"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629971133.658557/events.out.tfevents.1629971133.8e89bd551565.924.141 +3 -0
- model-bin/finetune/base/log/1629971573.5504477/events.out.tfevents.1629971573.8e89bd551565.924.143 +3 -0
- model-bin/finetune/base/log/1629972008.5623612/events.out.tfevents.1629972008.8e89bd551565.924.145 +3 -0
- model-bin/finetune/base/log/1629972456.7531004/events.out.tfevents.1629972456.8e89bd551565.924.147 +3 -0
- model-bin/finetune/base/log/1629972904.2005453/events.out.tfevents.1629972904.8e89bd551565.924.149 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629971133.8e89bd551565.924.140 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629971573.8e89bd551565.924.142 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629972006.8e89bd551565.924.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629972456.8e89bd551565.924.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629972904.8e89bd551565.924.148 +3 -0
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/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:fa0edaaaa03517b3f1b3434e7b65727cf52a203ac9559e485dd50109700d00b0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/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 sha256:89fdb8822d04c0e776c6618e1fbb3f7c8f6fc46a341224093650210e6dbf5e95
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/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:fee4da5b6558c360e1c85595c190ce22dc3af5fcb8457a8bff78bcb52d054734
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/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 sha256:1f3fa3d51212d4b89bdf2da68a546d3928c0338347821085cd5a2d6170f0f82e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/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 sha256:5ee3a48684f7142013dfb78e8ee48790be9791cb13103a47502dfd62d8c47af1
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -222870,11 +222870,806 @@
|
|
| 222870 |
"eval_steps_per_second": 0.673,
|
| 222871 |
"eval_wer": 0.1876936697653829,
|
| 222872 |
"step": 97072
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 222873 |
}
|
| 222874 |
],
|
| 222875 |
-
"max_steps":
|
| 222876 |
"num_train_epochs": 5000,
|
| 222877 |
-
"total_flos": 2.
|
| 222878 |
"trial_name": null,
|
| 222879 |
"trial_params": null
|
| 222880 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 786.995983935743,
|
| 5 |
+
"global_step": 97696,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 222870 |
"eval_steps_per_second": 0.673,
|
| 222871 |
"eval_wer": 0.1876936697653829,
|
| 222872 |
"step": 97072
|
| 222873 |
+
},
|
| 222874 |
+
{
|
| 222875 |
+
"epoch": 776.02,
|
| 222876 |
+
"learning_rate": 8.460961538461539e-06,
|
| 222877 |
+
"loss": 0.412,
|
| 222878 |
+
"step": 97075
|
| 222879 |
+
},
|
| 222880 |
+
{
|
| 222881 |
+
"epoch": 776.06,
|
| 222882 |
+
"learning_rate": 8.46088141025641e-06,
|
| 222883 |
+
"loss": 0.2697,
|
| 222884 |
+
"step": 97080
|
| 222885 |
+
},
|
| 222886 |
+
{
|
| 222887 |
+
"epoch": 776.1,
|
| 222888 |
+
"learning_rate": 8.460801282051283e-06,
|
| 222889 |
+
"loss": 0.2728,
|
| 222890 |
+
"step": 97085
|
| 222891 |
+
},
|
| 222892 |
+
{
|
| 222893 |
+
"epoch": 776.14,
|
| 222894 |
+
"learning_rate": 8.460721153846155e-06,
|
| 222895 |
+
"loss": 0.3461,
|
| 222896 |
+
"step": 97090
|
| 222897 |
+
},
|
| 222898 |
+
{
|
| 222899 |
+
"epoch": 776.18,
|
| 222900 |
+
"learning_rate": 8.460641025641026e-06,
|
| 222901 |
+
"loss": 0.7381,
|
| 222902 |
+
"step": 97095
|
| 222903 |
+
},
|
| 222904 |
+
{
|
| 222905 |
+
"epoch": 776.22,
|
| 222906 |
+
"learning_rate": 8.460560897435897e-06,
|
| 222907 |
+
"loss": 1.052,
|
| 222908 |
+
"step": 97100
|
