"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630128637.686285/events.out.tfevents.1630128637.86bb0ddabf9b.4092.31 +3 -0
- model-bin/finetune/base/log/1630129022.8707244/events.out.tfevents.1630129022.86bb0ddabf9b.4092.33 +3 -0
- model-bin/finetune/base/log/1630129401.5085042/events.out.tfevents.1630129401.86bb0ddabf9b.4092.35 +3 -0
- model-bin/finetune/base/log/1630129786.576031/events.out.tfevents.1630129786.86bb0ddabf9b.4092.37 +3 -0
- model-bin/finetune/base/log/1630130174.479612/events.out.tfevents.1630130174.86bb0ddabf9b.4092.39 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630128637.86bb0ddabf9b.4092.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630129022.86bb0ddabf9b.4092.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630129401.86bb0ddabf9b.4092.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630129786.86bb0ddabf9b.4092.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630130174.86bb0ddabf9b.4092.38 +3 -0
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02d754e3e35b5bd4dcad1290ee437bc10dda642925d72fb3e1f864b94c718255
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/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:b0255865cadd0645b960c336eea8a46b1f0f5c05aedb2a71da5a163adb3073fb
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/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:aa7c7af68a6f1073c4850ff28a3ebe67c89df0455e85db0e2f01209cdcfbc270
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/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 sha256:0fb121336be771fd94b174473516525f5106a0131cfb91565c4039ac35f3471c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/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 sha256:0154ab06f351cde3389d68a9fd4cd15b57e588da1a39ef8fca1b9e44cd8cf2ee
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -251835,11 +251835,800 @@
|
|
| 251835 |
"eval_steps_per_second": 0.752,
|
| 251836 |
"eval_wer": 0.18405627198124266,
|
| 251837 |
"step": 119837
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251838 |
}
|
| 251839 |
],
|
| 251840 |
-
"max_steps":
|
| 251841 |
"num_train_epochs": 5000,
|
| 251842 |
-
"total_flos": 3.
|
| 251843 |
"trial_name": null,
|
| 251844 |
"trial_params": null
|
| 251845 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 970.995983935743,
|
| 5 |
+
"global_step": 120459,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 251835 |
"eval_steps_per_second": 0.752,
|
| 251836 |
"eval_wer": 0.18405627198124266,
|
| 251837 |
"step": 119837
|
| 251838 |
+
},
|
| 251839 |
+
{
|
| 251840 |
+
"epoch": 966.02,
|
| 251841 |
+
"learning_rate": 8.096266025641027e-06,
|
| 251842 |
+
"loss": 0.348,
|
| 251843 |
+
"step": 119840
|
| 251844 |
+
},
|
| 251845 |
+
{
|
| 251846 |
+
"epoch": 966.06,
|
| 251847 |
+
"learning_rate": 8.096185897435898e-06,
|
| 251848 |
+
"loss": 0.2931,
|
| 251849 |
+
"step": 119845
|
| 251850 |
+
},
|
| 251851 |
+
{
|
| 251852 |
+
"epoch": 966.1,
|
| 251853 |
+
"learning_rate": 8.09610576923077e-06,
|
| 251854 |
+
"loss": 0.2475,
|
| 251855 |
+
"step": 119850
|
| 251856 |
+
},
|
| 251857 |
+
{
|
| 251858 |
+
"epoch": 966.14,
|
| 251859 |
+
"learning_rate": 8.096025641025642e-06,
|
| 251860 |
+
"loss": 0.4117,
|
| 251861 |
+
"step": 119855
|
| 251862 |
+
},
|
| 251863 |
+
{
|
| 251864 |
+
"epoch": 966.18,
|
| 251865 |
+
"learning_rate": 8.095945512820514e-06,
|
| 251866 |
+
"loss": 0.696,
|
| 251867 |
+
"step": 119860
|
| 251868 |
+
},
|
| 251869 |
+
{
|
| 251870 |
+
"epoch": 966.22,
|
| 251871 |
+
"learning_rate": 8.095865384615385e-06,
|
| 251872 |
+
"loss": 0.8789,
|
| 251873 |
+
"step": 119865
|
| 251874 |
+
},
|
| 251875 |
+
{
|
| 251876 |
+
