"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629855529.4768627/events.out.tfevents.1629855529.7e498afd5545.905.13 +3 -0
- model-bin/finetune/base/log/1629856180.0808132/events.out.tfevents.1629856180.7e498afd5545.905.15 +3 -0
- model-bin/finetune/base/log/1629856939.1696703/events.out.tfevents.1629856939.7e498afd5545.905.17 +3 -0
- model-bin/finetune/base/log/1629857599.5043178/events.out.tfevents.1629857599.7e498afd5545.905.19 +3 -0
- model-bin/finetune/base/log/1629858276.3230329/events.out.tfevents.1629858276.7e498afd5545.905.21 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629855529.7e498afd5545.905.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629856180.7e498afd5545.905.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629856939.7e498afd5545.905.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629857599.7e498afd5545.905.18 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629858276.7e498afd5545.905.20 +3 -0
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/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 sha256:1238d3e0407c1b5eab603ddadea5d697c591b6ed7191dded1195fc755b757e8e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/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 sha256:254d4cf0f3862cae7773484aa9cfabaf779e738f449d472541eb9a091e3fb36a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/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 sha256:9c9688cedf4af3de5be5cd9e8d216e7a95ddb6ceceb3ddf456a28ff4bf1c92c9
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/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 sha256:70f0230abe0d9e7f4ba777c6d4ca204523033a231565d6e066189da086fa610e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/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:e207074215dc1be4bcfdc4da83cf51e498c21731dc990103b72674b1b112cc7a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -188664,11 +188664,800 @@
|
|
| 188664 |
"eval_steps_per_second": 0.67,
|
| 188665 |
"eval_wer": 0.19290399522957663,
|
| 188666 |
"step": 70187
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 188667 |
}
|
| 188668 |
],
|
| 188669 |
"max_steps": 620000,
|
| 188670 |
"num_train_epochs": 5000,
|
| 188671 |
-
"total_flos": 1.
|
| 188672 |
"trial_name": null,
|
| 188673 |
"trial_params": null
|
| 188674 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 570.995983935743,
|
| 5 |
+
"global_step": 70809,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 188664 |
"eval_steps_per_second": 0.67,
|
| 188665 |
"eval_wer": 0.19290399522957663,
|
| 188666 |
"step": 70187
|
| 188667 |
+
},
|
| 188668 |
+
{
|
| 188669 |
+
"epoch": 561.02,
|
| 188670 |
+
"learning_rate": 8.891602564102566e-06,
|
| 188671 |
+
"loss": 0.3236,
|
| 188672 |
+
"step": 70190
|
| 188673 |
+
},
|
| 188674 |
+
{
|
