"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630028869.9681413/events.out.tfevents.1630028869.52f5c7e305a3.886.1 +3 -0
- model-bin/finetune/base/log/1630029382.1230361/events.out.tfevents.1630029382.52f5c7e305a3.886.3 +3 -0
- model-bin/finetune/base/log/1630029880.354848/events.out.tfevents.1630029880.52f5c7e305a3.886.5 +3 -0
- model-bin/finetune/base/log/1630030344.963895/events.out.tfevents.1630030344.52f5c7e305a3.886.7 +3 -0
- model-bin/finetune/base/log/1630030813.4653862/events.out.tfevents.1630030813.52f5c7e305a3.886.9 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630028869.52f5c7e305a3.886.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630029382.52f5c7e305a3.886.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630029880.52f5c7e305a3.886.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630030344.52f5c7e305a3.886.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630030813.52f5c7e305a3.886.8 +3 -0
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/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:a5c95e29b9a11f0df90c496d284a209b4d4bb87c82f4ebbdc10fedfcb7f4e9f2
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/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 sha256:0968d7acadf12d6158da8c9173813456d82600089e426dbb1667783d46e2094d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/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 sha256:446c8d332b18a83a20f05add4da282bc0603f62a7bfa6fb27fc318e926ab7177
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/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 sha256:74fdc36ad24f6dd920a90851f0154bac2c24bed22311d0ffda23ca06b7bf1327
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/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 sha256:064d7547eb00c8ba8fbfe85fae82d91a126055bdaaf847fc479e7655312359af
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -234747,11 +234747,800 @@
|
|
| 234747 |
"eval_steps_per_second": 0.702,
|
| 234748 |
"eval_wer": 0.18673609604397193,
|
| 234749 |
"step": 106405
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 234750 |
}
|
| 234751 |
],
|
| 234752 |
"max_steps": 625000,
|
| 234753 |
"num_train_epochs": 5000,
|
| 234754 |
-
"total_flos":
|
| 234755 |
"trial_name": null,
|
| 234756 |
"trial_params": null
|
| 234757 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 856.0,
|
| 5 |
+
"global_step": 107027,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 234747 |
"eval_steps_per_second": 0.702,
|
| 234748 |
"eval_wer": 0.18673609604397193,
|
| 234749 |
"step": 106405
|
| 234750 |
+
},
|
| 234751 |
+
{
|
| 234752 |
+
"epoch": 858.04,
|
| 234753 |
+
"learning_rate": 8.297770597738288e-06,
|
| 234754 |
+
"loss": 0.3207,
|
| 234755 |
+
"step": 106410
|
| 234756 |
+
},
|
| 234757 |
+
{
|
| 234758 |
+
"epoch": 858.08,
|
| 234759 |
+
"learning_rate": 8.297689822294022e-06,
|
| 234760 |
+
"loss": 0.2787,
|
| 234761 |
+
"step": 106415
|
| 234762 |
+
},
|
| 234763 |
+
{
|
| 234764 |
+
"epoch": 858.12,
|
| 234765 |
+
"learning_rate": 8.297609046849758e-06,
|
| 234766 |
+
"loss": 0.3789,
|
| 234767 |
+
"step": 106420
|
| 234768 |
+
},
|
| 234769 |
+
{
|
| 234770 |
+
"epoch": 858.16,
|
| 234771 |
+
"learning_rate": 8.297528271405494e-06,
|
| 234772 |
+
"loss": 0.5362,
|
| 234773 |
+
"step": 106425
|
| 234774 |
+
},
|
| 234775 |
+
{
|
| 234776 |
+
"epoch": 858.2,
|
| 234777 |
+
"learning_rate": 8.297447495961228e-06,
|
| 234778 |
+
"loss": 1.1553,
|
| 234779 |
+
"step": 106430
|
| 234780 |
+
},
|
| 234781 |
+
{
|
| 234782 |
+
"epoch": 858.24,
|
| 234783 |
+
"learning_rate": 8.297366720516964e-06,
|
| 234784 |
+
"loss": 0.2614,
|
| 234785 |
+
"step": 106435
|
| 234786 |
