"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/trainer_state.json +641 -5
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/training_args.bin +0 -0
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/trainer_state.json +0 -0
- model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629939465.6417572/events.out.tfevents.1629939465.8e89bd551565.924.1 +3 -0
- model-bin/finetune/base/log/1629939905.367717/events.out.tfevents.1629939905.8e89bd551565.924.3 +3 -0
- model-bin/finetune/base/log/1629940450.426258/events.out.tfevents.1629940450.8e89bd551565.924.5 +3 -0
- model-bin/finetune/base/log/1629940922.8454914/events.out.tfevents.1629940922.8e89bd551565.924.7 +3 -0
- model-bin/finetune/base/log/1629941350.450414/events.out.tfevents.1629941350.8e89bd551565.924.9 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629939465.8e89bd551565.924.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629939905.8e89bd551565.924.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629940450.8e89bd551565.924.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629940922.8e89bd551565.924.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629941350.8e89bd551565.924.8 +3 -0
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/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 sha256:03ab89beca2f81624334757a47196f2700ffb64a09d11ad89369efc03d60446c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/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:f743c4a7f5e90dc62c3a3443673fcb9e237965bd397b6783bf1f839fc38b5c10
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/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 sha256:12bb072db3e1f88523968032baaced85ff7adeedb4a8a975c73385b0edf9e365
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/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:f4cd92943fc5a165b57777669567f5fddb9a9bb4e6eef2e77618b7c86a10f5fc
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/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 sha256:320a18356ec6893f73250de1ff44f61371d24c656b47dfc9e100b1a509e1b7ee
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88857}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -211782,11 +211782,647 @@
|
|
| 211782 |
"eval_steps_per_second": 0.656,
|
| 211783 |
"eval_wer": 0.189195053156867,
|
| 211784 |
"step": 88359
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 211785 |
}
|
| 211786 |
],
|
| 211787 |
"max_steps": 620000,
|
| 211788 |
"num_train_epochs": 5000,
|
| 211789 |
-
"total_flos": 2.
|
| 211790 |
"trial_name": null,
|
| 211791 |
"trial_params": null
|
| 211792 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.17731766728757703,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 715.995983935743,
|
| 5 |
+
"global_step": 88857,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 211782 |
"eval_steps_per_second": 0.656,
|
| 211783 |
"eval_wer": 0.189195053156867,
|
| 211784 |
"step": 88359
|
| 211785 |
+
},
|
| 211786 |
+
{
|
| 211787 |
+
"epoch": 706.01,
|
| 211788 |
+
"learning_rate": 8.600576923076924e-06,
|
| 211789 |
+
"loss": 0.3692,
|
| 211790 |
+
"step": 88360
|
| 211791 |
+
},
|
| 211792 |
+
{
|
| 211793 |
+
"epoch": 706.05,
|
| 211794 |
+
"learning_rate": 8.600496794871795e-06,
|