| 188675 |
+
"epoch": 561.06,
|
| 188676 |
+
"learning_rate": 8.891522435897435e-06,
|
| 188677 |
+
"loss": 0.3191,
|
| 188678 |
+
"step": 70195
|
| 188679 |
+
},
|
| 188680 |
+
{
|
| 188681 |
+
"epoch": 561.1,
|
| 188682 |
+
"learning_rate": 8.891442307692308e-06,
|
| 188683 |
+
"loss": 0.3566,
|
| 188684 |
+
"step": 70200
|
| 188685 |
+
},
|
| 188686 |
+
{
|
| 188687 |
+
"epoch": 561.14,
|
| 188688 |
+
"learning_rate": 8.891362179487181e-06,
|
| 188689 |
+
"loss": 0.3839,
|
| 188690 |
+
"step": 70205
|
| 188691 |
+
},
|
| 188692 |
+
{
|
| 188693 |
+
"epoch": 561.18,
|
| 188694 |
+
"learning_rate": 8.891282051282051e-06,
|
| 188695 |
+
"loss": 0.7562,
|
| 188696 |
+
"step": 70210
|
| 188697 |
+
},
|
| 188698 |
+
{
|
| 188699 |
+
"epoch": 561.22,
|
| 188700 |
+
"learning_rate": 8.891201923076924e-06,
|
| 188701 |
+
"loss": 0.8426,
|
| 188702 |
+
"step": 70215
|
| 188703 |
+
},
|
| 188704 |
+
{
|
| 188705 |
+
"epoch": 561.26,
|
| 188706 |
+
"learning_rate": 8.891121794871795e-06,
|
| 188707 |
+
"loss": 0.2785,
|
| 188708 |
+
"step": 70220
|
| 188709 |
+
},
|
| 188710 |
+
{
|
| 188711 |
+
"epoch": 561.3,
|
| 188712 |
+
"learning_rate": 8.891041666666667e-06,
|
| 188713 |
+
"loss": 0.2801,
|
| 188714 |
+
"step": 70225
|
| 188715 |
+
},
|
| 188716 |
+
{
|
| 188717 |
+
"epoch": 561.34,
|
| 188718 |
+
"learning_rate": 8.890961538461538e-06,
|
| 188719 |
+
"loss": 0.442,
|
| 188720 |
+
"step": 70230
|
| 188721 |
+
},
|
| 188722 |
+
{
|
| 188723 |
+
"epoch": 561.38,
|
| 188724 |
+
"learning_rate": 8.890881410256411e-06,
|
| 188725 |
+
"loss": 0.7326,
|
| 188726 |
+
"step": 70235
|
| 188727 |
+
},
|
| 188728 |
+
{
|
| 188729 |
+
"epoch": 561.42,
|
| 188730 |
+
"learning_rate": 8.890801282051283e-06,
|
| 188731 |
+
"loss": 1.0664,
|
| 188732 |
+
"step": 70240
|
| 188733 |
+
},
|
| 188734 |
+
{
|
| 188735 |
+
"epoch": 561.46,
|
| 188736 |
+
"learning_rate": 8.890721153846154e-06,
|
| 188737 |
+
"loss": 0.2819,
|
| 188738 |
+
"step": 70245
|
| 188739 |
+
},
|
| 188740 |
+
{
|
| 188741 |
+
"epoch": 561.5,
|
| 188742 |
+
"learning_rate": 8.890641025641025e-06,
|
| 188743 |
+
"loss": 0.3275,
|
| 188744 |
+
"step": 70250
|
| 188745 |
+
},
|
| 188746 |
+
{
|
| 188747 |
+
"epoch": 561.54,
|
| 188748 |
+
"learning_rate": 8.890560897435898e-06,
|
| 188749 |
+
"loss": 0.4213,
|
| 188750 |
+
"step": 70255
|
| 188751 |
+
},
|
| 188752 |
+
{
|
| 188753 |
+
"epoch": 561.58,
|
| 188754 |
+
"learning_rate": 8.89048076923077e-06,
|
| 188755 |
+
"loss": 0.908,
|
| 188756 |
+
"step": 70260
|
| 188757 |
+
},
|
| 188758 |
+
{
|
| 188759 |
+
"epoch": 561.62,
|
| 188760 |
+
"learning_rate": 8.890400641025641e-06,
|
| 188761 |
+
"loss": 0.8696,
|
| 188762 |
+
"step": 70265
|
| 188763 |
+
},
|
| 188764 |
+
{
|
| 188765 |
+
"epoch": 561.66,
|
| 188766 |
+
"learning_rate": 8.890320512820514e-06,