+
},
|
| 234787 |
+
{
|
| 234788 |
+
"epoch": 858.28,
|
| 234789 |
+
"learning_rate": 8.297285945072698e-06,
|
| 234790 |
+
"loss": 0.2563,
|
| 234791 |
+
"step": 106440
|
| 234792 |
+
},
|
| 234793 |
+
{
|
| 234794 |
+
"epoch": 858.32,
|
| 234795 |
+
"learning_rate": 8.297205169628434e-06,
|
| 234796 |
+
"loss": 0.3186,
|
| 234797 |
+
"step": 106445
|
| 234798 |
+
},
|
| 234799 |
+
{
|
| 234800 |
+
"epoch": 858.36,
|
| 234801 |
+
"learning_rate": 8.297124394184168e-06,
|
| 234802 |
+
"loss": 0.3874,
|
| 234803 |
+
"step": 106450
|
| 234804 |
+
},
|
| 234805 |
+
{
|
| 234806 |
+
"epoch": 858.4,
|
| 234807 |
+
"learning_rate": 8.297043618739904e-06,
|
| 234808 |
+
"loss": 1.1646,
|
| 234809 |
+
"step": 106455
|
| 234810 |
+
},
|
| 234811 |
+
{
|
| 234812 |
+
"epoch": 858.44,
|
| 234813 |
+
"learning_rate": 8.296962843295638e-06,
|
| 234814 |
+
"loss": 0.3417,
|
| 234815 |
+
"step": 106460
|
| 234816 |
+
},
|
| 234817 |
+
{
|
| 234818 |
+
"epoch": 858.48,
|
| 234819 |
+
"learning_rate": 8.296882067851374e-06,
|
| 234820 |
+
"loss": 0.2837,
|
| 234821 |
+
"step": 106465
|
| 234822 |
+
},
|
| 234823 |
+
{
|
| 234824 |
+
"epoch": 858.52,
|
| 234825 |
+
"learning_rate": 8.296801292407108e-06,
|
| 234826 |
+
"loss": 0.2927,
|
| 234827 |
+
"step": 106470
|
| 234828 |
+
},
|
| 234829 |
+
{
|
| 234830 |
+
"epoch": 858.56,
|
| 234831 |
+
"learning_rate": 8.296720516962844e-06,
|
| 234832 |
+
"loss": 0.495,
|
| 234833 |
+
"step": 106475
|
| 234834 |
+
},
|
| 234835 |
+
{
|
| 234836 |
+
"epoch": 858.6,
|
| 234837 |
+
"learning_rate": 8.29663974151858e-06,
|
| 234838 |
+
"loss": 1.4269,
|
| 234839 |
+
"step": 106480
|
| 234840 |
+
},
|
| 234841 |
+
{
|
| 234842 |
+
"epoch": 858.65,
|
| 234843 |
+
"learning_rate": 8.296558966074314e-06,
|
| 234844 |
+
"loss": 0.2961,
|
| 234845 |
+
"step": 106485
|
| 234846 |
+
},
|
| 234847 |
+
{
|
| 234848 |
+
"epoch": 858.69,
|
| 234849 |
+
"learning_rate": 8.29647819063005e-06,
|
| 234850 |
+
"loss": 0.3464,
|
| 234851 |
+
"step": 106490
|
| 234852 |
+
},
|
| 234853 |
+
{
|
| 234854 |
+
"epoch": 858.73,
|
| 234855 |
+
"learning_rate": 8.296397415185783e-06,
|
| 234856 |
+
"loss": 0.3163,
|
| 234857 |
+
"step": 106495
|
| 234858 |
+
},
|
| 234859 |
+
{
|
| 234860 |
+
"epoch": 858.77,
|
| 234861 |
+
"learning_rate": 8.29631663974152e-06,
|
| 234862 |
+
"loss": 0.4696,
|
| 234863 |
+
"step": 106500
|
| 234864 |
+
},
|
| 234865 |
+
{
|
| 234866 |
+
"epoch": 858.81,
|
| 234867 |
+
"learning_rate": 8.296235864297253e-06,
|
| 234868 |
+
"loss": 1.1545,
|
| 234869 |
+
"step": 106505
|
| 234870 |
+
},
|
| 234871 |
+
{
|
| 234872 |
+
"epoch": 858.85,
|
| 234873 |
+
"learning_rate": 8.29615508885299e-06,
|
| 234874 |
+
"loss": 0.3433,
|
| 234875 |
+
"step": 106510
|
| 234876 |
+
},
|
| 234877 |
+
{
|
| 234878 |
+
"epoch": 858.89,
|
| 234879 |
+
"learning_rate": 8.296074313408723e-06,
|
| 234880 |
+
"loss": 0.3527,
|
| 234881 |
+
"step": 106515
|
| 234882 |
+
},
|
| 234883 |
+
{
|
| 234884 |
+
"epoch": 858.93,
|
| 234885 |
+
"learning_rate": 8.29599353796446e-06,
|
| 234886 |
+
"loss": 0.3019,
|
| 234887 |
+
"step": 106520
|
| 234888 |
+
},
|
| 234889 |
+
{
|
| 234890 |
+
"epoch": 858.97,
|
| 234891 |
+
"learning_rate": 8.295912762520193e-06,
|
| 234892 |
+
"loss": 0.4983,
|
| 234893 |
+
"step": 106525
|
| 234894 |
+
},
|
| 234895 |
+
{
|
| 234896 |
+
"epoch": 859.0,
|
| 234897 |
+
"eval_loss": 0.39661577343940735,
|
| 234898 |
+
"eval_runtime": 42.8738,
|
| 234899 |
+
"eval_samples_per_second": 19.546,
|
| 234900 |
+
"eval_steps_per_second": 0.63,
|
| 234901 |
+
"eval_wer": 0.19128400116993272,
|
| 234902 |
+
"step": 106529
|
| 234903 |
+
},
|
| 234904 |
+
{
|
| 234905 |
+
"epoch": 852.01,