| 211795 |
+
"loss": 0.2983,
|
| 211796 |
+
"step": 88365
|
| 211797 |
+
},
|
| 211798 |
+
{
|
| 211799 |
+
"epoch": 706.09,
|
| 211800 |
+
"learning_rate": 8.600416666666668e-06,
|
| 211801 |
+
"loss": 0.3386,
|
| 211802 |
+
"step": 88370
|
| 211803 |
+
},
|
| 211804 |
+
{
|
| 211805 |
+
"epoch": 706.13,
|
| 211806 |
+
"learning_rate": 8.60033653846154e-06,
|
| 211807 |
+
"loss": 0.3999,
|
| 211808 |
+
"step": 88375
|
| 211809 |
+
},
|
| 211810 |
+
{
|
| 211811 |
+
"epoch": 706.17,
|
| 211812 |
+
"learning_rate": 8.600256410256411e-06,
|
| 211813 |
+
"loss": 0.5919,
|
| 211814 |
+
"step": 88380
|
| 211815 |
+
},
|
| 211816 |
+
{
|
| 211817 |
+
"epoch": 706.21,
|
| 211818 |
+
"learning_rate": 8.600176282051282e-06,
|
| 211819 |
+
"loss": 1.0348,
|
| 211820 |
+
"step": 88385
|
| 211821 |
+
},
|
| 211822 |
+
{
|
| 211823 |
+
"epoch": 706.25,
|
| 211824 |
+
"learning_rate": 8.600096153846155e-06,
|
| 211825 |
+
"loss": 0.2688,
|
| 211826 |
+
"step": 88390
|
| 211827 |
+
},
|
| 211828 |
+
{
|
| 211829 |
+
"epoch": 706.29,
|
| 211830 |
+
"learning_rate": 8.600016025641025e-06,
|
| 211831 |
+
"loss": 0.2694,
|
| 211832 |
+
"step": 88395
|
| 211833 |
+
},
|
| 211834 |
+
{
|
| 211835 |
+
"epoch": 706.33,
|
| 211836 |
+
"learning_rate": 8.599935897435898e-06,
|
| 211837 |
+
"loss": 0.2873,
|
| 211838 |
+
"step": 88400
|
| 211839 |
+
},
|
| 211840 |
+
{
|
| 211841 |
+
"epoch": 706.37,
|
| 211842 |
+
"learning_rate": 8.599855769230771e-06,
|
| 211843 |
+
"loss": 0.5365,
|
| 211844 |
+
"step": 88405
|
| 211845 |
+
},
|
| 211846 |
+
{
|
| 211847 |
+
"epoch": 706.41,
|
| 211848 |
+
"learning_rate": 8.599775641025643e-06,
|
| 211849 |
+
"loss": 1.1916,
|
| 211850 |
+
"step": 88410
|
| 211851 |
+
},
|
| 211852 |
+
{
|
| 211853 |
+
"epoch": 706.45,
|
| 211854 |
+
"learning_rate": 8.599695512820514e-06,
|
| 211855 |
+
"loss": 0.3537,
|
| 211856 |
+
"step": 88415
|
| 211857 |
+
},
|
| 211858 |
+
{
|
| 211859 |
+
"epoch": 706.49,
|
| 211860 |
+
"learning_rate": 8.599615384615385e-06,
|
| 211861 |
+
"loss": 0.334,
|
| 211862 |
+
"step": 88420
|
| 211863 |
+
},
|
| 211864 |
+
{
|
| 211865 |
+
"epoch": 706.53,
|
| 211866 |
+
"learning_rate": 8.599535256410258e-06,
|
| 211867 |
+
"loss": 0.3377,
|
| 211868 |
+
"step": 88425
|
| 211869 |
+
},
|
| 211870 |
+
{
|
| 211871 |
+
"epoch": 706.57,
|
| 211872 |
+
"learning_rate": 8.599455128205128e-06,
|
| 211873 |
+
"loss": 0.5144,
|
| 211874 |
+
"step": 88430
|
| 211875 |
+
},
|
| 211876 |
+
{
|
| 211877 |
+
"epoch": 706.61,
|
| 211878 |
+
"learning_rate": 8.599375000000001e-06,
|
| 211879 |
+
"loss": 1.357,
|
| 211880 |
+
"step": 88435
|
| 211881 |
+
},
|
| 211882 |
+
{
|
| 211883 |
+
"epoch": 706.65,
|
| 211884 |
+
"learning_rate": 8.599294871794872e-06,
|
| 211885 |
+
"loss": 0.3338,
|
| 211886 |
+
"step": 88440
|
| 211887 |
+
},
|
| 211888 |
+
{
|
| 211889 |
+
"epoch": 706.69,
|
| 211890 |
+
"learning_rate": 8.599214743589744e-06,
|
| 211891 |
+
"loss": 0.3426,
|
| 211892 |
+
"step": 88445
|
| 211893 |
+
},
|
| 211894 |
+
{
|
| 211895 |
+
"epoch": 706.73,
|
| 211896 |
+
"learning_rate": 8.599134615384615e-06,
|
| 211897 |
+
"loss": 0.3479,
|
| 211898 |
+
"step": 88450
|
| 211899 |
+
},
|
| 211900 |
+
{
|
| 211901 |
+
"epoch": 706.77,
|
| 211902 |
+
