"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630219370.8727126/events.out.tfevents.1630219370.cc93b136ebf5.1086.69 +3 -0
- model-bin/finetune/base/log/1630219805.8800578/events.out.tfevents.1630219805.cc93b136ebf5.1086.71 +3 -0
- model-bin/finetune/base/log/1630220232.2200906/events.out.tfevents.1630220232.cc93b136ebf5.1086.73 +3 -0
- model-bin/finetune/base/log/1630220657.4988012/events.out.tfevents.1630220657.cc93b136ebf5.1086.75 +3 -0
- model-bin/finetune/base/log/1630221086.942379/events.out.tfevents.1630221086.cc93b136ebf5.1086.77 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630219370.cc93b136ebf5.1086.68 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630219805.cc93b136ebf5.1086.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630220232.cc93b136ebf5.1086.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630220657.cc93b136ebf5.1086.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630221086.cc93b136ebf5.1086.76 +3 -0
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/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:c32f27f34285961ef1c2d17b72932d97df3c85fe34c1af4c98848e52dacd451e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/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 sha256:8351e640c66ad22ff88b540b1f5148ac45011e53894dd0e5a6aca65953a11ab1
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/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:b21f79b4647567ac18c5ddf4705dd7757597b5aede8466bbbba681d49a722784
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/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 sha256:2e170cd10e7cac1057972a08b3ceff6564b37bdd2d3e49b5793664b44700242a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/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:e4f6586054aebf82ecfbabb74061ff8b26d79c6c14d172d2bd7f7bb893ae4c32
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.16819527695529718,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-143240",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -281772,11 +281772,800 @@
|
|
| 281772 |
"eval_steps_per_second": 0.651,
|
| 281773 |
"eval_wer": 0.18056155507559396,
|
| 281774 |
"step": 143365
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 281775 |
}
|
| 281776 |
],
|
| 281777 |
"max_steps": 625000,
|
| 281778 |
"num_train_epochs": 5000,
|
| 281779 |
-
"total_flos": 4.
|
| 281780 |
"trial_name": null,
|
| 281781 |
"trial_params": null
|
| 281782 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.16819527695529718,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-143240",
|
| 4 |
+
"epoch": 1151.0,
|
| 5 |
+
"global_step": 143987,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 281772 |
"eval_steps_per_second": 0.651,
|
| 281773 |
"eval_wer": 0.18056155507559396,
|
| 281774 |
"step": 143365
|
| 281775 |
+
},
|
| 281776 |
+
{
|
| 281777 |
+
"epoch": 1156.04,
|
| 281778 |
+
"learning_rate": 7.700920840064621e-06,
|
| 281779 |
+
"loss": 0.2962,
|
| 281780 |
+
"step": 143370
|
| 281781 |
+
},
|
| 281782 |
+
{
|
| 281783 |
+
"epoch": 1156.08,
|
| 281784 |
+
"learning_rate": 7.700840064620356e-06,
|
| 281785 |
+
"loss": 0.2587,
|
| 281786 |
+
"step": 143375
|
| 281787 |
+
},
|
| 281788 |
+
{
|
| 281789 |
+
"epoch": 1156.12,
|
| 281790 |
+
"learning_rate": 7.700759289176091e-06,
|
| 281791 |
+
"loss": 0.3334,
|
| 281792 |
+
"step": 143380
|
| 281793 |
+
},
|
| 281794 |
+
{
|
| 281795 |
+
"epoch": 1156.16,
|
| 281796 |
+
"learning_rate": 7.700678513731826e-06,
|
| 281797 |
+
"loss": 0.3843,
|
| 281798 |
+
"step": 143385
|
| 281799 |
+
},
|
| 281800 |
+
{
|
| 281801 |
+
"epoch": 1156.2,
|
| 281802 |
+
"learning_rate": 7.700597738287561e-06,
|
| 281803 |
+
"loss": 1.2568,
|
| 281804 |
+