| 222909 |
+
},
|
| 222910 |
+
{
|
| 222911 |
+
"epoch": 776.26,
|
| 222912 |
+
"learning_rate": 8.46048076923077e-06,
|
| 222913 |
+
"loss": 0.2764,
|
| 222914 |
+
"step": 97105
|
| 222915 |
+
},
|
| 222916 |
+
{
|
| 222917 |
+
"epoch": 776.3,
|
| 222918 |
+
"learning_rate": 8.460400641025642e-06,
|
| 222919 |
+
"loss": 0.3675,
|
| 222920 |
+
"step": 97110
|
| 222921 |
+
},
|
| 222922 |
+
{
|
| 222923 |
+
"epoch": 776.34,
|
| 222924 |
+
"learning_rate": 8.460320512820513e-06,
|
| 222925 |
+
"loss": 0.4231,
|
| 222926 |
+
"step": 97115
|
| 222927 |
+
},
|
| 222928 |
+
{
|
| 222929 |
+
"epoch": 776.38,
|
| 222930 |
+
"learning_rate": 8.460240384615386e-06,
|
| 222931 |
+
"loss": 0.6782,
|
| 222932 |
+
"step": 97120
|
| 222933 |
+
},
|
| 222934 |
+
{
|
| 222935 |
+
"epoch": 776.42,
|
| 222936 |
+
"learning_rate": 8.460160256410258e-06,
|
| 222937 |
+
"loss": 0.9475,
|
| 222938 |
+
"step": 97125
|
| 222939 |
+
},
|
| 222940 |
+
{
|
| 222941 |
+
"epoch": 776.46,
|
| 222942 |
+
"learning_rate": 8.460080128205129e-06,
|
| 222943 |
+
"loss": 0.2851,
|
| 222944 |
+
"step": 97130
|
| 222945 |
+
},
|
| 222946 |
+
{
|
| 222947 |
+
"epoch": 776.5,
|
| 222948 |
+
"learning_rate": 8.46e-06,
|
| 222949 |
+
"loss": 0.31,
|
| 222950 |
+
"step": 97135
|
| 222951 |
+
},
|
| 222952 |
+
{
|
| 222953 |
+
"epoch": 776.54,
|
| 222954 |
+
"learning_rate": 8.459919871794873e-06,
|
| 222955 |
+
"loss": 0.4253,
|
| 222956 |
+
"step": 97140
|
| 222957 |
+
},
|
| 222958 |
+
{
|
| 222959 |
+
"epoch": 776.58,
|
| 222960 |
+
"learning_rate": 8.459839743589745e-06,
|
| 222961 |
+
"loss": 0.802,
|
| 222962 |
+
"step": 97145
|
| 222963 |
+
},
|
| 222964 |
+
{
|
| 222965 |
+
"epoch": 776.62,
|
| 222966 |
+
"learning_rate": 8.459759615384616e-06,
|
| 222967 |
+
"loss": 0.8742,
|
| 222968 |
+
"step": 97150
|
| 222969 |
+
},
|
| 222970 |
+
{
|
| 222971 |
+
"epoch": 776.66,
|
| 222972 |
+
"learning_rate": 8.459679487179487e-06,
|
| 222973 |
+
"loss": 0.2785,
|
| 222974 |
+
"step": 97155
|
| 222975 |
+
},
|
| 222976 |
+
{
|
| 222977 |
+
"epoch": 776.7,
|
| 222978 |
+
"learning_rate": 8.45959935897436e-06,
|
| 222979 |
+
"loss": 0.2911,
|
| 222980 |
+
"step": 97160
|
| 222981 |
+
},
|
| 222982 |
+
{
|
| 222983 |
+
"epoch": 776.74,
|
| 222984 |
+
"learning_rate": 8.459519230769232e-06,
|
| 222985 |
+
"loss": 0.3901,
|
| 222986 |
+
"step": 97165
|
| 222987 |
+
},
|
| 222988 |
+
{
|
| 222989 |
+
"epoch": 776.78,
|
| 222990 |
+
"learning_rate": 8.459439102564103e-06,
|
| 222991 |
+
"loss": 0.7846,
|
| 222992 |
+
"step": 97170
|
| 222993 |
+
},
|
| 222994 |
+
{
|
| 222995 |
+
"epoch": 776.82,
|
| 222996 |
+
"learning_rate": 8.459358974358976e-06,
|
| 222997 |
+
"loss": 0.8061,
|
| 222998 |
+
"step": 97175
|
| 222999 |
+
},
|
| 223000 |
+
{
|
| 223001 |
+
"epoch": 776.86,
|
| 223002 |
+
"learning_rate": 8.459278846153846e-06,
|
| 223003 |
+
"loss": 0.3152,
|
| 223004 |
+
"step": 97180
|
| 223005 |
+
},
|
| 223006 |
+
{
|
| 223007 |
+
"epoch": 776.9,
|
| 223008 |
+
"learning_rate": 8.459198717948719e-06,
|
| 223009 |
+
"loss": 0.3018,
|
| 223010 |
+
"step": 97185
|
| 223011 |
+
},
|
| 223012 |
+
{
|
| 223013 |
+
"epoch": 776.94,
|
| 223014 |
+
"learning_rate": 8.45911858974359e-06,
|
| 223015 |
+
"loss": 0.3541,
|
| 223016 |
+
"step": 97190
|
| 223017 |
+
},
|
| 223018 |
+
{
|
| 223019 |
+
"epoch": 776.98,
|
| 223020 |
+
"learning_rate": 8.459038461538462e-06,
|
| 223021 |
+
"loss": 0.8178,
|
| 223022 |
+
"step": 97195
|
| 223023 |
+
},
|
| 223024 |
+
{
|
| 223025 |
+
"epoch": 777.0,
|
| 223026 |
+
"eval_loss": 0.38707292079925537,
|
| 223027 |
+
"eval_runtime": 40.7286,
|
| 223028 |
+
"eval_samples_per_second": 20.624,
|
| 223029 |
+
"eval_steps_per_second": 0.663,