"learning_rate": 8.599054487179488e-06,
|
| 211903 |
+
"loss": 0.584,
|
| 211904 |
+
"step": 88455
|
| 211905 |
+
},
|
| 211906 |
+
{
|
| 211907 |
+
"epoch": 706.81,
|
| 211908 |
+
"learning_rate": 8.59897435897436e-06,
|
| 211909 |
+
"loss": 1.4261,
|
| 211910 |
+
"step": 88460
|
| 211911 |
+
},
|
| 211912 |
+
{
|
| 211913 |
+
"epoch": 706.85,
|
| 211914 |
+
"learning_rate": 8.59889423076923e-06,
|
| 211915 |
+
"loss": 0.2559,
|
| 211916 |
+
"step": 88465
|
| 211917 |
+
},
|
| 211918 |
+
{
|
| 211919 |
+
"epoch": 706.89,
|
| 211920 |
+
"learning_rate": 8.598814102564104e-06,
|
| 211921 |
+
"loss": 0.3294,
|
| 211922 |
+
"step": 88470
|
| 211923 |
+
},
|
| 211924 |
+
{
|
| 211925 |
+
"epoch": 706.93,
|
| 211926 |
+
"learning_rate": 8.598733974358975e-06,
|
| 211927 |
+
"loss": 0.3859,
|
| 211928 |
+
"step": 88475
|
| 211929 |
+
},
|
| 211930 |
+
{
|
| 211931 |
+
"epoch": 706.97,
|
| 211932 |
+
"learning_rate": 8.598653846153847e-06,
|
| 211933 |
+
"loss": 0.592,
|
| 211934 |
+
"step": 88480
|
| 211935 |
+
},
|
| 211936 |
+
{
|
| 211937 |
+
"epoch": 707.0,
|
| 211938 |
+
"eval_loss": 0.40724536776542664,
|
| 211939 |
+
"eval_runtime": 38.5536,
|
| 211940 |
+
"eval_samples_per_second": 21.762,
|
| 211941 |
+
"eval_steps_per_second": 0.7,
|
| 211942 |
+
"eval_wer": 0.1872142696545005,
|
| 211943 |
+
"step": 88484
|
| 211944 |
+
},
|
| 211945 |
+
{
|
| 211946 |
+
"epoch": 713.01,
|
| 211947 |
+
"learning_rate": 8.598573717948718e-06,
|
| 211948 |
+
"loss": 0.3631,
|
| 211949 |
+
"step": 88485
|
| 211950 |
+
},
|
| 211951 |
+
{
|
| 211952 |
+
"epoch": 713.05,
|
| 211953 |
+
"learning_rate": 8.598493589743591e-06,
|
| 211954 |
+
"loss": 0.3241,
|
| 211955 |
+
"step": 88490
|
| 211956 |
+
},
|
| 211957 |
+
{
|
| 211958 |
+
"epoch": 713.09,
|
| 211959 |
+
"learning_rate": 8.598413461538462e-06,
|
| 211960 |
+
"loss": 0.3114,
|
| 211961 |
+
"step": 88495
|
| 211962 |
+
},
|
| 211963 |
+
{
|
| 211964 |
+
"epoch": 713.13,
|
| 211965 |
+
"learning_rate": 8.598333333333334e-06,
|
| 211966 |
+
"loss": 0.3329,
|
| 211967 |
+
"step": 88500
|
| 211968 |
+
},
|
| 211969 |
+
{
|
| 211970 |
+
"epoch": 713.17,
|
| 211971 |
+
"learning_rate": 8.598253205128207e-06,
|
| 211972 |
+
"loss": 0.5927,
|
| 211973 |
+
"step": 88505
|
| 211974 |
+
},
|
| 211975 |
+
{
|
| 211976 |
+
"epoch": 713.21,
|
| 211977 |
+
"learning_rate": 8.598173076923078e-06,
|
| 211978 |
+
"loss": 1.2067,
|
| 211979 |
+
"step": 88510
|
| 211980 |
+
},
|
| 211981 |
+
{
|
| 211982 |
+
"epoch": 713.25,
|
| 211983 |
+
"learning_rate": 8.59809294871795e-06,
|
| 211984 |
+
"loss": 0.3103,
|
| 211985 |
+
"step": 88515
|
| 211986 |
+
},
|
| 211987 |
+
{
|
| 211988 |
+
"epoch": 713.29,
|
| 211989 |
+
"learning_rate": 8.598012820512821e-06,
|
| 211990 |
+
"loss": 0.3437,
|
| 211991 |
+
"step": 88520
|
| 211992 |
+
},
|
| 211993 |
+
{
|
| 211994 |
+
"epoch": 713.33,
|
| 211995 |
+
"learning_rate": 8.597932692307694e-06,
|
| 211996 |
+
"loss": 0.3895,
|
| 211997 |
+
"step": 88525
|
| 211998 |
+
},
|
| 211999 |
+
{
|
| 212000 |
+
"epoch": 713.37,
|
| 212001 |
+
"learning_rate": 8.597852564102565e-06,
|
| 212002 |
+
"loss": 0.6844,
|
| 212003 |
+
"step": 88530
|
| 212004 |
+
},
|
| 212005 |
+
{
|
| 212006 |
+
"epoch": 713.41,
|
| 212007 |
+
"learning_rate": 8.597772435897437e-06,