|
| 234906 |
+
"learning_rate": 8.29583198707593e-06,
|
| 234907 |
+
"loss": 0.499,
|
| 234908 |
+
"step": 106530
|
| 234909 |
+
},
|
| 234910 |
+
{
|
| 234911 |
+
"epoch": 852.05,
|
| 234912 |
+
"learning_rate": 8.295751211631665e-06,
|
| 234913 |
+
"loss": 0.3251,
|
| 234914 |
+
"step": 106535
|
| 234915 |
+
},
|
| 234916 |
+
{
|
| 234917 |
+
"epoch": 852.09,
|
| 234918 |
+
"learning_rate": 8.2956704361874e-06,
|
| 234919 |
+
"loss": 0.3215,
|
| 234920 |
+
"step": 106540
|
| 234921 |
+
},
|
| 234922 |
+
{
|
| 234923 |
+
"epoch": 852.13,
|
| 234924 |
+
"learning_rate": 8.295589660743135e-06,
|
| 234925 |
+
"loss": 0.3227,
|
| 234926 |
+
"step": 106545
|
| 234927 |
+
},
|
| 234928 |
+
{
|
| 234929 |
+
"epoch": 852.17,
|
| 234930 |
+
"learning_rate": 8.29550888529887e-06,
|
| 234931 |
+
"loss": 0.5501,
|
| 234932 |
+
"step": 106550
|
| 234933 |
+
},
|
| 234934 |
+
{
|
| 234935 |
+
"epoch": 852.21,
|
| 234936 |
+
"learning_rate": 8.295428109854605e-06,
|
| 234937 |
+
"loss": 1.1344,
|
| 234938 |
+
"step": 106555
|
| 234939 |
+
},
|
| 234940 |
+
{
|
| 234941 |
+
"epoch": 852.25,
|
| 234942 |
+
"learning_rate": 8.29534733441034e-06,
|
| 234943 |
+
"loss": 0.3162,
|
| 234944 |
+
"step": 106560
|
| 234945 |
+
},
|
| 234946 |
+
{
|
| 234947 |
+
"epoch": 852.29,
|
| 234948 |
+
"learning_rate": 8.295266558966075e-06,
|
| 234949 |
+
"loss": 0.2899,
|
| 234950 |
+
"step": 106565
|
| 234951 |
+
},
|
| 234952 |
+
{
|
| 234953 |
+
"epoch": 852.33,
|
| 234954 |
+
"learning_rate": 8.295185783521809e-06,
|
| 234955 |
+
"loss": 0.3417,
|
| 234956 |
+
"step": 106570
|
| 234957 |
+
},
|
| 234958 |
+
{
|
| 234959 |
+
"epoch": 852.37,
|
| 234960 |
+
"learning_rate": 8.295105008077545e-06,
|
| 234961 |
+
"loss": 0.6559,
|
| 234962 |
+
"step": 106575
|
| 234963 |
+
},
|
| 234964 |
+
{
|
| 234965 |
+
"epoch": 852.41,
|
| 234966 |
+
"learning_rate": 8.295024232633279e-06,
|
| 234967 |
+
"loss": 1.1389,
|
| 234968 |
+
"step": 106580
|
| 234969 |
+
},
|
| 234970 |
+
{
|
| 234971 |
+
"epoch": 852.45,
|
| 234972 |
+
"learning_rate": 8.294943457189015e-06,
|
| 234973 |
+
"loss": 0.2899,
|
| 234974 |
+
"step": 106585
|
| 234975 |
+
},
|
| 234976 |
+
{
|
| 234977 |
+
"epoch": 852.49,
|
| 234978 |
+
"learning_rate": 8.294862681744749e-06,
|
| 234979 |
+
"loss": 0.31,
|
| 234980 |
+
"step": 106590
|
| 234981 |
+
},
|
| 234982 |
+
{
|
| 234983 |
+
"epoch": 852.53,
|
| 234984 |
+
"learning_rate": 8.294781906300485e-06,
|
| 234985 |
+
"loss": 0.291,
|
| 234986 |
+
"step": 106595
|
| 234987 |
+
},
|
| 234988 |
+
{
|
| 234989 |
+
"epoch": 852.57,
|
| 234990 |
+
"learning_rate": 8.29470113085622e-06,
|
| 234991 |
+
"loss": 0.5147,
|
| 234992 |
+
"step": 106600
|
| 234993 |
+
},
|
| 234994 |
+
{
|
| 234995 |
+
"epoch": 852.61,
|
| 234996 |
+
"learning_rate": 8.294620355411955e-06,
|
| 234997 |
+
"loss": 1.1312,
|
| 234998 |
+
"step": 106605
|
| 234999 |
+
},
|
| 235000 |
+
{
|
| 235001 |
+
"epoch": 852.65,
|
| 235002 |
+
"learning_rate": 8.29453957996769e-06,
|
| 235003 |
+
"loss": 0.293,
|
| 235004 |
+
"step": 106610
|
| 235005 |
+
},
|
| 235006 |
+
{
|
| 235007 |
+
"epoch": 852.69,
|
| 235008 |
+
"learning_rate": 8.294458804523425e-06,
|
| 235009 |
+
"loss": 0.3372,
|
| 235010 |
+
"step": 106615
|
| 235011 |
+
},
|
| 235012 |
+
{
|
| 235013 |
+
"epoch": 852.73,
|
| 235014 |
+
"learning_rate": 8.29437802907916e-06,
|
| 235015 |
+
"loss": 0.3311,
|
| 235016 |
+
"step": 106620
|
| 235017 |
+
},
|
| 235018 |
+
{
|
| 235019 |
+
"epoch": 852.77,
|
| 235020 |
+
"learning_rate": 8.294297253634895e-06,
|
| 235021 |
+
"loss": 0.5603,
|
| 235022 |
+
"step": 106625
|
| 235023 |
+
},
|
| 235024 |
+
{
|
| 235025 |
+
"epoch": 852.81,
|
| 235026 |
+
