"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/trainer_state.json +427 -4
- model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629658404.8958828/events.out.tfevents.1629658404.91f414e9573a.903.201 +3 -0
- model-bin/finetune/base/log/1629659053.497968/events.out.tfevents.1629659053.91f414e9573a.903.203 +3 -0
- model-bin/finetune/base/log/1629659698.488905/events.out.tfevents.1629659698.91f414e9573a.903.205 +3 -0
- model-bin/finetune/base/log/1629660352.110468/events.out.tfevents.1629660352.91f414e9573a.903.207 +3 -0
- model-bin/finetune/base/log/1629661015.2787066/events.out.tfevents.1629661015.91f414e9573a.903.209 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629658404.91f414e9573a.903.200 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629659053.91f414e9573a.903.202 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629659698.91f414e9573a.903.204 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629660352.91f414e9573a.903.206 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629661015.91f414e9573a.903.208 +3 -0
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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 sha256:606b047c853de40071331b7796efbb6f776249cd1bff23dab600fee2875913d3
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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:aa86806dab10d349f92a5da6af1c2b2fbbc296f6a6e1a44ece0fae95a377d2e5
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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:af7b465a3c87e65e3c2b48872e0975904be02b62b7a24c569f86fdff1488d5bc
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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 sha256:60a77c8dd462fc3b4be2ff5da7bab717e71264b6dc82fc18d24d6930b1315c4f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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:dfe2c837bbdd7d4a3fe4962a3f90e4af8486c244dd16628508c17e17338fa67f
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/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,
|
|
@@ -145872,11 +145872,434 @@
|
|
| 145872 |
"eval_steps_per_second": 0.657,
|
| 145873 |
"eval_wer": 0.22463930627195933,
|
| 145874 |
"step": 35089
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145875 |
}
|
| 145876 |
],
|
| 145877 |
-
"max_steps":
|
| 145878 |
"num_train_epochs": 5000,
|
| 145879 |
-
"total_flos":
|
| 145880 |
"trial_name": null,
|
| 145881 |
"trial_params": null
|
| 145882 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 286.99598393574297,
|
| 5 |
+
"global_step": 35710,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 145872 |
"eval_steps_per_second": 0.657,
|
| 145873 |
"eval_wer": 0.22463930627195933,
|
| 145874 |
"step": 35089
|
| 145875 |
+
},
|
| 145876 |
+
{
|
| 145877 |
+
"epoch": 282.01,
|
| 145878 |
+
"learning_rate": 9.449483037156705e-06,
|
| 145879 |
+
"loss": 0.5899,
|
| 145880 |
+
"step": 35090
|
| 145881 |
+
},
|
| 145882 |
+
{
|
| 145883 |
+
"epoch": 282.09,
|
| 145884 |
+
"learning_rate": 9.449321486268175e-06,
|
| 145885 |
+
"loss": 0.3611,
