"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629923470.4368432/events.out.tfevents.1629923470.7e498afd5545.7645.145 +3 -0
- model-bin/finetune/base/log/1629923933.9843855/events.out.tfevents.1629923933.7e498afd5545.7645.147 +3 -0
- model-bin/finetune/base/log/1629924482.2018788/events.out.tfevents.1629924482.7e498afd5545.7645.149 +3 -0
- model-bin/finetune/base/log/1629924979.6126964/events.out.tfevents.1629924979.7e498afd5545.7645.151 +3 -0
- model-bin/finetune/base/log/1629925434.4852555/events.out.tfevents.1629925434.7e498afd5545.7645.153 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629923470.7e498afd5545.7645.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629923933.7e498afd5545.7645.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629924482.7e498afd5545.7645.148 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629924979.7e498afd5545.7645.150 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629925434.7e498afd5545.7645.152 +3 -0
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/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 sha256:8cb828e60d114ad9dc3ceadb37368203bd7eecf6bfd146372b381c4c22e7bfd6
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/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 sha256:7210d43a584dc7110bb013457c8bde2d07cac2ead64b2d96061683da7c32196f
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/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 sha256:483bbb4122a05e42b5f275d32a45398659861e698779087c7b6c1925b5ee3053
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/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:d7124f58fe067811067c4ed41a5b8e4aa29a7baa624a2afb0e6cbdd6a7f276b1
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/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 sha256:8a9aaf9f33c3d8265d80f1e4b5cf1209f542c0719612971f697bb6b8e1783bdb
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -207825,11 +207825,800 @@
|
|
| 207825 |
"eval_steps_per_second": 0.652,
|
| 207826 |
"eval_wer": 0.1900681168973852,
|
| 207827 |
"step": 85246
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 207828 |
}
|
| 207829 |
],
|
| 207830 |
-
"max_steps":
|
| 207831 |
"num_train_epochs": 5000,
|
| 207832 |
-
"total_flos": 2.
|
| 207833 |
"trial_name": null,
|
| 207834 |
"trial_params": null
|
| 207835 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 692.0,
|
| 5 |
+
"global_step": 85866,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 207825 |
"eval_steps_per_second": 0.652,
|
| 207826 |
"eval_wer": 0.1900681168973852,
|
| 207827 |
"step": 85246
|
| 207828 |
+
},
|
| 207829 |
+
{
|
| 207830 |
+
"epoch": 687.03,
|
| 207831 |
+
"learning_rate": 8.639483037156706e-06,
|
| 207832 |
+
"loss": 0.3456,
|
| 207833 |
+
"step": 85250
|
| 207834 |
+
},
|
| 207835 |
+
{
|
| 207836 |
+
"epoch": 687.07,
|
| 207837 |
+
"learning_rate": 8.639402261712441e-06,
|
| 207838 |
+
"loss": 0.3289,
|
| 207839 |
+
"step": 85255
|
| 207840 |
+
},
|
| 207841 |
+
{
|
| 207842 |
+
"epoch": 687.11,
|
| 207843 |
+
"learning_rate": 8.639321486268176e-06,
|
| 207844 |
+
"loss": 0.2793,
|
| 207845 |
+
"step": 85260
|
| 207846 |
+
},
|
| 207847 |
+
{
|
| 207848 |
+
"epoch": 687.15,
|
| 207849 |
+
"learning_rate": 8.639240710823911e-06,
|
| 207850 |
+
"loss": 0.4921,
|
| 207851 |
+
"step": 85265
|
| 207852 |
+
},
|
| 207853 |
+
{
|
| 207854 |
+
"epoch": 687.19,
|
| 207855 |
+
"learning_rate": 8.639159935379646e-06,
|
| 207856 |
+
"loss": 0.9617,
|
| 207857 |
+
"step": 85270
|
| 207858 |
+
},
|
| 207859 |
+
{
|
| 207860 |
+
"epoch": 687.23,
|
| 207861 |
+
"learning_rate": 8.639079159935381e-06,
|
| 207862 |
+
"loss": 0.7473,
|