|
| 223030 |
+
"eval_wer": 0.18352890669719518,
|
| 223031 |
+
"step": 97197
|
| 223032 |
+
},
|
| 223033 |
+
{
|
| 223034 |
+
"epoch": 777.02,
|
| 223035 |
+
"learning_rate": 8.458958333333333e-06,
|
| 223036 |
+
"loss": 0.4278,
|
| 223037 |
+
"step": 97200
|
| 223038 |
+
},
|
| 223039 |
+
{
|
| 223040 |
+
"epoch": 777.06,
|
| 223041 |
+
"learning_rate": 8.458878205128206e-06,
|
| 223042 |
+
"loss": 0.265,
|
| 223043 |
+
"step": 97205
|
| 223044 |
+
},
|
| 223045 |
+
{
|
| 223046 |
+
"epoch": 777.1,
|
| 223047 |
+
"learning_rate": 8.458798076923077e-06,
|
| 223048 |
+
"loss": 0.3068,
|
| 223049 |
+
"step": 97210
|
| 223050 |
+
},
|
| 223051 |
+
{
|
| 223052 |
+
"epoch": 777.14,
|
| 223053 |
+
"learning_rate": 8.458717948717949e-06,
|
| 223054 |
+
"loss": 0.333,
|
| 223055 |
+
"step": 97215
|
| 223056 |
+
},
|
| 223057 |
+
{
|
| 223058 |
+
"epoch": 777.18,
|
| 223059 |
+
"learning_rate": 8.458637820512822e-06,
|
| 223060 |
+
"loss": 0.7217,
|
| 223061 |
+
"step": 97220
|
| 223062 |
+
},
|
| 223063 |
+
{
|
| 223064 |
+
"epoch": 777.22,
|
| 223065 |
+
"learning_rate": 8.458557692307693e-06,
|
| 223066 |
+
"loss": 0.8617,
|
| 223067 |
+
"step": 97225
|
| 223068 |
+
},
|
| 223069 |
+
{
|
| 223070 |
+
"epoch": 777.26,
|
| 223071 |
+
"learning_rate": 8.458477564102565e-06,
|
| 223072 |
+
"loss": 0.292,
|
| 223073 |
+
"step": 97230
|
| 223074 |
+
},
|
| 223075 |
+
{
|
| 223076 |
+
"epoch": 777.3,
|
| 223077 |
+
"learning_rate": 8.458397435897436e-06,
|
| 223078 |
+
"loss": 0.2605,
|
| 223079 |
+
"step": 97235
|
| 223080 |
+
},
|
| 223081 |
+
{
|
| 223082 |
+
"epoch": 777.34,
|
| 223083 |
+
"learning_rate": 8.458317307692309e-06,
|
| 223084 |
+
"loss": 0.3936,
|
| 223085 |
+
"step": 97240
|
| 223086 |
+
},
|
| 223087 |
+
{
|
| 223088 |
+
"epoch": 777.38,
|
| 223089 |
+
"learning_rate": 8.45823717948718e-06,
|
| 223090 |
+
"loss": 0.6877,
|
| 223091 |
+
"step": 97245
|
| 223092 |
+
},
|
| 223093 |
+
{
|
| 223094 |
+
"epoch": 777.42,
|
| 223095 |
+
"learning_rate": 8.458157051282052e-06,
|
| 223096 |
+
"loss": 0.8796,
|
| 223097 |
+
"step": 97250
|
| 223098 |
+
},
|
| 223099 |
+
{
|
| 223100 |
+
"epoch": 777.46,
|
| 223101 |
+
"learning_rate": 8.458076923076923e-06,
|
| 223102 |
+
"loss": 0.2542,
|
| 223103 |
+
"step": 97255
|
| 223104 |
+
},
|
| 223105 |
+
{
|
| 223106 |
+
"epoch": 777.5,
|
| 223107 |
+
"learning_rate": 8.457996794871796e-06,
|
| 223108 |
+
"loss": 0.271,
|
| 223109 |
+
"step": 97260
|
| 223110 |
+
},
|
| 223111 |
+
{
|
| 223112 |
+
"epoch": 777.54,
|
| 223113 |
+
"learning_rate": 8.457916666666667e-06,
|
| 223114 |
+
"loss": 0.3593,
|
| 223115 |
+
"step": 97265
|
| 223116 |
+
},
|
| 223117 |
+
{
|
| 223118 |
+
"epoch": 777.58,
|
| 223119 |
+
"learning_rate": 8.457836538461539e-06,
|
| 223120 |
+
"loss": 0.7809,
|
| 223121 |
+
"step": 97270
|
| 223122 |
+
},
|
| 223123 |
+
{
|
| 223124 |
+
"epoch": 777.62,
|
| 223125 |
+
"learning_rate": 8.457756410256412e-06,
|
| 223126 |
+
"loss": 0.8645,
|
| 223127 |
+
"step": 97275
|
| 223128 |
+
},
|
| 223129 |
+
{
|
| 223130 |
+
"epoch": 777.66,
|
| 223131 |
+
"learning_rate": 8.457676282051283e-06,
|
| 223132 |
+
"loss": 0.3124,
|
| 223133 |
+
"step": 97280
|
| 223134 |
+
},
|
| 223135 |
+
{
|
| 223136 |
+
"epoch": 777.7,
|
| 223137 |
+
"learning_rate": 8.457596153846155e-06,
|
| 223138 |
+
"loss": 0.3244,
|
| 223139 |
+
"step": 97285
|
| 223140 |
+
},
|
| 223141 |
+
{
|
| 223142 |
+
"epoch": 777.74,
|
| 223143 |
+
"learning_rate": 8.457516025641026e-06,
|
| 223144 |
+
"loss": 0.3759,
|
| 223145 |
+
"step": 97290
|
| 223146 |
+
},
|
| 223147 |
+
{
|
| 223148 |
+
"epoch": 777.78,
|
| 223149 |
+
"learning_rate": 8.457435897435899e-06,
|
| 223150 |