"step": 143390
|
| 281805 |
+
},
|
| 281806 |
+
{
|
| 281807 |
+
"epoch": 1156.24,
|
| 281808 |
+
"learning_rate": 7.700516962843296e-06,
|
| 281809 |
+
"loss": 0.3173,
|
| 281810 |
+
"step": 143395
|
| 281811 |
+
},
|
| 281812 |
+
{
|
| 281813 |
+
"epoch": 1156.28,
|
| 281814 |
+
"learning_rate": 7.700436187399031e-06,
|
| 281815 |
+
"loss": 0.2457,
|
| 281816 |
+
"step": 143400
|
| 281817 |
+
},
|
| 281818 |
+
{
|
| 281819 |
+
"epoch": 1156.32,
|
| 281820 |
+
"learning_rate": 7.700355411954766e-06,
|
| 281821 |
+
"loss": 0.3265,
|
| 281822 |
+
"step": 143405
|
| 281823 |
+
},
|
| 281824 |
+
{
|
| 281825 |
+
"epoch": 1156.36,
|
| 281826 |
+
"learning_rate": 7.700274636510501e-06,
|
| 281827 |
+
"loss": 0.4942,
|
| 281828 |
+
"step": 143410
|
| 281829 |
+
},
|
| 281830 |
+
{
|
| 281831 |
+
"epoch": 1156.4,
|
| 281832 |
+
"learning_rate": 7.700193861066237e-06,
|
| 281833 |
+
"loss": 1.1454,
|
| 281834 |
+
"step": 143415
|
| 281835 |
+
},
|
| 281836 |
+
{
|
| 281837 |
+
"epoch": 1156.44,
|
| 281838 |
+
"learning_rate": 7.700113085621971e-06,
|
| 281839 |
+
"loss": 0.3211,
|
| 281840 |
+
"step": 143420
|
| 281841 |
+
},
|
| 281842 |
+
{
|
| 281843 |
+
"epoch": 1156.48,
|
| 281844 |
+
"learning_rate": 7.700032310177707e-06,
|
| 281845 |
+
"loss": 0.2822,
|
| 281846 |
+
"step": 143425
|
| 281847 |
+
},
|
| 281848 |
+
{
|
| 281849 |
+
"epoch": 1156.52,
|
| 281850 |
+
"learning_rate": 7.699951534733441e-06,
|
| 281851 |
+
"loss": 0.3959,
|
| 281852 |
+
"step": 143430
|
| 281853 |
+
},
|
| 281854 |
+
{
|
| 281855 |
+
"epoch": 1156.56,
|
| 281856 |
+
"learning_rate": 7.699870759289177e-06,
|
| 281857 |
+
"loss": 0.4006,
|
| 281858 |
+
"step": 143435
|
| 281859 |
+
},
|
| 281860 |
+
{
|
| 281861 |
+
"epoch": 1156.6,
|
| 281862 |
+
"learning_rate": 7.699789983844911e-06,
|
| 281863 |
+
"loss": 1.0403,
|
| 281864 |
+
"step": 143440
|
| 281865 |
+
},
|
| 281866 |
+
{
|
| 281867 |
+
"epoch": 1156.65,
|
| 281868 |
+
"learning_rate": 7.699709208400647e-06,
|
| 281869 |
+
"loss": 0.3076,
|
| 281870 |
+
"step": 143445
|
| 281871 |
+
},
|
| 281872 |
+
{
|
| 281873 |
+
"epoch": 1156.69,
|
| 281874 |
+
"learning_rate": 7.699628432956381e-06,
|
| 281875 |
+
"loss": 0.3066,
|
| 281876 |
+
"step": 143450
|
| 281877 |
+
},
|
| 281878 |
+
{
|
| 281879 |
+
"epoch": 1156.73,
|
| 281880 |
+
"learning_rate": 7.699547657512117e-06,
|
| 281881 |
+
"loss": 0.3324,
|
| 281882 |
+
"step": 143455
|
| 281883 |
+
},
|
| 281884 |
+
{
|
| 281885 |
+
"epoch": 1156.77,
|
| 281886 |
+
"learning_rate": 7.699466882067851e-06,
|
| 281887 |
+
"loss": 0.441,
|
| 281888 |
+
"step": 143460
|
| 281889 |
+
},
|
| 281890 |
+
{
|
| 281891 |
+
"epoch": 1156.81,
|
| 281892 |
+
"learning_rate": 7.699386106623587e-06,
|
| 281893 |
+
"loss": 1.2269,
|
| 281894 |
+
"step": 143465
|
| 281895 |
+
},
|
| 281896 |
+
{
|
| 281897 |
+
"epoch": 1156.85,
|
| 281898 |
+
"learning_rate": 7.699305331179321e-06,
|
| 281899 |
+
"loss": 0.2736,
|
| 281900 |
+
"step": 143470
|
| 281901 |
+
},
|
| 281902 |
+
{
|
| 281903 |
+
"epoch": 1156.89,
|
| 281904 |
+
"learning_rate": 7.699224555735057e-06,
|
| 281905 |
+
"loss": 0.2975,
|
| 281906 |
+
"step": 143475
|
| 281907 |
+
},
|
| 281908 |
+
{
|
| 281909 |
+
"epoch": 1156.93,
|
| 281910 |
+
"learning_rate": 7.699143780290791e-06,
|
| 281911 |
+
"loss": 0.2988,
|
| 281912 |
+
"step": 143480
|
| 281913 |
+
},
|
| 281914 |
+
{
|
| 281915 |
+
"epoch": 1156.97,
|
| 281916 |
+
"learning_rate": 7.699063004846527e-06,
|
| 281917 |
+
"loss": 0.4613,
|
| 281918 |
+
"step": 143485
|
| 281919 |
+
},
|
| 281920 |
+
{
|
| 281921 |
+
"epoch": 1157.0,
|
| 281922 |
+
"eval_loss": 0.3576335608959198,
|
| 281923 |
+
"eval_runtime": 41.3609,
|
| 281924 |
+
"eval_samples_per_second": 20.236,
|
| 281925 |