|
| 145886 |
+
"step": 35100
|
| 145887 |
+
},
|
| 145888 |
+
{
|
| 145889 |
+
"epoch": 282.17,
|
| 145890 |
+
"learning_rate": 9.449159935379645e-06,
|
| 145891 |
+
"loss": 0.4706,
|
| 145892 |
+
"step": 35110
|
| 145893 |
+
},
|
| 145894 |
+
{
|
| 145895 |
+
"epoch": 282.25,
|
| 145896 |
+
"learning_rate": 9.448998384491116e-06,
|
| 145897 |
+
"loss": 0.8606,
|
| 145898 |
+
"step": 35120
|
| 145899 |
+
},
|
| 145900 |
+
{
|
| 145901 |
+
"epoch": 282.33,
|
| 145902 |
+
"learning_rate": 9.448836833602586e-06,
|
| 145903 |
+
"loss": 0.4232,
|
| 145904 |
+
"step": 35130
|
| 145905 |
+
},
|
| 145906 |
+
{
|
| 145907 |
+
"epoch": 282.41,
|
| 145908 |
+
"learning_rate": 9.448675282714056e-06,
|
| 145909 |
+
"loss": 1.0112,
|
| 145910 |
+
"step": 35140
|
| 145911 |
+
},
|
| 145912 |
+
{
|
| 145913 |
+
"epoch": 282.49,
|
| 145914 |
+
"learning_rate": 9.448513731825526e-06,
|
| 145915 |
+
"loss": 0.5655,
|
| 145916 |
+
"step": 35150
|
| 145917 |
+
},
|
| 145918 |
+
{
|
| 145919 |
+
"epoch": 282.57,
|
| 145920 |
+
"learning_rate": 9.448352180936996e-06,
|
| 145921 |
+
"loss": 0.5542,
|
| 145922 |
+
"step": 35160
|
| 145923 |
+
},
|
| 145924 |
+
{
|
| 145925 |
+
"epoch": 282.65,
|
| 145926 |
+
"learning_rate": 9.448190630048466e-06,
|
| 145927 |
+
"loss": 0.8786,
|
| 145928 |
+
"step": 35170
|
| 145929 |
+
},
|
| 145930 |
+
{
|
| 145931 |
+
"epoch": 282.73,
|
| 145932 |
+
"learning_rate": 9.448029079159936e-06,
|
| 145933 |
+
"loss": 0.3751,
|
| 145934 |
+
"step": 35180
|
| 145935 |
+
},
|
| 145936 |
+
{
|
| 145937 |
+
"epoch": 282.81,
|
| 145938 |
+
"learning_rate": 9.447867528271406e-06,
|
| 145939 |
+
"loss": 0.927,
|
| 145940 |
+
"step": 35190
|
| 145941 |
+
},
|
| 145942 |
+
{
|
| 145943 |
+
"epoch": 282.9,
|
| 145944 |
+
"learning_rate": 9.447705977382876e-06,
|
| 145945 |
+
"loss": 0.5187,
|
| 145946 |
+
"step": 35200
|
| 145947 |
+
},
|
| 145948 |
+
{
|
| 145949 |
+
"epoch": 282.98,
|
| 145950 |
+
"learning_rate": 9.447544426494346e-06,
|
| 145951 |
+
"loss": 0.5656,
|
| 145952 |
+
"step": 35210
|
| 145953 |
+
},
|
| 145954 |
+
{
|
| 145955 |
+
"epoch": 283.0,
|
| 145956 |
+
"eval_loss": 0.49615463614463806,
|
| 145957 |
+
"eval_runtime": 41.7942,
|
| 145958 |
+
"eval_samples_per_second": 20.122,
|
| 145959 |
+
"eval_steps_per_second": 0.646,
|
| 145960 |
+
"eval_wer": 0.21229090909090909,
|
| 145961 |
+
"step": 35213
|
| 145962 |
+
},
|
| 145963 |
+
{
|
| 145964 |
+
"epoch": 281.06,
|
| 145965 |
+
"learning_rate": 9.447382875605816e-06,
|
| 145966 |
+
"loss": 0.3457,
|
| 145967 |
+
"step": 35220
|
| 145968 |
+
},
|
| 145969 |
+
{
|
| 145970 |
+
"epoch": 281.14,
|
| 145971 |
+
"learning_rate": 9.447221324717288e-06,
|
| 145972 |
+
"loss": 0.5896,
|
| 145973 |
+
"step": 35230
|
| 145974 |
+
},
|
| 145975 |
+
{
|
| 145976 |
+
"epoch": 281.22,
|
| 145977 |
+
"learning_rate": 9.447059773828758e-06,
|
| 145978 |
+
"loss": 0.9601,
|
| 145979 |
+
"step": 35240
|
| 145980 |
+
},
|
| 145981 |
+
{
|
| 145982 |
+
"epoch": 281.3,
|
| 145983 |
+
"learning_rate": 9.446898222940228e-06,
|
| 145984 |
+
"loss": 0.3643,