"epoch": 966.27,
|
| 251877 |
+
"learning_rate": 8.095785256410257e-06,
|
| 251878 |
+
"loss": 0.286,
|
| 251879 |
+
"step": 119870
|
| 251880 |
+
},
|
| 251881 |
+
{
|
| 251882 |
+
"epoch": 966.31,
|
| 251883 |
+
"learning_rate": 8.09570512820513e-06,
|
| 251884 |
+
"loss": 0.2694,
|
| 251885 |
+
"step": 119875
|
| 251886 |
+
},
|
| 251887 |
+
{
|
| 251888 |
+
"epoch": 966.35,
|
| 251889 |
+
"learning_rate": 8.095625000000001e-06,
|
| 251890 |
+
"loss": 0.3881,
|
| 251891 |
+
"step": 119880
|
| 251892 |
+
},
|
| 251893 |
+
{
|
| 251894 |
+
"epoch": 966.39,
|
| 251895 |
+
"learning_rate": 8.095544871794872e-06,
|
| 251896 |
+
"loss": 0.7294,
|
| 251897 |
+
"step": 119885
|
| 251898 |
+
},
|
| 251899 |
+
{
|
| 251900 |
+
"epoch": 966.43,
|
| 251901 |
+
"learning_rate": 8.095464743589744e-06,
|
| 251902 |
+
"loss": 0.6951,
|
| 251903 |
+
"step": 119890
|
| 251904 |
+
},
|
| 251905 |
+
{
|
| 251906 |
+
"epoch": 966.47,
|
| 251907 |
+
"learning_rate": 8.095384615384617e-06,
|
| 251908 |
+
"loss": 0.2959,
|
| 251909 |
+
"step": 119895
|
| 251910 |
+
},
|
| 251911 |
+
{
|
| 251912 |
+
"epoch": 966.51,
|
| 251913 |
+
"learning_rate": 8.095304487179488e-06,
|
| 251914 |
+
"loss": 0.2939,
|
| 251915 |
+
"step": 119900
|
| 251916 |
+
},
|
| 251917 |
+
{
|
| 251918 |
+
"epoch": 966.55,
|
| 251919 |
+
"learning_rate": 8.09522435897436e-06,
|
| 251920 |
+
"loss": 0.3478,
|
| 251921 |
+
"step": 119905
|
| 251922 |
+
},
|
| 251923 |
+
{
|
| 251924 |
+
"epoch": 966.59,
|
| 251925 |
+
"learning_rate": 8.095144230769232e-06,
|
| 251926 |
+
"loss": 0.6793,
|
| 251927 |
+
"step": 119910
|
| 251928 |
+
},
|
| 251929 |
+
{
|
| 251930 |
+
"epoch": 966.63,
|
| 251931 |
+
"learning_rate": 8.095064102564104e-06,
|
| 251932 |
+
"loss": 0.8692,
|
| 251933 |
+
"step": 119915
|
| 251934 |
+
},
|
| 251935 |
+
{
|
| 251936 |
+
"epoch": 966.67,
|
| 251937 |
+
"learning_rate": 8.094983974358975e-06,
|
| 251938 |
+
"loss": 0.2593,
|
| 251939 |
+
"step": 119920
|
| 251940 |
+
},
|
| 251941 |
+
{
|
| 251942 |
+
"epoch": 966.71,
|
| 251943 |
+
"learning_rate": 8.094903846153847e-06,
|
| 251944 |
+
"loss": 0.2772,
|
| 251945 |
+
"step": 119925
|
| 251946 |
+
},
|
| 251947 |
+
{
|
| 251948 |
+
"epoch": 966.75,
|
| 251949 |
+
"learning_rate": 8.09482371794872e-06,
|
| 251950 |
+
"loss": 0.4432,
|
| 251951 |
+
"step": 119930
|
| 251952 |
+
},
|
| 251953 |
+
{
|
| 251954 |
+
"epoch": 966.79,
|
| 251955 |
+
"learning_rate": 8.09474358974359e-06,
|
| 251956 |
+
"loss": 0.7561,
|
| 251957 |
+
"step": 119935
|
| 251958 |
+
},
|
| 251959 |
+
{
|
| 251960 |
+
"epoch": 966.83,
|
| 251961 |
+
"learning_rate": 8.094663461538462e-06,
|
| 251962 |
+
"loss": 0.885,
|
| 251963 |
+
"step": 119940
|
| 251964 |
+
},
|
| 251965 |
+
{
|
| 251966 |
+
"epoch": 966.87,
|
| 251967 |
+
"learning_rate": 8.094583333333334e-06,
|
| 251968 |
+
"loss": 0.2437,
|
| 251969 |
+
"step": 119945
|
| 251970 |
+
},
|
| 251971 |
+
{
|
| 251972 |
+
"epoch": 966.91,
|
| 251973 |
+
"learning_rate": 8.094503205128205e-06,
|
| 251974 |
+
"loss": 0.3621,
|
| 251975 |
+
"step": 119950
|
| 251976 |
+
},
|
| 251977 |
+
{
|
| 251978 |
+
"epoch": 966.95,
|
| 251979 |
+
"learning_rate": 8.094423076923078e-06,
|
| 251980 |
+
"loss": 0.4108,
|
| 251981 |
+
"step": 119955
|
| 251982 |
+
},
|
| 251983 |
+
{
|
| 251984 |
+
"epoch": 966.99,
|
| 251985 |
+
"learning_rate": 8.09434294871795e-06,
|
| 251986 |
+
"loss": 0.8868,
|
| 251987 |
+
"step": 119960
|
| 251988 |
+
},
|
| 251989 |
+
{
|
| 251990 |
+
"epoch": 967.0,
|
| 251991 |
+
"eval_loss": 0.39156973361968994,
|
| 251992 |
+
"eval_runtime": 35.5899,
|
| 251993 |
+
"eval_samples_per_second": 23.63,
|
| 251994 |
+
"eval_steps_per_second": 0.759,