+
"eval_steps_per_second": 0.653,
|
| 281926 |
+
"eval_wer": 0.1753503413582465,
|
| 281927 |
+
"step": 143489
|
| 281928 |
+
},
|
| 281929 |
+
{
|
| 281930 |
+
"epoch": 1157.01,
|
| 281931 |
+
"learning_rate": 7.698982229402263e-06,
|
| 281932 |
+
"loss": 0.3329,
|
| 281933 |
+
"step": 143490
|
| 281934 |
+
},
|
| 281935 |
+
{
|
| 281936 |
+
"epoch": 1157.05,
|
| 281937 |
+
"learning_rate": 7.698901453957997e-06,
|
| 281938 |
+
"loss": 0.288,
|
| 281939 |
+
"step": 143495
|
| 281940 |
+
},
|
| 281941 |
+
{
|
| 281942 |
+
"epoch": 1157.09,
|
| 281943 |
+
"learning_rate": 7.698820678513733e-06,
|
| 281944 |
+
"loss": 0.3328,
|
| 281945 |
+
"step": 143500
|
| 281946 |
+
},
|
| 281947 |
+
{
|
| 281948 |
+
"epoch": 1157.13,
|
| 281949 |
+
"learning_rate": 7.698739903069467e-06,
|
| 281950 |
+
"loss": 0.3706,
|
| 281951 |
+
"step": 143505
|
| 281952 |
+
},
|
| 281953 |
+
{
|
| 281954 |
+
"epoch": 1157.17,
|
| 281955 |
+
"learning_rate": 7.698659127625203e-06,
|
| 281956 |
+
"loss": 0.5247,
|
| 281957 |
+
"step": 143510
|
| 281958 |
+
},
|
| 281959 |
+
{
|
| 281960 |
+
"epoch": 1157.21,
|
| 281961 |
+
"learning_rate": 7.698578352180937e-06,
|
| 281962 |
+
"loss": 1.1031,
|
| 281963 |
+
"step": 143515
|
| 281964 |
+
},
|
| 281965 |
+
{
|
| 281966 |
+
"epoch": 1157.25,
|
| 281967 |
+
"learning_rate": 7.698497576736673e-06,
|
| 281968 |
+
"loss": 0.3005,
|
| 281969 |
+
"step": 143520
|
| 281970 |
+
},
|
| 281971 |
+
{
|
| 281972 |
+
"epoch": 1157.29,
|
| 281973 |
+
"learning_rate": 7.698416801292407e-06,
|
| 281974 |
+
"loss": 0.3484,
|
| 281975 |
+
"step": 143525
|
| 281976 |
+
},
|
| 281977 |
+
{
|
| 281978 |
+
"epoch": 1157.33,
|
| 281979 |
+
"learning_rate": 7.698336025848143e-06,
|
| 281980 |
+
"loss": 0.2836,
|
| 281981 |
+
"step": 143530
|
| 281982 |
+
},
|
| 281983 |
+
{
|
| 281984 |
+
"epoch": 1157.37,
|
| 281985 |
+
"learning_rate": 7.698255250403877e-06,
|
| 281986 |
+
"loss": 0.5206,
|
| 281987 |
+
"step": 143535
|
| 281988 |
+
},
|
| 281989 |
+
{
|
| 281990 |
+
"epoch": 1157.41,
|
| 281991 |
+
"learning_rate": 7.698174474959613e-06,
|
| 281992 |
+
"loss": 1.0886,
|
| 281993 |
+
"step": 143540
|
| 281994 |
+
},
|
| 281995 |
+
{
|
| 281996 |
+
"epoch": 1157.45,
|
| 281997 |
+
"learning_rate": 7.698093699515347e-06,
|
| 281998 |
+
"loss": 0.3511,
|
| 281999 |
+
"step": 143545
|
| 282000 |
+
},
|
| 282001 |
+
{
|
| 282002 |
+
"epoch": 1157.49,
|
| 282003 |
+
"learning_rate": 7.698012924071083e-06,
|
| 282004 |
+
"loss": 0.286,
|
| 282005 |
+
"step": 143550
|
| 282006 |
+
},
|
| 282007 |
+
{
|
| 282008 |
+
"epoch": 1157.53,
|
| 282009 |
+
"learning_rate": 7.697932148626819e-06,
|
| 282010 |
+
"loss": 0.3322,
|
| 282011 |
+
"step": 143555
|
| 282012 |
+
},
|
| 282013 |
+
{
|
| 282014 |
+
"epoch": 1157.57,
|
| 282015 |
+
"learning_rate": 7.697851373182553e-06,
|
| 282016 |
+
"loss": 0.5271,
|
| 282017 |
+
"step": 143560
|
| 282018 |
+
},
|
| 282019 |
+
{
|
| 282020 |
+
"epoch": 1157.61,
|
| 282021 |
+
"learning_rate": 7.697770597738289e-06,
|
| 282022 |
+
"loss": 1.0804,
|
| 282023 |
+
"step": 143565
|
| 282024 |
+
},
|
| 282025 |
+
{
|
| 282026 |
+
"epoch": 1157.65,
|
| 282027 |
+
"learning_rate": 7.697689822294023e-06,
|
| 282028 |
+
"loss": 0.2825,
|
| 282029 |
+
"step": 143570
|
| 282030 |
+
},
|
| 282031 |
+
{
|
| 282032 |
+
"epoch": 1157.69,
|
| 282033 |
+
"learning_rate": 7.697609046849758e-06,
|
| 282034 |
+
"loss": 0.2583,
|
| 282035 |
+
"step": 143575
|
| 282036 |
+
},
|
| 282037 |
+
{
|
| 282038 |
+
"epoch": 1157.73,
|
| 282039 |
+
"learning_rate": 7.697528271405493e-06,
|
| 282040 |
+
"loss": 0.3537,
|
| 282041 |
+
"step": 143580
|
| 282042 |
+
},
|
| 282043 |
+
{
|
| 282044 |
+
"epoch": 1157.77,
|
| 282045 |
+
"learning_rate": 7.697447495961228e-06,
