"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629716349.3372576/events.out.tfevents.1629716349.74272264b15c.932.113 +3 -0
- model-bin/finetune/base/log/1629716986.7392747/events.out.tfevents.1629716986.74272264b15c.932.115 +3 -0
- model-bin/finetune/base/log/1629717626.7822158/events.out.tfevents.1629717626.74272264b15c.932.117 +3 -0
- model-bin/finetune/base/log/1629718254.5594208/events.out.tfevents.1629718254.74272264b15c.932.119 +3 -0
- model-bin/finetune/base/log/1629718890.2393265/events.out.tfevents.1629718890.74272264b15c.932.121 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629716349.74272264b15c.932.112 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629716986.74272264b15c.932.114 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629717626.74272264b15c.932.116 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629718254.74272264b15c.932.118 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629718890.74272264b15c.932.120 +3 -0
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/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 sha256:323678391e8aadc9775828253e4c8733881ea240ec250e28265f08422dc1a380
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/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 sha256:457fa8e92b572612d4a32aa7e0dbf42c00e47254b696c96af3d62dc6fc6344b9
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/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 sha256:1e9b91ab2837003f2b1e91b19cf6006c8c2c0a0984748dadf42405e34c592d4a
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/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 sha256:8f01615f858d0e4986be58d3b1cf8e6bbe2178c7de9f5b765fd848a808957723
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/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:e83611e9c1dc8bb2d6bf8327aaeef82558ee28c64fa698894b902b504ed5edd4
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -156837,11 +156837,806 @@
|
|
| 156837 |
"eval_steps_per_second": 0.686,
|
| 156838 |
"eval_wer": 0.2054794520547945,
|
| 156839 |
"step": 45173
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156840 |
}
|
| 156841 |
],
|
| 156842 |
-
"max_steps":
|
| 156843 |
"num_train_epochs": 5000,
|
| 156844 |
-
"total_flos": 1.
|
| 156845 |
"trial_name": null,
|
| 156846 |
"trial_params": null
|
| 156847 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
+
"epoch": 366.0,
|
| 5 |
+
"global_step": 45796,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 156837 |
"eval_steps_per_second": 0.686,
|
| 156838 |
"eval_wer": 0.2054794520547945,
|
| 156839 |
"step": 45173
|
| 156840 |
+
},
|
| 156841 |
+
{
|
| 156842 |
+
"epoch": 361.02,
|
| 156843 |
+
"learning_rate": 9.292339743589744e-06,
|
| 156844 |
+
"loss": 0.3835,
|
| 156845 |
+
"step": 45175
|
| 156846 |
+
},
|
| 156847 |
+
{
|
| 156848 |
+
"epoch": 361.06,
|
| 156849 |
+
"learning_rate": 9.292259615384615e-06,
|
| 156850 |
+
"loss": 0.3602,
|
| 156851 |
+
"step": 45180
|
| 156852 |
+
},
|
| 156853 |
+
{
|
| 156854 |
+
"epoch": 361.1,
|
| 156855 |
+
"learning_rate": 9.292179487179488e-06,
|
| 156856 |
+
"loss": 0.3974,
|
| 156857 |
+
"step": 45185
|
| 156858 |
+
},
|
| 156859 |
+
{
|
| 156860 |
+
"epoch": 361.14,
|
| 156861 |
+
"learning_rate": 9.29209935897436e-06,
|
| 156862 |
+
"loss": 0.4718,
|
| 156863 |
+
"step": 45190
|
| 156864 |
+
},
|
| 156865 |
+
{
|
| 156866 |
+
"epoch": 361.18,
|
| 156867 |
+
"learning_rate": 9.292019230769231e-06,
|
| 156868 |
+
"loss": 0.5956,
|
| 156869 |
+
"step": 45195
|
| 156870 |
+
},
|
| 156871 |
+
{
|
| 156872 |
+
"epoch": 361.22,
|
| 156873 |
+
"learning_rate": 9.291939102564104e-06,
|
| 156874 |
+
"loss": 1.0405,
|
| 156875 |
+
"step": 45200