"learning_rate": 8.29421647819063e-06,
|
| 235027 |
+
"loss": 1.0863,
|
| 235028 |
+
"step": 106630
|
| 235029 |
+
},
|
| 235030 |
+
{
|
| 235031 |
+
"epoch": 852.85,
|
| 235032 |
+
"learning_rate": 8.294135702746365e-06,
|
| 235033 |
+
"loss": 0.307,
|
| 235034 |
+
"step": 106635
|
| 235035 |
+
},
|
| 235036 |
+
{
|
| 235037 |
+
"epoch": 852.89,
|
| 235038 |
+
"learning_rate": 8.2940549273021e-06,
|
| 235039 |
+
"loss": 0.2969,
|
| 235040 |
+
"step": 106640
|
| 235041 |
+
},
|
| 235042 |
+
{
|
| 235043 |
+
"epoch": 852.93,
|
| 235044 |
+
"learning_rate": 8.293974151857835e-06,
|
| 235045 |
+
"loss": 0.3318,
|
| 235046 |
+
"step": 106645
|
| 235047 |
+
},
|
| 235048 |
+
{
|
| 235049 |
+
"epoch": 852.97,
|
| 235050 |
+
"learning_rate": 8.29389337641357e-06,
|
| 235051 |
+
"loss": 0.5622,
|
| 235052 |
+
"step": 106650
|
| 235053 |
+
},
|
| 235054 |
+
{
|
| 235055 |
+
"epoch": 853.0,
|
| 235056 |
+
"eval_loss": 0.35763418674468994,
|
| 235057 |
+
"eval_runtime": 43.1239,
|
| 235058 |
+
"eval_samples_per_second": 19.432,
|
| 235059 |
+
"eval_steps_per_second": 0.626,
|
| 235060 |
+
"eval_wer": 0.1841034582970067,
|
| 235061 |
+
"step": 106654
|
| 235062 |
+
},
|
| 235063 |
+
{
|
| 235064 |
+
"epoch": 860.01,
|
| 235065 |
+
"learning_rate": 8.293828756058158e-06,
|
| 235066 |
+
"loss": 0.3261,
|
| 235067 |
+
"step": 106655
|
| 235068 |
+
},
|
| 235069 |
+
{
|
| 235070 |
+
"epoch": 860.05,
|
| 235071 |
+
"learning_rate": 8.293747980613894e-06,
|
| 235072 |
+
"loss": 0.3114,
|
| 235073 |
+
"step": 106660
|
| 235074 |
+
},
|
| 235075 |
+
{
|
| 235076 |
+
"epoch": 860.09,
|
| 235077 |
+
"learning_rate": 8.293667205169628e-06,
|
| 235078 |
+
"loss": 0.3144,
|
| 235079 |
+
"step": 106665
|
| 235080 |
+
},
|
| 235081 |
+
{
|
| 235082 |
+
"epoch": 860.13,
|
| 235083 |
+
"learning_rate": 8.293586429725364e-06,
|
| 235084 |
+
"loss": 0.3585,
|
| 235085 |
+
"step": 106670
|
| 235086 |
+
},
|
| 235087 |
+
{
|
| 235088 |
+
"epoch": 860.17,
|
| 235089 |
+
"learning_rate": 8.2935056542811e-06,
|
| 235090 |
+
"loss": 0.5961,
|
| 235091 |
+
"step": 106675
|
| 235092 |
+
},
|
| 235093 |
+
{
|
| 235094 |
+
"epoch": 860.21,
|
| 235095 |
+
"learning_rate": 8.293424878836834e-06,
|
| 235096 |
+
"loss": 1.0914,
|
| 235097 |
+
"step": 106680
|
| 235098 |
+
},
|
| 235099 |
+
{
|
| 235100 |
+
"epoch": 860.25,
|
| 235101 |
+
"learning_rate": 8.29334410339257e-06,
|
| 235102 |
+
"loss": 0.3309,
|
| 235103 |
+
"step": 106685
|
| 235104 |
+
},
|
| 235105 |
+
{
|
| 235106 |
+
"epoch": 860.29,
|
| 235107 |
+
"learning_rate": 8.293263327948304e-06,
|
| 235108 |
+
"loss": 0.2584,
|
| 235109 |
+
"step": 106690
|
| 235110 |
+
},
|
| 235111 |
+
{
|
| 235112 |
+
"epoch": 860.33,
|
| 235113 |
+
"learning_rate": 8.29318255250404e-06,
|
| 235114 |
+
"loss": 0.3574,
|
| 235115 |
+
"step": 106695
|
| 235116 |
+
},
|
| 235117 |
+
{
|
| 235118 |
+
"epoch": 860.37,
|
| 235119 |
+
"learning_rate": 8.293101777059774e-06,
|
| 235120 |
+
"loss": 0.5731,
|
| 235121 |
+
"step": 106700
|
| 235122 |
+
},
|
| 235123 |
+
{
|
| 235124 |
+
"epoch": 860.41,
|
| 235125 |
+
"learning_rate": 8.29302100161551e-06,
|
| 235126 |
+
"loss": 1.1228,
|
| 235127 |
+
"step": 106705
|
| 235128 |
+
},
|
| 235129 |
+
{
|
| 235130 |
+
"epoch": 860.45,
|
| 235131 |
+
"learning_rate": 8.292940226171244e-06,
|
| 235132 |
+
"loss": 0.3394,
|
| 235133 |
+
"step": 106710
|
| 235134 |
+
},
|
| 235135 |
+
{
|
| 235136 |
+
"epoch": 860.49,
|
| 235137 |
+
"learning_rate": 8.29285945072698e-06,
|
| 235138 |
+
"loss": 0.302,
|
| 235139 |
+
"step": 106715
|
| 235140 |
+
},
|
| 235141 |
+
{
|
| 235142 |
+
"epoch": 860.53,
|
| 235143 |
+
"learning_rate": 8.292778675282714e-06,
|
| 235144 |
+