|
| 282046 |
+
"loss": 0.5304,
|
| 282047 |
+
"step": 143585
|
| 282048 |
+
},
|
| 282049 |
+
{
|
| 282050 |
+
"epoch": 1157.81,
|
| 282051 |
+
"learning_rate": 7.697366720516963e-06,
|
| 282052 |
+
"loss": 1.2032,
|
| 282053 |
+
"step": 143590
|
| 282054 |
+
},
|
| 282055 |
+
{
|
| 282056 |
+
"epoch": 1157.85,
|
| 282057 |
+
"learning_rate": 7.697285945072698e-06,
|
| 282058 |
+
"loss": 0.3729,
|
| 282059 |
+
"step": 143595
|
| 282060 |
+
},
|
| 282061 |
+
{
|
| 282062 |
+
"epoch": 1157.9,
|
| 282063 |
+
"learning_rate": 7.697205169628433e-06,
|
| 282064 |
+
"loss": 0.291,
|
| 282065 |
+
"step": 143600
|
| 282066 |
+
},
|
| 282067 |
+
{
|
| 282068 |
+
"epoch": 1157.94,
|
| 282069 |
+
"learning_rate": 7.697124394184168e-06,
|
| 282070 |
+
"loss": 0.2936,
|
| 282071 |
+
"step": 143605
|
| 282072 |
+
},
|
| 282073 |
+
{
|
| 282074 |
+
"epoch": 1157.98,
|
| 282075 |
+
"learning_rate": 7.697043618739903e-06,
|
| 282076 |
+
"loss": 0.5296,
|
| 282077 |
+
"step": 143610
|
| 282078 |
+
},
|
| 282079 |
+
{
|
| 282080 |
+
"epoch": 1158.0,
|
| 282081 |
+
"eval_loss": 0.39611005783081055,
|
| 282082 |
+
"eval_runtime": 40.5454,
|
| 282083 |
+
"eval_samples_per_second": 20.644,
|
| 282084 |
+
"eval_steps_per_second": 0.666,
|
| 282085 |
+
"eval_wer": 0.18108326596604687,
|
| 282086 |
+
"step": 143613
|
| 282087 |
+
},
|
| 282088 |
+
{
|
| 282089 |
+
"epoch": 1158.02,
|
| 282090 |
+
"learning_rate": 7.696962843295638e-06,
|
| 282091 |
+
"loss": 0.3619,
|
| 282092 |
+
"step": 143615
|
| 282093 |
+
},
|
| 282094 |
+
{
|
| 282095 |
+
"epoch": 1158.06,
|
| 282096 |
+
"learning_rate": 7.696882067851374e-06,
|
| 282097 |
+
"loss": 0.3366,
|
| 282098 |
+
"step": 143620
|
| 282099 |
+
},
|
| 282100 |
+
{
|
| 282101 |
+
"epoch": 1158.1,
|
| 282102 |
+
"learning_rate": 7.696801292407108e-06,
|
| 282103 |
+
"loss": 0.2358,
|
| 282104 |
+
"step": 143625
|
| 282105 |
+
},
|
| 282106 |
+
{
|
| 282107 |
+
"epoch": 1158.14,
|
| 282108 |
+
"learning_rate": 7.696720516962844e-06,
|
| 282109 |
+
"loss": 0.3511,
|
| 282110 |
+
"step": 143630
|
| 282111 |
+
},
|
| 282112 |
+
{
|
| 282113 |
+
"epoch": 1158.18,
|
| 282114 |
+
"learning_rate": 7.696639741518578e-06,
|
| 282115 |
+
"loss": 0.493,
|
| 282116 |
+
"step": 143635
|
| 282117 |
+
},
|
| 282118 |
+
{
|
| 282119 |
+
"epoch": 1158.22,
|
| 282120 |
+
"learning_rate": 7.696558966074314e-06,
|
| 282121 |
+
"loss": 0.991,
|
| 282122 |
+
"step": 143640
|
| 282123 |
+
},
|
| 282124 |
+
{
|
| 282125 |
+
"epoch": 1158.26,
|
| 282126 |
+
"learning_rate": 7.696478190630048e-06,
|
| 282127 |
+
"loss": 0.3034,
|
| 282128 |
+
"step": 143645
|
| 282129 |
+
},
|
| 282130 |
+
{
|
| 282131 |
+
"epoch": 1158.3,
|
| 282132 |
+
"learning_rate": 7.696397415185784e-06,
|
| 282133 |
+
"loss": 0.314,
|
| 282134 |
+
"step": 143650
|
| 282135 |
+
},
|
| 282136 |
+
{
|
| 282137 |
+
"epoch": 1158.34,
|
| 282138 |
+
"learning_rate": 7.696316639741518e-06,
|
| 282139 |
+
"loss": 0.3291,
|
| 282140 |
+
"step": 143655
|
| 282141 |
+
},
|
| 282142 |
+
{
|
| 282143 |
+
"epoch": 1158.38,
|
| 282144 |
+
"learning_rate": 7.696235864297254e-06,
|
| 282145 |
+
"loss": 0.693,
|
| 282146 |
+
"step": 143660
|
| 282147 |
+
},
|
| 282148 |
+
{
|
| 282149 |
+
"epoch": 1158.42,
|
| 282150 |
+
"learning_rate": 7.696155088852988e-06,
|
| 282151 |
+
"loss": 0.9516,
|
| 282152 |
+
"step": 143665
|
| 282153 |
+
},
|
| 282154 |
+
{
|
| 282155 |
+
"epoch": 1158.46,
|
| 282156 |
+
"learning_rate": 7.696074313408724e-06,
|
| 282157 |
+
"loss": 0.2922,
|
| 282158 |
+
"step": 143670
|
| 282159 |
+
},
|
| 282160 |
+
{
|
| 282161 |
+
"epoch": 1158.5,
|
| 282162 |
+
"learning_rate": 7.695993537964458e-06,
|
| 282163 |
+
"loss": 0.2584,
|
| 282164 |
+
"step": 143675
|