|
| 156876 |
+
},
|
| 156877 |
+
{
|
| 156878 |
+
"epoch": 361.26,
|
| 156879 |
+
"learning_rate": 9.291858974358975e-06,
|
| 156880 |
+
"loss": 0.3789,
|
| 156881 |
+
"step": 45205
|
| 156882 |
+
},
|
| 156883 |
+
{
|
| 156884 |
+
"epoch": 361.3,
|
| 156885 |
+
"learning_rate": 9.291778846153847e-06,
|
| 156886 |
+
"loss": 0.3426,
|
| 156887 |
+
"step": 45210
|
| 156888 |
+
},
|
| 156889 |
+
{
|
| 156890 |
+
"epoch": 361.34,
|
| 156891 |
+
"learning_rate": 9.291698717948718e-06,
|
| 156892 |
+
"loss": 0.4088,
|
| 156893 |
+
"step": 45215
|
| 156894 |
+
},
|
| 156895 |
+
{
|
| 156896 |
+
"epoch": 361.38,
|
| 156897 |
+
"learning_rate": 9.291618589743591e-06,
|
| 156898 |
+
"loss": 0.7672,
|
| 156899 |
+
"step": 45220
|
| 156900 |
+
},
|
| 156901 |
+
{
|
| 156902 |
+
"epoch": 361.42,
|
| 156903 |
+
"learning_rate": 9.291538461538463e-06,
|
| 156904 |
+
"loss": 1.1723,
|
| 156905 |
+
"step": 45225
|
| 156906 |
+
},
|
| 156907 |
+
{
|
| 156908 |
+
"epoch": 361.46,
|
| 156909 |
+
"learning_rate": 9.291458333333334e-06,
|
| 156910 |
+
"loss": 0.3007,
|
| 156911 |
+
"step": 45230
|
| 156912 |
+
},
|
| 156913 |
+
{
|
| 156914 |
+
"epoch": 361.5,
|
| 156915 |
+
"learning_rate": 9.291378205128207e-06,
|
| 156916 |
+
"loss": 0.3785,
|
| 156917 |
+
"step": 45235
|
| 156918 |
+
},
|
| 156919 |
+
{
|
| 156920 |
+
"epoch": 361.54,
|
| 156921 |
+
"learning_rate": 9.291298076923078e-06,
|
| 156922 |
+
"loss": 0.3804,
|
| 156923 |
+
"step": 45240
|
| 156924 |
+
},
|
| 156925 |
+
{
|
| 156926 |
+
"epoch": 361.58,
|
| 156927 |
+
"learning_rate": 9.29121794871795e-06,
|
| 156928 |
+
"loss": 0.75,
|
| 156929 |
+
"step": 45245
|
| 156930 |
+
},
|
| 156931 |
+
{
|
| 156932 |
+
"epoch": 361.62,
|
| 156933 |
+
"learning_rate": 9.291137820512821e-06,
|
| 156934 |
+
"loss": 1.0444,
|
| 156935 |
+
"step": 45250
|
| 156936 |
+
},
|
| 156937 |
+
{
|
| 156938 |
+
"epoch": 361.66,
|
| 156939 |
+
"learning_rate": 9.291057692307694e-06,
|
| 156940 |
+
"loss": 0.3557,
|
| 156941 |
+
"step": 45255
|
| 156942 |
+
},
|
| 156943 |
+
{
|
| 156944 |
+
"epoch": 361.7,
|
| 156945 |
+
"learning_rate": 9.290977564102564e-06,
|
| 156946 |
+
"loss": 0.3839,
|
| 156947 |
+
"step": 45260
|
| 156948 |
+
},
|
| 156949 |
+
{
|
| 156950 |
+
"epoch": 361.74,
|
| 156951 |
+
"learning_rate": 9.290897435897437e-06,
|
| 156952 |
+
"loss": 0.4331,
|
| 156953 |
+
"step": 45265
|
| 156954 |
+
},
|
| 156955 |
+
{
|
| 156956 |
+
"epoch": 361.78,
|
| 156957 |
+
"learning_rate": 9.290817307692308e-06,
|
| 156958 |
+
"loss": 0.742,
|
| 156959 |
+
"step": 45270
|
| 156960 |
+
},
|
| 156961 |
+
{
|
| 156962 |
+
"epoch": 361.82,
|
| 156963 |
+
"learning_rate": 9.29073717948718e-06,
|
| 156964 |
+
"loss": 1.2639,
|
| 156965 |
+
"step": 45275
|
| 156966 |
+
},
|
| 156967 |
+
{
|
| 156968 |
+
"epoch": 361.86,
|
| 156969 |
+
"learning_rate": 9.290657051282051e-06,
|
| 156970 |
+
"loss": 0.3131,
|
| 156971 |
+
"step": 45280
|
| 156972 |
+
},
|
| 156973 |
+
{
|
| 156974 |
+
"epoch": 361.9,
|
| 156975 |
+
"learning_rate": 9.290576923076924e-06,
|
| 156976 |
+
"loss": 0.3454,
|
| 156977 |
+
"step": 45285
|
| 156978 |
+
},
|
| 156979 |
+
{
|
| 156980 |
+
"epoch": 361.94,
|
| 156981 |
+
"learning_rate": 9.290496794871795e-06,
|
| 156982 |
+
"loss": 0.4017,
|
| 156983 |
+
"step": 45290
|
| 156984 |
+
},
|
| 156985 |
+
{
|
| 156986 |
+
"epoch": 361.98,
|
| 156987 |
+
"learning_rate": 9.290416666666667e-06,
|
| 156988 |
+
"loss": 1.0668,
|
| 156989 |
+
"step": 45295
|
| 156990 |
+
},
|
| 156991 |
+
{
|
| 156992 |
+
"epoch": 362.0,
|
| 156993 |
+
"eval_loss": 0.41494983434677124,
|
| 156994 |
+
"eval_runtime": 39.7716,
|
| 156995 |
+
"eval_samples_per_second": 21.221,
|
| 156996 |
+