"loss": 0.3569,
|
| 235145 |
+
"step": 106720
|
| 235146 |
+
},
|
| 235147 |
+
{
|
| 235148 |
+
"epoch": 860.57,
|
| 235149 |
+
"learning_rate": 8.29269789983845e-06,
|
| 235150 |
+
"loss": 0.6229,
|
| 235151 |
+
"step": 106725
|
| 235152 |
+
},
|
| 235153 |
+
{
|
| 235154 |
+
"epoch": 860.61,
|
| 235155 |
+
"learning_rate": 8.292617124394184e-06,
|
| 235156 |
+
"loss": 1.1973,
|
| 235157 |
+
"step": 106730
|
| 235158 |
+
},
|
| 235159 |
+
{
|
| 235160 |
+
"epoch": 860.65,
|
| 235161 |
+
"learning_rate": 8.29253634894992e-06,
|
| 235162 |
+
"loss": 0.2899,
|
| 235163 |
+
"step": 106735
|
| 235164 |
+
},
|
| 235165 |
+
{
|
| 235166 |
+
"epoch": 860.69,
|
| 235167 |
+
"learning_rate": 8.292455573505655e-06,
|
| 235168 |
+
"loss": 0.265,
|
| 235169 |
+
"step": 106740
|
| 235170 |
+
},
|
| 235171 |
+
{
|
| 235172 |
+
"epoch": 860.73,
|
| 235173 |
+
"learning_rate": 8.29237479806139e-06,
|
| 235174 |
+
"loss": 0.2899,
|
| 235175 |
+
"step": 106745
|
| 235176 |
+
},
|
| 235177 |
+
{
|
| 235178 |
+
"epoch": 860.77,
|
| 235179 |
+
"learning_rate": 8.292294022617125e-06,
|
| 235180 |
+
"loss": 0.5312,
|
| 235181 |
+
"step": 106750
|
| 235182 |
+
},
|
| 235183 |
+
{
|
| 235184 |
+
"epoch": 860.81,
|
| 235185 |
+
"learning_rate": 8.29221324717286e-06,
|
| 235186 |
+
"loss": 1.1628,
|
| 235187 |
+
"step": 106755
|
| 235188 |
+
},
|
| 235189 |
+
{
|
| 235190 |
+
"epoch": 860.85,
|
| 235191 |
+
"learning_rate": 8.292132471728595e-06,
|
| 235192 |
+
"loss": 0.2695,
|
| 235193 |
+
"step": 106760
|
| 235194 |
+
},
|
| 235195 |
+
{
|
| 235196 |
+
"epoch": 860.89,
|
| 235197 |
+
"learning_rate": 8.29205169628433e-06,
|
| 235198 |
+
"loss": 0.3333,
|
| 235199 |
+
"step": 106765
|
| 235200 |
+
},
|
| 235201 |
+
{
|
| 235202 |
+
"epoch": 860.93,
|
| 235203 |
+
"learning_rate": 8.291970920840065e-06,
|
| 235204 |
+
"loss": 0.2823,
|
| 235205 |
+
"step": 106770
|
| 235206 |
+
},
|
| 235207 |
+
{
|
| 235208 |
+
"epoch": 860.97,
|
| 235209 |
+
"learning_rate": 8.2918901453958e-06,
|
| 235210 |
+
"loss": 0.6549,
|
| 235211 |
+
"step": 106775
|
| 235212 |
+
},
|
| 235213 |
+
{
|
| 235214 |
+
"epoch": 861.0,
|
| 235215 |
+
"eval_loss": 0.3717738687992096,
|
| 235216 |
+
"eval_runtime": 43.2755,
|
| 235217 |
+
"eval_samples_per_second": 19.364,
|
| 235218 |
+
"eval_steps_per_second": 0.624,
|
| 235219 |
+
"eval_wer": 0.18219537050516815,
|
| 235220 |
+
"step": 106778
|
| 235221 |
+
},
|
| 235222 |
+
{
|
| 235223 |
+
"epoch": 861.02,
|
| 235224 |
+
"learning_rate": 8.291809369951535e-06,
|
| 235225 |
+
"loss": 0.4225,
|
| 235226 |
+
"step": 106780
|
| 235227 |
+
},
|
| 235228 |
+
{
|
| 235229 |
+
"epoch": 861.06,
|
| 235230 |
+
"learning_rate": 8.29172859450727e-06,
|
| 235231 |
+
"loss": 0.2655,
|
| 235232 |
+
"step": 106785
|
| 235233 |
+
},
|
| 235234 |
+
{
|
| 235235 |
+
"epoch": 861.1,
|
| 235236 |
+
"learning_rate": 8.291647819063005e-06,
|
| 235237 |
+
"loss": 0.3107,
|
| 235238 |
+
"step": 106790
|
| 235239 |
+
},
|
| 235240 |
+
{
|
| 235241 |
+
"epoch": 861.14,
|
| 235242 |
+
"learning_rate": 8.29156704361874e-06,
|
| 235243 |
+
"loss": 0.319,
|
| 235244 |
+
"step": 106795
|
| 235245 |
+
},
|
| 235246 |
+
{
|
| 235247 |
+
"epoch": 861.18,
|
| 235248 |
+
"learning_rate": 8.291486268174475e-06,
|
| 235249 |
+
"loss": 0.5923,
|
| 235250 |
+
"step": 106800
|
| 235251 |
+
},
|
| 235252 |
+
{
|
| 235253 |
+
"epoch": 861.22,
|
| 235254 |
+
"learning_rate": 8.291405492730211e-06,
|
| 235255 |
+
"loss": 0.9439,
|
| 235256 |
+
"step": 106805
|
| 235257 |
+
},
|
| 235258 |
+
{
|
| 235259 |
+
"epoch": 861.26,
|
| 235260 |
+
"learning_rate": 8.291324717285945e-06,
|
| 235261 |
+
"loss": 0.2706,
|
| 235262 |
+
"step": 106810
|
| 235263 |