|
| 188767 |
+
"loss": 0.3694,
|
| 188768 |
+
"step": 70270
|
| 188769 |
+
},
|
| 188770 |
+
{
|
| 188771 |
+
"epoch": 561.7,
|
| 188772 |
+
"learning_rate": 8.890240384615386e-06,
|
| 188773 |
+
"loss": 0.2928,
|
| 188774 |
+
"step": 70275
|
| 188775 |
+
},
|
| 188776 |
+
{
|
| 188777 |
+
"epoch": 561.74,
|
| 188778 |
+
"learning_rate": 8.890160256410257e-06,
|
| 188779 |
+
"loss": 0.3682,
|
| 188780 |
+
"step": 70280
|
| 188781 |
+
},
|
| 188782 |
+
{
|
| 188783 |
+
"epoch": 561.78,
|
| 188784 |
+
"learning_rate": 8.890080128205128e-06,
|
| 188785 |
+
"loss": 0.8639,
|
| 188786 |
+
"step": 70285
|
| 188787 |
+
},
|
| 188788 |
+
{
|
| 188789 |
+
"epoch": 561.82,
|
| 188790 |
+
"learning_rate": 8.890000000000001e-06,
|
| 188791 |
+
"loss": 0.9488,
|
| 188792 |
+
"step": 70290
|
| 188793 |
+
},
|
| 188794 |
+
{
|
| 188795 |
+
"epoch": 561.86,
|
| 188796 |
+
"learning_rate": 8.889919871794873e-06,
|
| 188797 |
+
"loss": 0.3665,
|
| 188798 |
+
"step": 70295
|
| 188799 |
+
},
|
| 188800 |
+
{
|
| 188801 |
+
"epoch": 561.9,
|
| 188802 |
+
"learning_rate": 8.889839743589744e-06,
|
| 188803 |
+
"loss": 0.3282,
|
| 188804 |
+
"step": 70300
|
| 188805 |
+
},
|
| 188806 |
+
{
|
| 188807 |
+
"epoch": 561.94,
|
| 188808 |
+
"learning_rate": 8.889759615384617e-06,
|
| 188809 |
+
"loss": 0.4235,
|
| 188810 |
+
"step": 70305
|
| 188811 |
+
},
|
| 188812 |
+
{
|
| 188813 |
+
"epoch": 561.98,
|
| 188814 |
+
"learning_rate": 8.889679487179488e-06,
|
| 188815 |
+
"loss": 0.8138,
|
| 188816 |
+
"step": 70310
|
| 188817 |
+
},
|
| 188818 |
+
{
|
| 188819 |
+
"epoch": 562.0,
|
| 188820 |
+
"eval_loss": 0.3786197006702423,
|
| 188821 |
+
"eval_runtime": 39.7295,
|
| 188822 |
+
"eval_samples_per_second": 21.118,
|
| 188823 |
+
"eval_steps_per_second": 0.68,
|
| 188824 |
+
"eval_wer": 0.197922893006575,
|
| 188825 |
+
"step": 70312
|
| 188826 |
+
},
|
| 188827 |
+
{
|
| 188828 |
+
"epoch": 562.02,
|
| 188829 |
+
"learning_rate": 8.88959935897436e-06,
|
| 188830 |
+
"loss": 0.3524,
|
| 188831 |
+
"step": 70315
|
| 188832 |
+
},
|
| 188833 |
+
{
|
| 188834 |
+
"epoch": 562.06,
|
| 188835 |
+
"learning_rate": 8.889519230769231e-06,
|
| 188836 |
+
"loss": 0.3185,
|
| 188837 |
+
"step": 70320
|
| 188838 |
+
},
|
| 188839 |
+
{
|
| 188840 |
+
"epoch": 562.1,
|
| 188841 |
+
"learning_rate": 8.889439102564104e-06,
|
| 188842 |
+
"loss": 0.3284,
|
| 188843 |
+
"step": 70325
|
| 188844 |
+
},
|
| 188845 |
+
{
|
| 188846 |
+
"epoch": 562.14,
|
| 188847 |
+
"learning_rate": 8.889358974358976e-06,
|
| 188848 |
+
"loss": 0.388,
|
| 188849 |
+
"step": 70330
|
| 188850 |
+
},
|
| 188851 |
+
{
|
| 188852 |
+
"epoch": 562.18,
|
| 188853 |
+
"learning_rate": 8.889278846153847e-06,
|
| 188854 |
+
"loss": 0.8204,
|
| 188855 |
+
"step": 70335
|
| 188856 |
+
},
|
| 188857 |
+
{
|
| 188858 |
+