+
"loss": 0.7656,
|
| 223151 |
+
"step": 97295
|
| 223152 |
+
},
|
| 223153 |
+
{
|
| 223154 |
+
"epoch": 777.82,
|
| 223155 |
+
"learning_rate": 8.457355769230769e-06,
|
| 223156 |
+
"loss": 0.8684,
|
| 223157 |
+
"step": 97300
|
| 223158 |
+
},
|
| 223159 |
+
{
|
| 223160 |
+
"epoch": 777.86,
|
| 223161 |
+
"learning_rate": 8.457275641025642e-06,
|
| 223162 |
+
"loss": 0.2972,
|
| 223163 |
+
"step": 97305
|
| 223164 |
+
},
|
| 223165 |
+
{
|
| 223166 |
+
"epoch": 777.9,
|
| 223167 |
+
"learning_rate": 8.457195512820513e-06,
|
| 223168 |
+
"loss": 0.3105,
|
| 223169 |
+
"step": 97310
|
| 223170 |
+
},
|
| 223171 |
+
{
|
| 223172 |
+
"epoch": 777.94,
|
| 223173 |
+
"learning_rate": 8.457115384615384e-06,
|
| 223174 |
+
"loss": 0.3688,
|
| 223175 |
+
"step": 97315
|
| 223176 |
+
},
|
| 223177 |
+
{
|
| 223178 |
+
"epoch": 777.98,
|
| 223179 |
+
"learning_rate": 8.457035256410258e-06,
|
| 223180 |
+
"loss": 0.696,
|
| 223181 |
+
"step": 97320
|
| 223182 |
+
},
|
| 223183 |
+
{
|
| 223184 |
+
"epoch": 778.0,
|
| 223185 |
+
"eval_loss": 0.3924795985221863,
|
| 223186 |
+
"eval_runtime": 38.7952,
|
| 223187 |
+
"eval_samples_per_second": 21.652,
|
| 223188 |
+
"eval_steps_per_second": 0.696,
|
| 223189 |
+
"eval_wer": 0.19125919515361314,
|
| 223190 |
+
"step": 97322
|
| 223191 |
+
},
|
| 223192 |
+
{
|
| 223193 |
+
"epoch": 778.02,
|
| 223194 |
+
"learning_rate": 8.456955128205129e-06,
|
| 223195 |
+
"loss": 0.3614,
|
| 223196 |
+
"step": 97325
|
| 223197 |
+
},
|
| 223198 |
+
{
|
| 223199 |
+
"epoch": 778.06,
|
| 223200 |
+
"learning_rate": 8.456875e-06,
|
| 223201 |
+
"loss": 0.3109,
|
| 223202 |
+
"step": 97330
|
| 223203 |
+
},
|
| 223204 |
+
{
|
| 223205 |
+
"epoch": 778.1,
|
| 223206 |
+
"learning_rate": 8.456794871794872e-06,
|
| 223207 |
+
"loss": 0.3762,
|
| 223208 |
+
"step": 97335
|
| 223209 |
+
},
|
| 223210 |
+
{
|
| 223211 |
+
"epoch": 778.14,
|
| 223212 |
+
"learning_rate": 8.456714743589745e-06,
|
| 223213 |
+
"loss": 0.36,
|
| 223214 |
+
"step": 97340
|
| 223215 |
+
},
|
| 223216 |
+
{
|
| 223217 |
+
"epoch": 778.18,
|
| 223218 |
+
"learning_rate": 8.456634615384616e-06,
|
| 223219 |
+
"loss": 0.8091,
|
| 223220 |
+
"step": 97345
|
| 223221 |
+
},
|
| 223222 |
+
{
|
| 223223 |
+
"epoch": 778.22,
|
| 223224 |
+
"learning_rate": 8.456554487179487e-06,
|
| 223225 |
+
"loss": 0.8861,
|
| 223226 |
+
"step": 97350
|
| 223227 |
+
},
|
| 223228 |
+
{
|
| 223229 |
+
"epoch": 778.26,
|
| 223230 |
+
"learning_rate": 8.456474358974359e-06,
|
| 223231 |
+
"loss": 0.3471,
|
| 223232 |
+
"step": 97355
|
| 223233 |
+
},
|
| 223234 |
+
{
|
| 223235 |
+
"epoch": 778.3,
|
| 223236 |
+
"learning_rate": 8.456394230769232e-06,
|
| 223237 |
+
"loss": 0.3243,
|
| 223238 |
+
"step": 97360
|
| 223239 |
+
},
|
| 223240 |
+
{
|
| 223241 |
+
"epoch": 778.34,
|
| 223242 |
+
"learning_rate": 8.456314102564103e-06,
|
| 223243 |
+
"loss": 0.4022,
|
| 223244 |
+
"step": 97365
|
| 223245 |
+
},
|
| 223246 |
+
{
|
| 223247 |
+
"epoch": 778.38,
|
| 223248 |
+
"learning_rate": 8.456233974358974e-06,
|
| 223249 |
+
"loss": 0.7714,
|
| 223250 |
+
"step": 97370
|
| 223251 |
+
},
|
| 223252 |
+
{
|
| 223253 |
+
"epoch": 778.42,
|
| 223254 |
+
"learning_rate": 8.456153846153848e-06,
|
| 223255 |
+
"loss": 0.9058,
|
| 223256 |
+
"step": 97375
|
| 223257 |
+
},
|
| 223258 |
+
{
|
| 223259 |
+
"epoch": 778.46,
|
| 223260 |
+
"learning_rate": 8.456073717948719e-06,
|
| 223261 |
+
"loss": 0.2574,
|
| 223262 |
+
"step": 97380
|
| 223263 |
+
},
|
| 223264 |
+
{
|
| 223265 |
+
"epoch": 778.5,
|
| 223266 |
+
"learning_rate": 8.45599358974359e-06,
|
| 223267 |
+
"loss": 0.3452,
|
| 223268 |
+
"step": 97385
|
| 223269 |
+
},
|
| 223270 |
+
{
|