+
},
|
| 235264 |
+
{
|
| 235265 |
+
"epoch": 861.3,
|
| 235266 |
+
"learning_rate": 8.291243941841681e-06,
|
| 235267 |
+
"loss": 0.2889,
|
| 235268 |
+
"step": 106815
|
| 235269 |
+
},
|
| 235270 |
+
{
|
| 235271 |
+
"epoch": 861.34,
|
| 235272 |
+
"learning_rate": 8.291163166397415e-06,
|
| 235273 |
+
"loss": 0.3736,
|
| 235274 |
+
"step": 106820
|
| 235275 |
+
},
|
| 235276 |
+
{
|
| 235277 |
+
"epoch": 861.38,
|
| 235278 |
+
"learning_rate": 8.291082390953151e-06,
|
| 235279 |
+
"loss": 0.6033,
|
| 235280 |
+
"step": 106825
|
| 235281 |
+
},
|
| 235282 |
+
{
|
| 235283 |
+
"epoch": 861.42,
|
| 235284 |
+
"learning_rate": 8.291001615508885e-06,
|
| 235285 |
+
"loss": 1.0259,
|
| 235286 |
+
"step": 106830
|
| 235287 |
+
},
|
| 235288 |
+
{
|
| 235289 |
+
"epoch": 861.46,
|
| 235290 |
+
"learning_rate": 8.290920840064621e-06,
|
| 235291 |
+
"loss": 0.2559,
|
| 235292 |
+
"step": 106835
|
| 235293 |
+
},
|
| 235294 |
+
{
|
| 235295 |
+
"epoch": 861.5,
|
| 235296 |
+
"learning_rate": 8.290840064620355e-06,
|
| 235297 |
+
"loss": 0.3113,
|
| 235298 |
+
"step": 106840
|
| 235299 |
+
},
|
| 235300 |
+
{
|
| 235301 |
+
"epoch": 861.54,
|
| 235302 |
+
"learning_rate": 8.290759289176091e-06,
|
| 235303 |
+
"loss": 0.3365,
|
| 235304 |
+
"step": 106845
|
| 235305 |
+
},
|
| 235306 |
+
{
|
| 235307 |
+
"epoch": 861.58,
|
| 235308 |
+
"learning_rate": 8.290678513731825e-06,
|
| 235309 |
+
"loss": 0.7247,
|
| 235310 |
+
"step": 106850
|
| 235311 |
+
},
|
| 235312 |
+
{
|
| 235313 |
+
"epoch": 861.62,
|
| 235314 |
+
"learning_rate": 8.29059773828756e-06,
|
| 235315 |
+
"loss": 1.118,
|
| 235316 |
+
"step": 106855
|
| 235317 |
+
},
|
| 235318 |
+
{
|
| 235319 |
+
"epoch": 861.66,
|
| 235320 |
+
"learning_rate": 8.290516962843295e-06,
|
| 235321 |
+
"loss": 0.3131,
|
| 235322 |
+
"step": 106860
|
| 235323 |
+
},
|
| 235324 |
+
{
|
| 235325 |
+
"epoch": 861.7,
|
| 235326 |
+
"learning_rate": 8.29043618739903e-06,
|
| 235327 |
+
"loss": 0.3114,
|
| 235328 |
+
"step": 106865
|
| 235329 |
+
},
|
| 235330 |
+
{
|
| 235331 |
+
"epoch": 861.74,
|
| 235332 |
+
"learning_rate": 8.290355411954767e-06,
|
| 235333 |
+
"loss": 0.295,
|
| 235334 |
+
"step": 106870
|
| 235335 |
+
},
|
| 235336 |
+
{
|
| 235337 |
+
"epoch": 861.78,
|
| 235338 |
+
"learning_rate": 8.2902746365105e-06,
|
| 235339 |
+
"loss": 0.6526,
|
| 235340 |
+
"step": 106875
|
| 235341 |
+
},
|
| 235342 |
+
{
|
| 235343 |
+
"epoch": 861.82,
|
| 235344 |
+
"learning_rate": 8.290193861066237e-06,
|
| 235345 |
+
"loss": 1.0912,
|
| 235346 |
+
"step": 106880
|
| 235347 |
+
},
|
| 235348 |
+
{
|
| 235349 |
+
"epoch": 861.86,
|
| 235350 |
+
"learning_rate": 8.29011308562197e-06,
|
| 235351 |
+
"loss": 0.3106,
|
| 235352 |
+
"step": 106885
|
| 235353 |
+
},
|
| 235354 |
+
{
|
| 235355 |
+
"epoch": 861.9,
|
| 235356 |
+
"learning_rate": 8.290032310177707e-06,
|
| 235357 |
+
"loss": 0.3466,
|
| 235358 |
+
"step": 106890
|
| 235359 |
+
},
|
| 235360 |
+
{
|
| 235361 |
+
"epoch": 861.94,
|
| 235362 |
+
"learning_rate": 8.28995153473344e-06,
|
| 235363 |
+
"loss": 0.3238,
|
| 235364 |
+
"step": 106895
|
| 235365 |
+
},
|
| 235366 |
+
{
|
| 235367 |
+
"epoch": 861.98,
|
| 235368 |
+
"learning_rate": 8.289870759289177e-06,
|
| 235369 |
+
"loss": 0.7486,
|
| 235370 |
+
"step": 106900
|
| 235371 |
+
},
|
| 235372 |
+
{
|
| 235373 |
+
"epoch": 862.0,
|
| 235374 |
+
"eval_loss": 0.3670842945575714,
|
| 235375 |
+
"eval_runtime": 42.8155,
|
| 235376 |
+
"eval_samples_per_second": 19.572,
|
| 235377 |
+
"eval_steps_per_second": 0.631,
|
| 235378 |
+
"eval_wer": 0.18326255674330064,
|
| 235379 |
+
"step": 106902
|
| 235380 |
+
},
|
| 235381 |
+
{
|
| 235382 |
+
"epoch": 855.02,
|
