"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/trainer_state.json +420 -3
- model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629641729.7589967/events.out.tfevents.1629641729.91f414e9573a.903.151 +3 -0
- model-bin/finetune/base/log/1629642395.5619562/events.out.tfevents.1629642395.91f414e9573a.903.153 +3 -0
- model-bin/finetune/base/log/1629643049.6534138/events.out.tfevents.1629643049.91f414e9573a.903.155 +3 -0
- model-bin/finetune/base/log/1629643711.158498/events.out.tfevents.1629643711.91f414e9573a.903.157 +3 -0
- model-bin/finetune/base/log/1629644375.2361584/events.out.tfevents.1629644375.91f414e9573a.903.159 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629641729.91f414e9573a.903.150 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629642395.91f414e9573a.903.152 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629643049.91f414e9573a.903.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629643711.91f414e9573a.903.156 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629644375.91f414e9573a.903.158 +3 -0
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/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:b9655c22548477bd8db92e500d00db6994f40d41184760e3b7232d6ee9e37982
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/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:d43c49f0b82b4116b0fb2bbb8f4db14722be977b2b98819268e884e26453e4be
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/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 sha256:4ce047836f61c617f7ec78fed13cf89b8a6ea57449d4b2039f5a971c82f84c35
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/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 sha256:5ef5b8afec8dd73df557d64a8e09416284e50252e3dfee8bc69adce3918b5579
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/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 sha256:3f4d23d3df4aff006b7f746f1ccaabf858487e6ea990eca0c6fd32ff36f5a367
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -143781,11 +143781,428 @@
|
|
| 143781 |
"eval_steps_per_second": 0.663,
|
| 143782 |
"eval_wer": 0.2137329286798179,
|
| 143783 |
"step": 31975
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 143784 |
}
|
| 143785 |
],
|
| 143786 |
"max_steps": 625000,
|
| 143787 |
"num_train_epochs": 5000,
|
| 143788 |
-
"total_flos":
|
| 143789 |
"trial_name": null,
|
| 143790 |
"trial_params": null
|
| 143791 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 260.0,
|
| 5 |
+
"global_step": 32599,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 143781 |
"eval_steps_per_second": 0.663,
|
| 143782 |
"eval_wer": 0.2137329286798179,
|
| 143783 |
"step": 31975
|
| 143784 |
+
},
|
| 143785 |
+
{
|
| 143786 |
+
"epoch": 257.04,
|
| 143787 |
+
"learning_rate": 9.4997253634895e-06,
|
| 143788 |
+
"loss": 0.4018,
|
| 143789 |
+
"step": 31980
|
| 143790 |
+
},
|
| 143791 |
+
{
|
| 143792 |
+
"epoch": 257.12,
|
| 143793 |
+
"learning_rate": 9.49956381260097e-06,
|
| 143794 |
+
"loss": 0.3734,
|
| 143795 |