|
| 145985 |
+
"step": 35250
|
| 145986 |
+
},
|
| 145987 |
+
{
|
| 145988 |
+
"epoch": 281.38,
|
| 145989 |
+
"learning_rate": 9.446736672051698e-06,
|
| 145990 |
+
"loss": 0.6521,
|
| 145991 |
+
"step": 35260
|
| 145992 |
+
},
|
| 145993 |
+
{
|
| 145994 |
+
"epoch": 281.46,
|
| 145995 |
+
"learning_rate": 9.446575121163168e-06,
|
| 145996 |
+
"loss": 0.7445,
|
| 145997 |
+
"step": 35270
|
| 145998 |
+
},
|
| 145999 |
+
{
|
| 146000 |
+
"epoch": 281.54,
|
| 146001 |
+
"learning_rate": 9.446413570274637e-06,
|
| 146002 |
+
"loss": 0.3963,
|
| 146003 |
+
"step": 35280
|
| 146004 |
+
},
|
| 146005 |
+
{
|
| 146006 |
+
"epoch": 281.62,
|
| 146007 |
+
"learning_rate": 9.446252019386107e-06,
|
| 146008 |
+
"loss": 1.0211,
|
| 146009 |
+
"step": 35290
|
| 146010 |
+
},
|
| 146011 |
+
{
|
| 146012 |
+
"epoch": 281.7,
|
| 146013 |
+
"learning_rate": 9.446090468497577e-06,
|
| 146014 |
+
"loss": 0.3496,
|
| 146015 |
+
"step": 35300
|
| 146016 |
+
},
|
| 146017 |
+
{
|
| 146018 |
+
"epoch": 281.78,
|
| 146019 |
+
"learning_rate": 9.445928917609047e-06,
|
| 146020 |
+
"loss": 0.5626,
|
| 146021 |
+
"step": 35310
|
| 146022 |
+
},
|
| 146023 |
+
{
|
| 146024 |
+
"epoch": 281.86,
|
| 146025 |
+
"learning_rate": 9.445767366720517e-06,
|
| 146026 |
+
"loss": 0.8037,
|
| 146027 |
+
"step": 35320
|
| 146028 |
+
},
|
| 146029 |
+
{
|
| 146030 |
+
"epoch": 281.94,
|
| 146031 |
+
"learning_rate": 9.445605815831987e-06,
|
| 146032 |
+
"loss": 0.3913,
|
| 146033 |
+
"step": 35330
|
| 146034 |
+
},
|
| 146035 |
+
{
|
| 146036 |
+
"epoch": 282.0,
|
| 146037 |
+
"eval_loss": 0.44793474674224854,
|
| 146038 |
+
"eval_runtime": 39.1099,
|
| 146039 |
+
"eval_samples_per_second": 21.478,
|
| 146040 |
+
"eval_steps_per_second": 0.69,
|
| 146041 |
+
"eval_wer": 0.21840472226426516,
|
| 146042 |
+
"step": 35338
|
| 146043 |
+
},
|
| 146044 |
+
{
|
| 146045 |
+
"epoch": 284.02,
|
| 146046 |
+
"learning_rate": 9.445444264943457e-06,
|
| 146047 |
+
"loss": 0.4271,
|
| 146048 |
+
"step": 35340
|
| 146049 |
+
},
|
| 146050 |
+
{
|
| 146051 |
+
"epoch": 284.1,
|
| 146052 |
+
"learning_rate": 9.445282714054929e-06,
|
| 146053 |
+
"loss": 0.3659,
|
| 146054 |
+
"step": 35350
|
| 146055 |
+
},
|
| 146056 |
+
{
|
| 146057 |
+
"epoch": 284.18,
|
| 146058 |
+
"learning_rate": 9.445121163166399e-06,
|
| 146059 |
+
"loss": 0.5429,
|
| 146060 |
+
"step": 35360
|
| 146061 |
+
},
|
| 146062 |
+
{
|
| 146063 |
+
"epoch": 284.26,
|
| 146064 |
+
"learning_rate": 9.444959612277869e-06,
|
| 146065 |
+
"loss": 0.76,
|
| 146066 |
+
"step": 35370
|
| 146067 |
+
},
|
| 146068 |
+
{
|
| 146069 |
+
"epoch": 284.34,
|
| 146070 |
+
"learning_rate": 9.444798061389339e-06,
|
| 146071 |
+
"loss": 0.452,
|
| 146072 |
+
"step": 35380
|
| 146073 |
+
},
|
| 146074 |
+
{
|
| 146075 |
+
"epoch": 284.42,
|
| 146076 |
+
"learning_rate": 9.444636510500809e-06,
|
| 146077 |
+
"loss": 1.0696,
|
| 146078 |
+
"step": 35390
|
| 146079 |
+
},
|
| 146080 |
+
{
|
| 146081 |
+
"epoch": 284.5,
|
| 146082 |
+
"learning_rate": 9.444474959612279e-06,
|
| 146083 |
+
"loss": 0.4338,
|