|
| 251995 |
+
"eval_wer": 0.17923625928923553,
|
| 251996 |
+
"step": 119961
|
| 251997 |
+
},
|
| 251998 |
+
{
|
| 251999 |
+
"epoch": 959.03,
|
| 252000 |
+
"learning_rate": 8.09426282051282e-06,
|
| 252001 |
+
"loss": 0.2988,
|
| 252002 |
+
"step": 119965
|
| 252003 |
+
},
|
| 252004 |
+
{
|
| 252005 |
+
"epoch": 959.07,
|
| 252006 |
+
"learning_rate": 8.094182692307692e-06,
|
| 252007 |
+
"loss": 0.2924,
|
| 252008 |
+
"step": 119970
|
| 252009 |
+
},
|
| 252010 |
+
{
|
| 252011 |
+
"epoch": 959.11,
|
| 252012 |
+
"learning_rate": 8.094102564102565e-06,
|
| 252013 |
+
"loss": 0.2814,
|
| 252014 |
+
"step": 119975
|
| 252015 |
+
},
|
| 252016 |
+
{
|
| 252017 |
+
"epoch": 959.15,
|
| 252018 |
+
"learning_rate": 8.094022435897437e-06,
|
| 252019 |
+
"loss": 0.4033,
|
| 252020 |
+
"step": 119980
|
| 252021 |
+
},
|
| 252022 |
+
{
|
| 252023 |
+
"epoch": 959.19,
|
| 252024 |
+
"learning_rate": 8.093942307692308e-06,
|
| 252025 |
+
"loss": 0.977,
|
| 252026 |
+
"step": 119985
|
| 252027 |
+
},
|
| 252028 |
+
{
|
| 252029 |
+
"epoch": 959.23,
|
| 252030 |
+
"learning_rate": 8.09386217948718e-06,
|
| 252031 |
+
"loss": 0.643,
|
| 252032 |
+
"step": 119990
|
| 252033 |
+
},
|
| 252034 |
+
{
|
| 252035 |
+
"epoch": 959.27,
|
| 252036 |
+
"learning_rate": 8.093782051282052e-06,
|
| 252037 |
+
"loss": 0.3368,
|
| 252038 |
+
"step": 119995
|
| 252039 |
+
},
|
| 252040 |
+
{
|
| 252041 |
+
"epoch": 959.31,
|
| 252042 |
+
"learning_rate": 8.093701923076924e-06,
|
| 252043 |
+
"loss": 0.3089,
|
| 252044 |
+
"step": 120000
|
| 252045 |
+
},
|
| 252046 |
+
{
|
| 252047 |
+
"epoch": 959.35,
|
| 252048 |
+
"learning_rate": 8.093621794871795e-06,
|
| 252049 |
+
"loss": 0.4179,
|
| 252050 |
+
"step": 120005
|
| 252051 |
+
},
|
| 252052 |
+
{
|
| 252053 |
+
"epoch": 959.39,
|
| 252054 |
+
"learning_rate": 8.093541666666668e-06,
|
| 252055 |
+
"loss": 0.9183,
|
| 252056 |
+
"step": 120010
|
| 252057 |
+
},
|
| 252058 |
+
{
|
| 252059 |
+
"epoch": 959.43,
|
| 252060 |
+
"learning_rate": 8.09346153846154e-06,
|
| 252061 |
+
"loss": 0.7414,
|
| 252062 |
+
"step": 120015
|
| 252063 |
+
},
|
| 252064 |
+
{
|
| 252065 |
+
"epoch": 959.47,
|
| 252066 |
+
"learning_rate": 8.09338141025641e-06,
|
| 252067 |
+
"loss": 0.2578,
|
| 252068 |
+
"step": 120020
|
| 252069 |
+
},
|
| 252070 |
+
{
|
| 252071 |
+
"epoch": 959.51,
|
| 252072 |
+
"learning_rate": 8.093301282051282e-06,
|
| 252073 |
+
"loss": 0.281,
|
| 252074 |
+
"step": 120025
|
| 252075 |
+
},
|
| 252076 |
+
{
|
| 252077 |
+
"epoch": 959.55,
|
| 252078 |
+
"learning_rate": 8.093221153846155e-06,
|
| 252079 |
+
"loss": 0.4018,
|
| 252080 |
+
"step": 120030
|
| 252081 |
+
},
|
| 252082 |
+
{
|
| 252083 |
+
"epoch": 959.59,
|
| 252084 |
+
"learning_rate": 8.093141025641027e-06,
|
| 252085 |
+
"loss": 0.9263,
|
| 252086 |
+
"step": 120035
|
| 252087 |
+
},
|
| 252088 |
+
{
|
| 252089 |
+
"epoch": 959.63,
|
| 252090 |
+
"learning_rate": 8.093060897435898e-06,
|
| 252091 |
+
"loss": 0.6744,
|
| 252092 |
+
"step": 120040
|
| 252093 |
+
},
|
| 252094 |
+
{
|
| 252095 |
+
"epoch": 959.67,
|
| 252096 |
+
"learning_rate": 8.09298076923077e-06,
|
| 252097 |
+
"loss": 0.2588,
|
| 252098 |
+
"step": 120045
|
| 252099 |
+
},
|
| 252100 |
+
{
|
| 252101 |
+
"epoch": 959.71,
|
| 252102 |
+
"learning_rate": 8.092900641025642e-06,
|
| 252103 |
+
"loss": 0.2629,
|
| 252104 |
+
"step": 120050
|
| 252105 |
+
},
|
| 252106 |
+
{
|
| 252107 |
+
"epoch": 959.75,
|
| 252108 |
+
"learning_rate": 8.092820512820514e-06,
|
| 252109 |
+
"loss": 0.3581,
|
| 252110 |
+
"step": 120055
|
| 252111 |
+
},
|
| 252112 |
+
{
|
| 252113 |
+
"epoch": 959.79,
|
| 252114 |
+
"learning_rate": 8.092740384615385e-06,