"eval_steps_per_second": 0.679,
|
| 156997 |
+
"eval_wer": 0.2089183094493344,
|
| 156998 |
+
"step": 45298
|
| 156999 |
+
},
|
| 157000 |
+
{
|
| 157001 |
+
"epoch": 362.02,
|
| 157002 |
+
"learning_rate": 9.29033653846154e-06,
|
| 157003 |
+
"loss": 0.3509,
|
| 157004 |
+
"step": 45300
|
| 157005 |
+
},
|
| 157006 |
+
{
|
| 157007 |
+
"epoch": 362.06,
|
| 157008 |
+
"learning_rate": 9.290256410256411e-06,
|
| 157009 |
+
"loss": 0.3,
|
| 157010 |
+
"step": 45305
|
| 157011 |
+
},
|
| 157012 |
+
{
|
| 157013 |
+
"epoch": 362.1,
|
| 157014 |
+
"learning_rate": 9.290176282051282e-06,
|
| 157015 |
+
"loss": 0.3262,
|
| 157016 |
+
"step": 45310
|
| 157017 |
+
},
|
| 157018 |
+
{
|
| 157019 |
+
"epoch": 362.14,
|
| 157020 |
+
"learning_rate": 9.290096153846154e-06,
|
| 157021 |
+
"loss": 0.3682,
|
| 157022 |
+
"step": 45315
|
| 157023 |
+
},
|
| 157024 |
+
{
|
| 157025 |
+
"epoch": 362.18,
|
| 157026 |
+
"learning_rate": 9.290016025641027e-06,
|
| 157027 |
+
"loss": 0.6127,
|
| 157028 |
+
"step": 45320
|
| 157029 |
+
},
|
| 157030 |
+
{
|
| 157031 |
+
"epoch": 362.22,
|
| 157032 |
+
"learning_rate": 9.289935897435898e-06,
|
| 157033 |
+
"loss": 1.1148,
|
| 157034 |
+
"step": 45325
|
| 157035 |
+
},
|
| 157036 |
+
{
|
| 157037 |
+
"epoch": 362.25,
|
| 157038 |
+
"learning_rate": 9.28985576923077e-06,
|
| 157039 |
+
"loss": 0.3755,
|
| 157040 |
+
"step": 45330
|
| 157041 |
+
},
|
| 157042 |
+
{
|
| 157043 |
+
"epoch": 362.29,
|
| 157044 |
+
"learning_rate": 9.289775641025643e-06,
|
| 157045 |
+
"loss": 0.3959,
|
| 157046 |
+
"step": 45335
|
| 157047 |
+
},
|
| 157048 |
+
{
|
| 157049 |
+
"epoch": 362.33,
|
| 157050 |
+
"learning_rate": 9.289695512820514e-06,
|
| 157051 |
+
"loss": 0.4437,
|
| 157052 |
+
"step": 45340
|
| 157053 |
+
},
|
| 157054 |
+
{
|
| 157055 |
+
"epoch": 362.37,
|
| 157056 |
+
"learning_rate": 9.289615384615385e-06,
|
| 157057 |
+
"loss": 0.71,
|
| 157058 |
+
"step": 45345
|
| 157059 |
+
},
|
| 157060 |
+
{
|
| 157061 |
+
"epoch": 362.41,
|
| 157062 |
+
"learning_rate": 9.289535256410257e-06,
|
| 157063 |
+
"loss": 1.124,
|
| 157064 |
+
"step": 45350
|
| 157065 |
+
},
|
| 157066 |
+
{
|
| 157067 |
+
"epoch": 362.45,
|
| 157068 |
+
"learning_rate": 9.28945512820513e-06,
|
| 157069 |
+
"loss": 0.3581,
|
| 157070 |
+
"step": 45355
|
| 157071 |
+
},
|
| 157072 |
+
{
|
| 157073 |
+
"epoch": 362.49,
|
| 157074 |
+
"learning_rate": 9.289375000000001e-06,
|
| 157075 |
+
"loss": 0.36,
|
| 157076 |
+
"step": 45360
|
| 157077 |
+
},
|
| 157078 |
+
{
|
| 157079 |
+
"epoch": 362.53,
|
| 157080 |
+
"learning_rate": 9.289294871794873e-06,
|
| 157081 |
+
"loss": 0.3602,
|
| 157082 |
+
"step": 45365
|
| 157083 |
+
},
|
| 157084 |
+
{
|
| 157085 |
+
"epoch": 362.57,
|
| 157086 |
+
"learning_rate": 9.289214743589744e-06,
|
| 157087 |
+
"loss": 0.7426,
|
| 157088 |
+
"step": 45370
|
| 157089 |
+
},
|
| 157090 |
+
{
|
| 157091 |
+
"epoch": 362.61,
|
| 157092 |
+
"learning_rate": 9.289134615384617e-06,
|
| 157093 |
+
"loss": 1.2191,
|
| 157094 |
+
"step": 45375
|
| 157095 |
+
},
|
| 157096 |
+
{
|
| 157097 |
+
"epoch": 362.65,
|
| 157098 |
+
"learning_rate": 9.289054487179487e-06,
|
| 157099 |
+
"loss": 0.4004,
|
| 157100 |
+
"step": 45380
|
| 157101 |
+
},
|
| 157102 |
+
{
|
| 157103 |
+
"epoch": 362.69,
|
| 157104 |
+
"learning_rate": 9.28897435897436e-06,
|
| 157105 |
+
"loss": 0.4166,
|
| 157106 |
+
"step": 45385
|
| 157107 |
+
},
|
| 157108 |
+
{
|
| 157109 |
+
"epoch": 362.73,
|
| 157110 |
+
"learning_rate": 9.288894230769233e-06,
|
| 157111 |
+
"loss": 0.4326,
|
| 157112 |
+
"step": 45390
|
| 157113 |
+
},
|
| 157114 |
+
{
|
| 157115 |
+
"epoch": 362.77,
|
| 157116 |
+
"learning_rate": 9.288814102564102e-06,
|
| 157117 |
+