| 282165 |
+
},
|
| 282166 |
+
{
|
| 282167 |
+
"epoch": 1158.54,
|
| 282168 |
+
"learning_rate": 7.695912762520194e-06,
|
| 282169 |
+
"loss": 0.313,
|
| 282170 |
+
"step": 143680
|
| 282171 |
+
},
|
| 282172 |
+
{
|
| 282173 |
+
"epoch": 1158.58,
|
| 282174 |
+
"learning_rate": 7.695831987075928e-06,
|
| 282175 |
+
"loss": 0.5341,
|
| 282176 |
+
"step": 143685
|
| 282177 |
+
},
|
| 282178 |
+
{
|
| 282179 |
+
"epoch": 1158.62,
|
| 282180 |
+
"learning_rate": 7.695751211631664e-06,
|
| 282181 |
+
"loss": 1.107,
|
| 282182 |
+
"step": 143690
|
| 282183 |
+
},
|
| 282184 |
+
{
|
| 282185 |
+
"epoch": 1158.66,
|
| 282186 |
+
"learning_rate": 7.6956704361874e-06,
|
| 282187 |
+
"loss": 0.4665,
|
| 282188 |
+
"step": 143695
|
| 282189 |
+
},
|
| 282190 |
+
{
|
| 282191 |
+
"epoch": 1158.7,
|
| 282192 |
+
"learning_rate": 7.695589660743134e-06,
|
| 282193 |
+
"loss": 0.2928,
|
| 282194 |
+
"step": 143700
|
| 282195 |
+
},
|
| 282196 |
+
{
|
| 282197 |
+
"epoch": 1158.74,
|
| 282198 |
+
"learning_rate": 7.69550888529887e-06,
|
| 282199 |
+
"loss": 0.3576,
|
| 282200 |
+
"step": 143705
|
| 282201 |
+
},
|
| 282202 |
+
{
|
| 282203 |
+
"epoch": 1158.78,
|
| 282204 |
+
"learning_rate": 7.695428109854604e-06,
|
| 282205 |
+
"loss": 0.6097,
|
| 282206 |
+
"step": 143710
|
| 282207 |
+
},
|
| 282208 |
+
{
|
| 282209 |
+
"epoch": 1158.82,
|
| 282210 |
+
"learning_rate": 7.69534733441034e-06,
|
| 282211 |
+
"loss": 0.9703,
|
| 282212 |
+
"step": 143715
|
| 282213 |
+
},
|
| 282214 |
+
{
|
| 282215 |
+
"epoch": 1158.86,
|
| 282216 |
+
"learning_rate": 7.695266558966074e-06,
|
| 282217 |
+
"loss": 0.2794,
|
| 282218 |
+
"step": 143720
|
| 282219 |
+
},
|
| 282220 |
+
{
|
| 282221 |
+
"epoch": 1158.9,
|
| 282222 |
+
"learning_rate": 7.69518578352181e-06,
|
| 282223 |
+
"loss": 0.2705,
|
| 282224 |
+
"step": 143725
|
| 282225 |
+
},
|
| 282226 |
+
{
|
| 282227 |
+
"epoch": 1158.94,
|
| 282228 |
+
"learning_rate": 7.695105008077544e-06,
|
| 282229 |
+
"loss": 0.299,
|
| 282230 |
+
"step": 143730
|
| 282231 |
+
},
|
| 282232 |
+
{
|
| 282233 |
+
"epoch": 1158.98,
|
| 282234 |
+
"learning_rate": 7.69502423263328e-06,
|
| 282235 |
+
"loss": 0.7515,
|
| 282236 |
+
"step": 143735
|
| 282237 |
+
},
|
| 282238 |
+
{
|
| 282239 |
+
"epoch": 1159.0,
|
| 282240 |
+
"eval_loss": 0.38360095024108887,
|
| 282241 |
+
"eval_runtime": 40.7226,
|
| 282242 |
+
"eval_samples_per_second": 20.554,
|
| 282243 |
+
"eval_steps_per_second": 0.663,
|
| 282244 |
+
"eval_wer": 0.1902169915410077,
|
| 282245 |
+
"step": 143737
|
| 282246 |
+
},
|
| 282247 |
+
{
|
| 282248 |
+
"epoch": 1149.02,
|
| 282249 |
+
"learning_rate": 7.694943457189014e-06,
|
| 282250 |
+
"loss": 0.3411,
|
| 282251 |
+
"step": 143740
|
| 282252 |
+
},
|
| 282253 |
+
{
|
| 282254 |
+
"epoch": 1149.06,
|
| 282255 |
+
"learning_rate": 7.69486268174475e-06,
|
| 282256 |
+
"loss": 0.2825,
|
| 282257 |
+
"step": 143745
|
| 282258 |
+
},
|
| 282259 |
+
{
|
| 282260 |
+
"epoch": 1149.1,
|
| 282261 |
+
"learning_rate": 7.694781906300486e-06,
|
| 282262 |
+
"loss": 0.3384,
|
| 282263 |
+
"step": 143750
|
| 282264 |
+
},
|
| 282265 |
+
{
|
| 282266 |
+
"epoch": 1149.14,
|
| 282267 |
+
"learning_rate": 7.69470113085622e-06,
|
| 282268 |
+
"loss": 0.3251,
|
| 282269 |
+
"step": 143755
|
| 282270 |
+
},
|
| 282271 |
+
{
|
| 282272 |
+
"epoch": 1149.18,
|
| 282273 |
+
"learning_rate": 7.694620355411956e-06,
|
| 282274 |
+
"loss": 0.6553,
|
| 282275 |
+
"step": 143760
|
| 282276 |
+
},
|
| 282277 |
+
{
|
| 282278 |
+
"epoch": 1149.22,
|
| 282279 |
+
"learning_rate": 7.69453957996769e-06,
|
| 282280 |
+
"loss": 0.8217,
|
| 282281 |
+
"step": 143765
|
| 282282 |
+
},
|
| 282283 |
+
{
|
| 282284 |
+
"epoch": 1149.26,
|
| 282285 |
+