| 223271 |
+
"epoch": 778.54,
|
| 223272 |
+
"learning_rate": 8.455913461538462e-06,
|
| 223273 |
+
"loss": 0.3814,
|
| 223274 |
+
"step": 97390
|
| 223275 |
+
},
|
| 223276 |
+
{
|
| 223277 |
+
"epoch": 778.58,
|
| 223278 |
+
"learning_rate": 8.455833333333335e-06,
|
| 223279 |
+
"loss": 0.807,
|
| 223280 |
+
"step": 97395
|
| 223281 |
+
},
|
| 223282 |
+
{
|
| 223283 |
+
"epoch": 778.62,
|
| 223284 |
+
"learning_rate": 8.455753205128206e-06,
|
| 223285 |
+
"loss": 0.7239,
|
| 223286 |
+
"step": 97400
|
| 223287 |
+
},
|
| 223288 |
+
{
|
| 223289 |
+
"epoch": 778.66,
|
| 223290 |
+
"learning_rate": 8.455673076923077e-06,
|
| 223291 |
+
"loss": 0.3245,
|
| 223292 |
+
"step": 97405
|
| 223293 |
+
},
|
| 223294 |
+
{
|
| 223295 |
+
"epoch": 778.7,
|
| 223296 |
+
"learning_rate": 8.455592948717949e-06,
|
| 223297 |
+
"loss": 0.2848,
|
| 223298 |
+
"step": 97410
|
| 223299 |
+
},
|
| 223300 |
+
{
|
| 223301 |
+
"epoch": 778.74,
|
| 223302 |
+
"learning_rate": 8.455512820512822e-06,
|
| 223303 |
+
"loss": 0.3232,
|
| 223304 |
+
"step": 97415
|
| 223305 |
+
},
|
| 223306 |
+
{
|
| 223307 |
+
"epoch": 778.78,
|
| 223308 |
+
"learning_rate": 8.455432692307693e-06,
|
| 223309 |
+
"loss": 0.8057,
|
| 223310 |
+
"step": 97420
|
| 223311 |
+
},
|
| 223312 |
+
{
|
| 223313 |
+
"epoch": 778.82,
|
| 223314 |
+
"learning_rate": 8.455352564102565e-06,
|
| 223315 |
+
"loss": 0.9114,
|
| 223316 |
+
"step": 97425
|
| 223317 |
+
},
|
| 223318 |
+
{
|
| 223319 |
+
"epoch": 778.86,
|
| 223320 |
+
"learning_rate": 8.455272435897438e-06,
|
| 223321 |
+
"loss": 0.3572,
|
| 223322 |
+
"step": 97430
|
| 223323 |
+
},
|
| 223324 |
+
{
|
| 223325 |
+
"epoch": 778.9,
|
| 223326 |
+
"learning_rate": 8.455192307692309e-06,
|
| 223327 |
+
"loss": 0.4777,
|
| 223328 |
+
"step": 97435
|
| 223329 |
+
},
|
| 223330 |
+
{
|
| 223331 |
+
"epoch": 778.94,
|
| 223332 |
+
"learning_rate": 8.45511217948718e-06,
|
| 223333 |
+
"loss": 0.4918,
|
| 223334 |
+
"step": 97440
|
| 223335 |
+
},
|
| 223336 |
+
{
|
| 223337 |
+
"epoch": 778.98,
|
| 223338 |
+
"learning_rate": 8.455032051282052e-06,
|
| 223339 |
+
"loss": 0.6985,
|
| 223340 |
+
"step": 97445
|
| 223341 |
+
},
|
| 223342 |
+
{
|
| 223343 |
+
"epoch": 779.0,
|
| 223344 |
+
"eval_loss": 0.34349197149276733,
|
| 223345 |
+
"eval_runtime": 37.9448,
|
| 223346 |
+
"eval_samples_per_second": 22.137,
|
| 223347 |
+
"eval_steps_per_second": 0.712,
|
| 223348 |
+
"eval_wer": 0.18190318636265554,
|
| 223349 |
+
"step": 97447
|
| 223350 |
+
},
|
| 223351 |
+
{
|
| 223352 |
+
"epoch": 779.02,
|
| 223353 |
+
"learning_rate": 8.454951923076925e-06,
|
| 223354 |
+
"loss": 0.3218,
|
| 223355 |
+
"step": 97450
|
| 223356 |
+
},
|
| 223357 |
+
{
|
| 223358 |
+
"epoch": 779.06,
|
| 223359 |
+
"learning_rate": 8.454871794871794e-06,
|
| 223360 |
+
"loss": 0.2532,
|
| 223361 |
+
"step": 97455
|
| 223362 |
+
},
|
| 223363 |
+
{
|
| 223364 |
+
"epoch": 779.1,
|
| 223365 |
+
"learning_rate": 8.454791666666667e-06,
|
| 223366 |
+
"loss": 0.3338,
|
| 223367 |
+
"step": 97460
|
| 223368 |
+
},
|
| 223369 |
+
{
|
| 223370 |
+
"epoch": 779.14,
|
| 223371 |
+
"learning_rate": 8.45471153846154e-06,
|
| 223372 |
+
"loss": 0.3953,
|
| 223373 |
+
"step": 97465
|
| 223374 |
+
},
|
| 223375 |
+
{
|
| 223376 |
+
"epoch": 779.18,
|
| 223377 |
+
"learning_rate": 8.45463141025641e-06,
|
| 223378 |
+
"loss": 0.6509,
|
| 223379 |
+
"step": 97470
|
| 223380 |
+
},
|
| 223381 |
+
{
|
| 223382 |
+
"epoch": 779.22,
|
| 223383 |
+
"learning_rate": 8.454551282051283e-06,
|
| 223384 |
+
"loss": 0.8978,
|
| 223385 |
+
"step": 97475
|
| 223386 |
+
},
|
| 223387 |
+
{
|
| 223388 |
+
"epoch": 779.26,
|
| 223389 |
+
"learning_rate": 8.454471153846155e-06,