"epoch": 562.22,
|
| 188859 |
+
"learning_rate": 8.889198717948718e-06,
|
| 188860 |
+
"loss": 0.8117,
|
| 188861 |
+
"step": 70340
|
| 188862 |
+
},
|
| 188863 |
+
{
|
| 188864 |
+
"epoch": 562.26,
|
| 188865 |
+
"learning_rate": 8.889118589743591e-06,
|
| 188866 |
+
"loss": 0.3556,
|
| 188867 |
+
"step": 70345
|
| 188868 |
+
},
|
| 188869 |
+
{
|
| 188870 |
+
"epoch": 562.3,
|
| 188871 |
+
"learning_rate": 8.889038461538461e-06,
|
| 188872 |
+
"loss": 0.3245,
|
| 188873 |
+
"step": 70350
|
| 188874 |
+
},
|
| 188875 |
+
{
|
| 188876 |
+
"epoch": 562.34,
|
| 188877 |
+
"learning_rate": 8.888958333333334e-06,
|
| 188878 |
+
"loss": 0.3523,
|
| 188879 |
+
"step": 70355
|
| 188880 |
+
},
|
| 188881 |
+
{
|
| 188882 |
+
"epoch": 562.38,
|
| 188883 |
+
"learning_rate": 8.888878205128207e-06,
|
| 188884 |
+
"loss": 0.7863,
|
| 188885 |
+
"step": 70360
|
| 188886 |
+
},
|
| 188887 |
+
{
|
| 188888 |
+
"epoch": 562.42,
|
| 188889 |
+
"learning_rate": 8.888798076923077e-06,
|
| 188890 |
+
"loss": 0.8844,
|
| 188891 |
+
"step": 70365
|
| 188892 |
+
},
|
| 188893 |
+
{
|
| 188894 |
+
"epoch": 562.46,
|
| 188895 |
+
"learning_rate": 8.88871794871795e-06,
|
| 188896 |
+
"loss": 0.362,
|
| 188897 |
+
"step": 70370
|
| 188898 |
+
},
|
| 188899 |
+
{
|
| 188900 |
+
"epoch": 562.5,
|
| 188901 |
+
"learning_rate": 8.888637820512821e-06,
|
| 188902 |
+
"loss": 0.3011,
|
| 188903 |
+
"step": 70375
|
| 188904 |
+
},
|
| 188905 |
+
{
|
| 188906 |
+
"epoch": 562.54,
|
| 188907 |
+
"learning_rate": 8.888557692307693e-06,
|
| 188908 |
+
"loss": 0.37,
|
| 188909 |
+
"step": 70380
|
| 188910 |
+
},
|
| 188911 |
+
{
|
| 188912 |
+
"epoch": 562.58,
|
| 188913 |
+
"learning_rate": 8.888477564102564e-06,
|
| 188914 |
+
"loss": 0.6502,
|
| 188915 |
+
"step": 70385
|
| 188916 |
+
},
|
| 188917 |
+
{
|
| 188918 |
+
"epoch": 562.62,
|
| 188919 |
+
"learning_rate": 8.888397435897437e-06,
|
| 188920 |
+
"loss": 0.8631,
|
| 188921 |
+
"step": 70390
|
| 188922 |
+
},
|
| 188923 |
+
{
|
| 188924 |
+
"epoch": 562.66,
|
| 188925 |
+
"learning_rate": 8.888317307692308e-06,
|
| 188926 |
+
"loss": 0.4869,
|
| 188927 |
+
"step": 70395
|
| 188928 |
+
},
|
| 188929 |
+
{
|
| 188930 |
+
"epoch": 562.7,
|
| 188931 |
+
"learning_rate": 8.88823717948718e-06,
|
| 188932 |
+
"loss": 0.3648,
|
| 188933 |
+
"step": 70400
|
| 188934 |
+
},
|
| 188935 |
+
{
|
| 188936 |
+
"epoch": 562.74,
|
| 188937 |
+
"learning_rate": 8.888157051282053e-06,
|
| 188938 |
+
"loss": 0.3954,
|
| 188939 |
+
"step": 70405
|
| 188940 |
+
},
|
| 188941 |
+
{
|
| 188942 |
+
"epoch": 562.78,
|
| 188943 |
+
"learning_rate": 8.888076923076924e-06,
|
| 188944 |
+
"loss": 0.7799,
|
| 188945 |
+
"step": 70410
|
| 188946 |
+
},
|
| 188947 |
+
{
|
| 188948 |
+
"epoch": 562.82,
|
| 188949 |
+
"learning_rate": 8.887996794871795e-06,
|
| 188950 |