"learning_rate": 7.694458804523426e-06,
|
| 282286 |
+
"loss": 0.3223,
|
| 282287 |
+
"step": 143770
|
| 282288 |
+
},
|
| 282289 |
+
{
|
| 282290 |
+
"epoch": 1149.3,
|
| 282291 |
+
"learning_rate": 7.69437802907916e-06,
|
| 282292 |
+
"loss": 0.2594,
|
| 282293 |
+
"step": 143775
|
| 282294 |
+
},
|
| 282295 |
+
{
|
| 282296 |
+
"epoch": 1149.34,
|
| 282297 |
+
"learning_rate": 7.694297253634896e-06,
|
| 282298 |
+
"loss": 0.3364,
|
| 282299 |
+
"step": 143780
|
| 282300 |
+
},
|
| 282301 |
+
{
|
| 282302 |
+
"epoch": 1149.38,
|
| 282303 |
+
"learning_rate": 7.69421647819063e-06,
|
| 282304 |
+
"loss": 0.7022,
|
| 282305 |
+
"step": 143785
|
| 282306 |
+
},
|
| 282307 |
+
{
|
| 282308 |
+
"epoch": 1149.42,
|
| 282309 |
+
"learning_rate": 7.694135702746365e-06,
|
| 282310 |
+
"loss": 0.7457,
|
| 282311 |
+
"step": 143790
|
| 282312 |
+
},
|
| 282313 |
+
{
|
| 282314 |
+
"epoch": 1149.46,
|
| 282315 |
+
"learning_rate": 7.6940549273021e-06,
|
| 282316 |
+
"loss": 0.2573,
|
| 282317 |
+
"step": 143795
|
| 282318 |
+
},
|
| 282319 |
+
{
|
| 282320 |
+
"epoch": 1149.5,
|
| 282321 |
+
"learning_rate": 7.693974151857835e-06,
|
| 282322 |
+
"loss": 0.2289,
|
| 282323 |
+
"step": 143800
|
| 282324 |
+
},
|
| 282325 |
+
{
|
| 282326 |
+
"epoch": 1149.54,
|
| 282327 |
+
"learning_rate": 7.69389337641357e-06,
|
| 282328 |
+
"loss": 0.3103,
|
| 282329 |
+
"step": 143805
|
| 282330 |
+
},
|
| 282331 |
+
{
|
| 282332 |
+
"epoch": 1149.58,
|
| 282333 |
+
"learning_rate": 7.693812600969305e-06,
|
| 282334 |
+
"loss": 0.6486,
|
| 282335 |
+
"step": 143810
|
| 282336 |
+
},
|
| 282337 |
+
{
|
| 282338 |
+
"epoch": 1149.62,
|
| 282339 |
+
"learning_rate": 7.693731825525041e-06,
|
| 282340 |
+
"loss": 0.7897,
|
| 282341 |
+
"step": 143815
|
| 282342 |
+
},
|
| 282343 |
+
{
|
| 282344 |
+
"epoch": 1149.66,
|
| 282345 |
+
"learning_rate": 7.693651050080775e-06,
|
| 282346 |
+
"loss": 0.2758,
|
| 282347 |
+
"step": 143820
|
| 282348 |
+
},
|
| 282349 |
+
{
|
| 282350 |
+
"epoch": 1149.7,
|
| 282351 |
+
"learning_rate": 7.693570274636511e-06,
|
| 282352 |
+
"loss": 0.2628,
|
| 282353 |
+
"step": 143825
|
| 282354 |
+
},
|
| 282355 |
+
{
|
| 282356 |
+
"epoch": 1149.74,
|
| 282357 |
+
"learning_rate": 7.693489499192245e-06,
|
| 282358 |
+
"loss": 0.4422,
|
| 282359 |
+
"step": 143830
|
| 282360 |
+
},
|
| 282361 |
+
{
|
| 282362 |
+
"epoch": 1149.78,
|
| 282363 |
+
"learning_rate": 7.693408723747981e-06,
|
| 282364 |
+
"loss": 0.7036,
|
| 282365 |
+
"step": 143835
|
| 282366 |
+
},
|
| 282367 |
+
{
|
| 282368 |
+
"epoch": 1149.82,
|
| 282369 |
+
"learning_rate": 7.693327948303715e-06,
|
| 282370 |
+
"loss": 0.7535,
|
| 282371 |
+
"step": 143840
|
| 282372 |
+
},
|
| 282373 |
+
{
|
| 282374 |
+
"epoch": 1149.86,
|
| 282375 |
+
"learning_rate": 7.693247172859451e-06,
|
| 282376 |
+
"loss": 0.2703,
|
| 282377 |
+
"step": 143845
|
| 282378 |
+
},
|
| 282379 |
+
{
|
| 282380 |
+
"epoch": 1149.9,
|
| 282381 |
+
"learning_rate": 7.693166397415185e-06,
|
| 282382 |
+
"loss": 0.3293,
|
| 282383 |
+
"step": 143850
|
| 282384 |
+
},
|
| 282385 |
+
{
|
| 282386 |
+
"epoch": 1149.94,
|
| 282387 |
+
"learning_rate": 7.693085621970921e-06,
|
| 282388 |
+
"loss": 0.2901,
|
| 282389 |
+
"step": 143855
|
| 282390 |
+
},
|
| 282391 |
+
{
|
| 282392 |
+
"epoch": 1149.98,
|
| 282393 |
+
"learning_rate": 7.693004846526655e-06,
|
| 282394 |
+
"loss": 0.7575,
|
| 282395 |
+
"step": 143860
|
| 282396 |
+
},
|
| 282397 |
+
{
|
| 282398 |
+
"epoch": 1150.0,
|
| 282399 |
+
"eval_loss": 0.330931693315506,
|
| 282400 |
+
"eval_runtime": 40.935,
|
| 282401 |
+
"eval_samples_per_second": 20.447,
|
| 282402 |
+
"eval_steps_per_second": 0.66,
|
| 282403 |
+
"eval_wer": 0.18041988628079894,
|
| 282404 |
+
"step": 143862