|
| 223390 |
+
"loss": 0.3133,
|
| 223391 |
+
"step": 97480
|
| 223392 |
+
},
|
| 223393 |
+
{
|
| 223394 |
+
"epoch": 779.3,
|
| 223395 |
+
"learning_rate": 8.454391025641026e-06,
|
| 223396 |
+
"loss": 0.3016,
|
| 223397 |
+
"step": 97485
|
| 223398 |
+
},
|
| 223399 |
+
{
|
| 223400 |
+
"epoch": 779.34,
|
| 223401 |
+
"learning_rate": 8.454310897435897e-06,
|
| 223402 |
+
"loss": 0.35,
|
| 223403 |
+
"step": 97490
|
| 223404 |
+
},
|
| 223405 |
+
{
|
| 223406 |
+
"epoch": 779.38,
|
| 223407 |
+
"learning_rate": 8.45423076923077e-06,
|
| 223408 |
+
"loss": 0.7728,
|
| 223409 |
+
"step": 97495
|
| 223410 |
+
},
|
| 223411 |
+
{
|
| 223412 |
+
"epoch": 779.42,
|
| 223413 |
+
"learning_rate": 8.454150641025642e-06,
|
| 223414 |
+
"loss": 0.755,
|
| 223415 |
+
"step": 97500
|
| 223416 |
+
},
|
| 223417 |
+
{
|
| 223418 |
+
"epoch": 779.46,
|
| 223419 |
+
"learning_rate": 8.454070512820513e-06,
|
| 223420 |
+
"loss": 0.2735,
|
| 223421 |
+
"step": 97505
|
| 223422 |
+
},
|
| 223423 |
+
{
|
| 223424 |
+
"epoch": 779.5,
|
| 223425 |
+
"learning_rate": 8.453990384615384e-06,
|
| 223426 |
+
"loss": 0.3361,
|
| 223427 |
+
"step": 97510
|
| 223428 |
+
},
|
| 223429 |
+
{
|
| 223430 |
+
"epoch": 779.54,
|
| 223431 |
+
"learning_rate": 8.453910256410257e-06,
|
| 223432 |
+
"loss": 0.3407,
|
| 223433 |
+
"step": 97515
|
| 223434 |
+
},
|
| 223435 |
+
{
|
| 223436 |
+
"epoch": 779.58,
|
| 223437 |
+
"learning_rate": 8.453830128205129e-06,
|
| 223438 |
+
"loss": 0.7513,
|
| 223439 |
+
"step": 97520
|
| 223440 |
+
},
|
| 223441 |
+
{
|
| 223442 |
+
"epoch": 779.62,
|
| 223443 |
+
"learning_rate": 8.45375e-06,
|
| 223444 |
+
"loss": 0.914,
|
| 223445 |
+
"step": 97525
|
| 223446 |
+
},
|
| 223447 |
+
{
|
| 223448 |
+
"epoch": 779.66,
|
| 223449 |
+
"learning_rate": 8.453669871794873e-06,
|
| 223450 |
+
"loss": 0.2733,
|
| 223451 |
+
"step": 97530
|
| 223452 |
+
},
|
| 223453 |
+
{
|
| 223454 |
+
"epoch": 779.7,
|
| 223455 |
+
"learning_rate": 8.453589743589745e-06,
|
| 223456 |
+
"loss": 0.2845,
|
| 223457 |
+
"step": 97535
|
| 223458 |
+
},
|
| 223459 |
+
{
|
| 223460 |
+
"epoch": 779.74,
|
| 223461 |
+
"learning_rate": 8.453509615384616e-06,
|
| 223462 |
+
"loss": 0.2958,
|
| 223463 |
+
"step": 97540
|
| 223464 |
+
},
|
| 223465 |
+
{
|
| 223466 |
+
"epoch": 779.78,
|
| 223467 |
+
"learning_rate": 8.453429487179487e-06,
|
| 223468 |
+
"loss": 0.712,
|
| 223469 |
+
"step": 97545
|
| 223470 |
+
},
|
| 223471 |
+
{
|
| 223472 |
+
"epoch": 779.82,
|
| 223473 |
+
"learning_rate": 8.45334935897436e-06,
|
| 223474 |
+
"loss": 0.8571,
|
| 223475 |
+
"step": 97550
|
| 223476 |
+
},
|
| 223477 |
+
{
|
| 223478 |
+
"epoch": 779.86,
|
| 223479 |
+
"learning_rate": 8.453269230769232e-06,
|
| 223480 |
+
"loss": 0.2719,
|
| 223481 |
+
"step": 97555
|
| 223482 |
+
},
|
| 223483 |
+
{
|
| 223484 |
+
"epoch": 779.9,
|
| 223485 |
+
"learning_rate": 8.453189102564103e-06,
|
| 223486 |
+
"loss": 0.2724,
|
| 223487 |
+
"step": 97560
|
| 223488 |
+
},
|
| 223489 |
+
{
|
| 223490 |
+
"epoch": 779.94,
|
| 223491 |
+
"learning_rate": 8.453108974358976e-06,
|
| 223492 |
+
"loss": 0.3582,
|
| 223493 |
+
"step": 97565
|
| 223494 |
+
},
|
| 223495 |
+
{
|
| 223496 |
+
"epoch": 779.98,
|
| 223497 |
+
"learning_rate": 8.453028846153847e-06,
|
| 223498 |
+
"loss": 0.8448,
|
| 223499 |
+
"step": 97570
|
| 223500 |
+
},
|
| 223501 |
+
{
|
| 223502 |
+
"epoch": 780.0,
|
| 223503 |
+
"eval_loss": 0.388048380613327,
|
| 223504 |
+
"eval_runtime": 39.4215,
|
| 223505 |
+
"eval_samples_per_second": 21.334,
|
| 223506 |
+
"eval_steps_per_second": 0.685,
|
| 223507 |
+
"eval_wer": 0.18076225045372052,
|
| 223508 |
+
"step": 97572
|
| 223509 |
+
},
|
