
+
"step": 48780
|
| 161424 |
+
},
|
| 161425 |
+
{
|
| 161426 |
+
"epoch": 393.0,
|
| 161427 |
+
"learning_rate": 9.234503205128206e-06,
|
| 161428 |
+
"loss": 1.3269,
|
| 161429 |
+
"step": 48785
|
| 161430 |
+
},
|
| 161431 |
+
{
|
| 161432 |
+
"epoch": 393.0,
|
| 161433 |
+
"eval_loss": 0.46593815088272095,
|
| 161434 |
+
"eval_runtime": 39.2623,
|
| 161435 |
+
"eval_samples_per_second": 21.267,
|
| 161436 |
+
"eval_steps_per_second": 0.688,
|
| 161437 |
+
"eval_wer": 0.20493316594047706,
|
| 161438 |
+
"step": 48785
|
| 161439 |
+
},
|
| 161440 |
+
{
|
| 161441 |
+
"epoch": 393.04,
|
| 161442 |
+
"learning_rate": 9.234423076923077e-06,
|
| 161443 |
+
"loss": 0.3957,
|
| 161444 |
+
"step": 48790
|
| 161445 |
+
},
|
| 161446 |
+
{
|
| 161447 |
+
"epoch": 393.08,
|
| 161448 |
+
"learning_rate": 9.234342948717949e-06,
|
| 161449 |
+
"loss": 0.4109,
|
| 161450 |
+
"step": 48795
|
| 161451 |
+
},
|
| 161452 |
+
{
|
| 161453 |
+
"epoch": 393.12,
|
| 161454 |
+
"learning_rate": 9.234262820512822e-06,
|
| 161455 |
+
"loss": 0.3512,
|
| 161456 |
+
"step": 48800
|
| 161457 |
+
},
|
| 161458 |
+
{
|
| 161459 |
+
"epoch": 393.16,
|
| 161460 |
+
"learning_rate": 9.234182692307693e-06,
|
| 161461 |
+
"loss": 0.5299,
|
| 161462 |
+
"step": 48805
|
| 161463 |
+
},
|
| 161464 |
+
{
|
| 161465 |
+
"epoch": 393.2,
|
| 161466 |
+
"learning_rate": 9.234102564102564e-06,
|
| 161467 |
+
"loss": 1.2309,
|
| 161468 |
+
"step": 48810
|
| 161469 |
+
},
|
| 161470 |
+
{
|
| 161471 |
+
"epoch": 393.24,
|
| 161472 |
+
"learning_rate": 9.234022435897436e-06,
|
| 161473 |
+
"loss": 0.3406,
|
| 161474 |
+
"step": 48815
|
| 161475 |
+
},
|
| 161476 |
+
{
|
| 161477 |
+
"epoch": 393.28,
|
| 161478 |
+
"learning_rate": 9.233942307692309e-06,
|
| 161479 |
+
"loss": 0.3598,
|
| 161480 |
+
"step": 48820
|
| 161481 |
+
},
|
| 161482 |
+
{
|
| 161483 |
+
"epoch": 393.32,
|
| 161484 |
+
"learning_rate": 9.23386217948718e-06,
|
| 161485 |
+
"loss": 0.3676,
|
| 161486 |
+
"step": 48825
|
| 161487 |
+
},
|
| 161488 |
+
{
|
| 161489 |
+
"epoch": 393.36,
|
| 161490 |
+
"learning_rate": 9.233782051282052e-06,
|
| 161491 |
+
"loss": 0.5236,
|
| 161492 |
+
"step": 48830
|
| 161493 |
+
},
|
| 161494 |
+
{
|
| 161495 |
+
"epoch": 393.4,
|
| 161496 |
+
"learning_rate": 9.233701923076925e-06,
|
| 161497 |
+
"loss": 1.3985,
|
| 161498 |
+
"step": 48835
|
| 161499 |
+
},
|
| 161500 |
+
{
|
| 161501 |
+
"epoch": 393.44,
|
| 161502 |
+
"learning_rate": 9.233621794871796e-06,
|
| 161503 |
+
"loss": 0.4045,
|
| 161504 |
+
"step": 48840
|
| 161505 |
+
},
|
| 161506 |
+
{
|
| 161507 |
+
"epoch": 393.48,
|
| 161508 |
+
"learning_rate": 9.233541666666667e-06,
|
| 161509 |
+
"loss": 0.3575,
|
| 161510 |
+
"step": 48845
|
| 161511 |
+
},
|
| 161512 |
+
{
|
| 161513 |
+
"epoch": 393.52,
|
| 161514 |
+
"learning_rate": 9.233461538461539e-06,
|
| 161515 |
+
"loss": 0.3358,
|
| 161516 |
+
"step": 48850
|
| 161517 |
+
},
|
| 161518 |
+
{
|
| 161519 |
+
"epoch": 393.56,
|
| 161520 |
+
"learning_rate": 9.233381410256412e-06,
|
| 161521 |
+
"loss": 0.5091,
|
| 161522 |
+
"step": 48855
|
| 161523 |
+
},
|
| 161524 |
+
{
|
| 161525 |
+
"epoch": 393.6,
|
| 161526 |
+
"learning_rate": 9.233301282051283e-06,
|
| 161527 |
+
"loss": 1.3725,
|
| 161528 |
+
"step": 48860
|
| 161529 |
+
},
|
| 161530 |
+
{
|
| 161531 |
+
"epoch": 393.64,
|
| 161532 |
+
"learning_rate": 9.233221153846155e-06,
|
| 161533 |
+
"loss": 0.3965,
|
| 161534 |
+
"step": 48865
|
| 161535 |
+
},
|
| 161536 |
+
{
|
| 161537 |
+
"epoch": 393.68,
|
| 161538 |
+
"learning_rate": 9.233141025641026e-06,
|
| 161539 |
+
"loss": 0.3452,
|
| 161540 |
+
"step": 48870
|
| 161541 |
+
},
|
| 161542 |
+
{
|
| 161543 |
+
"epoch": 393.72,
|
| 161544 |
+
"learning_rate": 9.233060897435899e-06,
|
| 161545 |
+
"loss": 0.3911,
|
| 161546 |
+
"step": 48875
|
| 161547 |
+
},
|
| 161548 |
+
{
|
| 161549 |
+
"epoch": 393.76,