| 207863 |
+
"step": 85275
|
| 207864 |
+
},
|
| 207865 |
+
{
|
| 207866 |
+
"epoch": 687.27,
|
| 207867 |
+
"learning_rate": 8.638998384491115e-06,
|
| 207868 |
+
"loss": 0.3558,
|
| 207869 |
+
"step": 85280
|
| 207870 |
+
},
|
| 207871 |
+
{
|
| 207872 |
+
"epoch": 687.31,
|
| 207873 |
+
"learning_rate": 8.638917609046851e-06,
|
| 207874 |
+
"loss": 0.3373,
|
| 207875 |
+
"step": 85285
|
| 207876 |
+
},
|
| 207877 |
+
{
|
| 207878 |
+
"epoch": 687.35,
|
| 207879 |
+
"learning_rate": 8.638836833602585e-06,
|
| 207880 |
+
"loss": 0.4039,
|
| 207881 |
+
"step": 85290
|
| 207882 |
+
},
|
| 207883 |
+
{
|
| 207884 |
+
"epoch": 687.39,
|
| 207885 |
+
"learning_rate": 8.638756058158321e-06,
|
| 207886 |
+
"loss": 0.8949,
|
| 207887 |
+
"step": 85295
|
| 207888 |
+
},
|
| 207889 |
+
{
|
| 207890 |
+
"epoch": 687.43,
|
| 207891 |
+
"learning_rate": 8.638675282714055e-06,
|
| 207892 |
+
"loss": 0.5874,
|
| 207893 |
+
"step": 85300
|
| 207894 |
+
},
|
| 207895 |
+
{
|
| 207896 |
+
"epoch": 687.47,
|
| 207897 |
+
"learning_rate": 8.638594507269791e-06,
|
| 207898 |
+
"loss": 0.3444,
|
| 207899 |
+
"step": 85305
|
| 207900 |
+
},
|
| 207901 |
+
{
|
| 207902 |
+
"epoch": 687.51,
|
| 207903 |
+
"learning_rate": 8.638513731825525e-06,
|
| 207904 |
+
"loss": 0.3351,
|
| 207905 |
+
"step": 85310
|
| 207906 |
+
},
|
| 207907 |
+
{
|
| 207908 |
+
"epoch": 687.55,
|
| 207909 |
+
"learning_rate": 8.638432956381261e-06,
|
| 207910 |
+
"loss": 0.4849,
|
| 207911 |
+
"step": 85315
|
| 207912 |
+
},
|
| 207913 |
+
{
|
| 207914 |
+
"epoch": 687.59,
|
| 207915 |
+
"learning_rate": 8.638352180936997e-06,
|
| 207916 |
+
"loss": 0.7937,
|
| 207917 |
+
"step": 85320
|
| 207918 |
+
},
|
| 207919 |
+
{
|
| 207920 |
+
"epoch": 687.63,
|
| 207921 |
+
"learning_rate": 8.638271405492731e-06,
|
| 207922 |
+
"loss": 0.7336,
|
| 207923 |
+
"step": 85325
|
| 207924 |
+
},
|
| 207925 |
+
{
|
| 207926 |
+
"epoch": 687.67,
|
| 207927 |
+
"learning_rate": 8.638190630048467e-06,
|
| 207928 |
+
"loss": 0.2841,
|
| 207929 |
+
"step": 85330
|
| 207930 |
+
},
|
| 207931 |
+
{
|
| 207932 |
+
"epoch": 687.71,
|
| 207933 |
+
"learning_rate": 8.638109854604201e-06,
|
| 207934 |
+
"loss": 0.2988,
|
| 207935 |
+
"step": 85335
|
| 207936 |
+
},
|
| 207937 |
+
{
|
| 207938 |
+
"epoch": 687.76,
|
| 207939 |
+
"learning_rate": 8.638029079159937e-06,
|
| 207940 |
+
"loss": 0.4396,
|
| 207941 |
+
"step": 85340
|
| 207942 |
+
},
|
| 207943 |
+
{
|
| 207944 |
+
"epoch": 687.8,
|
| 207945 |
+
"learning_rate": 8.637948303715671e-06,
|
| 207946 |
+
"loss": 0.8903,
|
| 207947 |
+
"step": 85345
|
| 207948 |
+
},
|
| 207949 |
+
{
|
| 207950 |
+
"epoch": 687.84,
|
| 207951 |
+
"learning_rate": 8.637867528271407e-06,
|
| 207952 |
+
"loss": 0.8414,
|
| 207953 |
+
"step": 85350
|
| 207954 |
+
},
|
| 207955 |
+
{
|
| 207956 |
+
"epoch": 687.88,
|
| 207957 |
+
"learning_rate": 8.637786752827141e-06,
|
| 207958 |
+
"loss": 0.4558,
|
| 207959 |
+
"step": 85355
|
| 207960 |
+
},
|
| 207961 |
+
{
|
| 207962 |
+
"epoch": 687.92,
|
| 207963 |
+
"learning_rate": 8.637705977382877e-06,
|
| 207964 |
+
"loss": 0.4324,
|
| 207965 |
+
"step": 85360
|
| 207966 |
+
},
|
| 207967 |
+
{
|
| 207968 |
+
"epoch": 687.96,
|
| 207969 |
+
"learning_rate": 8.637625201938611e-06,
|
| 207970 |
+
"loss": 0.373,
|
| 207971 |
+
"step": 85365
|
| 207972 |
+
},
|
| 207973 |
+
{
|
| 207974 |
+
"epoch": 688.0,
|
| 207975 |
+
"learning_rate": 8.637544426494347e-06,
|
| 207976 |
+
"loss": 1.1997,
|
| 207977 |
+
"step": 85370
|
| 207978 |
+
},
|
| 207979 |
+
{
|
| 207980 |
+
"epoch": 688.0,
|
| 207981 |
+
"eval_loss": 0.38581112027168274,
|
| 207982 |
+
"eval_runtime": 41.9661,
|
| 207983 |
+