+
"step": 31990
|
| 143796 |
+
},
|
| 143797 |
+
{
|
| 143798 |
+
"epoch": 257.2,
|
| 143799 |
+
"learning_rate": 9.49940226171244e-06,
|
| 143800 |
+
"loss": 0.9871,
|
| 143801 |
+
"step": 32000
|
| 143802 |
+
},
|
| 143803 |
+
{
|
| 143804 |
+
"epoch": 257.28,
|
| 143805 |
+
"learning_rate": 9.49924071082391e-06,
|
| 143806 |
+
"loss": 0.4114,
|
| 143807 |
+
"step": 32010
|
| 143808 |
+
},
|
| 143809 |
+
{
|
| 143810 |
+
"epoch": 257.36,
|
| 143811 |
+
"learning_rate": 9.49907915993538e-06,
|
| 143812 |
+
"loss": 0.622,
|
| 143813 |
+
"step": 32020
|
| 143814 |
+
},
|
| 143815 |
+
{
|
| 143816 |
+
"epoch": 257.44,
|
| 143817 |
+
"learning_rate": 9.49891760904685e-06,
|
| 143818 |
+
"loss": 0.8742,
|
| 143819 |
+
"step": 32030
|
| 143820 |
+
},
|
| 143821 |
+
{
|
| 143822 |
+
"epoch": 257.52,
|
| 143823 |
+
"learning_rate": 9.49875605815832e-06,
|
| 143824 |
+
"loss": 0.4095,
|
| 143825 |
+
"step": 32040
|
| 143826 |
+
},
|
| 143827 |
+
{
|
| 143828 |
+
"epoch": 257.6,
|
| 143829 |
+
"learning_rate": 9.49859450726979e-06,
|
| 143830 |
+
"loss": 1.0621,
|
| 143831 |
+
"step": 32050
|
| 143832 |
+
},
|
| 143833 |
+
{
|
| 143834 |
+
"epoch": 257.69,
|
| 143835 |
+
"learning_rate": 9.49843295638126e-06,
|
| 143836 |
+
"loss": 0.4454,
|
| 143837 |
+
"step": 32060
|
| 143838 |
+
},
|
| 143839 |
+
{
|
| 143840 |
+
"epoch": 257.77,
|
| 143841 |
+
"learning_rate": 9.49827140549273e-06,
|
| 143842 |
+
"loss": 0.5362,
|
| 143843 |
+
"step": 32070
|
| 143844 |
+
},
|
| 143845 |
+
{
|
| 143846 |
+
"epoch": 257.85,
|
| 143847 |
+
"learning_rate": 9.498109854604201e-06,
|
| 143848 |
+
"loss": 0.8852,
|
| 143849 |
+
"step": 32080
|
| 143850 |
+
},
|
| 143851 |
+
{
|
| 143852 |
+
"epoch": 257.93,
|
| 143853 |
+
"learning_rate": 9.497948303715671e-06,
|
| 143854 |
+
"loss": 0.4239,
|
| 143855 |
+
"step": 32090
|
| 143856 |
+
},
|
| 143857 |
+
{
|
| 143858 |
+
"epoch": 258.0,
|
| 143859 |
+
"eval_loss": 0.4601561725139618,
|
| 143860 |
+
"eval_runtime": 39.1812,
|
| 143861 |
+
"eval_samples_per_second": 21.439,
|
| 143862 |
+
"eval_steps_per_second": 0.689,
|
| 143863 |
+
"eval_wer": 0.21824480369515012,
|
| 143864 |
+
"step": 32099
|
| 143865 |
+
},
|
| 143866 |
+
{
|
| 143867 |
+
"epoch": 256.01,
|
| 143868 |
+
"learning_rate": 9.497786752827141e-06,
|
| 143869 |
+
"loss": 0.6621,
|
| 143870 |
+
"step": 32100
|
| 143871 |
+
},
|
| 143872 |
+
{
|
| 143873 |
+
"epoch": 256.09,
|
| 143874 |
+
"learning_rate": 9.497625201938611e-06,
|
| 143875 |
+
"loss": 0.3827,
|
| 143876 |
+
"step": 32110
|
| 143877 |
+
},
|
| 143878 |
+
{
|
| 143879 |
+
"epoch": 256.17,
|
| 143880 |
+
"learning_rate": 9.497463651050081e-06,
|
| 143881 |
+
"loss": 0.5514,
|
| 143882 |
+
"step": 32120
|
| 143883 |
+
},
|
| 143884 |
+
{
|
| 143885 |
+
"epoch": 256.25,
|
| 143886 |
+
"learning_rate": 9.497302100161551e-06,
|
| 143887 |
+
"loss": 0.7821,
|
| 143888 |
+
"step": 32130
|
| 143889 |
+
},
|
| 143890 |
+
{
|
| 143891 |
+
"epoch": 256.33,
|
| 143892 |
+
"learning_rate": 9.497140549273021e-06,
|