| 223510 |
+
{
|
| 223511 |
+
"epoch": 786.02,
|
| 223512 |
+
"learning_rate": 8.452948717948719e-06,
|
| 223513 |
+
"loss": 0.3889,
|
| 223514 |
+
"step": 97575
|
| 223515 |
+
},
|
| 223516 |
+
{
|
| 223517 |
+
"epoch": 786.06,
|
| 223518 |
+
"learning_rate": 8.45286858974359e-06,
|
| 223519 |
+
"loss": 0.3254,
|
| 223520 |
+
"step": 97580
|
| 223521 |
+
},
|
| 223522 |
+
{
|
| 223523 |
+
"epoch": 786.1,
|
| 223524 |
+
"learning_rate": 8.452788461538463e-06,
|
| 223525 |
+
"loss": 0.2856,
|
| 223526 |
+
"step": 97585
|
| 223527 |
+
},
|
| 223528 |
+
{
|
| 223529 |
+
"epoch": 786.14,
|
| 223530 |
+
"learning_rate": 8.452708333333333e-06,
|
| 223531 |
+
"loss": 0.384,
|
| 223532 |
+
"step": 97590
|
| 223533 |
+
},
|
| 223534 |
+
{
|
| 223535 |
+
"epoch": 786.18,
|
| 223536 |
+
"learning_rate": 8.452628205128206e-06,
|
| 223537 |
+
"loss": 0.8424,
|
| 223538 |
+
"step": 97595
|
| 223539 |
+
},
|
| 223540 |
+
{
|
| 223541 |
+
"epoch": 786.22,
|
| 223542 |
+
"learning_rate": 8.452548076923077e-06,
|
| 223543 |
+
"loss": 0.9895,
|
| 223544 |
+
"step": 97600
|
| 223545 |
+
},
|
| 223546 |
+
{
|
| 223547 |
+
"epoch": 786.27,
|
| 223548 |
+
"learning_rate": 8.452467948717949e-06,
|
| 223549 |
+
"loss": 0.2796,
|
| 223550 |
+
"step": 97605
|
| 223551 |
+
},
|
| 223552 |
+
{
|
| 223553 |
+
"epoch": 786.31,
|
| 223554 |
+
"learning_rate": 8.45238782051282e-06,
|
| 223555 |
+
"loss": 0.3126,
|
| 223556 |
+
"step": 97610
|
| 223557 |
+
},
|
| 223558 |
+
{
|
| 223559 |
+
"epoch": 786.35,
|
| 223560 |
+
"learning_rate": 8.452307692307693e-06,
|
| 223561 |
+
"loss": 0.419,
|
| 223562 |
+
"step": 97615
|
| 223563 |
+
},
|
| 223564 |
+
{
|
| 223565 |
+
"epoch": 786.39,
|
| 223566 |
+
"learning_rate": 8.452227564102564e-06,
|
| 223567 |
+
"loss": 0.7767,
|
| 223568 |
+
"step": 97620
|
| 223569 |
+
},
|
| 223570 |
+
{
|
| 223571 |
+
"epoch": 786.43,
|
| 223572 |
+
"learning_rate": 8.452147435897436e-06,
|
| 223573 |
+
"loss": 0.8357,
|
| 223574 |
+
"step": 97625
|
| 223575 |
+
},
|
| 223576 |
+
{
|
| 223577 |
+
"epoch": 786.47,
|
| 223578 |
+
"learning_rate": 8.452067307692309e-06,
|
| 223579 |
+
"loss": 0.3138,
|
| 223580 |
+
"step": 97630
|
| 223581 |
+
},
|
| 223582 |
+
{
|
| 223583 |
+
"epoch": 786.51,
|
| 223584 |
+
"learning_rate": 8.45198717948718e-06,
|
| 223585 |
+
"loss": 0.3118,
|
| 223586 |
+
"step": 97635
|
| 223587 |
+
},
|
| 223588 |
+
{
|
| 223589 |
+
"epoch": 786.55,
|
| 223590 |
+
"learning_rate": 8.451907051282052e-06,
|
| 223591 |
+
"loss": 0.3321,
|
| 223592 |
+
"step": 97640
|
| 223593 |
+
},
|
| 223594 |
+
{
|
| 223595 |
+
"epoch": 786.59,
|
| 223596 |
+
"learning_rate": 8.451826923076923e-06,
|
| 223597 |
+
"loss": 0.7731,
|
| 223598 |
+
"step": 97645
|
| 223599 |
+
},
|
| 223600 |
+
{
|
| 223601 |
+
"epoch": 786.63,
|
| 223602 |
+
"learning_rate": 8.451746794871796e-06,
|
| 223603 |
+
"loss": 0.7596,
|
| 223604 |
+
"step": 97650
|
| 223605 |
+
},
|
| 223606 |
+
{
|
| 223607 |
+
"epoch": 786.67,
|
| 223608 |
+
"learning_rate": 8.451666666666667e-06,
|
| 223609 |
+
"loss": 0.31,
|
| 223610 |
+
"step": 97655
|
| 223611 |
+
},
|
| 223612 |
+
{
|
| 223613 |
+
"epoch": 786.71,
|
| 223614 |
+
"learning_rate": 8.451586538461539e-06,
|
| 223615 |
+
"loss": 0.362,
|
| 223616 |
+
"step": 97660
|
| 223617 |
+
},
|
| 223618 |
+
{
|
| 223619 |
+
"epoch": 786.75,
|
| 223620 |
+
"learning_rate": 8.451506410256412e-06,
|
| 223621 |
+
"loss": 0.3394,
|
| 223622 |
+
"step": 97665
|
| 223623 |
+
},
|
| 223624 |
+
{
|
| 223625 |
+
"epoch": 786.79,
|
| 223626 |
+
"learning_rate": 8.451426282051283e-06,
|
| 223627 |
+
"loss": 0.7482,
|
| 223628 |
+
"step": 97670
|
| 223629 |
+
},
|
| 223630 |
+
{
|
| 223631 |
+
"epoch": 786.83,