|
| 252115 |
+
"loss": 0.7867,
|
| 252116 |
+
"step": 120060
|
| 252117 |
+
},
|
| 252118 |
+
{
|
| 252119 |
+
"epoch": 959.83,
|
| 252120 |
+
"learning_rate": 8.092660256410258e-06,
|
| 252121 |
+
"loss": 0.6652,
|
| 252122 |
+
"step": 120065
|
| 252123 |
+
},
|
| 252124 |
+
{
|
| 252125 |
+
"epoch": 959.87,
|
| 252126 |
+
"learning_rate": 8.092580128205128e-06,
|
| 252127 |
+
"loss": 0.2852,
|
| 252128 |
+
"step": 120070
|
| 252129 |
+
},
|
| 252130 |
+
{
|
| 252131 |
+
"epoch": 959.91,
|
| 252132 |
+
"learning_rate": 8.092500000000001e-06,
|
| 252133 |
+
"loss": 0.3501,
|
| 252134 |
+
"step": 120075
|
| 252135 |
+
},
|
| 252136 |
+
{
|
| 252137 |
+
"epoch": 959.95,
|
| 252138 |
+
"learning_rate": 8.092419871794872e-06,
|
| 252139 |
+
"loss": 0.3466,
|
| 252140 |
+
"step": 120080
|
| 252141 |
+
},
|
| 252142 |
+
{
|
| 252143 |
+
"epoch": 959.99,
|
| 252144 |
+
"learning_rate": 8.092339743589744e-06,
|
| 252145 |
+
"loss": 0.9262,
|
| 252146 |
+
"step": 120085
|
| 252147 |
+
},
|
| 252148 |
+
{
|
| 252149 |
+
"epoch": 960.0,
|
| 252150 |
+
"eval_loss": 0.40789806842803955,
|
| 252151 |
+
"eval_runtime": 35.1638,
|
| 252152 |
+
"eval_samples_per_second": 23.917,
|
| 252153 |
+
"eval_steps_per_second": 0.768,
|
| 252154 |
+
"eval_wer": 0.18865858998686325,
|
| 252155 |
+
"step": 120086
|
| 252156 |
+
},
|
| 252157 |
+
{
|
| 252158 |
+
"epoch": 960.03,
|
| 252159 |
+
"learning_rate": 8.092259615384615e-06,
|
| 252160 |
+
"loss": 0.2871,
|
| 252161 |
+
"step": 120090
|
| 252162 |
+
},
|
| 252163 |
+
{
|
| 252164 |
+
"epoch": 960.07,
|
| 252165 |
+
"learning_rate": 8.092179487179488e-06,
|
| 252166 |
+
"loss": 0.2907,
|
| 252167 |
+
"step": 120095
|
| 252168 |
+
},
|
| 252169 |
+
{
|
| 252170 |
+
"epoch": 960.11,
|
| 252171 |
+
"learning_rate": 8.09209935897436e-06,
|
| 252172 |
+
"loss": 0.3301,
|
| 252173 |
+
"step": 120100
|
| 252174 |
+
},
|
| 252175 |
+
{
|
| 252176 |
+
"epoch": 960.15,
|
| 252177 |
+
"learning_rate": 8.09201923076923e-06,
|
| 252178 |
+
"loss": 0.4019,
|
| 252179 |
+
"step": 120105
|
| 252180 |
+
},
|
| 252181 |
+
{
|
| 252182 |
+
"epoch": 960.19,
|
| 252183 |
+
"learning_rate": 8.091939102564104e-06,
|
| 252184 |
+
"loss": 0.966,
|
| 252185 |
+
"step": 120110
|
| 252186 |
+
},
|
| 252187 |
+
{
|
| 252188 |
+
"epoch": 960.23,
|
| 252189 |
+
"learning_rate": 8.091858974358975e-06,
|
| 252190 |
+
"loss": 0.6316,
|
| 252191 |
+
"step": 120115
|
| 252192 |
+
},
|
| 252193 |
+
{
|
| 252194 |
+
"epoch": 960.27,
|
| 252195 |
+
"learning_rate": 8.091778846153846e-06,
|
| 252196 |
+
"loss": 0.2679,
|
| 252197 |
+
"step": 120120
|
| 252198 |
+
},
|
| 252199 |
+
{
|
| 252200 |
+
"epoch": 960.31,
|
| 252201 |
+
"learning_rate": 8.091698717948718e-06,
|
| 252202 |
+
"loss": 0.2956,
|
| 252203 |
+
"step": 120125
|
| 252204 |
+
},
|
| 252205 |
+
{
|
| 252206 |
+
"epoch": 960.35,
|
| 252207 |
+
"learning_rate": 8.091618589743591e-06,
|
| 252208 |
+
"loss": 0.3808,
|
| 252209 |
+
"step": 120130
|
| 252210 |
+
},
|
| 252211 |
+
{
|
| 252212 |
+
"epoch": 960.39,
|
| 252213 |
+
"learning_rate": 8.091538461538462e-06,
|
| 252214 |
+
"loss": 0.9202,
|
| 252215 |
+
"step": 120135
|
| 252216 |
+
},
|
| 252217 |
+
{
|
| 252218 |
+
"epoch": 960.43,
|
| 252219 |
+
"learning_rate": 8.091458333333334e-06,
|
| 252220 |
+
"loss": 0.6906,
|
| 252221 |
+
"step": 120140
|
| 252222 |
+
},
|
| 252223 |
+
{
|
| 252224 |
+
"epoch": 960.47,
|
| 252225 |
+
"learning_rate": 8.091378205128205e-06,
|
| 252226 |
+
"loss": 0.2824,
|
| 252227 |
+
"step": 120145
|
| 252228 |
+
},
|
| 252229 |
+
{
|
| 252230 |
+
"epoch": 960.51,
|
| 252231 |
+
"learning_rate": 8.091298076923078e-06,
|
| 252232 |
+
"loss": 0.3033,