+
"loss": 0.7664,
|
| 188951 |
+
"step": 70415
|
| 188952 |
+
},
|
| 188953 |
+
{
|
| 188954 |
+
"epoch": 562.86,
|
| 188955 |
+
"learning_rate": 8.887916666666667e-06,
|
| 188956 |
+
"loss": 0.316,
|
| 188957 |
+
"step": 70420
|
| 188958 |
+
},
|
| 188959 |
+
{
|
| 188960 |
+
"epoch": 562.9,
|
| 188961 |
+
"learning_rate": 8.88783653846154e-06,
|
| 188962 |
+
"loss": 0.3008,
|
| 188963 |
+
"step": 70425
|
| 188964 |
+
},
|
| 188965 |
+
{
|
| 188966 |
+
"epoch": 562.94,
|
| 188967 |
+
"learning_rate": 8.887756410256411e-06,
|
| 188968 |
+
"loss": 0.4025,
|
| 188969 |
+
"step": 70430
|
| 188970 |
+
},
|
| 188971 |
+
{
|
| 188972 |
+
"epoch": 562.98,
|
| 188973 |
+
"learning_rate": 8.887676282051283e-06,
|
| 188974 |
+
"loss": 0.7471,
|
| 188975 |
+
"step": 70435
|
| 188976 |
+
},
|
| 188977 |
+
{
|
| 188978 |
+
"epoch": 563.0,
|
| 188979 |
+
"eval_loss": 0.46410617232322693,
|
| 188980 |
+
"eval_runtime": 42.712,
|
| 188981 |
+
"eval_samples_per_second": 19.643,
|
| 188982 |
+
"eval_steps_per_second": 0.632,
|
| 188983 |
+
"eval_wer": 0.19668026042784575,
|
| 188984 |
+
"step": 70437
|
| 188985 |
+
},
|
| 188986 |
+
{
|
| 188987 |
+
"epoch": 568.02,
|
| 188988 |
+
"learning_rate": 8.887596153846154e-06,
|
| 188989 |
+
"loss": 0.3198,
|
| 188990 |
+
"step": 70440
|
| 188991 |
+
},
|
| 188992 |
+
{
|
| 188993 |
+
"epoch": 568.06,
|
| 188994 |
+
"learning_rate": 8.887516025641027e-06,
|
| 188995 |
+
"loss": 0.3777,
|
| 188996 |
+
"step": 70445
|
| 188997 |
+
},
|
| 188998 |
+
{
|
| 188999 |
+
"epoch": 568.1,
|
| 189000 |
+
"learning_rate": 8.887435897435898e-06,
|
| 189001 |
+
"loss": 0.365,
|
| 189002 |
+
"step": 70450
|
| 189003 |
+
},
|
| 189004 |
+
{
|
| 189005 |
+
"epoch": 568.14,
|
| 189006 |
+
"learning_rate": 8.88735576923077e-06,
|
| 189007 |
+
"loss": 0.3966,
|
| 189008 |
+
"step": 70455
|
| 189009 |
+
},
|
| 189010 |
+
{
|
| 189011 |
+
"epoch": 568.18,
|
| 189012 |
+
"learning_rate": 8.887275641025643e-06,
|
| 189013 |
+
"loss": 0.8013,
|
| 189014 |
+
"step": 70460
|
| 189015 |
+
},
|
| 189016 |
+
{
|
| 189017 |
+
"epoch": 568.22,
|
| 189018 |
+
"learning_rate": 8.887195512820514e-06,
|
| 189019 |
+
"loss": 0.8617,
|
| 189020 |
+
"step": 70465
|
| 189021 |
+
},
|
| 189022 |
+
{
|
| 189023 |
+
"epoch": 568.27,
|
| 189024 |
+
"learning_rate": 8.887115384615385e-06,
|
| 189025 |
+
"loss": 0.3124,
|
| 189026 |
+
"step": 70470
|
| 189027 |
+
},
|
| 189028 |
+
{
|
| 189029 |
+
"epoch": 568.31,
|
| 189030 |
+
"learning_rate": 8.887035256410257e-06,
|
| 189031 |
+
"loss": 0.3463,
|
| 189032 |
+
"step": 70475
|
| 189033 |
+
},
|
| 189034 |
+
{
|
| 189035 |
+
"epoch": 568.35,
|
| 189036 |
+
"learning_rate": 8.88695512820513e-06,
|
| 189037 |
+
"loss": 0.3506,
|
| 189038 |
+
"step": 70480
|
| 189039 |
+
},
|
| 189040 |
+
{
|
| 189041 |
+
"epoch": 568.39,
|