| 143893 |
+
"loss": 0.411,
|
| 143894 |
+
"step": 32140
|
| 143895 |
+
},
|
| 143896 |
+
{
|
| 143897 |
+
"epoch": 256.41,
|
| 143898 |
+
"learning_rate": 9.496978998384491e-06,
|
| 143899 |
+
"loss": 1.0059,
|
| 143900 |
+
"step": 32150
|
| 143901 |
+
},
|
| 143902 |
+
{
|
| 143903 |
+
"epoch": 256.49,
|
| 143904 |
+
"learning_rate": 9.496817447495961e-06,
|
| 143905 |
+
"loss": 0.3535,
|
| 143906 |
+
"step": 32160
|
| 143907 |
+
},
|
| 143908 |
+
{
|
| 143909 |
+
"epoch": 256.57,
|
| 143910 |
+
"learning_rate": 9.496655896607431e-06,
|
| 143911 |
+
"loss": 0.5797,
|
| 143912 |
+
"step": 32170
|
| 143913 |
+
},
|
| 143914 |
+
{
|
| 143915 |
+
"epoch": 256.65,
|
| 143916 |
+
"learning_rate": 9.496494345718901e-06,
|
| 143917 |
+
"loss": 0.8309,
|
| 143918 |
+
"step": 32180
|
| 143919 |
+
},
|
| 143920 |
+
{
|
| 143921 |
+
"epoch": 256.73,
|
| 143922 |
+
"learning_rate": 9.496332794830371e-06,
|
| 143923 |
+
"loss": 0.4263,
|
| 143924 |
+
"step": 32190
|
| 143925 |
+
},
|
| 143926 |
+
{
|
| 143927 |
+
"epoch": 256.81,
|
| 143928 |
+
"learning_rate": 9.496171243941843e-06,
|
| 143929 |
+
"loss": 0.9757,
|
| 143930 |
+
"step": 32200
|
| 143931 |
+
},
|
| 143932 |
+
{
|
| 143933 |
+
"epoch": 256.89,
|
| 143934 |
+
"learning_rate": 9.496009693053313e-06,
|
| 143935 |
+
"loss": 0.3924,
|
| 143936 |
+
"step": 32210
|
| 143937 |
+
},
|
| 143938 |
+
{
|
| 143939 |
+
"epoch": 256.97,
|
| 143940 |
+
"learning_rate": 9.495848142164783e-06,
|
| 143941 |
+
"loss": 0.4806,
|
| 143942 |
+
"step": 32220
|
| 143943 |
+
},
|
| 143944 |
+
{
|
| 143945 |
+
"epoch": 257.0,
|
| 143946 |
+
"eval_loss": 0.4139370620250702,
|
| 143947 |
+
"eval_runtime": 41.1864,
|
| 143948 |
+
"eval_samples_per_second": 20.395,
|
| 143949 |
+
"eval_steps_per_second": 0.656,
|
| 143950 |
+
"eval_wer": 0.2141655081583376,
|
| 143951 |
+
"step": 32224
|
| 143952 |
+
},
|
| 143953 |
+
{
|
| 143954 |
+
"epoch": 257.05,
|
| 143955 |
+
"learning_rate": 9.495686591276253e-06,
|
| 143956 |
+
"loss": 0.413,
|
| 143957 |
+
"step": 32230
|
| 143958 |
+
},
|
| 143959 |
+
{
|
| 143960 |
+
"epoch": 257.13,
|
| 143961 |
+
"learning_rate": 9.495525040387723e-06,
|
| 143962 |
+
"loss": 0.3938,
|
| 143963 |
+
"step": 32240
|
| 143964 |
+
},
|
| 143965 |
+
{
|
| 143966 |
+
"epoch": 257.21,
|
| 143967 |
+
"learning_rate": 9.495363489499193e-06,
|
| 143968 |
+
"loss": 0.9424,
|
| 143969 |
+
"step": 32250
|
| 143970 |
+
},
|
| 143971 |
+
{
|
| 143972 |
+
"epoch": 257.29,
|
| 143973 |
+
"learning_rate": 9.495201938610663e-06,
|
| 143974 |
+
"loss": 0.3502,
|
| 143975 |
+
"step": 32260
|
| 143976 |
+
},
|
| 143977 |
+
{
|
| 143978 |
+
"epoch": 257.37,
|
| 143979 |
+
"learning_rate": 9.495040387722133e-06,
|
| 143980 |
+
"loss": 0.5737,
|
| 143981 |
+
"step": 32270
|
| 143982 |
+
},
|
| 143983 |
+
{
|
| 143984 |
+
"epoch": 257.45,
|
| 143985 |
+
"learning_rate": 9.494878836833603e-06,
|
| 143986 |
+
"loss": 0.8254,
|
| 143987 |
+
"step": 32280
|
| 143988 |
+
},
|
| 143989 |
+
{
|
| 143990 |
+
"epoch": 257.53,
|
| 143991 |
+