"loss": 0.7307,
|
| 157118 |
+
"step": 45395
|
| 157119 |
+
},
|
| 157120 |
+
{
|
| 157121 |
+
"epoch": 362.81,
|
| 157122 |
+
"learning_rate": 9.288733974358975e-06,
|
| 157123 |
+
"loss": 1.0404,
|
| 157124 |
+
"step": 45400
|
| 157125 |
+
},
|
| 157126 |
+
{
|
| 157127 |
+
"epoch": 362.85,
|
| 157128 |
+
"learning_rate": 9.288653846153847e-06,
|
| 157129 |
+
"loss": 0.3228,
|
| 157130 |
+
"step": 45405
|
| 157131 |
+
},
|
| 157132 |
+
{
|
| 157133 |
+
"epoch": 362.89,
|
| 157134 |
+
"learning_rate": 9.288573717948718e-06,
|
| 157135 |
+
"loss": 0.4026,
|
| 157136 |
+
"step": 45410
|
| 157137 |
+
},
|
| 157138 |
+
{
|
| 157139 |
+
"epoch": 362.93,
|
| 157140 |
+
"learning_rate": 9.28849358974359e-06,
|
| 157141 |
+
"loss": 0.369,
|
| 157142 |
+
"step": 45415
|
| 157143 |
+
},
|
| 157144 |
+
{
|
| 157145 |
+
"epoch": 362.97,
|
| 157146 |
+
"learning_rate": 9.288413461538463e-06,
|
| 157147 |
+
"loss": 0.7,
|
| 157148 |
+
"step": 45420
|
| 157149 |
+
},
|
| 157150 |
+
{
|
| 157151 |
+
"epoch": 363.0,
|
| 157152 |
+
"eval_loss": 0.4071485996246338,
|
| 157153 |
+
"eval_runtime": 38.9922,
|
| 157154 |
+
"eval_samples_per_second": 21.645,
|
| 157155 |
+
"eval_steps_per_second": 0.692,
|
| 157156 |
+
"eval_wer": 0.213078618662748,
|
| 157157 |
+
"step": 45423
|
| 157158 |
+
},
|
| 157159 |
+
{
|
| 157160 |
+
"epoch": 366.02,
|
| 157161 |
+
"learning_rate": 9.288333333333334e-06,
|
| 157162 |
+
"loss": 0.4014,
|
| 157163 |
+
"step": 45425
|
| 157164 |
+
},
|
| 157165 |
+
{
|
| 157166 |
+
"epoch": 366.06,
|
| 157167 |
+
"learning_rate": 9.288253205128205e-06,
|
| 157168 |
+
"loss": 0.3133,
|
| 157169 |
+
"step": 45430
|
| 157170 |
+
},
|
| 157171 |
+
{
|
| 157172 |
+
"epoch": 366.1,
|
| 157173 |
+
"learning_rate": 9.288173076923078e-06,
|
| 157174 |
+
"loss": 0.3504,
|
| 157175 |
+
"step": 45435
|
| 157176 |
+
},
|
| 157177 |
+
{
|
| 157178 |
+
"epoch": 366.14,
|
| 157179 |
+
"learning_rate": 9.28809294871795e-06,
|
| 157180 |
+
"loss": 0.3995,
|
| 157181 |
+
"step": 45440
|
| 157182 |
+
},
|
| 157183 |
+
{
|
| 157184 |
+
"epoch": 366.18,
|
| 157185 |
+
"learning_rate": 9.288012820512821e-06,
|
| 157186 |
+
"loss": 0.6585,
|
| 157187 |
+
"step": 45445
|
| 157188 |
+
},
|
| 157189 |
+
{
|
| 157190 |
+
"epoch": 366.22,
|
| 157191 |
+
"learning_rate": 9.287932692307692e-06,
|
| 157192 |
+
"loss": 1.3226,
|
| 157193 |
+
"step": 45450
|
| 157194 |
+
},
|
| 157195 |
+
{
|
| 157196 |
+
"epoch": 366.26,
|
| 157197 |
+
"learning_rate": 9.287852564102565e-06,
|
| 157198 |
+
"loss": 0.6021,
|
| 157199 |
+
"step": 45455
|
| 157200 |
+
},
|
| 157201 |
+
{
|
| 157202 |
+
"epoch": 366.3,
|
| 157203 |
+
"learning_rate": 9.287772435897437e-06,
|
| 157204 |
+
"loss": 0.3472,
|
| 157205 |
+
"step": 45460
|
| 157206 |
+
},
|
| 157207 |
+
{
|
| 157208 |
+
"epoch": 366.34,
|
| 157209 |
+
"learning_rate": 9.287692307692308e-06,
|
| 157210 |
+
"loss": 0.471,
|
| 157211 |
+
"step": 45465
|
| 157212 |
+
},
|
| 157213 |
+
{
|
| 157214 |
+
"epoch": 366.38,
|
| 157215 |
+
"learning_rate": 9.28761217948718e-06,
|
| 157216 |
+
"loss": 0.7762,
|
| 157217 |
+
"step": 45470
|
| 157218 |
+
},
|
| 157219 |
+
{
|
| 157220 |
+
"epoch": 366.42,
|
| 157221 |
+
"learning_rate": 9.287532051282053e-06,
|
| 157222 |
+
"loss": 1.0868,
|
| 157223 |
+
"step": 45475
|
| 157224 |
+
},
|
| 157225 |
+
{
|
| 157226 |
+
"epoch": 366.46,
|
| 157227 |
+
"learning_rate": 9.287451923076924e-06,
|
| 157228 |
+
"loss": 0.3655,
|
| 157229 |
+
"step": 45480
|
| 157230 |
+
},
|
| 157231 |
+
{
|
| 157232 |
+
"epoch": 366.5,
|
| 157233 |
+
"learning_rate": 9.287371794871795e-06,
|
| 157234 |
+
"loss": 0.3321,
|
| 157235 |
+
"step": 45485
|
| 157236 |
+
},
|
| 157237 |
+
{
|
| 157238 |
+