|
| 252233 |
+
"step": 120150
|
| 252234 |
+
},
|
| 252235 |
+
{
|
| 252236 |
+
"epoch": 960.55,
|
| 252237 |
+
"learning_rate": 8.09121794871795e-06,
|
| 252238 |
+
"loss": 0.4007,
|
| 252239 |
+
"step": 120155
|
| 252240 |
+
},
|
| 252241 |
+
{
|
| 252242 |
+
"epoch": 960.59,
|
| 252243 |
+
"learning_rate": 8.09113782051282e-06,
|
| 252244 |
+
"loss": 0.9142,
|
| 252245 |
+
"step": 120160
|
| 252246 |
+
},
|
| 252247 |
+
{
|
| 252248 |
+
"epoch": 960.63,
|
| 252249 |
+
"learning_rate": 8.091057692307694e-06,
|
| 252250 |
+
"loss": 0.6788,
|
| 252251 |
+
"step": 120165
|
| 252252 |
+
},
|
| 252253 |
+
{
|
| 252254 |
+
"epoch": 960.67,
|
| 252255 |
+
"learning_rate": 8.090977564102565e-06,
|
| 252256 |
+
"loss": 0.2561,
|
| 252257 |
+
"step": 120170
|
| 252258 |
+
},
|
| 252259 |
+
{
|
| 252260 |
+
"epoch": 960.71,
|
| 252261 |
+
"learning_rate": 8.090897435897437e-06,
|
| 252262 |
+
"loss": 0.3775,
|
| 252263 |
+
"step": 120175
|
| 252264 |
+
},
|
| 252265 |
+
{
|
| 252266 |
+
"epoch": 960.75,
|
| 252267 |
+
"learning_rate": 8.090817307692308e-06,
|
| 252268 |
+
"loss": 0.3481,
|
| 252269 |
+
"step": 120180
|
| 252270 |
+
},
|
| 252271 |
+
{
|
| 252272 |
+
"epoch": 960.79,
|
| 252273 |
+
"learning_rate": 8.090737179487181e-06,
|
| 252274 |
+
"loss": 0.8537,
|
| 252275 |
+
"step": 120185
|
| 252276 |
+
},
|
| 252277 |
+
{
|
| 252278 |
+
"epoch": 960.83,
|
| 252279 |
+
"learning_rate": 8.090657051282052e-06,
|
| 252280 |
+
"loss": 0.685,
|
| 252281 |
+
"step": 120190
|
| 252282 |
+
},
|
| 252283 |
+
{
|
| 252284 |
+
"epoch": 960.87,
|
| 252285 |
+
"learning_rate": 8.090576923076924e-06,
|
| 252286 |
+
"loss": 0.3219,
|
| 252287 |
+
"step": 120195
|
| 252288 |
+
},
|
| 252289 |
+
{
|
| 252290 |
+
"epoch": 960.91,
|
| 252291 |
+
"learning_rate": 8.090496794871797e-06,
|
| 252292 |
+
"loss": 0.3084,
|
| 252293 |
+
"step": 120200
|
| 252294 |
+
},
|
| 252295 |
+
{
|
| 252296 |
+
"epoch": 960.95,
|
| 252297 |
+
"learning_rate": 8.090416666666668e-06,
|
| 252298 |
+
"loss": 0.3571,
|
| 252299 |
+
"step": 120205
|
| 252300 |
+
},
|
| 252301 |
+
{
|
| 252302 |
+
"epoch": 960.99,
|
| 252303 |
+
"learning_rate": 8.09033653846154e-06,
|
| 252304 |
+
"loss": 0.957,
|
| 252305 |
+
"step": 120210
|
| 252306 |
+
},
|
| 252307 |
+
{
|
| 252308 |
+
"epoch": 961.0,
|
| 252309 |
+
"eval_loss": 0.4401938319206238,
|
| 252310 |
+
"eval_runtime": 35.346,
|
| 252311 |
+
"eval_samples_per_second": 23.822,
|
| 252312 |
+
"eval_steps_per_second": 0.764,
|
| 252313 |
+
"eval_wer": 0.1752412742181264,
|
| 252314 |
+
"step": 120211
|
| 252315 |
+
},
|
| 252316 |
+
{
|
| 252317 |
+
"epoch": 969.03,
|
| 252318 |
+
"learning_rate": 8.09025641025641e-06,
|
| 252319 |
+
"loss": 0.3214,
|
| 252320 |
+
"step": 120215
|
| 252321 |
+
},
|
| 252322 |
+
{
|
| 252323 |
+
"epoch": 969.07,
|
| 252324 |
+
"learning_rate": 8.090176282051284e-06,
|
| 252325 |
+
"loss": 0.3152,
|
| 252326 |
+
"step": 120220
|
| 252327 |
+
},
|
| 252328 |
+
{
|
| 252329 |
+
"epoch": 969.11,
|
| 252330 |
+
"learning_rate": 8.090096153846153e-06,
|
| 252331 |
+
"loss": 0.2879,
|
| 252332 |
+
"step": 120225
|
| 252333 |
+
},
|
| 252334 |
+
{
|
| 252335 |
+
"epoch": 969.15,
|
| 252336 |
+
"learning_rate": 8.090016025641027e-06,
|
| 252337 |
+
"loss": 0.5024,
|
| 252338 |
+
"step": 120230
|
| 252339 |
+
},
|
| 252340 |
+
{
|
| 252341 |
+
"epoch": 969.19,
|
| 252342 |
+
"learning_rate": 8.089935897435898e-06,
|
| 252343 |
+
"loss": 0.8369,
|
| 252344 |
+
"step": 120235
|
| 252345 |
+
},
|
| 252346 |
+
{
|
| 252347 |
+
"epoch": 969.23,
|
| 252348 |
+
"learning_rate": 8.08985576923077e-06,
|
| 252349 |
+
"loss": 0.7349,
|
| 252350 |
+
"step": 120240
|
| 252351 |