"learning_rate": 9.494717285945073e-06,
|
| 143992 |
+
"loss": 0.3682,
|
| 143993 |
+
"step": 32290
|
| 143994 |
+
},
|
| 143995 |
+
{
|
| 143996 |
+
"epoch": 257.61,
|
| 143997 |
+
"learning_rate": 9.494555735056543e-06,
|
| 143998 |
+
"loss": 0.908,
|
| 143999 |
+
"step": 32300
|
| 144000 |
+
},
|
| 144001 |
+
{
|
| 144002 |
+
"epoch": 257.69,
|
| 144003 |
+
"learning_rate": 9.494394184168013e-06,
|
| 144004 |
+
"loss": 0.3974,
|
| 144005 |
+
"step": 32310
|
| 144006 |
+
},
|
| 144007 |
+
{
|
| 144008 |
+
"epoch": 257.77,
|
| 144009 |
+
"learning_rate": 9.494232633279484e-06,
|
| 144010 |
+
"loss": 0.5497,
|
| 144011 |
+
"step": 32320
|
| 144012 |
+
},
|
| 144013 |
+
{
|
| 144014 |
+
"epoch": 257.85,
|
| 144015 |
+
"learning_rate": 9.494071082390954e-06,
|
| 144016 |
+
"loss": 0.7748,
|
| 144017 |
+
"step": 32330
|
| 144018 |
+
},
|
| 144019 |
+
{
|
| 144020 |
+
"epoch": 257.93,
|
| 144021 |
+
"learning_rate": 9.493909531502424e-06,
|
| 144022 |
+
"loss": 0.4176,
|
| 144023 |
+
"step": 32340
|
| 144024 |
+
},
|
| 144025 |
+
{
|
| 144026 |
+
"epoch": 258.0,
|
| 144027 |
+
"eval_loss": 0.517149806022644,
|
| 144028 |
+
"eval_runtime": 39.7322,
|
| 144029 |
+
"eval_samples_per_second": 21.142,
|
| 144030 |
+
"eval_steps_per_second": 0.68,
|
| 144031 |
+
"eval_wer": 0.22357875814545727,
|
| 144032 |
+
"step": 32349
|
| 144033 |
+
},
|
| 144034 |
+
{
|
| 144035 |
+
"epoch": 258.01,
|
| 144036 |
+
"learning_rate": 9.493747980613894e-06,
|
| 144037 |
+
"loss": 0.5777,
|
| 144038 |
+
"step": 32350
|
| 144039 |
+
},
|
| 144040 |
+
{
|
| 144041 |
+
"epoch": 258.09,
|
| 144042 |
+
"learning_rate": 9.493586429725364e-06,
|
| 144043 |
+
"loss": 0.6018,
|
| 144044 |
+
"step": 32360
|
| 144045 |
+
},
|
| 144046 |
+
{
|
| 144047 |
+
"epoch": 258.17,
|
| 144048 |
+
"learning_rate": 9.493424878836834e-06,
|
| 144049 |
+
"loss": 0.4925,
|
| 144050 |
+
"step": 32370
|
| 144051 |
+
},
|
| 144052 |
+
{
|
| 144053 |
+
"epoch": 258.25,
|
| 144054 |
+
"learning_rate": 9.493263327948304e-06,
|
| 144055 |
+
"loss": 0.9365,
|
| 144056 |
+
"step": 32380
|
| 144057 |
+
},
|
| 144058 |
+
{
|
| 144059 |
+
"epoch": 258.33,
|
| 144060 |
+
"learning_rate": 9.493101777059774e-06,
|
| 144061 |
+
"loss": 0.4282,
|
| 144062 |
+
"step": 32390
|
| 144063 |
+
},
|
| 144064 |
+
{
|
| 144065 |
+
"epoch": 258.41,
|
| 144066 |
+
"learning_rate": 9.492940226171244e-06,
|
| 144067 |
+
"loss": 0.9752,
|
| 144068 |
+
"step": 32400
|
| 144069 |
+
},
|
| 144070 |
+
{
|
| 144071 |
+
"epoch": 258.49,
|
| 144072 |
+
"learning_rate": 9.492778675282714e-06,
|
| 144073 |
+
"loss": 0.4168,
|
| 144074 |
+
"step": 32410
|
| 144075 |
+
},
|
| 144076 |
+
{
|
| 144077 |
+
"epoch": 258.57,
|
| 144078 |
+
"learning_rate": 9.492617124394184e-06,
|
| 144079 |
+
"loss": 0.548,
|
| 144080 |
+
"step": 32420
|
| 144081 |
+
},
|
| 144082 |
+
{
|
| 144083 |
+
"epoch": 258.65,
|
| 144084 |
+
"learning_rate": 9.492455573505656e-06,
|
| 144085 |
+
"loss": 0.8692,
|
| 144086 |
+
"step": 32430
|
| 144087 |
+
},
|
| 144088 |
+
{
|