"eval_samples_per_second": 20.016,
|
| 207984 |
+
"eval_steps_per_second": 0.643,
|
| 207985 |
+
"eval_wer": 0.18613762644251317,
|
| 207986 |
+
"step": 85370
|
| 207987 |
+
},
|
| 207988 |
+
{
|
| 207989 |
+
"epoch": 688.04,
|
| 207990 |
+
"learning_rate": 8.637463651050083e-06,
|
| 207991 |
+
"loss": 0.3324,
|
| 207992 |
+
"step": 85375
|
| 207993 |
+
},
|
| 207994 |
+
{
|
| 207995 |
+
"epoch": 688.08,
|
| 207996 |
+
"learning_rate": 8.637382875605817e-06,
|
| 207997 |
+
"loss": 0.3157,
|
| 207998 |
+
"step": 85380
|
| 207999 |
+
},
|
| 208000 |
+
{
|
| 208001 |
+
"epoch": 688.12,
|
| 208002 |
+
"learning_rate": 8.637302100161553e-06,
|
| 208003 |
+
"loss": 0.3456,
|
| 208004 |
+
"step": 85385
|
| 208005 |
+
},
|
| 208006 |
+
{
|
| 208007 |
+
"epoch": 688.16,
|
| 208008 |
+
"learning_rate": 8.637221324717287e-06,
|
| 208009 |
+
"loss": 0.4304,
|
| 208010 |
+
"step": 85390
|
| 208011 |
+
},
|
| 208012 |
+
{
|
| 208013 |
+
"epoch": 688.2,
|
| 208014 |
+
"learning_rate": 8.637140549273023e-06,
|
| 208015 |
+
"loss": 1.0546,
|
| 208016 |
+
"step": 85395
|
| 208017 |
+
},
|
| 208018 |
+
{
|
| 208019 |
+
"epoch": 688.24,
|
| 208020 |
+
"learning_rate": 8.637059773828757e-06,
|
| 208021 |
+
"loss": 0.3361,
|
| 208022 |
+
"step": 85400
|
| 208023 |
+
},
|
| 208024 |
+
{
|
| 208025 |
+
"epoch": 688.28,
|
| 208026 |
+
"learning_rate": 8.636978998384493e-06,
|
| 208027 |
+
"loss": 0.2779,
|
| 208028 |
+
"step": 85405
|
| 208029 |
+
},
|
| 208030 |
+
{
|
| 208031 |
+
"epoch": 688.32,
|
| 208032 |
+
"learning_rate": 8.636898222940227e-06,
|
| 208033 |
+
"loss": 0.3086,
|
| 208034 |
+
"step": 85410
|
| 208035 |
+
},
|
| 208036 |
+
{
|
| 208037 |
+
"epoch": 688.36,
|
| 208038 |
+
"learning_rate": 8.636817447495963e-06,
|
| 208039 |
+
"loss": 0.4391,
|
| 208040 |
+
"step": 85415
|
| 208041 |
+
},
|
| 208042 |
+
{
|
| 208043 |
+
"epoch": 688.4,
|
| 208044 |
+
"learning_rate": 8.636736672051697e-06,
|
| 208045 |
+
"loss": 1.2883,
|
| 208046 |
+
"step": 85420
|
| 208047 |
+
},
|
| 208048 |
+
{
|
| 208049 |
+
"epoch": 688.44,
|
| 208050 |
+
"learning_rate": 8.636655896607433e-06,
|
| 208051 |
+
"loss": 0.3306,
|
| 208052 |
+
"step": 85425
|
| 208053 |
+
},
|
| 208054 |
+
{
|
| 208055 |
+
"epoch": 688.48,
|
| 208056 |
+
"learning_rate": 8.636575121163167e-06,
|
| 208057 |
+
"loss": 0.3151,
|
| 208058 |
+
"step": 85430
|
| 208059 |
+
},
|
| 208060 |
+
{
|
| 208061 |
+
"epoch": 688.52,
|
| 208062 |
+
"learning_rate": 8.636494345718903e-06,
|
| 208063 |
+
"loss": 0.3301,
|
| 208064 |
+
"step": 85435
|
| 208065 |
+
},
|
| 208066 |
+
{
|
| 208067 |
+
"epoch": 688.56,
|
| 208068 |
+
"learning_rate": 8.636413570274638e-06,
|
| 208069 |
+
"loss": 0.4831,
|
| 208070 |
+
"step": 85440
|
| 208071 |
+
},
|
| 208072 |
+
{
|
| 208073 |
+
"epoch": 688.6,
|
| 208074 |
+
"learning_rate": 8.636332794830373e-06,
|
| 208075 |
+
"loss": 1.2196,
|
| 208076 |
+
"step": 85445
|
| 208077 |
+
},
|
| 208078 |
+
{
|
| 208079 |
+
"epoch": 688.64,
|
| 208080 |
+
"learning_rate": 8.636252019386108e-06,
|
| 208081 |
+
"loss": 0.3267,
|
| 208082 |
+
"step": 85450
|
| 208083 |
+
},
|
| 208084 |
+
{
|
| 208085 |
+
"epoch": 688.68,
|
| 208086 |
+
"learning_rate": 8.636171243941843e-06,
|
| 208087 |
+
"loss": 0.3363,
|
| 208088 |
+
"step": 85455
|
| 208089 |
+
},
|
| 208090 |
+
{
|
| 208091 |
+
"epoch": 688.72,
|
| 208092 |
+
"learning_rate": 8.636090468497578e-06,
|
| 208093 |
+
"loss": 0.3013,
|
| 208094 |
+
"step": 85460
|
| 208095 |
+
},
|
| 208096 |
+
{
|
| 208097 |
+
"epoch": 688.76,
|
| 208098 |
+
"learning_rate": 8.636009693053313e-06,
|
| 208099 |
+
"loss": 0.5165,
|
| 208100 |
+
"step": 85465
|
| 208101 |
+
},
|
| 208102 |
+
{
|
| 208103 |
+