+
},
|
| 252352 |
+
{
|
| 252353 |
+
"epoch": 969.27,
|
| 252354 |
+
"learning_rate": 8.08977564102564e-06,
|
| 252355 |
+
"loss": 0.3077,
|
| 252356 |
+
"step": 120245
|
| 252357 |
+
},
|
| 252358 |
+
{
|
| 252359 |
+
"epoch": 969.31,
|
| 252360 |
+
"learning_rate": 8.089695512820514e-06,
|
| 252361 |
+
"loss": 0.2623,
|
| 252362 |
+
"step": 120250
|
| 252363 |
+
},
|
| 252364 |
+
{
|
| 252365 |
+
"epoch": 969.35,
|
| 252366 |
+
"learning_rate": 8.089615384615385e-06,
|
| 252367 |
+
"loss": 0.368,
|
| 252368 |
+
"step": 120255
|
| 252369 |
+
},
|
| 252370 |
+
{
|
| 252371 |
+
"epoch": 969.39,
|
| 252372 |
+
"learning_rate": 8.089535256410256e-06,
|
| 252373 |
+
"loss": 0.7865,
|
| 252374 |
+
"step": 120260
|
| 252375 |
+
},
|
| 252376 |
+
{
|
| 252377 |
+
"epoch": 969.43,
|
| 252378 |
+
"learning_rate": 8.08945512820513e-06,
|
| 252379 |
+
"loss": 0.5965,
|
| 252380 |
+
"step": 120265
|
| 252381 |
+
},
|
| 252382 |
+
{
|
| 252383 |
+
"epoch": 969.47,
|
| 252384 |
+
"learning_rate": 8.089375e-06,
|
| 252385 |
+
"loss": 0.2959,
|
| 252386 |
+
"step": 120270
|
| 252387 |
+
},
|
| 252388 |
+
{
|
| 252389 |
+
"epoch": 969.51,
|
| 252390 |
+
"learning_rate": 8.089294871794872e-06,
|
| 252391 |
+
"loss": 0.2584,
|
| 252392 |
+
"step": 120275
|
| 252393 |
+
},
|
| 252394 |
+
{
|
| 252395 |
+
"epoch": 969.55,
|
| 252396 |
+
"learning_rate": 8.089214743589744e-06,
|
| 252397 |
+
"loss": 0.396,
|
| 252398 |
+
"step": 120280
|
| 252399 |
+
},
|
| 252400 |
+
{
|
| 252401 |
+
"epoch": 969.59,
|
| 252402 |
+
"learning_rate": 8.089134615384617e-06,
|
| 252403 |
+
"loss": 0.9179,
|
| 252404 |
+
"step": 120285
|
| 252405 |
+
},
|
| 252406 |
+
{
|
| 252407 |
+
"epoch": 969.63,
|
| 252408 |
+
"learning_rate": 8.089054487179488e-06,
|
| 252409 |
+
"loss": 0.7094,
|
| 252410 |
+
"step": 120290
|
| 252411 |
+
},
|
| 252412 |
+
{
|
| 252413 |
+
"epoch": 969.67,
|
| 252414 |
+
"learning_rate": 8.08897435897436e-06,
|
| 252415 |
+
"loss": 0.258,
|
| 252416 |
+
"step": 120295
|
| 252417 |
+
},
|
| 252418 |
+
{
|
| 252419 |
+
"epoch": 969.71,
|
| 252420 |
+
"learning_rate": 8.088894230769232e-06,
|
| 252421 |
+
"loss": 0.3283,
|
| 252422 |
+
"step": 120300
|
| 252423 |
+
},
|
| 252424 |
+
{
|
| 252425 |
+
"epoch": 969.76,
|
| 252426 |
+
"learning_rate": 8.088814102564104e-06,
|
| 252427 |
+
"loss": 0.4285,
|
| 252428 |
+
"step": 120305
|
| 252429 |
+
},
|
| 252430 |
+
{
|
| 252431 |
+
"epoch": 969.8,
|
| 252432 |
+
"learning_rate": 8.088733974358975e-06,
|
| 252433 |
+
"loss": 0.9096,
|
| 252434 |
+
"step": 120310
|
| 252435 |
+
},
|
| 252436 |
+
{
|
| 252437 |
+
"epoch": 969.84,
|
| 252438 |
+
"learning_rate": 8.088653846153846e-06,
|
| 252439 |
+
"loss": 0.6961,
|
| 252440 |
+
"step": 120315
|
| 252441 |
+
},
|
| 252442 |
+
{
|
| 252443 |
+
"epoch": 969.88,
|
| 252444 |
+
"learning_rate": 8.08857371794872e-06,
|
| 252445 |
+
"loss": 0.3139,
|
| 252446 |
+
"step": 120320
|
| 252447 |
+
},
|
| 252448 |
+
{
|
| 252449 |
+
"epoch": 969.92,
|
| 252450 |
+
"learning_rate": 8.08849358974359e-06,
|
| 252451 |
+
"loss": 0.3107,
|
| 252452 |
+
"step": 120325
|
| 252453 |
+
},
|
| 252454 |
+
{
|
| 252455 |
+
"epoch": 969.96,
|
| 252456 |
+
"learning_rate": 8.088413461538462e-06,
|
| 252457 |
+
"loss": 0.4235,
|
| 252458 |
+
"step": 120330
|
| 252459 |
+
},
|
| 252460 |
+
{
|
| 252461 |
+
"epoch": 970.0,
|
| 252462 |
+
"learning_rate": 8.088333333333334e-06,
|
| 252463 |
+
"loss": 1.1324,
|
| 252464 |
+
"step": 120335
|
| 252465 |
+
},
|
| 252466 |
+
{
|
| 252467 |
+
"epoch": 970.0,
|
| 252468 |
+
"eval_loss": 0.3457880914211273,
|
| 252469 |
+
"eval_runtime": 35.8797,
|
| 252470 |
+
"eval_samples_per_second": 23.467,
|
| 252471 |