| 235383 |
+
"learning_rate": 8.28978998384491e-06,
|
| 235384 |
+
"loss": 0.3439,
|
| 235385 |
+
"step": 106905
|
| 235386 |
+
},
|
| 235387 |
+
{
|
| 235388 |
+
"epoch": 855.06,
|
| 235389 |
+
"learning_rate": 8.289709208400647e-06,
|
| 235390 |
+
"loss": 0.294,
|
| 235391 |
+
"step": 106910
|
| 235392 |
+
},
|
| 235393 |
+
{
|
| 235394 |
+
"epoch": 855.1,
|
| 235395 |
+
"learning_rate": 8.28962843295638e-06,
|
| 235396 |
+
"loss": 0.2934,
|
| 235397 |
+
"step": 106915
|
| 235398 |
+
},
|
| 235399 |
+
{
|
| 235400 |
+
"epoch": 855.14,
|
| 235401 |
+
"learning_rate": 8.289547657512117e-06,
|
| 235402 |
+
"loss": 0.3603,
|
| 235403 |
+
"step": 106920
|
| 235404 |
+
},
|
| 235405 |
+
{
|
| 235406 |
+
"epoch": 855.18,
|
| 235407 |
+
"learning_rate": 8.289466882067852e-06,
|
| 235408 |
+
"loss": 0.763,
|
| 235409 |
+
"step": 106925
|
| 235410 |
+
},
|
| 235411 |
+
{
|
| 235412 |
+
"epoch": 855.22,
|
| 235413 |
+
"learning_rate": 8.289386106623586e-06,
|
| 235414 |
+
"loss": 0.8405,
|
| 235415 |
+
"step": 106930
|
| 235416 |
+
},
|
| 235417 |
+
{
|
| 235418 |
+
"epoch": 855.26,
|
| 235419 |
+
"learning_rate": 8.289305331179322e-06,
|
| 235420 |
+
"loss": 0.3088,
|
| 235421 |
+
"step": 106935
|
| 235422 |
+
},
|
| 235423 |
+
{
|
| 235424 |
+
"epoch": 855.3,
|
| 235425 |
+
"learning_rate": 8.289224555735056e-06,
|
| 235426 |
+
"loss": 0.2922,
|
| 235427 |
+
"step": 106940
|
| 235428 |
+
},
|
| 235429 |
+
{
|
| 235430 |
+
"epoch": 855.34,
|
| 235431 |
+
"learning_rate": 8.289143780290792e-06,
|
| 235432 |
+
"loss": 0.4262,
|
| 235433 |
+
"step": 106945
|
| 235434 |
+
},
|
| 235435 |
+
{
|
| 235436 |
+
"epoch": 855.38,
|
| 235437 |
+
"learning_rate": 8.289063004846526e-06,
|
| 235438 |
+
"loss": 0.7446,
|
| 235439 |
+
"step": 106950
|
| 235440 |
+
},
|
| 235441 |
+
{
|
| 235442 |
+
"epoch": 855.42,
|
| 235443 |
+
"learning_rate": 8.288982229402262e-06,
|
| 235444 |
+
"loss": 0.8839,
|
| 235445 |
+
"step": 106955
|
| 235446 |
+
},
|
| 235447 |
+
{
|
| 235448 |
+
"epoch": 855.46,
|
| 235449 |
+
"learning_rate": 8.288901453957996e-06,
|
| 235450 |
+
"loss": 0.2554,
|
| 235451 |
+
"step": 106960
|
| 235452 |
+
},
|
| 235453 |
+
{
|
| 235454 |
+
"epoch": 855.5,
|
| 235455 |
+
"learning_rate": 8.288820678513732e-06,
|
| 235456 |
+
"loss": 0.2864,
|
| 235457 |
+
"step": 106965
|
| 235458 |
+
},
|
| 235459 |
+
{
|
| 235460 |
+
"epoch": 855.54,
|
| 235461 |
+
"learning_rate": 8.288739903069466e-06,
|
| 235462 |
+
"loss": 0.3054,
|
| 235463 |
+
"step": 106970
|
| 235464 |
+
},
|
| 235465 |
+
{
|
| 235466 |
+
"epoch": 855.58,
|
| 235467 |
+
"learning_rate": 8.288659127625202e-06,
|
| 235468 |
+
"loss": 0.7656,
|
| 235469 |
+
"step": 106975
|
| 235470 |
+
},
|
| 235471 |
+
{
|
| 235472 |
+
"epoch": 855.62,
|
| 235473 |
+
"learning_rate": 8.288578352180938e-06,
|
| 235474 |
+
"loss": 0.9497,
|
| 235475 |
+
"step": 106980
|
| 235476 |
+
},
|
| 235477 |
+
{
|
| 235478 |
+
"epoch": 855.66,
|
| 235479 |
+
"learning_rate": 8.288497576736672e-06,
|
| 235480 |
+
"loss": 0.2572,
|
| 235481 |
+
"step": 106985
|
| 235482 |
+
},
|
| 235483 |
+
{
|
| 235484 |
+
"epoch": 855.7,
|
| 235485 |
+
"learning_rate": 8.288416801292408e-06,
|
| 235486 |
+
"loss": 0.3056,
|
| 235487 |
+
"step": 106990
|
| 235488 |
+
},
|
| 235489 |
+
{
|
| 235490 |
+
"epoch": 855.74,
|
| 235491 |
+
"learning_rate": 8.288336025848142e-06,
|
| 235492 |
+
"loss": 0.3994,
|
| 235493 |
+
"step": 106995
|
| 235494 |
+
},
|
| 235495 |
+
{
|
| 235496 |
+
"epoch": 855.78,
|
| 235497 |
+
"learning_rate": 8.288255250403878e-06,
|
| 235498 |
+
"loss": 0.7842,
|
| 235499 |
+
"step": 107000
|
| 235500 |
+
},
|
| 235501 |
+
{
|
| 235502 |
+
"epoch": 855.82,
|
| 235503 |
+