"epoch": 688.8,
|
| 208104 |
+
"learning_rate": 8.635928917609048e-06,
|
| 208105 |
+
"loss": 1.227,
|
| 208106 |
+
"step": 85470
|
| 208107 |
+
},
|
| 208108 |
+
{
|
| 208109 |
+
"epoch": 688.84,
|
| 208110 |
+
"learning_rate": 8.635848142164783e-06,
|
| 208111 |
+
"loss": 0.354,
|
| 208112 |
+
"step": 85475
|
| 208113 |
+
},
|
| 208114 |
+
{
|
| 208115 |
+
"epoch": 688.88,
|
| 208116 |
+
"learning_rate": 8.635767366720518e-06,
|
| 208117 |
+
"loss": 0.2623,
|
| 208118 |
+
"step": 85480
|
| 208119 |
+
},
|
| 208120 |
+
{
|
| 208121 |
+
"epoch": 688.92,
|
| 208122 |
+
"learning_rate": 8.635686591276253e-06,
|
| 208123 |
+
"loss": 0.3408,
|
| 208124 |
+
"step": 85485
|
| 208125 |
+
},
|
| 208126 |
+
{
|
| 208127 |
+
"epoch": 688.96,
|
| 208128 |
+
"learning_rate": 8.635605815831988e-06,
|
| 208129 |
+
"loss": 0.4418,
|
| 208130 |
+
"step": 85490
|
| 208131 |
+
},
|
| 208132 |
+
{
|
| 208133 |
+
"epoch": 689.0,
|
| 208134 |
+
"eval_loss": 0.3666008412837982,
|
| 208135 |
+
"eval_runtime": 43.6666,
|
| 208136 |
+
"eval_samples_per_second": 19.237,
|
| 208137 |
+
"eval_steps_per_second": 0.618,
|
| 208138 |
+
"eval_wer": 0.19521510412913454,
|
| 208139 |
+
"step": 85494
|
| 208140 |
+
},
|
| 208141 |
+
{
|
| 208142 |
+
"epoch": 689.01,
|
| 208143 |
+
"learning_rate": 8.635525040387722e-06,
|
| 208144 |
+
"loss": 0.5303,
|
| 208145 |
+
"step": 85495
|
| 208146 |
+
},
|
| 208147 |
+
{
|
| 208148 |
+
"epoch": 689.05,
|
| 208149 |
+
"learning_rate": 8.635444264943458e-06,
|
| 208150 |
+
"loss": 0.3356,
|
| 208151 |
+
"step": 85500
|
| 208152 |
+
},
|
| 208153 |
+
{
|
| 208154 |
+
"epoch": 689.09,
|
| 208155 |
+
"learning_rate": 8.635363489499194e-06,
|
| 208156 |
+
"loss": 0.2876,
|
| 208157 |
+
"step": 85505
|
| 208158 |
+
},
|
| 208159 |
+
{
|
| 208160 |
+
"epoch": 689.13,
|
| 208161 |
+
"learning_rate": 8.635282714054928e-06,
|
| 208162 |
+
"loss": 0.3518,
|
| 208163 |
+
"step": 85510
|
| 208164 |
+
},
|
| 208165 |
+
{
|
| 208166 |
+
"epoch": 689.17,
|
| 208167 |
+
"learning_rate": 8.635201938610664e-06,
|
| 208168 |
+
"loss": 0.6885,
|
| 208169 |
+
"step": 85515
|
| 208170 |
+
},
|
| 208171 |
+
{
|
| 208172 |
+
"epoch": 689.21,
|
| 208173 |
+
"learning_rate": 8.635121163166398e-06,
|
| 208174 |
+
"loss": 1.1915,
|
| 208175 |
+
"step": 85520
|
| 208176 |
+
},
|
| 208177 |
+
{
|
| 208178 |
+
"epoch": 689.25,
|
| 208179 |
+
"learning_rate": 8.635040387722134e-06,
|
| 208180 |
+
"loss": 0.3036,
|
| 208181 |
+
"step": 85525
|
| 208182 |
+
},
|
| 208183 |
+
{
|
| 208184 |
+
"epoch": 689.29,
|
| 208185 |
+
"learning_rate": 8.634959612277868e-06,
|
| 208186 |
+
"loss": 0.3395,
|
| 208187 |
+
"step": 85530
|
| 208188 |
+
},
|
| 208189 |
+
{
|
| 208190 |
+
"epoch": 689.33,
|
| 208191 |
+
"learning_rate": 8.634878836833604e-06,
|
| 208192 |
+
"loss": 0.3868,
|
| 208193 |
+
"step": 85535
|
| 208194 |
+
},
|
| 208195 |
+
{
|
| 208196 |
+
"epoch": 689.37,
|
| 208197 |
+
"learning_rate": 8.634798061389338e-06,
|
| 208198 |
+
"loss": 0.5856,
|
| 208199 |
+
"step": 85540
|
| 208200 |
+
},
|
| 208201 |
+
{
|
| 208202 |
+
"epoch": 689.41,
|
| 208203 |
+
"learning_rate": 8.634717285945074e-06,
|
| 208204 |
+
"loss": 1.1731,
|
| 208205 |
+
"step": 85545
|
| 208206 |
+
},
|
| 208207 |
+
{
|
| 208208 |
+
"epoch": 689.45,
|
| 208209 |
+
"learning_rate": 8.634636510500808e-06,
|
| 208210 |
+
"loss": 0.3879,
|
| 208211 |
+
"step": 85550
|
| 208212 |
+
},
|
| 208213 |
+
{
|
| 208214 |
+
"epoch": 689.49,
|
| 208215 |
+
"learning_rate": 8.634555735056544e-06,
|
| 208216 |
+
"loss": 0.3396,
|
| 208217 |
+
"step": 85555
|
| 208218 |
+
},
|
| 208219 |
+
{
|
| 208220 |
+
"epoch": 689.53,
|
| 208221 |
+
"learning_rate": 8.634474959612278e-06,