+
"eval_steps_per_second": 0.753,
|
| 252472 |
+
"eval_wer": 0.1774146905725853,
|
| 252473 |
+
"step": 120335
|
| 252474 |
+
},
|
| 252475 |
+
{
|
| 252476 |
+
"epoch": 970.04,
|
| 252477 |
+
"learning_rate": 8.088253205128207e-06,
|
| 252478 |
+
"loss": 0.2901,
|
| 252479 |
+
"step": 120340
|
| 252480 |
+
},
|
| 252481 |
+
{
|
| 252482 |
+
"epoch": 970.08,
|
| 252483 |
+
"learning_rate": 8.088173076923076e-06,
|
| 252484 |
+
"loss": 0.2719,
|
| 252485 |
+
"step": 120345
|
| 252486 |
+
},
|
| 252487 |
+
{
|
| 252488 |
+
"epoch": 970.12,
|
| 252489 |
+
"learning_rate": 8.08809294871795e-06,
|
| 252490 |
+
"loss": 0.348,
|
| 252491 |
+
"step": 120350
|
| 252492 |
+
},
|
| 252493 |
+
{
|
| 252494 |
+
"epoch": 970.16,
|
| 252495 |
+
"learning_rate": 8.088012820512822e-06,
|
| 252496 |
+
"loss": 0.3995,
|
| 252497 |
+
"step": 120355
|
| 252498 |
+
},
|
| 252499 |
+
{
|
| 252500 |
+
"epoch": 970.2,
|
| 252501 |
+
"learning_rate": 8.087932692307692e-06,
|
| 252502 |
+
"loss": 1.1411,
|
| 252503 |
+
"step": 120360
|
| 252504 |
+
},
|
| 252505 |
+
{
|
| 252506 |
+
"epoch": 970.24,
|
| 252507 |
+
"learning_rate": 8.087852564102565e-06,
|
| 252508 |
+
"loss": 0.3333,
|
| 252509 |
+
"step": 120365
|
| 252510 |
+
},
|
| 252511 |
+
{
|
| 252512 |
+
"epoch": 970.28,
|
| 252513 |
+
"learning_rate": 8.087772435897436e-06,
|
| 252514 |
+
"loss": 0.2983,
|
| 252515 |
+
"step": 120370
|
| 252516 |
+
},
|
| 252517 |
+
{
|
| 252518 |
+
"epoch": 970.32,
|
| 252519 |
+
"learning_rate": 8.087692307692308e-06,
|
| 252520 |
+
"loss": 0.3089,
|
| 252521 |
+
"step": 120375
|
| 252522 |
+
},
|
| 252523 |
+
{
|
| 252524 |
+
"epoch": 970.36,
|
| 252525 |
+
"learning_rate": 8.087612179487179e-06,
|
| 252526 |
+
"loss": 0.4078,
|
| 252527 |
+
"step": 120380
|
| 252528 |
+
},
|
| 252529 |
+
{
|
| 252530 |
+
"epoch": 970.4,
|
| 252531 |
+
"learning_rate": 8.087532051282052e-06,
|
| 252532 |
+
"loss": 1.2498,
|
| 252533 |
+
"step": 120385
|
| 252534 |
+
},
|
| 252535 |
+
{
|
| 252536 |
+
"epoch": 970.44,
|
| 252537 |
+
"learning_rate": 8.087451923076924e-06,
|
| 252538 |
+
"loss": 0.3265,
|
| 252539 |
+
"step": 120390
|
| 252540 |
+
},
|
| 252541 |
+
{
|
| 252542 |
+
"epoch": 970.48,
|
| 252543 |
+
"learning_rate": 8.087371794871795e-06,
|
| 252544 |
+
"loss": 0.3281,
|
| 252545 |
+
"step": 120395
|
| 252546 |
+
},
|
| 252547 |
+
{
|
| 252548 |
+
"epoch": 970.52,
|
| 252549 |
+
"learning_rate": 8.087291666666668e-06,
|
| 252550 |
+
"loss": 0.2896,
|
| 252551 |
+
"step": 120400
|
| 252552 |
+
},
|
| 252553 |
+
{
|
| 252554 |
+
"epoch": 970.56,
|
| 252555 |
+
"learning_rate": 8.08721153846154e-06,
|
| 252556 |
+
"loss": 0.4595,
|
| 252557 |
+
"step": 120405
|
| 252558 |
+
},
|
| 252559 |
+
{
|
| 252560 |
+
"epoch": 970.6,
|
| 252561 |
+
"learning_rate": 8.08713141025641e-06,
|
| 252562 |
+
"loss": 1.323,
|
| 252563 |
+
"step": 120410
|
| 252564 |
+
},
|
| 252565 |
+
{
|
| 252566 |
+
"epoch": 970.64,
|
| 252567 |
+
"learning_rate": 8.087051282051282e-06,
|
| 252568 |
+
"loss": 0.3019,
|
| 252569 |
+
"step": 120415
|
| 252570 |
+
},
|
| 252571 |
+
{
|
| 252572 |
+
"epoch": 970.68,
|
| 252573 |
+
"learning_rate": 8.086971153846155e-06,
|
| 252574 |
+
"loss": 0.2593,
|
| 252575 |
+
"step": 120420
|
| 252576 |
+
},
|
| 252577 |
+
{
|
| 252578 |
+
"epoch": 970.72,
|
| 252579 |
+
"learning_rate": 8.086891025641026e-06,
|
| 252580 |
+
"loss": 0.3622,
|
| 252581 |
+
"step": 120425
|
| 252582 |
+
},
|
| 252583 |
+
{
|
| 252584 |
+
"epoch": 970.76,
|
| 252585 |
+
"learning_rate": 8.086810897435898e-06,
|
| 252586 |
+
"loss": 0.4652,
|
| 252587 |
+
"step": 120430
|
| 252588 |
+
},
|
| 252589 |
+
{
|
| 252590 |
+
"epoch": 970.8,
|
| 252591 |
+