|
| 212008 |
+
"loss": 1.1051,
|
| 212009 |
+
"step": 88535
|
| 212010 |
+
},
|
| 212011 |
+
{
|
| 212012 |
+
"epoch": 713.45,
|
| 212013 |
+
"learning_rate": 8.597692307692308e-06,
|
| 212014 |
+
"loss": 0.2844,
|
| 212015 |
+
"step": 88540
|
| 212016 |
+
},
|
| 212017 |
+
{
|
| 212018 |
+
"epoch": 713.49,
|
| 212019 |
+
"learning_rate": 8.597612179487181e-06,
|
| 212020 |
+
"loss": 0.2973,
|
| 212021 |
+
"step": 88545
|
| 212022 |
+
},
|
| 212023 |
+
{
|
| 212024 |
+
"epoch": 713.53,
|
| 212025 |
+
"learning_rate": 8.59753205128205e-06,
|
| 212026 |
+
"loss": 0.5913,
|
| 212027 |
+
"step": 88550
|
| 212028 |
+
},
|
| 212029 |
+
{
|
| 212030 |
+
"epoch": 713.57,
|
| 212031 |
+
"learning_rate": 8.597451923076924e-06,
|
| 212032 |
+
"loss": 0.5365,
|
| 212033 |
+
"step": 88555
|
| 212034 |
+
},
|
| 212035 |
+
{
|
| 212036 |
+
"epoch": 713.61,
|
| 212037 |
+
"learning_rate": 8.597371794871797e-06,
|
| 212038 |
+
"loss": 1.1011,
|
| 212039 |
+
"step": 88560
|
| 212040 |
+
},
|
| 212041 |
+
{
|
| 212042 |
+
"epoch": 713.65,
|
| 212043 |
+
"learning_rate": 8.597291666666666e-06,
|
| 212044 |
+
"loss": 0.304,
|
| 212045 |
+
"step": 88565
|
| 212046 |
+
},
|
| 212047 |
+
{
|
| 212048 |
+
"epoch": 713.69,
|
| 212049 |
+
"learning_rate": 8.59721153846154e-06,
|
| 212050 |
+
"loss": 0.3349,
|
| 212051 |
+
"step": 88570
|
| 212052 |
+
},
|
| 212053 |
+
{
|
| 212054 |
+
"epoch": 713.73,
|
| 212055 |
+
"learning_rate": 8.597131410256411e-06,
|
| 212056 |
+
"loss": 0.3495,
|
| 212057 |
+
"step": 88575
|
| 212058 |
+
},
|
| 212059 |
+
{
|
| 212060 |
+
"epoch": 713.77,
|
| 212061 |
+
"learning_rate": 8.597051282051282e-06,
|
| 212062 |
+
"loss": 0.5413,
|
| 212063 |
+
"step": 88580
|
| 212064 |
+
},
|
| 212065 |
+
{
|
| 212066 |
+
"epoch": 713.81,
|
| 212067 |
+
"learning_rate": 8.596971153846154e-06,
|
| 212068 |
+
"loss": 1.1389,
|
| 212069 |
+
"step": 88585
|
| 212070 |
+
},
|
| 212071 |
+
{
|
| 212072 |
+
"epoch": 713.85,
|
| 212073 |
+
"learning_rate": 8.596891025641027e-06,
|
| 212074 |
+
"loss": 0.3478,
|
| 212075 |
+
"step": 88590
|
| 212076 |
+
},
|
| 212077 |
+
{
|
| 212078 |
+
"epoch": 713.9,
|
| 212079 |
+
"learning_rate": 8.596810897435898e-06,
|
| 212080 |
+
"loss": 0.3048,
|
| 212081 |
+
"step": 88595
|
| 212082 |
+
},
|
| 212083 |
+
{
|
| 212084 |
+
"epoch": 713.94,
|
| 212085 |
+
"learning_rate": 8.59673076923077e-06,
|
| 212086 |
+
"loss": 0.3369,
|
| 212087 |
+
"step": 88600
|
| 212088 |
+
},
|
| 212089 |
+
{
|
| 212090 |
+
"epoch": 713.98,
|
| 212091 |
+
"learning_rate": 8.59665064102564e-06,
|
| 212092 |
+
"loss": 0.6376,
|
| 212093 |
+
"step": 88605
|
| 212094 |
+
},
|
| 212095 |
+
{
|
| 212096 |
+
"epoch": 714.0,
|
| 212097 |
+
"eval_loss": 0.38512417674064636,
|
| 212098 |
+
"eval_runtime": 39.6255,
|
| 212099 |
+
"eval_samples_per_second": 21.198,
|
| 212100 |
+
"eval_steps_per_second": 0.681,
|
| 212101 |
+
"eval_wer": 0.18599016080475805,
|
| 212102 |
+
"step": 88608
|
| 212103 |
+
},
|
| 212104 |
+
{
|
| 212105 |
+
"epoch": 708.02,
|
| 212106 |
+
"learning_rate": 8.596570512820514e-06,
|
| 212107 |
+
"loss": 0.3381,
|
| 212108 |
+
"step": 88610
|
| 212109 |
+
},
|
| 212110 |
+
{
|
| 212111 |
+
"epoch": 708.06,
|
| 212112 |
+
"learning_rate": 8.596490384615385e-06,