| 189042 |
+
"learning_rate": 8.886875000000001e-06,
|
| 189043 |
+
"loss": 0.8531,
|
| 189044 |
+
"step": 70485
|
| 189045 |
+
},
|
| 189046 |
+
{
|
| 189047 |
+
"epoch": 568.43,
|
| 189048 |
+
"learning_rate": 8.886794871794873e-06,
|
| 189049 |
+
"loss": 0.9379,
|
| 189050 |
+
"step": 70490
|
| 189051 |
+
},
|
| 189052 |
+
{
|
| 189053 |
+
"epoch": 568.47,
|
| 189054 |
+
"learning_rate": 8.886714743589744e-06,
|
| 189055 |
+
"loss": 0.3052,
|
| 189056 |
+
"step": 70495
|
| 189057 |
+
},
|
| 189058 |
+
{
|
| 189059 |
+
"epoch": 568.51,
|
| 189060 |
+
"learning_rate": 8.886634615384617e-06,
|
| 189061 |
+
"loss": 0.2875,
|
| 189062 |
+
"step": 70500
|
| 189063 |
+
},
|
| 189064 |
+
{
|
| 189065 |
+
"epoch": 568.55,
|
| 189066 |
+
"learning_rate": 8.886554487179488e-06,
|
| 189067 |
+
"loss": 0.3442,
|
| 189068 |
+
"step": 70505
|
| 189069 |
+
},
|
| 189070 |
+
{
|
| 189071 |
+
"epoch": 568.59,
|
| 189072 |
+
"learning_rate": 8.88647435897436e-06,
|
| 189073 |
+
"loss": 0.7231,
|
| 189074 |
+
"step": 70510
|
| 189075 |
+
},
|
| 189076 |
+
{
|
| 189077 |
+
"epoch": 568.63,
|
| 189078 |
+
"learning_rate": 8.886394230769233e-06,
|
| 189079 |
+
"loss": 0.8665,
|
| 189080 |
+
"step": 70515
|
| 189081 |
+
},
|
| 189082 |
+
{
|
| 189083 |
+
"epoch": 568.67,
|
| 189084 |
+
"learning_rate": 8.886314102564102e-06,
|
| 189085 |
+
"loss": 0.3137,
|
| 189086 |
+
"step": 70520
|
| 189087 |
+
},
|
| 189088 |
+
{
|
| 189089 |
+
"epoch": 568.71,
|
| 189090 |
+
"learning_rate": 8.886233974358975e-06,
|
| 189091 |
+
"loss": 0.3238,
|
| 189092 |
+
"step": 70525
|
| 189093 |
+
},
|
| 189094 |
+
{
|
| 189095 |
+
"epoch": 568.75,
|
| 189096 |
+
"learning_rate": 8.886153846153847e-06,
|
| 189097 |
+
"loss": 0.3674,
|
| 189098 |
+
"step": 70530
|
| 189099 |
+
},
|
| 189100 |
+
{
|
| 189101 |
+
"epoch": 568.79,
|
| 189102 |
+
"learning_rate": 8.886073717948718e-06,
|
| 189103 |
+
"loss": 0.7046,
|
| 189104 |
+
"step": 70535
|
| 189105 |
+
},
|
| 189106 |
+
{
|
| 189107 |
+
"epoch": 568.83,
|
| 189108 |
+
"learning_rate": 8.88599358974359e-06,
|
| 189109 |
+
"loss": 1.0743,
|
| 189110 |
+
"step": 70540
|
| 189111 |
+
},
|
| 189112 |
+
{
|
| 189113 |
+
"epoch": 568.87,
|
| 189114 |
+
"learning_rate": 8.885913461538463e-06,
|
| 189115 |
+
"loss": 0.2614,
|
| 189116 |
+
"step": 70545
|
| 189117 |
+
},
|
| 189118 |
+
{
|
| 189119 |
+
"epoch": 568.91,
|
| 189120 |
+
"learning_rate": 8.885833333333334e-06,
|
| 189121 |
+
"loss": 0.4457,
|
| 189122 |
+
"step": 70550
|
| 189123 |
+
},
|
| 189124 |
+
{
|
| 189125 |
+
"epoch": 568.95,
|
| 189126 |
+
"learning_rate": 8.885753205128205e-06,
|
| 189127 |
+
"loss": 0.399,
|
| 189128 |
+
"step": 70555
|
| 189129 |
+
},
|
| 189130 |
+
{
|
| 189131 |
+
"epoch": 568.99,
|
| 189132 |
+
"learning_rate": 8.885673076923078e-06,
|
| 189133 |
+
"loss": 0.9402,