| 146084 |
+
"step": 35400
|
| 146085 |
+
},
|
| 146086 |
+
{
|
| 146087 |
+
"epoch": 284.58,
|
| 146088 |
+
"learning_rate": 9.444313408723749e-06,
|
| 146089 |
+
"loss": 0.5894,
|
| 146090 |
+
"step": 35410
|
| 146091 |
+
},
|
| 146092 |
+
{
|
| 146093 |
+
"epoch": 284.66,
|
| 146094 |
+
"learning_rate": 9.444151857835219e-06,
|
| 146095 |
+
"loss": 0.7576,
|
| 146096 |
+
"step": 35420
|
| 146097 |
+
},
|
| 146098 |
+
{
|
| 146099 |
+
"epoch": 284.74,
|
| 146100 |
+
"learning_rate": 9.443990306946689e-06,
|
| 146101 |
+
"loss": 0.3937,
|
| 146102 |
+
"step": 35430
|
| 146103 |
+
},
|
| 146104 |
+
{
|
| 146105 |
+
"epoch": 284.82,
|
| 146106 |
+
"learning_rate": 9.443828756058159e-06,
|
| 146107 |
+
"loss": 0.9792,
|
| 146108 |
+
"step": 35440
|
| 146109 |
+
},
|
| 146110 |
+
{
|
| 146111 |
+
"epoch": 284.9,
|
| 146112 |
+
"learning_rate": 9.443667205169629e-06,
|
| 146113 |
+
"loss": 0.3537,
|
| 146114 |
+
"step": 35450
|
| 146115 |
+
},
|
| 146116 |
+
{
|
| 146117 |
+
"epoch": 284.98,
|
| 146118 |
+
"learning_rate": 9.443505654281099e-06,
|
| 146119 |
+
"loss": 0.6313,
|
| 146120 |
+
"step": 35460
|
| 146121 |
+
},
|
| 146122 |
+
{
|
| 146123 |
+
"epoch": 285.0,
|
| 146124 |
+
"eval_loss": 0.4291652739048004,
|
| 146125 |
+
"eval_runtime": 39.7896,
|
| 146126 |
+
"eval_samples_per_second": 21.111,
|
| 146127 |
+
"eval_steps_per_second": 0.679,
|
| 146128 |
+
"eval_wer": 0.2125739644970414,
|
| 146129 |
+
"step": 35462
|
| 146130 |
+
},
|
| 146131 |
+
{
|
| 146132 |
+
"epoch": 285.06,
|
| 146133 |
+
"learning_rate": 9.44334410339257e-06,
|
| 146134 |
+
"loss": 0.3841,
|
| 146135 |
+
"step": 35470
|
| 146136 |
+
},
|
| 146137 |
+
{
|
| 146138 |
+
"epoch": 285.14,
|
| 146139 |
+
"learning_rate": 9.44318255250404e-06,
|
| 146140 |
+
"loss": 0.3971,
|
| 146141 |
+
"step": 35480
|
| 146142 |
+
},
|
| 146143 |
+
{
|
| 146144 |
+
"epoch": 285.22,
|
| 146145 |
+
"learning_rate": 9.44302100161551e-06,
|
| 146146 |
+
"loss": 0.9095,
|
| 146147 |
+
"step": 35490
|
| 146148 |
+
},
|
| 146149 |
+
{
|
| 146150 |
+
"epoch": 285.31,
|
| 146151 |
+
"learning_rate": 9.44285945072698e-06,
|
| 146152 |
+
"loss": 0.3885,
|
| 146153 |
+
"step": 35500
|
| 146154 |
+
},
|
| 146155 |
+
{
|
| 146156 |
+
"epoch": 285.39,
|
| 146157 |
+
"learning_rate": 9.44269789983845e-06,
|
| 146158 |
+
"loss": 0.6077,
|
| 146159 |
+
"step": 35510
|
| 146160 |
+
},
|
| 146161 |
+
{
|
| 146162 |
+
"epoch": 285.47,
|
| 146163 |
+
"learning_rate": 9.44253634894992e-06,
|
| 146164 |
+
"loss": 0.7422,
|
| 146165 |
+
"step": 35520
|
| 146166 |
+
},
|
| 146167 |
+
{
|
| 146168 |
+
"epoch": 285.55,
|
| 146169 |
+
"learning_rate": 9.44237479806139e-06,
|
| 146170 |
+
"loss": 0.504,
|
| 146171 |
+
"step": 35530
|
| 146172 |
+
},
|
| 146173 |
+
{
|
| 146174 |
+
"epoch": 285.63,
|
| 146175 |
+
"learning_rate": 9.44221324717286e-06,
|
| 146176 |
+
"loss": 0.9686,
|
| 146177 |
+
"step": 35540
|
| 146178 |
+
},
|
| 146179 |
+
{
|
| 146180 |
+
"epoch": 285.71,
|
| 146181 |
+
"learning_rate": 9.44205169628433e-06,
|
| 146182 |
+
"loss": 0.3852,
|
| 146183 |