|
| 208222 |
+
"loss": 0.3003,
|
| 208223 |
+
"step": 85560
|
| 208224 |
+
},
|
| 208225 |
+
{
|
| 208226 |
+
"epoch": 689.57,
|
| 208227 |
+
"learning_rate": 8.634394184168014e-06,
|
| 208228 |
+
"loss": 0.485,
|
| 208229 |
+
"step": 85565
|
| 208230 |
+
},
|
| 208231 |
+
{
|
| 208232 |
+
"epoch": 689.61,
|
| 208233 |
+
"learning_rate": 8.63431340872375e-06,
|
| 208234 |
+
"loss": 1.1631,
|
| 208235 |
+
"step": 85570
|
| 208236 |
+
},
|
| 208237 |
+
{
|
| 208238 |
+
"epoch": 689.65,
|
| 208239 |
+
"learning_rate": 8.634232633279484e-06,
|
| 208240 |
+
"loss": 0.2812,
|
| 208241 |
+
"step": 85575
|
| 208242 |
+
},
|
| 208243 |
+
{
|
| 208244 |
+
"epoch": 689.69,
|
| 208245 |
+
"learning_rate": 8.63415185783522e-06,
|
| 208246 |
+
"loss": 0.3453,
|
| 208247 |
+
"step": 85580
|
| 208248 |
+
},
|
| 208249 |
+
{
|
| 208250 |
+
"epoch": 689.73,
|
| 208251 |
+
"learning_rate": 8.634071082390954e-06,
|
| 208252 |
+
"loss": 0.3241,
|
| 208253 |
+
"step": 85585
|
| 208254 |
+
},
|
| 208255 |
+
{
|
| 208256 |
+
"epoch": 689.77,
|
| 208257 |
+
"learning_rate": 8.63399030694669e-06,
|
| 208258 |
+
"loss": 0.5035,
|
| 208259 |
+
"step": 85590
|
| 208260 |
+
},
|
| 208261 |
+
{
|
| 208262 |
+
"epoch": 689.81,
|
| 208263 |
+
"learning_rate": 8.633909531502424e-06,
|
| 208264 |
+
"loss": 1.1485,
|
| 208265 |
+
"step": 85595
|
| 208266 |
+
},
|
| 208267 |
+
{
|
| 208268 |
+
"epoch": 689.85,
|
| 208269 |
+
"learning_rate": 8.63382875605816e-06,
|
| 208270 |
+
"loss": 0.3099,
|
| 208271 |
+
"step": 85600
|
| 208272 |
+
},
|
| 208273 |
+
{
|
| 208274 |
+
"epoch": 689.89,
|
| 208275 |
+
"learning_rate": 8.633747980613894e-06,
|
| 208276 |
+
"loss": 0.3697,
|
| 208277 |
+
"step": 85605
|
| 208278 |
+
},
|
| 208279 |
+
{
|
| 208280 |
+
"epoch": 689.93,
|
| 208281 |
+
"learning_rate": 8.63366720516963e-06,
|
| 208282 |
+
"loss": 0.5265,
|
| 208283 |
+
"step": 85610
|
| 208284 |
+
},
|
| 208285 |
+
{
|
| 208286 |
+
"epoch": 689.97,
|
| 208287 |
+
"learning_rate": 8.633586429725364e-06,
|
| 208288 |
+
"loss": 0.6628,
|
| 208289 |
+
"step": 85615
|
| 208290 |
+
},
|
| 208291 |
+
{
|
| 208292 |
+
"epoch": 690.0,
|
| 208293 |
+
"eval_loss": 0.3753606975078583,
|
| 208294 |
+
"eval_runtime": 43.1551,
|
| 208295 |
+
"eval_samples_per_second": 19.465,
|
| 208296 |
+
"eval_steps_per_second": 0.626,
|
| 208297 |
+
"eval_wer": 0.18599016080475805,
|
| 208298 |
+
"step": 85618
|
| 208299 |
+
},
|
| 208300 |
+
{
|
| 208301 |
+
"epoch": 690.02,
|
| 208302 |
+
"learning_rate": 8.6335056542811e-06,
|
| 208303 |
+
"loss": 0.3795,
|
| 208304 |
+
"step": 85620
|
| 208305 |
+
},
|
| 208306 |
+
{
|
| 208307 |
+
"epoch": 690.06,
|
| 208308 |
+
"learning_rate": 8.633424878836834e-06,
|
| 208309 |
+
"loss": 0.3151,
|
| 208310 |
+
"step": 85625
|
| 208311 |
+
},
|
| 208312 |
+
{
|
| 208313 |
+
"epoch": 690.1,
|
| 208314 |
+
"learning_rate": 8.63334410339257e-06,
|
| 208315 |
+
"loss": 0.3625,
|
| 208316 |
+
"step": 85630
|
| 208317 |
+
},
|
| 208318 |
+
{
|
| 208319 |
+
"epoch": 690.14,
|
| 208320 |
+
"learning_rate": 8.633263327948306e-06,
|
| 208321 |
+
"loss": 0.39,
|
| 208322 |
+
"step": 85635
|
| 208323 |
+
},
|
| 208324 |
+
{
|
| 208325 |
+
"epoch": 690.18,
|
| 208326 |
+
"learning_rate": 8.63318255250404e-06,
|
| 208327 |
+
"loss": 0.6459,
|
| 208328 |
+
"step": 85640
|
| 208329 |
+
},
|
| 208330 |
+
{
|
| 208331 |
+
"epoch": 690.22,
|
| 208332 |
+
"learning_rate": 8.633101777059776e-06,
|
| 208333 |
+
"loss": 0.9562,
|
| 208334 |
+
"step": 85645
|
| 208335 |
+
},
|
| 208336 |
+
{
|
| 208337 |
+
"epoch": 690.26,
|
| 208338 |
+
"learning_rate": 8.63302100161551e-06,
|
| 208339 |
+
"loss": 0.3422,
|
| 208340 |
+
"step": 85650