"epoch": 366.54,
|
| 157239 |
+
"learning_rate": 9.287291666666668e-06,
|
| 157240 |
+
"loss": 0.5079,
|
| 157241 |
+
"step": 45490
|
| 157242 |
+
},
|
| 157243 |
+
{
|
| 157244 |
+
"epoch": 366.58,
|
| 157245 |
+
"learning_rate": 9.28721153846154e-06,
|
| 157246 |
+
"loss": 0.766,
|
| 157247 |
+
"step": 45495
|
| 157248 |
+
},
|
| 157249 |
+
{
|
| 157250 |
+
"epoch": 366.62,
|
| 157251 |
+
"learning_rate": 9.287131410256411e-06,
|
| 157252 |
+
"loss": 1.1498,
|
| 157253 |
+
"step": 45500
|
| 157254 |
+
},
|
| 157255 |
+
{
|
| 157256 |
+
"epoch": 366.66,
|
| 157257 |
+
"learning_rate": 9.287051282051282e-06,
|
| 157258 |
+
"loss": 0.3382,
|
| 157259 |
+
"step": 45505
|
| 157260 |
+
},
|
| 157261 |
+
{
|
| 157262 |
+
"epoch": 366.7,
|
| 157263 |
+
"learning_rate": 9.286971153846155e-06,
|
| 157264 |
+
"loss": 0.3356,
|
| 157265 |
+
"step": 45510
|
| 157266 |
+
},
|
| 157267 |
+
{
|
| 157268 |
+
"epoch": 366.74,
|
| 157269 |
+
"learning_rate": 9.286891025641027e-06,
|
| 157270 |
+
"loss": 0.3995,
|
| 157271 |
+
"step": 45515
|
| 157272 |
+
},
|
| 157273 |
+
{
|
| 157274 |
+
"epoch": 366.78,
|
| 157275 |
+
"learning_rate": 9.286810897435898e-06,
|
| 157276 |
+
"loss": 0.6504,
|
| 157277 |
+
"step": 45520
|
| 157278 |
+
},
|
| 157279 |
+
{
|
| 157280 |
+
"epoch": 366.82,
|
| 157281 |
+
"learning_rate": 9.28673076923077e-06,
|
| 157282 |
+
"loss": 1.2519,
|
| 157283 |
+
"step": 45525
|
| 157284 |
+
},
|
| 157285 |
+
{
|
| 157286 |
+
"epoch": 366.86,
|
| 157287 |
+
"learning_rate": 9.286650641025643e-06,
|
| 157288 |
+
"loss": 0.3346,
|
| 157289 |
+
"step": 45530
|
| 157290 |
+
},
|
| 157291 |
+
{
|
| 157292 |
+
"epoch": 366.9,
|
| 157293 |
+
"learning_rate": 9.286570512820512e-06,
|
| 157294 |
+
"loss": 0.4175,
|
| 157295 |
+
"step": 45535
|
| 157296 |
+
},
|
| 157297 |
+
{
|
| 157298 |
+
"epoch": 366.94,
|
| 157299 |
+
"learning_rate": 9.286490384615385e-06,
|
| 157300 |
+
"loss": 0.4086,
|
| 157301 |
+
"step": 45540
|
| 157302 |
+
},
|
| 157303 |
+
{
|
| 157304 |
+
"epoch": 366.98,
|
| 157305 |
+
"learning_rate": 9.286410256410258e-06,
|
| 157306 |
+
"loss": 0.7602,
|
| 157307 |
+
"step": 45545
|
| 157308 |
+
},
|
| 157309 |
+
{
|
| 157310 |
+
"epoch": 367.0,
|
| 157311 |
+
"eval_loss": 0.4091971218585968,
|
| 157312 |
+
"eval_runtime": 39.6231,
|
| 157313 |
+
"eval_samples_per_second": 21.301,
|
| 157314 |
+
"eval_steps_per_second": 0.681,
|
| 157315 |
+
"eval_wer": 0.20468468468468468,
|
| 157316 |
+
"step": 45547
|
| 157317 |
+
},
|
| 157318 |
+
{
|
| 157319 |
+
"epoch": 367.02,
|
| 157320 |
+
"learning_rate": 9.286330128205128e-06,
|
| 157321 |
+
"loss": 0.406,
|
| 157322 |
+
"step": 45550
|
| 157323 |
+
},
|
| 157324 |
+
{
|
| 157325 |
+
"epoch": 367.06,
|
| 157326 |
+
"learning_rate": 9.286250000000001e-06,
|
| 157327 |
+
"loss": 0.3524,
|
| 157328 |
+
"step": 45555
|
| 157329 |
+
},
|
| 157330 |
+
{
|
| 157331 |
+
"epoch": 367.1,
|
| 157332 |
+
"learning_rate": 9.286169871794872e-06,
|
| 157333 |
+
"loss": 0.3638,
|
| 157334 |
+
"step": 45560
|
| 157335 |
+
},
|
| 157336 |
+
{
|
| 157337 |
+
"epoch": 367.14,
|
| 157338 |
+
"learning_rate": 9.286089743589744e-06,
|
| 157339 |
+
"loss": 0.3811,
|
| 157340 |
+
"step": 45565
|
| 157341 |
+
},
|
| 157342 |
+
{
|
| 157343 |
+
"epoch": 367.18,
|
| 157344 |
+
"learning_rate": 9.286009615384615e-06,
|
| 157345 |
+
"loss": 0.8961,
|
| 157346 |
+
"step": 45570
|
| 157347 |
+
},
|
| 157348 |
+
{
|
| 157349 |
+
"epoch": 367.22,
|
| 157350 |
+
"learning_rate": 9.285929487179488e-06,
|
| 157351 |
+
"loss": 1.0123,
|
| 157352 |
+
"step": 45575
|
| 157353 |
+
},
|
| 157354 |
+
{
|
| 157355 |
+
"epoch": 367.27,
|
| 157356 |
+
"learning_rate": 9.28584935897436e-06,
|
| 157357 |
+