|
| 282405 |
+
},
|
| 282406 |
+
{
|
| 282407 |
+
"epoch": 1150.02,
|
| 282408 |
+
"learning_rate": 7.692924071082391e-06,
|
| 282409 |
+
"loss": 0.4115,
|
| 282410 |
+
"step": 143865
|
| 282411 |
+
},
|
| 282412 |
+
{
|
| 282413 |
+
"epoch": 1150.06,
|
| 282414 |
+
"learning_rate": 7.692843295638127e-06,
|
| 282415 |
+
"loss": 0.2711,
|
| 282416 |
+
"step": 143870
|
| 282417 |
+
},
|
| 282418 |
+
{
|
| 282419 |
+
"epoch": 1150.1,
|
| 282420 |
+
"learning_rate": 7.692762520193861e-06,
|
| 282421 |
+
"loss": 0.3022,
|
| 282422 |
+
"step": 143875
|
| 282423 |
+
},
|
| 282424 |
+
{
|
| 282425 |
+
"epoch": 1150.14,
|
| 282426 |
+
"learning_rate": 7.692681744749597e-06,
|
| 282427 |
+
"loss": 0.3806,
|
| 282428 |
+
"step": 143880
|
| 282429 |
+
},
|
| 282430 |
+
{
|
| 282431 |
+
"epoch": 1150.18,
|
| 282432 |
+
"learning_rate": 7.692600969305331e-06,
|
| 282433 |
+
"loss": 0.7295,
|
| 282434 |
+
"step": 143885
|
| 282435 |
+
},
|
| 282436 |
+
{
|
| 282437 |
+
"epoch": 1150.22,
|
| 282438 |
+
"learning_rate": 7.692520193861067e-06,
|
| 282439 |
+
"loss": 0.7231,
|
| 282440 |
+
"step": 143890
|
| 282441 |
+
},
|
| 282442 |
+
{
|
| 282443 |
+
"epoch": 1150.26,
|
| 282444 |
+
"learning_rate": 7.692439418416801e-06,
|
| 282445 |
+
"loss": 0.2712,
|
| 282446 |
+
"step": 143895
|
| 282447 |
+
},
|
| 282448 |
+
{
|
| 282449 |
+
"epoch": 1150.3,
|
| 282450 |
+
"learning_rate": 7.692358642972537e-06,
|
| 282451 |
+
"loss": 0.3067,
|
| 282452 |
+
"step": 143900
|
| 282453 |
+
},
|
| 282454 |
+
{
|
| 282455 |
+
"epoch": 1150.34,
|
| 282456 |
+
"learning_rate": 7.692277867528271e-06,
|
| 282457 |
+
"loss": 0.361,
|
| 282458 |
+
"step": 143905
|
| 282459 |
+
},
|
| 282460 |
+
{
|
| 282461 |
+
"epoch": 1150.38,
|
| 282462 |
+
"learning_rate": 7.692197092084007e-06,
|
| 282463 |
+
"loss": 0.7103,
|
| 282464 |
+
"step": 143910
|
| 282465 |
+
},
|
| 282466 |
+
{
|
| 282467 |
+
"epoch": 1150.42,
|
| 282468 |
+
"learning_rate": 7.692116316639741e-06,
|
| 282469 |
+
"loss": 0.8117,
|
| 282470 |
+
"step": 143915
|
| 282471 |
+
},
|
| 282472 |
+
{
|
| 282473 |
+
"epoch": 1150.46,
|
| 282474 |
+
"learning_rate": 7.692035541195477e-06,
|
| 282475 |
+
"loss": 0.2792,
|
| 282476 |
+
"step": 143920
|
| 282477 |
+
},
|
| 282478 |
+
{
|
| 282479 |
+
"epoch": 1150.5,
|
| 282480 |
+
"learning_rate": 7.691954765751213e-06,
|
| 282481 |
+
"loss": 0.2409,
|
| 282482 |
+
"step": 143925
|
| 282483 |
+
},
|
| 282484 |
+
{
|
| 282485 |
+
"epoch": 1150.54,
|
| 282486 |
+
"learning_rate": 7.691873990306947e-06,
|
| 282487 |
+
"loss": 0.4053,
|
| 282488 |
+
"step": 143930
|
| 282489 |
+
},
|
| 282490 |
+
{
|
| 282491 |
+
"epoch": 1150.58,
|
| 282492 |
+
"learning_rate": 7.691793214862683e-06,
|
| 282493 |
+
"loss": 0.7631,
|
| 282494 |
+
"step": 143935
|
| 282495 |
+
},
|
| 282496 |
+
{
|
| 282497 |
+
"epoch": 1150.62,
|
| 282498 |
+
"learning_rate": 7.691712439418417e-06,
|
| 282499 |
+
"loss": 0.817,
|
| 282500 |
+
"step": 143940
|
| 282501 |
+
},
|
| 282502 |
+
{
|
| 282503 |
+
"epoch": 1150.66,
|
| 282504 |
+
"learning_rate": 7.691631663974153e-06,
|
| 282505 |
+
"loss": 0.2555,
|
| 282506 |
+
"step": 143945
|
| 282507 |
+
},
|
| 282508 |
+
{
|
| 282509 |
+
"epoch": 1150.7,
|
| 282510 |
+
"learning_rate": 7.691550888529887e-06,
|
| 282511 |
+
"loss": 0.2972,
|
| 282512 |
+
"step": 143950
|
| 282513 |
+
},
|
| 282514 |
+
{
|
| 282515 |
+
"epoch": 1150.74,
|
| 282516 |
+
"learning_rate": 7.691470113085623e-06,
|
| 282517 |
+
"loss": 0.3903,
|
| 282518 |
+
"step": 143955
|
| 282519 |
+
},
|
| 282520 |
+
{
|
| 282521 |
+
"epoch": 1150.78,
|
| 282522 |
+
"learning_rate": 7.691389337641357e-06,
|
| 282523 |
+
"loss": 0.7996,
|
| 282524 |
+
"step": 143960
|
| 282525 |
+
},
|
| 282526 |
+
{