| 144089 |
+
"epoch": 258.73,
|
| 144090 |
+
"learning_rate": 9.492294022617126e-06,
|
| 144091 |
+
"loss": 0.3934,
|
| 144092 |
+
"step": 32440
|
| 144093 |
+
},
|
| 144094 |
+
{
|
| 144095 |
+
"epoch": 258.8,
|
| 144096 |
+
"learning_rate": 9.492132471728596e-06,
|
| 144097 |
+
"loss": 0.9343,
|
| 144098 |
+
"step": 32450
|
| 144099 |
+
},
|
| 144100 |
+
{
|
| 144101 |
+
"epoch": 258.88,
|
| 144102 |
+
"learning_rate": 9.491970920840066e-06,
|
| 144103 |
+
"loss": 0.4049,
|
| 144104 |
+
"step": 32460
|
| 144105 |
+
},
|
| 144106 |
+
{
|
| 144107 |
+
"epoch": 258.96,
|
| 144108 |
+
"learning_rate": 9.491809369951536e-06,
|
| 144109 |
+
"loss": 0.5295,
|
| 144110 |
+
"step": 32470
|
| 144111 |
+
},
|
| 144112 |
+
{
|
| 144113 |
+
"epoch": 259.0,
|
| 144114 |
+
"eval_loss": 0.39702287316322327,
|
| 144115 |
+
"eval_runtime": 41.6389,
|
| 144116 |
+
"eval_samples_per_second": 20.173,
|
| 144117 |
+
"eval_steps_per_second": 0.648,
|
| 144118 |
+
"eval_wer": 0.21267545725223308,
|
| 144119 |
+
"step": 32474
|
| 144120 |
+
},
|
| 144121 |
+
{
|
| 144122 |
+
"epoch": 259.05,
|
| 144123 |
+
"learning_rate": 9.491647819063006e-06,
|
| 144124 |
+
"loss": 0.4037,
|
| 144125 |
+
"step": 32480
|
| 144126 |
+
},
|
| 144127 |
+
{
|
| 144128 |
+
"epoch": 259.13,
|
| 144129 |
+
"learning_rate": 9.491486268174475e-06,
|
| 144130 |
+
"loss": 0.4094,
|
| 144131 |
+
"step": 32490
|
| 144132 |
+
},
|
| 144133 |
+
{
|
| 144134 |
+
"epoch": 259.21,
|
| 144135 |
+
"learning_rate": 9.491324717285945e-06,
|
| 144136 |
+
"loss": 0.9999,
|
| 144137 |
+
"step": 32500
|
| 144138 |
+
},
|
| 144139 |
+
{
|
| 144140 |
+
"epoch": 259.29,
|
| 144141 |
+
"learning_rate": 9.491163166397415e-06,
|
| 144142 |
+
"loss": 0.3856,
|
| 144143 |
+
"step": 32510
|
| 144144 |
+
},
|
| 144145 |
+
{
|
| 144146 |
+
"epoch": 259.37,
|
| 144147 |
+
"learning_rate": 9.491001615508885e-06,
|
| 144148 |
+
"loss": 0.5663,
|
| 144149 |
+
"step": 32520
|
| 144150 |
+
},
|
| 144151 |
+
{
|
| 144152 |
+
"epoch": 259.45,
|
| 144153 |
+
"learning_rate": 9.490840064620355e-06,
|
| 144154 |
+
"loss": 0.8961,
|
| 144155 |
+
"step": 32530
|
| 144156 |
+
},
|
| 144157 |
+
{
|
| 144158 |
+
"epoch": 259.53,
|
| 144159 |
+
"learning_rate": 9.490678513731825e-06,
|
| 144160 |
+
"loss": 0.381,
|
| 144161 |
+
"step": 32540
|
| 144162 |
+
},
|
| 144163 |
+
{
|
| 144164 |
+
"epoch": 259.61,
|
| 144165 |
+
"learning_rate": 9.490516962843297e-06,
|
| 144166 |
+
"loss": 1.0026,
|
| 144167 |
+
"step": 32550
|
| 144168 |
+
},
|
| 144169 |
+
{
|
| 144170 |
+
"epoch": 259.69,
|
| 144171 |
+
"learning_rate": 9.490355411954767e-06,
|
| 144172 |
+
"loss": 0.394,
|
| 144173 |
+
"step": 32560
|
| 144174 |
+
},
|
| 144175 |
+
{
|
| 144176 |
+
"epoch": 259.77,
|
| 144177 |
+
"learning_rate": 9.490193861066237e-06,
|
| 144178 |
+
"loss": 0.4965,
|
| 144179 |
+
"step": 32570
|
| 144180 |
+
},
|
| 144181 |
+
{
|
| 144182 |
+
"epoch": 259.85,
|
| 144183 |
+
"learning_rate": 9.490032310177707e-06,
|
| 144184 |
+
"loss": 0.8374,
|
| 144185 |
+