"learning_rate": 8.288174474959612e-06,
|
| 235504 |
+
"loss": 0.9429,
|
| 235505 |
+
"step": 107005
|
| 235506 |
+
},
|
| 235507 |
+
{
|
| 235508 |
+
"epoch": 855.86,
|
| 235509 |
+
"learning_rate": 8.288093699515348e-06,
|
| 235510 |
+
"loss": 0.3151,
|
| 235511 |
+
"step": 107010
|
| 235512 |
+
},
|
| 235513 |
+
{
|
| 235514 |
+
"epoch": 855.9,
|
| 235515 |
+
"learning_rate": 8.288012924071082e-06,
|
| 235516 |
+
"loss": 0.3058,
|
| 235517 |
+
"step": 107015
|
| 235518 |
+
},
|
| 235519 |
+
{
|
| 235520 |
+
"epoch": 855.94,
|
| 235521 |
+
"learning_rate": 8.287932148626818e-06,
|
| 235522 |
+
"loss": 0.4205,
|
| 235523 |
+
"step": 107020
|
| 235524 |
+
},
|
| 235525 |
+
{
|
| 235526 |
+
"epoch": 855.98,
|
| 235527 |
+
"learning_rate": 8.287851373182552e-06,
|
| 235528 |
+
"loss": 0.7296,
|
| 235529 |
+
"step": 107025
|
| 235530 |
+
},
|
| 235531 |
+
{
|
| 235532 |
+
"epoch": 856.0,
|
| 235533 |
+
"eval_loss": 0.35013189911842346,
|
| 235534 |
+
"eval_runtime": 42.4934,
|
| 235535 |
+
"eval_samples_per_second": 19.697,
|
| 235536 |
+
"eval_steps_per_second": 0.635,
|
| 235537 |
+
"eval_wer": 0.18572358429160338,
|
| 235538 |
+
"step": 107027
|
| 235539 |
}
|
| 235540 |
],
|
| 235541 |
"max_steps": 625000,
|
| 235542 |
"num_train_epochs": 5000,
|
| 235543 |
+
"total_flos": 3.011837425036952e+20,
|
| 235544 |
"trial_name": null,
|
| 235545 |
"trial_params": null
|
| 235546 |
}
|
model-bin/finetune/base/{checkpoint-106405 → checkpoint-107027}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630028869.9681413/events.out.tfevents.1630028869.52f5c7e305a3.886.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7905fbff6c690f998c8bea13c1564f5e3c3e273a9f4494e804e1a7da99fbfe1f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630029382.1230361/events.out.tfevents.1630029382.52f5c7e305a3.886.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af162c755a5006ae40cdfab7c477004c330ac4a10341f751f55c9917b9be81e5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630029880.354848/events.out.tfevents.1630029880.52f5c7e305a3.886.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1050f3858f2055b4f7be50c3dfcc2f2ec32458a237d12ef98a91c811d99deef9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630030344.963895/events.out.tfevents.1630030344.52f5c7e305a3.886.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83f2057fd326383a39100f3ca1056bf6945e381e177b4c25bc6cfb4c49c78abc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630030813.4653862/events.out.tfevents.1630030813.52f5c7e305a3.886.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1193c7e3d97e31e91de66f2404d79480c258bf23c45c214d7b796d20fb35254a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630028869.52f5c7e305a3.886.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6061c88e9c3db1efa1d809010bbf8c646521faee0f47b938b0553c43b05ff81
|
| 3 |
+
size 8470
|
model-bin/finetune/base/log/events.out.tfevents.1630029382.52f5c7e305a3.886.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f8e751ff11c41fd6f91046886ade6e0102baa1dbaf562d2c08cfb21b07a85936
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630029880.52f5c7e305a3.886.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:93245ec0dbefa3a8da066b350be38a9cbace47fed6a7daec6530ed22d4801781
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630030344.52f5c7e305a3.886.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe472bbe3db9cd39ecae52e0ac3b6631e09fea569cddc6b8c6c76dc237728775
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630030813.52f5c7e305a3.886.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52795d5429b46713a073e137337c1493d76752d723672c2cf5d29afdcfb83cd2
|
| 3 |
+
size 8622
|