|
| 212113 |
+
"loss": 0.3083,
|
| 212114 |
+
"step": 88615
|
| 212115 |
+
},
|
| 212116 |
+
{
|
| 212117 |
+
"epoch": 708.1,
|
| 212118 |
+
"learning_rate": 8.596410256410257e-06,
|
| 212119 |
+
"loss": 0.3253,
|
| 212120 |
+
"step": 88620
|
| 212121 |
+
},
|
| 212122 |
+
{
|
| 212123 |
+
"epoch": 708.14,
|
| 212124 |
+
"learning_rate": 8.59633012820513e-06,
|
| 212125 |
+
"loss": 0.4044,
|
| 212126 |
+
"step": 88625
|
| 212127 |
+
},
|
| 212128 |
+
{
|
| 212129 |
+
"epoch": 708.18,
|
| 212130 |
+
"learning_rate": 8.596250000000001e-06,
|
| 212131 |
+
"loss": 0.6699,
|
| 212132 |
+
"step": 88630
|
| 212133 |
+
},
|
| 212134 |
+
{
|
| 212135 |
+
"epoch": 708.22,
|
| 212136 |
+
"learning_rate": 8.596169871794872e-06,
|
| 212137 |
+
"loss": 0.958,
|
| 212138 |
+
"step": 88635
|
| 212139 |
+
},
|
| 212140 |
+
{
|
| 212141 |
+
"epoch": 708.26,
|
| 212142 |
+
"learning_rate": 8.596089743589744e-06,
|
| 212143 |
+
"loss": 0.3337,
|
| 212144 |
+
"step": 88640
|
| 212145 |
+
},
|
| 212146 |
+
{
|
| 212147 |
+
"epoch": 708.3,
|
| 212148 |
+
"learning_rate": 8.596009615384617e-06,
|
| 212149 |
+
"loss": 0.3107,
|
| 212150 |
+
"step": 88645
|
| 212151 |
+
},
|
| 212152 |
+
{
|
| 212153 |
+
"epoch": 708.34,
|
| 212154 |
+
"learning_rate": 8.595929487179488e-06,
|
| 212155 |
+
"loss": 0.4837,
|
| 212156 |
+
"step": 88650
|
| 212157 |
+
},
|
| 212158 |
+
{
|
| 212159 |
+
"epoch": 708.38,
|
| 212160 |
+
"learning_rate": 8.59584935897436e-06,
|
| 212161 |
+
"loss": 0.7526,
|
| 212162 |
+
"step": 88655
|
| 212163 |
+
},
|
| 212164 |
+
{
|
| 212165 |
+
"epoch": 708.42,
|
| 212166 |
+
"learning_rate": 8.595769230769232e-06,
|
| 212167 |
+
"loss": 1.1701,
|
| 212168 |
+
"step": 88660
|
| 212169 |
+
},
|
| 212170 |
+
{
|
| 212171 |
+
"epoch": 708.46,
|
| 212172 |
+
"learning_rate": 8.595689102564104e-06,
|
| 212173 |
+
"loss": 0.3475,
|
| 212174 |
+
"step": 88665
|
| 212175 |
+
},
|
| 212176 |
+
{
|
| 212177 |
+
"epoch": 708.5,
|
| 212178 |
+
"learning_rate": 8.595608974358975e-06,
|
| 212179 |
+
"loss": 0.2971,
|
| 212180 |
+
"step": 88670
|
| 212181 |
+
},
|
| 212182 |
+
{
|
| 212183 |
+
"epoch": 708.54,
|
| 212184 |
+
"learning_rate": 8.595528846153847e-06,
|
| 212185 |
+
"loss": 0.3562,
|
| 212186 |
+
"step": 88675
|
| 212187 |
+
},
|
| 212188 |
+
{
|
| 212189 |
+
"epoch": 708.58,
|
| 212190 |
+
"learning_rate": 8.59544871794872e-06,
|
| 212191 |
+
"loss": 0.6971,
|
| 212192 |
+
"step": 88680
|
| 212193 |
+
},
|
| 212194 |
+
{
|
| 212195 |
+
"epoch": 708.62,
|
| 212196 |
+
"learning_rate": 8.595368589743591e-06,
|
| 212197 |
+
"loss": 1.0363,
|
| 212198 |
+
"step": 88685
|
| 212199 |
+
},
|
| 212200 |
+
{
|
| 212201 |
+
"epoch": 708.66,
|
| 212202 |
+
"learning_rate": 8.595288461538462e-06,
|
| 212203 |
+
"loss": 0.3066,
|
| 212204 |
+
"step": 88690
|
| 212205 |
+
},
|
| 212206 |
+
{
|
| 212207 |
+
"epoch": 708.7,
|
| 212208 |
+
"learning_rate": 8.595208333333334e-06,
|
| 212209 |
+
"loss": 0.2913,
|
| 212210 |
+
"step": 88695
|
| 212211 |
+
},
|
| 212212 |
+
{
|
| 212213 |
+
"epoch": 708.74,
|
| 212214 |
+
"learning_rate": 8.595128205128207e-06,
|
| 212215 |
+
"loss": 0.341,
|
| 212216 |
+
"step": 88700
|
| 212217 |
+
},
|
| 212218 |
+
{
|
| 212219 |
+
"epoch": 708.78,
|
| 212220 |
+