|
| 189134 |
+
"step": 70560
|
| 189135 |
+
},
|
| 189136 |
+
{
|
| 189137 |
+
"epoch": 569.0,
|
| 189138 |
+
"eval_loss": 0.4022397994995117,
|
| 189139 |
+
"eval_runtime": 37.522,
|
| 189140 |
+
"eval_samples_per_second": 22.387,
|
| 189141 |
+
"eval_steps_per_second": 0.72,
|
| 189142 |
+
"eval_wer": 0.1899042004421518,
|
| 189143 |
+
"step": 70561
|
| 189144 |
+
},
|
| 189145 |
+
{
|
| 189146 |
+
"epoch": 569.03,
|
| 189147 |
+
"learning_rate": 8.88559294871795e-06,
|
| 189148 |
+
"loss": 0.4009,
|
| 189149 |
+
"step": 70565
|
| 189150 |
+
},
|
| 189151 |
+
{
|
| 189152 |
+
"epoch": 569.07,
|
| 189153 |
+
"learning_rate": 8.885512820512821e-06,
|
| 189154 |
+
"loss": 0.3283,
|
| 189155 |
+
"step": 70570
|
| 189156 |
+
},
|
| 189157 |
+
{
|
| 189158 |
+
"epoch": 569.11,
|
| 189159 |
+
"learning_rate": 8.885432692307692e-06,
|
| 189160 |
+
"loss": 0.3124,
|
| 189161 |
+
"step": 70575
|
| 189162 |
+
},
|
| 189163 |
+
{
|
| 189164 |
+
"epoch": 569.15,
|
| 189165 |
+
"learning_rate": 8.885352564102565e-06,
|
| 189166 |
+
"loss": 0.4535,
|
| 189167 |
+
"step": 70580
|
| 189168 |
+
},
|
| 189169 |
+
{
|
| 189170 |
+
"epoch": 569.19,
|
| 189171 |
+
"learning_rate": 8.885272435897437e-06,
|
| 189172 |
+
"loss": 0.9139,
|
| 189173 |
+
"step": 70585
|
| 189174 |
+
},
|
| 189175 |
+
{
|
| 189176 |
+
"epoch": 569.23,
|
| 189177 |
+
"learning_rate": 8.885192307692308e-06,
|
| 189178 |
+
"loss": 0.6242,
|
| 189179 |
+
"step": 70590
|
| 189180 |
+
},
|
| 189181 |
+
{
|
| 189182 |
+
"epoch": 569.27,
|
| 189183 |
+
"learning_rate": 8.88511217948718e-06,
|
| 189184 |
+
"loss": 0.3918,
|
| 189185 |
+
"step": 70595
|
| 189186 |
+
},
|
| 189187 |
+
{
|
| 189188 |
+
"epoch": 569.31,
|
| 189189 |
+
"learning_rate": 8.885032051282053e-06,
|
| 189190 |
+
"loss": 0.3409,
|
| 189191 |
+
"step": 70600
|
| 189192 |
+
},
|
| 189193 |
+
{
|
| 189194 |
+
"epoch": 569.35,
|
| 189195 |
+
"learning_rate": 8.884951923076924e-06,
|
| 189196 |
+
"loss": 0.427,
|
| 189197 |
+
"step": 70605
|
| 189198 |
+
},
|
| 189199 |
+
{
|
| 189200 |
+
"epoch": 569.39,
|
| 189201 |
+
"learning_rate": 8.884871794871795e-06,
|
| 189202 |
+
"loss": 0.9461,
|
| 189203 |
+
"step": 70610
|
| 189204 |
+
},
|
| 189205 |
+
{
|
| 189206 |
+
"epoch": 569.43,
|
| 189207 |
+
"learning_rate": 8.884791666666668e-06,
|
| 189208 |
+
"loss": 0.6715,
|
| 189209 |
+
"step": 70615
|
| 189210 |
+
},
|
| 189211 |
+
{
|
| 189212 |
+
"epoch": 569.47,
|
| 189213 |
+
"learning_rate": 8.88471153846154e-06,
|
| 189214 |
+
"loss": 0.3049,
|
| 189215 |
+
"step": 70620
|
| 189216 |
+
},
|
| 189217 |
+
{
|
| 189218 |
+
"epoch": 569.51,
|
| 189219 |
+
"learning_rate": 8.884631410256411e-06,
|
| 189220 |
+
"loss": 0.3306,
|
| 189221 |
+
"step": 70625
|
| 189222 |
+
},
|
| 189223 |
+
{
|
| 189224 |
+
"epoch": 569.55,
|
| 189225 |