+
"step": 35550
|
| 146184 |
+
},
|
| 146185 |
+
{
|
| 146186 |
+
"epoch": 285.79,
|
| 146187 |
+
"learning_rate": 9.4418901453958e-06,
|
| 146188 |
+
"loss": 0.6923,
|
| 146189 |
+
"step": 35560
|
| 146190 |
+
},
|
| 146191 |
+
{
|
| 146192 |
+
"epoch": 285.87,
|
| 146193 |
+
"learning_rate": 9.44172859450727e-06,
|
| 146194 |
+
"loss": 0.6688,
|
| 146195 |
+
"step": 35570
|
| 146196 |
+
},
|
| 146197 |
+
{
|
| 146198 |
+
"epoch": 285.95,
|
| 146199 |
+
"learning_rate": 9.441567043618742e-06,
|
| 146200 |
+
"loss": 0.4197,
|
| 146201 |
+
"step": 35580
|
| 146202 |
+
},
|
| 146203 |
+
{
|
| 146204 |
+
"epoch": 286.0,
|
| 146205 |
+
"eval_loss": 0.44989779591560364,
|
| 146206 |
+
"eval_runtime": 41.5608,
|
| 146207 |
+
"eval_samples_per_second": 20.211,
|
| 146208 |
+
"eval_steps_per_second": 0.65,
|
| 146209 |
+
"eval_wer": 0.21492472179794894,
|
| 146210 |
+
"step": 35586
|
| 146211 |
+
},
|
| 146212 |
+
{
|
| 146213 |
+
"epoch": 286.03,
|
| 146214 |
+
"learning_rate": 9.441405492730212e-06,
|
| 146215 |
+
"loss": 0.4786,
|
| 146216 |
+
"step": 35590
|
| 146217 |
+
},
|
| 146218 |
+
{
|
| 146219 |
+
"epoch": 286.11,
|
| 146220 |
+
"learning_rate": 9.441243941841682e-06,
|
| 146221 |
+
"loss": 0.3757,
|
| 146222 |
+
"step": 35600
|
| 146223 |
+
},
|
| 146224 |
+
{
|
| 146225 |
+
"epoch": 286.19,
|
| 146226 |
+
"learning_rate": 9.441082390953152e-06,
|
| 146227 |
+
"loss": 0.8507,
|
| 146228 |
+
"step": 35610
|
| 146229 |
+
},
|
| 146230 |
+
{
|
| 146231 |
+
"epoch": 286.27,
|
| 146232 |
+
"learning_rate": 9.440920840064622e-06,
|
| 146233 |
+
"loss": 0.656,
|
| 146234 |
+
"step": 35620
|
| 146235 |
+
},
|
| 146236 |
+
{
|
| 146237 |
+
"epoch": 286.35,
|
| 146238 |
+
"learning_rate": 9.440759289176092e-06,
|
| 146239 |
+
"loss": 0.4201,
|
| 146240 |
+
"step": 35630
|
| 146241 |
+
},
|
| 146242 |
+
{
|
| 146243 |
+
"epoch": 286.43,
|
| 146244 |
+
"learning_rate": 9.440597738287562e-06,
|
| 146245 |
+
"loss": 0.9412,
|
| 146246 |
+
"step": 35640
|
| 146247 |
+
},
|
| 146248 |
+
{
|
| 146249 |
+
"epoch": 286.51,
|
| 146250 |
+
"learning_rate": 9.440436187399032e-06,
|
| 146251 |
+
"loss": 0.3543,
|
| 146252 |
+
"step": 35650
|
| 146253 |
+
},
|
| 146254 |
+
{
|
| 146255 |
+
"epoch": 286.59,
|
| 146256 |
+
"learning_rate": 9.440274636510502e-06,
|
| 146257 |
+
"loss": 0.7653,
|
| 146258 |
+
"step": 35660
|
| 146259 |
+
},
|
| 146260 |
+
{
|
| 146261 |
+
"epoch": 286.67,
|
| 146262 |
+
"learning_rate": 9.440113085621972e-06,
|
| 146263 |
+
"loss": 0.5552,
|
| 146264 |
+
"step": 35670
|
| 146265 |
+
},
|
| 146266 |
+
{
|
| 146267 |
+
"epoch": 286.76,
|
| 146268 |
+
"learning_rate": 9.439951534733442e-06,
|
| 146269 |
+
"loss": 0.4265,
|
| 146270 |
+
"step": 35680
|
| 146271 |
+
},
|
| 146272 |
+
{
|
| 146273 |
+
"epoch": 286.84,
|
| 146274 |
+
"learning_rate": 9.439789983844912e-06,
|
| 146275 |
+
"loss": 0.982,
|
| 146276 |
+
"step": 35690
|
| 146277 |
+
},
|
| 146278 |
+
{
|
| 146279 |
+
"epoch": 286.92,
|
| 146280 |
+
"learning_rate": 9.439628432956383e-06,
|
| 146281 |
+
"loss": 0.4137,
|
| 146282 |
+