"learning_rate": 8.595048076923076e-06,
|
| 212221 |
+
"loss": 0.6731,
|
| 212222 |
+
"step": 88705
|
| 212223 |
+
},
|
| 212224 |
+
{
|
| 212225 |
+
"epoch": 708.82,
|
| 212226 |
+
"learning_rate": 8.59496794871795e-06,
|
| 212227 |
+
"loss": 1.0604,
|
| 212228 |
+
"step": 88710
|
| 212229 |
+
},
|
| 212230 |
+
{
|
| 212231 |
+
"epoch": 708.86,
|
| 212232 |
+
"learning_rate": 8.594887820512822e-06,
|
| 212233 |
+
"loss": 0.2956,
|
| 212234 |
+
"step": 88715
|
| 212235 |
+
},
|
| 212236 |
+
{
|
| 212237 |
+
"epoch": 708.9,
|
| 212238 |
+
"learning_rate": 8.594807692307692e-06,
|
| 212239 |
+
"loss": 0.2839,
|
| 212240 |
+
"step": 88720
|
| 212241 |
+
},
|
| 212242 |
+
{
|
| 212243 |
+
"epoch": 708.94,
|
| 212244 |
+
"learning_rate": 8.594727564102565e-06,
|
| 212245 |
+
"loss": 0.3427,
|
| 212246 |
+
"step": 88725
|
| 212247 |
+
},
|
| 212248 |
+
{
|
| 212249 |
+
"epoch": 708.98,
|
| 212250 |
+
"learning_rate": 8.594647435897437e-06,
|
| 212251 |
+
"loss": 0.7291,
|
| 212252 |
+
"step": 88730
|
| 212253 |
+
},
|
| 212254 |
+
{
|
| 212255 |
+
"epoch": 709.0,
|
| 212256 |
+
"eval_loss": 0.384115993976593,
|
| 212257 |
+
"eval_runtime": 39.7382,
|
| 212258 |
+
"eval_samples_per_second": 21.189,
|
| 212259 |
+
"eval_steps_per_second": 0.679,
|
| 212260 |
+
"eval_wer": 0.18778051252352684,
|
| 212261 |
+
"step": 88733
|
| 212262 |
+
},
|
| 212263 |
+
{
|
| 212264 |
+
"epoch": 715.02,
|
| 212265 |
+
"learning_rate": 8.594567307692308e-06,
|
| 212266 |
+
"loss": 0.3286,
|
| 212267 |
+
"step": 88735
|
| 212268 |
+
},
|
| 212269 |
+
{
|
| 212270 |
+
"epoch": 715.06,
|
| 212271 |
+
"learning_rate": 8.59448717948718e-06,
|
| 212272 |
+
"loss": 0.2882,
|
| 212273 |
+
"step": 88740
|
| 212274 |
+
},
|
| 212275 |
+
{
|
| 212276 |
+
"epoch": 715.1,
|
| 212277 |
+
"learning_rate": 8.594407051282052e-06,
|
| 212278 |
+
"loss": 0.3112,
|
| 212279 |
+
"step": 88745
|
| 212280 |
+
},
|
| 212281 |
+
{
|
| 212282 |
+
"epoch": 715.14,
|
| 212283 |
+
"learning_rate": 8.594326923076924e-06,
|
| 212284 |
+
"loss": 0.3241,
|
| 212285 |
+
"step": 88750
|
| 212286 |
+
},
|
| 212287 |
+
{
|
| 212288 |
+
"epoch": 715.18,
|
| 212289 |
+
"learning_rate": 8.594246794871795e-06,
|
| 212290 |
+
"loss": 0.5784,
|
| 212291 |
+
"step": 88755
|
| 212292 |
+
},
|
| 212293 |
+
{
|
| 212294 |
+
"epoch": 715.22,
|
| 212295 |
+
"learning_rate": 8.594166666666668e-06,
|
| 212296 |
+
"loss": 0.9482,
|
| 212297 |
+
"step": 88760
|
| 212298 |
+
},
|
| 212299 |
+
{
|
| 212300 |
+
"epoch": 715.26,
|
| 212301 |
+
"learning_rate": 8.59408653846154e-06,
|
| 212302 |
+
"loss": 0.2835,
|
| 212303 |
+
"step": 88765
|
| 212304 |
+
},
|
| 212305 |
+
{
|
| 212306 |
+
"epoch": 715.3,
|
| 212307 |
+
"learning_rate": 8.59400641025641e-06,
|
| 212308 |
+
"loss": 0.2901,
|
| 212309 |
+
"step": 88770
|
| 212310 |
+
},
|
| 212311 |
+
{
|
| 212312 |
+
"epoch": 715.34,
|
| 212313 |
+
"learning_rate": 8.593926282051282e-06,
|
| 212314 |
+
"loss": 0.3611,
|
| 212315 |
+
"step": 88775
|
| 212316 |
+
},
|
| 212317 |
+
{
|
| 212318 |
+
"epoch": 715.38,
|
| 212319 |
+
"learning_rate": 8.593846153846155e-06,
|
| 212320 |
+
"loss": 0.6981,
|
| 212321 |
+
"step": 88780
|
| 212322 |
+
},
|
| 212323 |
+
{
|
| 212324 |
+
"epoch": 715.42,
|
| 212325 |
+