|
| 161550 |
+
"learning_rate": 9.23298076923077e-06,
|
| 161551 |
+
"loss": 0.5074,
|
| 161552 |
+
"step": 48880
|
| 161553 |
+
},
|
| 161554 |
+
{
|
| 161555 |
+
"epoch": 393.8,
|
| 161556 |
+
"learning_rate": 9.232900641025642e-06,
|
| 161557 |
+
"loss": 1.3414,
|
| 161558 |
+
"step": 48885
|
| 161559 |
+
},
|
| 161560 |
+
{
|
| 161561 |
+
"epoch": 393.84,
|
| 161562 |
+
"learning_rate": 9.232820512820515e-06,
|
| 161563 |
+
"loss": 0.336,
|
| 161564 |
+
"step": 48890
|
| 161565 |
+
},
|
| 161566 |
+
{
|
| 161567 |
+
"epoch": 393.88,
|
| 161568 |
+
"learning_rate": 9.232740384615384e-06,
|
| 161569 |
+
"loss": 0.3504,
|
| 161570 |
+
"step": 48895
|
| 161571 |
+
},
|
| 161572 |
+
{
|
| 161573 |
+
"epoch": 393.92,
|
| 161574 |
+
"learning_rate": 9.232660256410257e-06,
|
| 161575 |
+
"loss": 0.3444,
|
| 161576 |
+
"step": 48900
|
| 161577 |
+
},
|
| 161578 |
+
{
|
| 161579 |
+
"epoch": 393.96,
|
| 161580 |
+
"learning_rate": 9.232580128205129e-06,
|
| 161581 |
+
"loss": 0.6403,
|
| 161582 |
+
"step": 48905
|
| 161583 |
+
},
|
| 161584 |
+
{
|
| 161585 |
+
"epoch": 394.0,
|
| 161586 |
+
"eval_loss": 0.4497189521789551,
|
| 161587 |
+
"eval_runtime": 38.7203,
|
| 161588 |
+
"eval_samples_per_second": 21.668,
|
| 161589 |
+
"eval_steps_per_second": 0.697,
|
| 161590 |
+
"eval_wer": 0.2061362152752593,
|
| 161591 |
+
"step": 48909
|
| 161592 |
}
|
| 161593 |
],
|
| 161594 |
+
"max_steps": 620000,
|
| 161595 |
"num_train_epochs": 5000,
|
| 161596 |
+
"total_flos": 1.3761440474553568e+20,
|
| 161597 |
"trial_name": null,
|
| 161598 |
"trial_params": null
|
| 161599 |
}
|
model-bin/finetune/base/{checkpoint-48288 → checkpoint-48909}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629732801.0318525/events.out.tfevents.1629732801.74272264b15c.932.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aba8c39840858da5f201e2f7408a67578c63d6df9b0092566d4c9d221b1e5cf5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629733448.706626/events.out.tfevents.1629733448.74272264b15c.932.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a472b033ce96c0bd6a0f8060530bf7d83e4fe6eb0e6af549c3db317c9339585e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629734099.0176368/events.out.tfevents.1629734099.74272264b15c.932.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89f7afb7fcfb6ec51d48c899bff3c9757345790d3f125f466f36a825f234cc93
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629734741.4134967/events.out.tfevents.1629734741.74272264b15c.932.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae022ea2d500cd12c05772ee93389d40f541de0126ded4d048c12d066b369f9a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629735490.0998623/events.out.tfevents.1629735490.74272264b15c.932.171
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:69012c6f3e379c3b8191cb281917566919db8c9ca698bccb394ddea32809c259
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629732801.74272264b15c.932.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:27d7917c6e798d63c0633940900a5c98817f2e26dfff1de436933b06111dc876
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629733448.74272264b15c.932.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88757493f5c75e4b133feb912bed6ae6245261256c533203525132988b9268f0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629734099.74272264b15c.932.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a24553b540095365f9d31dad7c7c6b29be93ed52662de3b5ecebc80840f87ec7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629734741.74272264b15c.932.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86d08f2a05c7e44931bc0bfb3b104accc921929ec6a6080d49e46fca0e1a6849
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629735490.74272264b15c.932.170
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:803fab0dbe29104ac46e639c4f5aa2a0fb08d00baf68021d4d01e4e60d40361d
|
| 3 |
+
size 8462
|