|
| 282527 |
+
"epoch": 1150.82,
|
| 282528 |
+
"learning_rate": 7.691308562197093e-06,
|
| 282529 |
+
"loss": 0.8304,
|
| 282530 |
+
"step": 143965
|
| 282531 |
+
},
|
| 282532 |
+
{
|
| 282533 |
+
"epoch": 1150.86,
|
| 282534 |
+
"learning_rate": 7.691227786752827e-06,
|
| 282535 |
+
"loss": 0.2814,
|
| 282536 |
+
"step": 143970
|
| 282537 |
+
},
|
| 282538 |
+
{
|
| 282539 |
+
"epoch": 1150.9,
|
| 282540 |
+
"learning_rate": 7.691147011308563e-06,
|
| 282541 |
+
"loss": 0.2936,
|
| 282542 |
+
"step": 143975
|
| 282543 |
+
},
|
| 282544 |
+
{
|
| 282545 |
+
"epoch": 1150.94,
|
| 282546 |
+
"learning_rate": 7.691066235864297e-06,
|
| 282547 |
+
"loss": 0.3127,
|
| 282548 |
+
"step": 143980
|
| 282549 |
+
},
|
| 282550 |
+
{
|
| 282551 |
+
"epoch": 1150.98,
|
| 282552 |
+
"learning_rate": 7.690985460420033e-06,
|
| 282553 |
+
"loss": 0.7676,
|
| 282554 |
+
"step": 143985
|
| 282555 |
+
},
|
| 282556 |
+
{
|
| 282557 |
+
"epoch": 1151.0,
|
| 282558 |
+
"eval_loss": 0.37501341104507446,
|
| 282559 |
+
"eval_runtime": 40.5276,
|
| 282560 |
+
"eval_samples_per_second": 20.677,
|
| 282561 |
+
"eval_steps_per_second": 0.666,
|
| 282562 |
+
"eval_wer": 0.1780933362916944,
|
| 282563 |
+
"step": 143987
|
| 282564 |
}
|
| 282565 |
],
|
| 282566 |
"max_steps": 625000,
|
| 282567 |
"num_train_epochs": 5000,
|
| 282568 |
+
"total_flos": 4.05173119543111e+20,
|
| 282569 |
"trial_name": null,
|
| 282570 |
"trial_params": null
|
| 282571 |
}
|
model-bin/finetune/base/{checkpoint-143365 → checkpoint-143987}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630219370.8727126/events.out.tfevents.1630219370.cc93b136ebf5.1086.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3e9963930e19fcf7fde95fee2c806784c2e18130b726ef3e534e856c8522589
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630219805.8800578/events.out.tfevents.1630219805.cc93b136ebf5.1086.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:354ee52e14108768f6fcf6e18c68aff134da7ed71a2ff7afdf79f38b15740e0d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630220232.2200906/events.out.tfevents.1630220232.cc93b136ebf5.1086.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d1b26e4921a3aeec4911825e77310d860f943327ef251e6952d37bc05fc12ddb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630220657.4988012/events.out.tfevents.1630220657.cc93b136ebf5.1086.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b18158e1b809c2b19d98ba9068a7a01e623f6b4281274ca2fbde57de0694d90
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630221086.942379/events.out.tfevents.1630221086.cc93b136ebf5.1086.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e6fd9e7909b83e79d5d09c906a6aba5d2b059eeb43a301e16b1f0d4250371093
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630219370.cc93b136ebf5.1086.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c01a680ede2394c62abbafe5b1f8f30ae8199af41c7a14a6b64367ca8755eff
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630219805.cc93b136ebf5.1086.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:27ccdc6d78b7de9ed5b697ebe15e42b840a736b1e3d3838614ff324afbb8c755
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630220232.cc93b136ebf5.1086.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5e5befcf045e5e180e4547159cc2b9f496ddd2897c9e29e51f6a1e474a58733
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630220657.cc93b136ebf5.1086.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fc53fee0e4db1347f03d8e5a6494fab6d8de4416b4188e98abf7b99f9f82979e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630221086.cc93b136ebf5.1086.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2fd6b1164acfdd5fcc8eb2bfbd62c9fc789572c1f0428edd3b1c72e5385c7599
|
| 3 |
+
size 8622
|