"learning_rate": 8.08673076923077e-06,
|
| 252592 |
+
"loss": 1.219,
|
| 252593 |
+
"step": 120435
|
| 252594 |
+
},
|
| 252595 |
+
{
|
| 252596 |
+
"epoch": 970.84,
|
| 252597 |
+
"learning_rate": 8.086650641025642e-06,
|
| 252598 |
+
"loss": 0.2825,
|
| 252599 |
+
"step": 120440
|
| 252600 |
+
},
|
| 252601 |
+
{
|
| 252602 |
+
"epoch": 970.88,
|
| 252603 |
+
"learning_rate": 8.086570512820514e-06,
|
| 252604 |
+
"loss": 0.2757,
|
| 252605 |
+
"step": 120445
|
| 252606 |
+
},
|
| 252607 |
+
{
|
| 252608 |
+
"epoch": 970.92,
|
| 252609 |
+
"learning_rate": 8.086490384615385e-06,
|
| 252610 |
+
"loss": 0.2955,
|
| 252611 |
+
"step": 120450
|
| 252612 |
+
},
|
| 252613 |
+
{
|
| 252614 |
+
"epoch": 970.96,
|
| 252615 |
+
"learning_rate": 8.086410256410258e-06,
|
| 252616 |
+
"loss": 0.5144,
|
| 252617 |
+
"step": 120455
|
| 252618 |
+
},
|
| 252619 |
+
{
|
| 252620 |
+
"epoch": 971.0,
|
| 252621 |
+
"eval_loss": 0.4524520933628082,
|
| 252622 |
+
"eval_runtime": 35.3415,
|
| 252623 |
+
"eval_samples_per_second": 23.825,
|
| 252624 |
+
"eval_steps_per_second": 0.764,
|
| 252625 |
+
"eval_wer": 0.18568594312587258,
|
| 252626 |
+
"step": 120459
|
| 252627 |
}
|
| 252628 |
],
|
| 252629 |
+
"max_steps": 620000,
|
| 252630 |
"num_train_epochs": 5000,
|
| 252631 |
+
"total_flos": 3.389937279198161e+20,
|
| 252632 |
"trial_name": null,
|
| 252633 |
"trial_params": null
|
| 252634 |
}
|
model-bin/finetune/base/{checkpoint-119837 → checkpoint-120459}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630128637.686285/events.out.tfevents.1630128637.86bb0ddabf9b.4092.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8629e25b2866bb91721c507417c5da84fec1f825a525ac155eb2fbe83349ca1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630129022.8707244/events.out.tfevents.1630129022.86bb0ddabf9b.4092.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3da616f0ad0b2a568ba716150120bc1e2a655f06531abf80682f7e01f332b41d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630129401.5085042/events.out.tfevents.1630129401.86bb0ddabf9b.4092.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30efa69d65c9de03c1d93a899048d06e2fc810649f2213c3d6e34a41f271d753
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630129786.576031/events.out.tfevents.1630129786.86bb0ddabf9b.4092.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c7ef6fc423c7aa4875f4dd537aca610cd055c7dd93fabb82f8efab510057763
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630130174.479612/events.out.tfevents.1630130174.86bb0ddabf9b.4092.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:decce1a8508f4477335b7508cc2940ca454cafeddf2bd4464db8e24c3ac3b6e7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630128637.86bb0ddabf9b.4092.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:889c5b5fe30ec571b4007894abe22c67a32d9a28873b616a40fe7e31afdd6062
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630129022.86bb0ddabf9b.4092.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c683b55c6c95cdb31e476e0e35aea043b8bdb6ebdd10499e87a4ffd58bd0ad1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630129401.86bb0ddabf9b.4092.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3358890093ccfcfd55a8bdcd7edf74b5823caa87989e8b8fda48afc4845df90d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630129786.86bb0ddabf9b.4092.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ee08e1168378209e1bef83b54005105504a705d5cdaf6b364021f18e1bf5529c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630130174.86bb0ddabf9b.4092.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4165172fd9dbeb2f86468eadd3af9ff577adcdf674e5044fe2e23b641db5b946
|
| 3 |
+
size 8462
|