"learning_rate": 8.593766025641027e-06,
|
| 212326 |
+
"loss": 0.9617,
|
| 212327 |
+
"step": 88785
|
| 212328 |
+
},
|
| 212329 |
+
{
|
| 212330 |
+
"epoch": 715.46,
|
| 212331 |
+
"learning_rate": 8.593685897435898e-06,
|
| 212332 |
+
"loss": 0.3106,
|
| 212333 |
+
"step": 88790
|
| 212334 |
+
},
|
| 212335 |
+
{
|
| 212336 |
+
"epoch": 715.5,
|
| 212337 |
+
"learning_rate": 8.59360576923077e-06,
|
| 212338 |
+
"loss": 0.2652,
|
| 212339 |
+
"step": 88795
|
| 212340 |
+
},
|
| 212341 |
+
{
|
| 212342 |
+
"epoch": 715.54,
|
| 212343 |
+
"learning_rate": 8.593525641025642e-06,
|
| 212344 |
+
"loss": 0.3255,
|
| 212345 |
+
"step": 88800
|
| 212346 |
+
},
|
| 212347 |
+
{
|
| 212348 |
+
"epoch": 715.58,
|
| 212349 |
+
"learning_rate": 8.593445512820514e-06,
|
| 212350 |
+
"loss": 0.6727,
|
| 212351 |
+
"step": 88805
|
| 212352 |
+
},
|
| 212353 |
+
{
|
| 212354 |
+
"epoch": 715.62,
|
| 212355 |
+
"learning_rate": 8.593365384615385e-06,
|
| 212356 |
+
"loss": 0.9983,
|
| 212357 |
+
"step": 88810
|
| 212358 |
+
},
|
| 212359 |
+
{
|
| 212360 |
+
"epoch": 715.66,
|
| 212361 |
+
"learning_rate": 8.593285256410258e-06,
|
| 212362 |
+
"loss": 0.3482,
|
| 212363 |
+
"step": 88815
|
| 212364 |
+
},
|
| 212365 |
+
{
|
| 212366 |
+
"epoch": 715.7,
|
| 212367 |
+
"learning_rate": 8.59320512820513e-06,
|
| 212368 |
+
"loss": 0.3422,
|
| 212369 |
+
"step": 88820
|
| 212370 |
+
},
|
| 212371 |
+
{
|
| 212372 |
+
"epoch": 715.74,
|
| 212373 |
+
"learning_rate": 8.593125e-06,
|
| 212374 |
+
"loss": 0.3654,
|
| 212375 |
+
"step": 88825
|
| 212376 |
+
},
|
| 212377 |
+
{
|
| 212378 |
+
"epoch": 715.78,
|
| 212379 |
+
"learning_rate": 8.593044871794872e-06,
|
| 212380 |
+
"loss": 0.7087,
|
| 212381 |
+
"step": 88830
|
| 212382 |
+
},
|
| 212383 |
+
{
|
| 212384 |
+
"epoch": 715.82,
|
| 212385 |
+
"learning_rate": 8.592964743589745e-06,
|
| 212386 |
+
"loss": 0.9036,
|
| 212387 |
+
"step": 88835
|
| 212388 |
+
},
|
| 212389 |
+
{
|
| 212390 |
+
"epoch": 715.86,
|
| 212391 |
+
"learning_rate": 8.592884615384615e-06,
|
| 212392 |
+
"loss": 0.3362,
|
| 212393 |
+
"step": 88840
|
| 212394 |
+
},
|
| 212395 |
+
{
|
| 212396 |
+
"epoch": 715.9,
|
| 212397 |
+
"learning_rate": 8.592804487179488e-06,
|
| 212398 |
+
"loss": 0.3543,
|
| 212399 |
+
"step": 88845
|
| 212400 |
+
},
|
| 212401 |
+
{
|
| 212402 |
+
"epoch": 715.94,
|
| 212403 |
+
"learning_rate": 8.59272435897436e-06,
|
| 212404 |
+
"loss": 0.3444,
|
| 212405 |
+
"step": 88850
|
| 212406 |
+
},
|
| 212407 |
+
{
|
| 212408 |
+
"epoch": 715.98,
|
| 212409 |
+
"learning_rate": 8.59264423076923e-06,
|
| 212410 |
+
"loss": 0.7275,
|
| 212411 |
+
"step": 88855
|
| 212412 |
+
},
|
| 212413 |
+
{
|
| 212414 |
+
"epoch": 716.0,
|
| 212415 |
+
"eval_loss": 0.42075592279434204,
|
| 212416 |
+
"eval_runtime": 40.0083,
|
| 212417 |
+
"eval_samples_per_second": 21.046,
|
| 212418 |
+
"eval_steps_per_second": 0.675,
|
| 212419 |
+
"eval_wer": 0.17731766728757703,
|
| 212420 |
+
"step": 88857
|
| 212421 |
}
|
| 212422 |
],
|
| 212423 |
"max_steps": 620000,
|
| 212424 |
"num_train_epochs": 5000,
|
| 212425 |
+
"total_flos": 2.5005701371269464e+20,
|
| 212426 |
"trial_name": null,
|
| 212427 |
"trial_params": null
|
| 212428 |
}