|
| 208341 |
+
},
|
| 208342 |
+
{
|
| 208343 |
+
"epoch": 690.3,
|
| 208344 |
+
"learning_rate": 8.632940226171245e-06,
|
| 208345 |
+
"loss": 0.3108,
|
| 208346 |
+
"step": 85655
|
| 208347 |
+
},
|
| 208348 |
+
{
|
| 208349 |
+
"epoch": 690.34,
|
| 208350 |
+
"learning_rate": 8.63285945072698e-06,
|
| 208351 |
+
"loss": 0.3851,
|
| 208352 |
+
"step": 85660
|
| 208353 |
+
},
|
| 208354 |
+
{
|
| 208355 |
+
"epoch": 690.38,
|
| 208356 |
+
"learning_rate": 8.632778675282715e-06,
|
| 208357 |
+
"loss": 0.5736,
|
| 208358 |
+
"step": 85665
|
| 208359 |
+
},
|
| 208360 |
+
{
|
| 208361 |
+
"epoch": 690.42,
|
| 208362 |
+
"learning_rate": 8.63269789983845e-06,
|
| 208363 |
+
"loss": 1.3345,
|
| 208364 |
+
"step": 85670
|
| 208365 |
+
},
|
| 208366 |
+
{
|
| 208367 |
+
"epoch": 690.46,
|
| 208368 |
+
"learning_rate": 8.632617124394185e-06,
|
| 208369 |
+
"loss": 0.3083,
|
| 208370 |
+
"step": 85675
|
| 208371 |
+
},
|
| 208372 |
+
{
|
| 208373 |
+
"epoch": 690.5,
|
| 208374 |
+
"learning_rate": 8.63253634894992e-06,
|
| 208375 |
+
"loss": 0.341,
|
| 208376 |
+
"step": 85680
|
| 208377 |
+
},
|
| 208378 |
+
{
|
| 208379 |
+
"epoch": 690.54,
|
| 208380 |
+
"learning_rate": 8.632455573505655e-06,
|
| 208381 |
+
"loss": 0.355,
|
| 208382 |
+
"step": 85685
|
| 208383 |
+
},
|
| 208384 |
+
{
|
| 208385 |
+
"epoch": 690.58,
|
| 208386 |
+
"learning_rate": 8.63237479806139e-06,
|
| 208387 |
+
"loss": 0.6607,
|
| 208388 |
+
"step": 85690
|
| 208389 |
+
},
|
| 208390 |
+
{
|
| 208391 |
+
"epoch": 690.62,
|
| 208392 |
+
"learning_rate": 8.632294022617125e-06,
|
| 208393 |
+
"loss": 1.0845,
|
| 208394 |
+
"step": 85695
|
| 208395 |
+
},
|
| 208396 |
+
{
|
| 208397 |
+
"epoch": 690.66,
|
| 208398 |
+
"learning_rate": 8.63221324717286e-06,
|
| 208399 |
+
"loss": 0.2937,
|
| 208400 |
+
"step": 85700
|
| 208401 |
+
},
|
| 208402 |
+
{
|
| 208403 |
+
"epoch": 690.7,
|
| 208404 |
+
"learning_rate": 8.632132471728595e-06,
|
| 208405 |
+
"loss": 0.3011,
|
| 208406 |
+
"step": 85705
|
| 208407 |
+
},
|
| 208408 |
+
{
|
| 208409 |
+
"epoch": 690.74,
|
| 208410 |
+
"learning_rate": 8.632051696284331e-06,
|
| 208411 |
+
"loss": 0.4066,
|
| 208412 |
+
"step": 85710
|
| 208413 |
+
},
|
| 208414 |
+
{
|
| 208415 |
+
"epoch": 690.78,
|
| 208416 |
+
"learning_rate": 8.631970920840065e-06,
|
| 208417 |
+
"loss": 0.6457,
|
| 208418 |
+
"step": 85715
|
| 208419 |
+
},
|
| 208420 |
+
{
|
| 208421 |
+
"epoch": 690.82,
|
| 208422 |
+
"learning_rate": 8.631890145395801e-06,
|
| 208423 |
+
"loss": 1.1609,
|
| 208424 |
+
"step": 85720
|
| 208425 |
+
},
|
| 208426 |
+
{
|
| 208427 |
+
"epoch": 690.86,
|
| 208428 |
+
"learning_rate": 8.631809369951535e-06,
|
| 208429 |
+
"loss": 0.2494,
|
| 208430 |
+
"step": 85725
|
| 208431 |
+
},
|
| 208432 |
+
{
|
| 208433 |
+
"epoch": 690.9,
|
| 208434 |
+
"learning_rate": 8.631728594507271e-06,
|
| 208435 |
+
"loss": 0.3139,
|
| 208436 |
+
"step": 85730
|
| 208437 |
+
},
|
| 208438 |
+
{
|
| 208439 |
+
"epoch": 690.94,
|
| 208440 |
+
"learning_rate": 8.631647819063005e-06,
|
| 208441 |
+
"loss": 0.4865,
|
| 208442 |
+
"step": 85735
|
| 208443 |
+
},
|
| 208444 |
+
{
|
| 208445 |
+
"epoch": 690.98,
|
| 208446 |
+
"learning_rate": 8.631567043618741e-06,
|
| 208447 |
+
"loss": 0.8141,
|
| 208448 |
+
"step": 85740
|
| 208449 |
+
},
|
| 208450 |
+
{
|
| 208451 |
+
"epoch": 691.0,
|
| 208452 |
+
"eval_loss": 0.38146573305130005,
|
| 208453 |
+
"eval_runtime": 40.5464,
|
| 208454 |
+
"eval_samples_per_second": 20.717,
|
| 208455 |
+
"eval_steps_per_second": 0.666,
|
| 208456 |
+
"eval_wer": 0.1951780415430267,
|
| 208457 |
+
"step": 85742
|
| 208458 |
+
},
|
| 208459 |
+
{
|
| 208460 |
+
"epoch": 691.02,