|
| 223632 |
+
"learning_rate": 8.451346153846154e-06,
|
| 223633 |
+
"loss": 0.8871,
|
| 223634 |
+
"step": 97675
|
| 223635 |
+
},
|
| 223636 |
+
{
|
| 223637 |
+
"epoch": 786.87,
|
| 223638 |
+
"learning_rate": 8.451266025641026e-06,
|
| 223639 |
+
"loss": 0.3006,
|
| 223640 |
+
"step": 97680
|
| 223641 |
+
},
|
| 223642 |
+
{
|
| 223643 |
+
"epoch": 786.91,
|
| 223644 |
+
"learning_rate": 8.451185897435899e-06,
|
| 223645 |
+
"loss": 0.3371,
|
| 223646 |
+
"step": 97685
|
| 223647 |
+
},
|
| 223648 |
+
{
|
| 223649 |
+
"epoch": 786.95,
|
| 223650 |
+
"learning_rate": 8.45110576923077e-06,
|
| 223651 |
+
"loss": 0.4676,
|
| 223652 |
+
"step": 97690
|
| 223653 |
+
},
|
| 223654 |
+
{
|
| 223655 |
+
"epoch": 786.99,
|
| 223656 |
+
"learning_rate": 8.451025641025642e-06,
|
| 223657 |
+
"loss": 0.9104,
|
| 223658 |
+
"step": 97695
|
| 223659 |
+
},
|
| 223660 |
+
{
|
| 223661 |
+
"epoch": 787.0,
|
| 223662 |
+
"eval_loss": 0.3587617874145508,
|
| 223663 |
+
"eval_runtime": 38.9877,
|
| 223664 |
+
"eval_samples_per_second": 21.545,
|
| 223665 |
+
"eval_steps_per_second": 0.693,
|
| 223666 |
+
"eval_wer": 0.1871849199970775,
|
| 223667 |
+
"step": 97696
|
| 223668 |
}
|
| 223669 |
],
|
| 223670 |
+
"max_steps": 620000,
|
| 223671 |
"num_train_epochs": 5000,
|
| 223672 |
+
"total_flos": 2.7493032883867953e+20,
|
| 223673 |
"trial_name": null,
|
| 223674 |
"trial_params": null
|
| 223675 |
}
|
model-bin/finetune/base/{checkpoint-97072 → checkpoint-97696}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629971133.658557/events.out.tfevents.1629971133.8e89bd551565.924.141
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:486f3035583b3105e248bf839b074df01d64778718f101ddfe068e772f6ade9b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629971573.5504477/events.out.tfevents.1629971573.8e89bd551565.924.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1ca9b1259961315e661d459efcfa01b16514d2bcd4bc47aed4d28e3ac89f019f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629972008.5623612/events.out.tfevents.1629972008.8e89bd551565.924.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b0d7e85e28d53c098dbbc3d31020b677c021f5713d193298dee1a38c7569ebf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629972456.7531004/events.out.tfevents.1629972456.8e89bd551565.924.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c057c32b335239a9df74880121e90b3592e136e7ab8412e2dbdf677dda47889
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629972904.2005453/events.out.tfevents.1629972904.8e89bd551565.924.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c6c9218f3f654d2f84ea4957f73a701f77ed65d4f7340c56671001f6e04389d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629971133.8e89bd551565.924.140
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c47b37185979306ac2dd84d59422a5b8baf336ef23a9c2af61d428bc0b33553c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629971573.8e89bd551565.924.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4b9ec2fc63e44df1c4bfef62b56084b0b5c49de32ab7376fb8e44d0a88e5c1b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629972006.8e89bd551565.924.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ee398d2176e54bc61286eacfa220199ab896d93eec6b97e4d269f8f1183c3fde
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629972456.8e89bd551565.924.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f88897eb50efae9607a5a26ccb8f2b45a0f02589f20d940cb04214f38234960f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629972904.8e89bd551565.924.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44efa315e11e341366ce605aa09324ddfad3d50ca5b56955995eec1addc13132
|
| 3 |
+
size 8622
|