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88857}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/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 sha256:85223ba53941b432c78346aab6f9972c74939487ac260fa93b1d6b1f037f2caa
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/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:a7d0ee63362b6103e3b14d409796c9f7d0158647e1d489337c6a63ebb1785554
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aa2d10ac6adbf42dc45f693ff91266acd0ee130d5f99d16aabf290292d077da1
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/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 sha256:9da1248dae22f22d368d318bd65aa7f85b60c95535b8a8cd70eeacd40df98fbe
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/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:ca8f1a5975684ec4e0bf01320f1c3c47ae2829cdaa18b8d590d05e5f125b3b39
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-79148 → checkpoint-88981}/trainer_state.json
RENAMED
|
The diff for this file is too large to render.
See raw diff
|
|
|
model-bin/finetune/base/{checkpoint-88359 → checkpoint-88981}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629939465.6417572/events.out.tfevents.1629939465.8e89bd551565.924.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e141ce9b4127e1770c55b27424b480f9c16cfa6bdddc9384d374ea0d3366cd5a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629939905.367717/events.out.tfevents.1629939905.8e89bd551565.924.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6ad0afe0573af339a0a61f1f61cb45e5da611c5fe3726ee9ae92cdb358076d0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629940450.426258/events.out.tfevents.1629940450.8e89bd551565.924.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abe17b1e4d81daacd59ebd5fc55b611e1ecbbe08c365049f1a9dca0363ccbde5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629940922.8454914/events.out.tfevents.1629940922.8e89bd551565.924.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3e7148ef8a4bb21a33b79e1754c1ce5d8745bba37d398773eb587100c19a3fb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629941350.450414/events.out.tfevents.1629941350.8e89bd551565.924.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9ee05a6111fcb4b236e602754c722306f188669fd805f372d1ec101cdae2803a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629939465.8e89bd551565.924.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1ec61a66682433e6a2c4c5ed8f52536e1bdadf14b2be8bf14989e4e914e480a
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1629939905.8e89bd551565.924.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:09a4bed9d1b47a2fe49ea61ab22b2e850eadbb3b56f5fbbacf3720eb199c4230
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629940450.8e89bd551565.924.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e6fcceaf09e59a28fc26f7a9881cc55849bdd514a9b0322a9089593c40cfb29
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629940922.8e89bd551565.924.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:33c176ebbf4c0baa5db92ea369c5070629ad1717dcf415f0901f471b8243bf9e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629941350.8e89bd551565.924.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14955db15f07f8b3d9576f0a5e03653799ac0382b970196223a0e10f1ecade56
|
| 3 |
+
size 8622
|