"step": 35700
|
| 146283 |
+
},
|
| 146284 |
+
{
|
| 146285 |
+
"epoch": 287.0,
|
| 146286 |
+
"learning_rate": 9.439466882067853e-06,
|
| 146287 |
+
"loss": 0.9714,
|
| 146288 |
+
"step": 35710
|
| 146289 |
+
},
|
| 146290 |
+
{
|
| 146291 |
+
"epoch": 287.0,
|
| 146292 |
+
"eval_loss": 0.4389137625694275,
|
| 146293 |
+
"eval_runtime": 40.2561,
|
| 146294 |
+
"eval_samples_per_second": 20.866,
|
| 146295 |
+
"eval_steps_per_second": 0.671,
|
| 146296 |
+
"eval_wer": 0.20595987699516766,
|
| 146297 |
+
"step": 35710
|
| 146298 |
}
|
| 146299 |
],
|
| 146300 |
+
"max_steps": 620000,
|
| 146301 |
"num_train_epochs": 5000,
|
| 146302 |
+
"total_flos": 1.004871656432552e+20,
|
| 146303 |
"trial_name": null,
|
| 146304 |
"trial_params": null
|
| 146305 |
}
|
model-bin/finetune/base/{checkpoint-35089 → checkpoint-35710}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629658404.8958828/events.out.tfevents.1629658404.91f414e9573a.903.201
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:79a4c92c34d40f149c8b8173e65a39681f651d68c367ed2f6846413fdc58e95b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629659053.497968/events.out.tfevents.1629659053.91f414e9573a.903.203
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3056d7860df15207738a6c49db08473ab5e3565d6bad0c237b61e3cb1786aaf7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629659698.488905/events.out.tfevents.1629659698.91f414e9573a.903.205
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:063b6328ba3b7bd3d3879dbdee36d733a0e9138b4ec89a26aec8c72c1da33fd1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629660352.110468/events.out.tfevents.1629660352.91f414e9573a.903.207
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b020253a3fa8178ca848b3771d29200e070deef6cf1c8068e251e7244d609768
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629661015.2787066/events.out.tfevents.1629661015.91f414e9573a.903.209
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:04da626bb8e2cd8ff7e485b07b84c6f9840d8eacce0312fe6daccdba9a12e5f0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629658404.91f414e9573a.903.200
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b3c1374fa827771d16221953489994fa234b02ba53c92cdde9209f6379ec01de
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629659053.91f414e9573a.903.202
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c4cc072c64283142cf32bfede61605378e9d83689a2514a32b9c31c725cb3ea
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629659698.91f414e9573a.903.204
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86256b6dcc49725b6114b500aad4bf305bc27bda3767ac91ee1fd71318e474f1
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629660352.91f414e9573a.903.206
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:35ae0ed26c4654e147f64a88d61bc87dc53d71342e787e3313d603b407647e90
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629661015.91f414e9573a.903.208
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb85473744ca507330c0dba76b16d5aec8a424958d57855819e4d35444fc3c0c
|
| 3 |
+
size 6703
|