"loss": 0.3578,
|
| 157358 |
+
"step": 45580
|
| 157359 |
+
},
|
| 157360 |
+
{
|
| 157361 |
+
"epoch": 367.31,
|
| 157362 |
+
"learning_rate": 9.285769230769231e-06,
|
| 157363 |
+
"loss": 0.3532,
|
| 157364 |
+
"step": 45585
|
| 157365 |
+
},
|
| 157366 |
+
{
|
| 157367 |
+
"epoch": 367.35,
|
| 157368 |
+
"learning_rate": 9.285689102564104e-06,
|
| 157369 |
+
"loss": 0.4196,
|
| 157370 |
+
"step": 45590
|
| 157371 |
+
},
|
| 157372 |
+
{
|
| 157373 |
+
"epoch": 367.39,
|
| 157374 |
+
"learning_rate": 9.285608974358975e-06,
|
| 157375 |
+
"loss": 0.8072,
|
| 157376 |
+
"step": 45595
|
| 157377 |
+
},
|
| 157378 |
+
{
|
| 157379 |
+
"epoch": 367.43,
|
| 157380 |
+
"learning_rate": 9.285528846153847e-06,
|
| 157381 |
+
"loss": 0.8262,
|
| 157382 |
+
"step": 45600
|
| 157383 |
+
},
|
| 157384 |
+
{
|
| 157385 |
+
"epoch": 367.47,
|
| 157386 |
+
"learning_rate": 9.285448717948718e-06,
|
| 157387 |
+
"loss": 0.3115,
|
| 157388 |
+
"step": 45605
|
| 157389 |
+
},
|
| 157390 |
+
{
|
| 157391 |
+
"epoch": 367.51,
|
| 157392 |
+
"learning_rate": 9.285368589743591e-06,
|
| 157393 |
+
"loss": 0.3387,
|
| 157394 |
+
"step": 45610
|
| 157395 |
+
},
|
| 157396 |
+
{
|
| 157397 |
+
"epoch": 367.55,
|
| 157398 |
+
"learning_rate": 9.285288461538462e-06,
|
| 157399 |
+
"loss": 0.4252,
|
| 157400 |
+
"step": 45615
|
| 157401 |
+
},
|
| 157402 |
+
{
|
| 157403 |
+
"epoch": 367.59,
|
| 157404 |
+
"learning_rate": 9.285208333333334e-06,
|
| 157405 |
+
"loss": 0.8838,
|
| 157406 |
+
"step": 45620
|
| 157407 |
+
},
|
| 157408 |
+
{
|
| 157409 |
+
"epoch": 367.63,
|
| 157410 |
+
"learning_rate": 9.285128205128205e-06,
|
| 157411 |
+
"loss": 0.8291,
|
| 157412 |
+
"step": 45625
|
| 157413 |
+
},
|
| 157414 |
+
{
|
| 157415 |
+
"epoch": 367.67,
|
| 157416 |
+
"learning_rate": 9.285048076923078e-06,
|
| 157417 |
+
"loss": 0.3431,
|
| 157418 |
+
"step": 45630
|
| 157419 |
+
},
|
| 157420 |
+
{
|
| 157421 |
+
"epoch": 367.71,
|
| 157422 |
+
"learning_rate": 9.28496794871795e-06,
|
| 157423 |
+
"loss": 0.3532,
|
| 157424 |
+
"step": 45635
|
| 157425 |
+
},
|
| 157426 |
+
{
|
| 157427 |
+
"epoch": 367.75,
|
| 157428 |
+
"learning_rate": 9.284887820512821e-06,
|
| 157429 |
+
"loss": 0.4254,
|
| 157430 |
+
"step": 45640
|
| 157431 |
+
},
|
| 157432 |
+
{
|
| 157433 |
+
"epoch": 367.79,
|
| 157434 |
+
"learning_rate": 9.284807692307694e-06,
|
| 157435 |
+
"loss": 0.8985,
|
| 157436 |
+
"step": 45645
|
| 157437 |
+
},
|
| 157438 |
+
{
|
| 157439 |
+
"epoch": 367.83,
|
| 157440 |
+
"learning_rate": 9.284727564102565e-06,
|
| 157441 |
+
"loss": 0.9471,
|
| 157442 |
+
"step": 45650
|
| 157443 |
+
},
|
| 157444 |
+
{
|
| 157445 |
+
"epoch": 367.87,
|
| 157446 |
+
"learning_rate": 9.284647435897437e-06,
|
| 157447 |
+
"loss": 0.3077,
|
| 157448 |
+
"step": 45655
|
| 157449 |
+
},
|
| 157450 |
+
{
|
| 157451 |
+
"epoch": 367.91,
|
| 157452 |
+
"learning_rate": 9.284567307692308e-06,
|
| 157453 |
+
"loss": 0.4046,
|
| 157454 |
+
"step": 45660
|
| 157455 |
+
},
|
| 157456 |
+
{
|
| 157457 |
+
"epoch": 367.95,
|
| 157458 |
+
"learning_rate": 9.284487179487181e-06,
|
| 157459 |
+
"loss": 0.4621,
|
| 157460 |
+
"step": 45665
|
| 157461 |
+
},
|
| 157462 |
+
{
|
| 157463 |
+
"epoch": 367.99,
|
| 157464 |
+
"learning_rate": 9.284407051282052e-06,
|
| 157465 |
+
"loss": 0.8468,
|
| 157466 |
+
"step": 45670
|
| 157467 |
+
},
|
| 157468 |
+
{
|
| 157469 |
+
"epoch": 368.0,
|
| 157470 |
+
"eval_loss": 0.43238022923469543,
|
| 157471 |
+
"eval_runtime": 39.5066,
|
| 157472 |
+
"eval_samples_per_second": 21.364,
|
| 157473 |
+
"eval_steps_per_second": 0.683,
|
| 157474 |
+
"eval_wer": 0.20307802433786687,
|
| 157475 |
+
"step": 45671
|
| 157476 |
+
},
|
| 157477 |
+
{
|