|
| 208461 |
+
"learning_rate": 8.631486268174475e-06,
|
| 208462 |
+
"loss": 0.3553,
|
| 208463 |
+
"step": 85745
|
| 208464 |
+
},
|
| 208465 |
+
{
|
| 208466 |
+
"epoch": 691.06,
|
| 208467 |
+
"learning_rate": 8.631405492730211e-06,
|
| 208468 |
+
"loss": 0.2489,
|
| 208469 |
+
"step": 85750
|
| 208470 |
+
},
|
| 208471 |
+
{
|
| 208472 |
+
"epoch": 691.1,
|
| 208473 |
+
"learning_rate": 8.631324717285945e-06,
|
| 208474 |
+
"loss": 0.375,
|
| 208475 |
+
"step": 85755
|
| 208476 |
+
},
|
| 208477 |
+
{
|
| 208478 |
+
"epoch": 691.15,
|
| 208479 |
+
"learning_rate": 8.631243941841681e-06,
|
| 208480 |
+
"loss": 0.3773,
|
| 208481 |
+
"step": 85760
|
| 208482 |
+
},
|
| 208483 |
+
{
|
| 208484 |
+
"epoch": 691.19,
|
| 208485 |
+
"learning_rate": 8.631163166397415e-06,
|
| 208486 |
+
"loss": 0.7523,
|
| 208487 |
+
"step": 85765
|
| 208488 |
+
},
|
| 208489 |
+
{
|
| 208490 |
+
"epoch": 691.23,
|
| 208491 |
+
"learning_rate": 8.631082390953151e-06,
|
| 208492 |
+
"loss": 0.848,
|
| 208493 |
+
"step": 85770
|
| 208494 |
+
},
|
| 208495 |
+
{
|
| 208496 |
+
"epoch": 691.27,
|
| 208497 |
+
"learning_rate": 8.631001615508887e-06,
|
| 208498 |
+
"loss": 0.292,
|
| 208499 |
+
"step": 85775
|
| 208500 |
+
},
|
| 208501 |
+
{
|
| 208502 |
+
"epoch": 691.31,
|
| 208503 |
+
"learning_rate": 8.630920840064621e-06,
|
| 208504 |
+
"loss": 0.2713,
|
| 208505 |
+
"step": 85780
|
| 208506 |
+
},
|
| 208507 |
+
{
|
| 208508 |
+
"epoch": 691.35,
|
| 208509 |
+
"learning_rate": 8.630840064620357e-06,
|
| 208510 |
+
"loss": 0.3566,
|
| 208511 |
+
"step": 85785
|
| 208512 |
+
},
|
| 208513 |
+
{
|
| 208514 |
+
"epoch": 691.39,
|
| 208515 |
+
"learning_rate": 8.630759289176091e-06,
|
| 208516 |
+
"loss": 0.5913,
|
| 208517 |
+
"step": 85790
|
| 208518 |
+
},
|
| 208519 |
+
{
|
| 208520 |
+
"epoch": 691.43,
|
| 208521 |
+
"learning_rate": 8.630678513731827e-06,
|
| 208522 |
+
"loss": 0.9146,
|
| 208523 |
+
"step": 85795
|
| 208524 |
+
},
|
| 208525 |
+
{
|
| 208526 |
+
"epoch": 691.47,
|
| 208527 |
+
"learning_rate": 8.630597738287561e-06,
|
| 208528 |
+
"loss": 0.329,
|
| 208529 |
+
"step": 85800
|
| 208530 |
+
},
|
| 208531 |
+
{
|
| 208532 |
+
"epoch": 691.51,
|
| 208533 |
+
"learning_rate": 8.630516962843297e-06,
|
| 208534 |
+
"loss": 0.2984,
|
| 208535 |
+
"step": 85805
|
| 208536 |
+
},
|
| 208537 |
+
{
|
| 208538 |
+
"epoch": 691.55,
|
| 208539 |
+
"learning_rate": 8.630436187399031e-06,
|
| 208540 |
+
"loss": 0.3816,
|
| 208541 |
+
"step": 85810
|
| 208542 |
+
},
|
| 208543 |
+
{
|
| 208544 |
+
"epoch": 691.59,
|
| 208545 |
+
"learning_rate": 8.630355411954767e-06,
|
| 208546 |
+
"loss": 0.8473,
|
| 208547 |
+
"step": 85815
|
| 208548 |
+
},
|
| 208549 |
+
{
|
| 208550 |
+
"epoch": 691.63,
|
| 208551 |
+
"learning_rate": 8.630274636510501e-06,
|
| 208552 |
+
"loss": 0.9378,
|
| 208553 |
+
"step": 85820
|
| 208554 |
+
},
|
| 208555 |
+
{
|
| 208556 |
+
"epoch": 691.67,
|
| 208557 |
+
"learning_rate": 8.630193861066237e-06,
|
| 208558 |
+
"loss": 0.3099,
|
| 208559 |
+
"step": 85825
|
| 208560 |
+
},
|
| 208561 |
+
{
|
| 208562 |
+
"epoch": 691.71,
|
| 208563 |
+
"learning_rate": 8.630113085621971e-06,
|
| 208564 |
+
"loss": 0.3133,
|
| 208565 |
+
"step": 85830
|
| 208566 |
+
},
|
| 208567 |
+
{
|
| 208568 |
+
"epoch": 691.75,
|
| 208569 |
+
"learning_rate": 8.630032310177707e-06,
|
| 208570 |
+
"loss": 0.376,
|
| 208571 |
+
"step": 85835
|
| 208572 |
+
},
|
| 208573 |
+
{
|
| 208574 |
+
"epoch": 691.79,
|
| 208575 |
+
"learning_rate": 8.629951534733443e-06,
|
| 208576 |
+
"loss": 0.7509,
|
| 208577 |
+
"step": 85840
|
| 208578 |
+
},
|
| 208579 |
+
{
|
| 208580 |
+
"epoch": 691.83,
|
| 208581 |
+