"step": 32580
|
| 144186 |
+
},
|
| 144187 |
+
{
|
| 144188 |
+
"epoch": 259.93,
|
| 144189 |
+
"learning_rate": 9.489870759289177e-06,
|
| 144190 |
+
"loss": 0.4129,
|
| 144191 |
+
"step": 32590
|
| 144192 |
+
},
|
| 144193 |
+
{
|
| 144194 |
+
"epoch": 260.0,
|
| 144195 |
+
"eval_loss": 0.45249393582344055,
|
| 144196 |
+
"eval_runtime": 42.1728,
|
| 144197 |
+
"eval_samples_per_second": 19.918,
|
| 144198 |
+
"eval_steps_per_second": 0.64,
|
| 144199 |
+
"eval_wer": 0.217536623524392,
|
| 144200 |
+
"step": 32599
|
| 144201 |
}
|
| 144202 |
],
|
| 144203 |
"max_steps": 625000,
|
| 144204 |
"num_train_epochs": 5000,
|
| 144205 |
+
"total_flos": 9.173029907579001e+19,
|
| 144206 |
"trial_name": null,
|
| 144207 |
"trial_params": null
|
| 144208 |
}
|
model-bin/finetune/base/{checkpoint-31975 → checkpoint-32599}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629641729.7589967/events.out.tfevents.1629641729.91f414e9573a.903.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ece58851f9242e30ff8a732e037a445b406ca4dcb1796f3b69f534969ead3cae
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629642395.5619562/events.out.tfevents.1629642395.91f414e9573a.903.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cdbef96050258edb3ab212a484fe520148a88db6cceec64efcf1061e558e2cd1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629643049.6534138/events.out.tfevents.1629643049.91f414e9573a.903.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2bb72da5b9431ee3536ab0a1cc5a2e154d6b01de8d5b247edccab83a7f9cdbac
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629643711.158498/events.out.tfevents.1629643711.91f414e9573a.903.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa4749e3bbac15abc474554a4edb932ed2e71f01984693590c2f555967bf5525
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629644375.2361584/events.out.tfevents.1629644375.91f414e9573a.903.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9663d900fcfccc93a0e695e2a9de137963e83bb44bb93ab23ecdbd978d21df71
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629641729.91f414e9573a.903.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3554f5be927311513fb5ff0caa49e870a7406b731b55d9c41ff6299aa9fa5d5c
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629642395.91f414e9573a.903.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:495755685d465bd7bc61c2ecbbed988e0a0ea2c9ccd3af3eb8aecf70847e59ea
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629643049.91f414e9573a.903.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a6ea32b8cbe2fb9710829f3dfd28f766375ba0c53476533d0134f4bf4cb3e91a
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629643711.91f414e9573a.903.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c1b17d3cb18159adce426a365e7f18c0c4bed4332534931520d2f58ca94b704
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629644375.91f414e9573a.903.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e626f5547e49a781ee6f4f682cf8659221f00b88b62142188ad120bed2ebeb60
|
| 3 |
+
size 6543
|