| 157478 |
+
"epoch": 365.03,
|
| 157479 |
+
"learning_rate": 9.284326923076924e-06,
|
| 157480 |
+
"loss": 0.4539,
|
| 157481 |
+
"step": 45675
|
| 157482 |
+
},
|
| 157483 |
+
{
|
| 157484 |
+
"epoch": 365.07,
|
| 157485 |
+
"learning_rate": 9.284246794871795e-06,
|
| 157486 |
+
"loss": 0.3296,
|
| 157487 |
+
"step": 45680
|
| 157488 |
+
},
|
| 157489 |
+
{
|
| 157490 |
+
"epoch": 365.11,
|
| 157491 |
+
"learning_rate": 9.284166666666668e-06,
|
| 157492 |
+
"loss": 0.3608,
|
| 157493 |
+
"step": 45685
|
| 157494 |
+
},
|
| 157495 |
+
{
|
| 157496 |
+
"epoch": 365.15,
|
| 157497 |
+
"learning_rate": 9.28408653846154e-06,
|
| 157498 |
+
"loss": 0.4521,
|
| 157499 |
+
"step": 45690
|
| 157500 |
+
},
|
| 157501 |
+
{
|
| 157502 |
+
"epoch": 365.19,
|
| 157503 |
+
"learning_rate": 9.284006410256411e-06,
|
| 157504 |
+
"loss": 1.0097,
|
| 157505 |
+
"step": 45695
|
| 157506 |
+
},
|
| 157507 |
+
{
|
| 157508 |
+
"epoch": 365.23,
|
| 157509 |
+
"learning_rate": 9.283926282051284e-06,
|
| 157510 |
+
"loss": 0.7506,
|
| 157511 |
+
"step": 45700
|
| 157512 |
+
},
|
| 157513 |
+
{
|
| 157514 |
+
"epoch": 365.27,
|
| 157515 |
+
"learning_rate": 9.283846153846154e-06,
|
| 157516 |
+
"loss": 0.3328,
|
| 157517 |
+
"step": 45705
|
| 157518 |
+
},
|
| 157519 |
+
{
|
| 157520 |
+
"epoch": 365.31,
|
| 157521 |
+
"learning_rate": 9.283766025641027e-06,
|
| 157522 |
+
"loss": 0.37,
|
| 157523 |
+
"step": 45710
|
| 157524 |
+
},
|
| 157525 |
+
{
|
| 157526 |
+
"epoch": 365.35,
|
| 157527 |
+
"learning_rate": 9.283685897435898e-06,
|
| 157528 |
+
"loss": 0.5818,
|
| 157529 |
+
"step": 45715
|
| 157530 |
+
},
|
| 157531 |
+
{
|
| 157532 |
+
"epoch": 365.39,
|
| 157533 |
+
"learning_rate": 9.28360576923077e-06,
|
| 157534 |
+
"loss": 0.9629,
|
| 157535 |
+
"step": 45720
|
| 157536 |
+
},
|
| 157537 |
+
{
|
| 157538 |
+
"epoch": 365.43,
|
| 157539 |
+
"learning_rate": 9.28352564102564e-06,
|
| 157540 |
+
"loss": 0.7497,
|
| 157541 |
+
"step": 45725
|
| 157542 |
+
},
|
| 157543 |
+
{
|
| 157544 |
+
"epoch": 365.47,
|
| 157545 |
+
"learning_rate": 9.283445512820514e-06,
|
| 157546 |
+
"loss": 0.3164,
|
| 157547 |
+
"step": 45730
|
| 157548 |
+
},
|
| 157549 |
+
{
|
| 157550 |
+
"epoch": 365.51,
|
| 157551 |
+
"learning_rate": 9.283365384615385e-06,
|
| 157552 |
+
"loss": 0.4934,
|
| 157553 |
+
"step": 45735
|
| 157554 |
+
},
|
| 157555 |
+
{
|
| 157556 |
+
"epoch": 365.55,
|
| 157557 |
+
"learning_rate": 9.283285256410257e-06,
|
| 157558 |
+
"loss": 0.5097,
|
| 157559 |
+
"step": 45740
|
| 157560 |
+
},
|
| 157561 |
+
{
|
| 157562 |
+
"epoch": 365.59,
|
| 157563 |
+
"learning_rate": 9.28320512820513e-06,
|
| 157564 |
+
"loss": 1.0495,
|
| 157565 |
+
"step": 45745
|
| 157566 |
+
},
|
| 157567 |
+
{
|
| 157568 |
+
"epoch": 365.63,
|
| 157569 |
+
"learning_rate": 9.283125000000001e-06,
|
| 157570 |
+
"loss": 0.804,
|
| 157571 |
+
"step": 45750
|
| 157572 |
+
},
|
| 157573 |
+
{
|
| 157574 |
+
"epoch": 365.67,
|
| 157575 |
+
"learning_rate": 9.283044871794872e-06,
|
| 157576 |
+
"loss": 0.3411,
|
| 157577 |
+
"step": 45755
|
| 157578 |
+
},
|
| 157579 |
+
{
|
| 157580 |
+
"epoch": 365.71,
|
| 157581 |
+
"learning_rate": 9.282964743589744e-06,
|
| 157582 |
+
"loss": 0.4465,
|
| 157583 |
+
"step": 45760
|
| 157584 |
+
},
|
| 157585 |
+
{
|
| 157586 |
+
"epoch": 365.75,
|
| 157587 |
+
"learning_rate": 9.282884615384617e-06,
|
| 157588 |
+
"loss": 0.516,
|
| 157589 |
+
"step": 45765
|
| 157590 |
+
},
|
| 157591 |
+
{
|
| 157592 |
+
"epoch": 365.79,
|
| 157593 |
+
"learning_rate": 9.282804487179488e-06,
|
| 157594 |
+
"loss": 1.0235,
|
| 157595 |
+
"step": 45770
|
| 157596 |
+
},
|
| 157597 |
+
{
|
| 157598 |
+
"epoch": 365.83,
|
| 157599 |
+