"learning_rate": 8.629870759289177e-06,
|
| 208582 |
+
"loss": 0.8782,
|
| 208583 |
+
"step": 85845
|
| 208584 |
+
},
|
| 208585 |
+
{
|
| 208586 |
+
"epoch": 691.87,
|
| 208587 |
+
"learning_rate": 8.629789983844913e-06,
|
| 208588 |
+
"loss": 0.3106,
|
| 208589 |
+
"step": 85850
|
| 208590 |
+
},
|
| 208591 |
+
{
|
| 208592 |
+
"epoch": 691.91,
|
| 208593 |
+
"learning_rate": 8.629709208400647e-06,
|
| 208594 |
+
"loss": 0.2805,
|
| 208595 |
+
"step": 85855
|
| 208596 |
+
},
|
| 208597 |
+
{
|
| 208598 |
+
"epoch": 691.95,
|
| 208599 |
+
"learning_rate": 8.629628432956383e-06,
|
| 208600 |
+
"loss": 0.4205,
|
| 208601 |
+
"step": 85860
|
| 208602 |
+
},
|
| 208603 |
+
{
|
| 208604 |
+
"epoch": 691.99,
|
| 208605 |
+
"learning_rate": 8.629547657512117e-06,
|
| 208606 |
+
"loss": 1.089,
|
| 208607 |
+
"step": 85865
|
| 208608 |
+
},
|
| 208609 |
+
{
|
| 208610 |
+
"epoch": 692.0,
|
| 208611 |
+
"eval_loss": 0.42601609230041504,
|
| 208612 |
+
"eval_runtime": 41.4153,
|
| 208613 |
+
"eval_samples_per_second": 20.282,
|
| 208614 |
+
"eval_steps_per_second": 0.652,
|
| 208615 |
+
"eval_wer": 0.1846578526805172,
|
| 208616 |
+
"step": 85866
|
| 208617 |
}
|
| 208618 |
],
|
| 208619 |
+
"max_steps": 620000,
|
| 208620 |
"num_train_epochs": 5000,
|
| 208621 |
+
"total_flos": 2.4164723863623284e+20,
|
| 208622 |
"trial_name": null,
|
| 208623 |
"trial_params": null
|
| 208624 |
}
|
model-bin/finetune/base/{checkpoint-85246 → checkpoint-85866}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629923470.4368432/events.out.tfevents.1629923470.7e498afd5545.7645.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ccc52b0b566a1817347ffec1d9fbac6470e31e06b7799e50d92325f2223080e1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629923933.9843855/events.out.tfevents.1629923933.7e498afd5545.7645.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:203e97e1577586182f844b1871ff2105010daf8c078409123d6f91802d37e092
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629924482.2018788/events.out.tfevents.1629924482.7e498afd5545.7645.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ba5cb2970def6c2ad06ab3ee0d5d3b6ac9981d7161ac367e55ad95afd838937
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629924979.6126964/events.out.tfevents.1629924979.7e498afd5545.7645.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:070269a6e50a6db19ef8f2101daaa99a151e215e8c06aafd98183d653b601ac9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629925434.4852555/events.out.tfevents.1629925434.7e498afd5545.7645.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1c952ea07bb5a5b932c9010dbc1ccaad8c3e7f4bb93cc0f4bc05bb282802f2c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629923470.7e498afd5545.7645.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16da3b1d595264e8f7aa6e87a8b3e900ff1b2e404d03e60bbe58f50388504e97
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629923933.7e498afd5545.7645.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b076666ad7c9560bad91f44ff57dde01e4cd5e59ddc5f2034bfcd2182a82adc
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629924482.7e498afd5545.7645.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:27c2b869fa7d3d9dc2411b139868232d1013f7063f9e65b61391c2d4e98d9853
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629924979.7e498afd5545.7645.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:773dd0c9b365a932168fcaa281be60b83d798b625954da0d38ac61fa8a07cb47
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629925434.7e498afd5545.7645.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:049a6413a36d98ea8de3fa5f0f455f478579aa3835b8fdcc4544c1ce58e74807
|
| 3 |
+
size 8622
|