"learning_rate": 9.28272435897436e-06,
|
| 157600 |
+
"loss": 0.8737,
|
| 157601 |
+
"step": 45775
|
| 157602 |
+
},
|
| 157603 |
+
{
|
| 157604 |
+
"epoch": 365.87,
|
| 157605 |
+
"learning_rate": 9.282644230769231e-06,
|
| 157606 |
+
"loss": 0.352,
|
| 157607 |
+
"step": 45780
|
| 157608 |
+
},
|
| 157609 |
+
{
|
| 157610 |
+
"epoch": 365.91,
|
| 157611 |
+
"learning_rate": 9.282564102564104e-06,
|
| 157612 |
+
"loss": 0.3786,
|
| 157613 |
+
"step": 45785
|
| 157614 |
+
},
|
| 157615 |
+
{
|
| 157616 |
+
"epoch": 365.95,
|
| 157617 |
+
"learning_rate": 9.282483974358975e-06,
|
| 157618 |
+
"loss": 0.4306,
|
| 157619 |
+
"step": 45790
|
| 157620 |
+
},
|
| 157621 |
+
{
|
| 157622 |
+
"epoch": 365.99,
|
| 157623 |
+
"learning_rate": 9.282403846153847e-06,
|
| 157624 |
+
"loss": 1.0093,
|
| 157625 |
+
"step": 45795
|
| 157626 |
+
},
|
| 157627 |
+
{
|
| 157628 |
+
"epoch": 366.0,
|
| 157629 |
+
"eval_loss": 0.5088355541229248,
|
| 157630 |
+
"eval_runtime": 38.1999,
|
| 157631 |
+
"eval_samples_per_second": 22.094,
|
| 157632 |
+
"eval_steps_per_second": 0.707,
|
| 157633 |
+
"eval_wer": 0.21347227391142537,
|
| 157634 |
+
"step": 45796
|
| 157635 |
}
|
| 157636 |
],
|
| 157637 |
+
"max_steps": 625000,
|
| 157638 |
"num_train_epochs": 5000,
|
| 157639 |
+
"total_flos": 1.2885530878567088e+20,
|
| 157640 |
"trial_name": null,
|
| 157641 |
"trial_params": null
|
| 157642 |
}
|
model-bin/finetune/base/{checkpoint-45173 → checkpoint-45796}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629716349.3372576/events.out.tfevents.1629716349.74272264b15c.932.113
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9fdfcf5ae9977582cb7d229e1a93a827bb1e46537e935c68c71c7a765d13922d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629716986.7392747/events.out.tfevents.1629716986.74272264b15c.932.115
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7e4a97bfe9be88c07acda451139a47eecb261c5e79d49acb35fe5d69e8b21213
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629717626.7822158/events.out.tfevents.1629717626.74272264b15c.932.117
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebfb387e03c60efba9c159abdb6c572b0645523354916240abc1e0a2c5f6615a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629718254.5594208/events.out.tfevents.1629718254.74272264b15c.932.119
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f45faf860c1c40e46eeb7ee328a1f5abb475208e18391bdd6d6a334643fcceb1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629718890.2393265/events.out.tfevents.1629718890.74272264b15c.932.121
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d19ea4700f1e680d54d613141a92a3a29aa9d6c422917a7a71bb63b5f9c3464f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629716349.74272264b15c.932.112
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c00a0b0aec25277d40f15bc33a14fd33525ae7bb8b106d9bb34f9116859cd10
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629716986.74272264b15c.932.114
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e79a408c1035141ebf6fa59d4f9c89568c9647f47a1d08fead70747547a9776
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629717626.74272264b15c.932.116
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2da48c63ea22386d7444e485477b0622434a7b46ac92484e83854532dfa704bf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629718254.74272264b15c.932.118
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:414249b55d36f0b0089686673da61791ad1b2dd6b996d80c6b17ec64192aa477
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629718890.74272264b15c.932.120
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:957eab446f8f452ff6b415b29ec3e7b8b0c56d4396c35eda153192914cbb8391
|
| 3 |
+
size 8622
|