+
"learning_rate": 8.884551282051282e-06,
|
| 189226 |
+
"loss": 0.3854,
|
| 189227 |
+
"step": 70630
|
| 189228 |
+
},
|
| 189229 |
+
{
|
| 189230 |
+
"epoch": 569.59,
|
| 189231 |
+
"learning_rate": 8.884471153846156e-06,
|
| 189232 |
+
"loss": 0.9237,
|
| 189233 |
+
"step": 70635
|
| 189234 |
+
},
|
| 189235 |
+
{
|
| 189236 |
+
"epoch": 569.63,
|
| 189237 |
+
"learning_rate": 8.884391025641025e-06,
|
| 189238 |
+
"loss": 0.6815,
|
| 189239 |
+
"step": 70640
|
| 189240 |
+
},
|
| 189241 |
+
{
|
| 189242 |
+
"epoch": 569.67,
|
| 189243 |
+
"learning_rate": 8.884310897435898e-06,
|
| 189244 |
+
"loss": 0.341,
|
| 189245 |
+
"step": 70645
|
| 189246 |
+
},
|
| 189247 |
+
{
|
| 189248 |
+
"epoch": 569.71,
|
| 189249 |
+
"learning_rate": 8.88423076923077e-06,
|
| 189250 |
+
"loss": 0.301,
|
| 189251 |
+
"step": 70650
|
| 189252 |
+
},
|
| 189253 |
+
{
|
| 189254 |
+
"epoch": 569.76,
|
| 189255 |
+
"learning_rate": 8.884150641025641e-06,
|
| 189256 |
+
"loss": 0.4376,
|
| 189257 |
+
"step": 70655
|
| 189258 |
+
},
|
| 189259 |
+
{
|
| 189260 |
+
"epoch": 569.8,
|
| 189261 |
+
"learning_rate": 8.884070512820514e-06,
|
| 189262 |
+
"loss": 0.9414,
|
| 189263 |
+
"step": 70660
|
| 189264 |
+
},
|
| 189265 |
+
{
|
| 189266 |
+
"epoch": 569.84,
|
| 189267 |
+
"learning_rate": 8.883990384615385e-06,
|
| 189268 |
+
"loss": 0.7538,
|
| 189269 |
+
"step": 70665
|
| 189270 |
+
},
|
| 189271 |
+
{
|
| 189272 |
+
"epoch": 569.88,
|
| 189273 |
+
"learning_rate": 8.883910256410257e-06,
|
| 189274 |
+
"loss": 0.2806,
|
| 189275 |
+
"step": 70670
|
| 189276 |
+
},
|
| 189277 |
+
{
|
| 189278 |
+
"epoch": 569.92,
|
| 189279 |
+
"learning_rate": 8.883830128205128e-06,
|
| 189280 |
+
"loss": 0.3797,
|
| 189281 |
+
"step": 70675
|
| 189282 |
+
},
|
| 189283 |
+
{
|
| 189284 |
+
"epoch": 569.96,
|
| 189285 |
+
"learning_rate": 8.883750000000001e-06,
|
| 189286 |
+
"loss": 0.4361,
|
| 189287 |
+
"step": 70680
|
| 189288 |
+
},
|
| 189289 |
+
{
|
| 189290 |
+
"epoch": 570.0,
|
| 189291 |
+
"learning_rate": 8.883669871794872e-06,
|
| 189292 |
+
"loss": 1.1359,
|
| 189293 |
+
"step": 70685
|
| 189294 |
+
},
|
| 189295 |
+
{
|
| 189296 |
+
"epoch": 570.0,
|
| 189297 |
+
"eval_loss": 0.4082900285720825,
|
| 189298 |
+
"eval_runtime": 39.7838,
|
| 189299 |
+
"eval_samples_per_second": 21.114,
|
| 189300 |
+
"eval_steps_per_second": 0.679,
|
| 189301 |
+
"eval_wer": 0.19334857311407713,
|
| 189302 |
+
"step": 70685
|
| 189303 |
+
},
|
| 189304 |
+
{
|
| 189305 |
+
"epoch": 570.04,
|
| 189306 |
+
"learning_rate": 8.883589743589744e-06,
|
| 189307 |
+
"loss": 0.3446,
|
| 189308 |
+
"step": 70690
|
| 189309 |
+
},
|
| 189310 |
+
{
|
| 189311 |
+
"epoch": 570.08,
|
| 189312 |
+
"learning_rate": 8.883509615384615e-06,
|
| 189313 |
+
"loss": 0.3753,
|
| 189314 |
+
"step": 70695
|
| 189315 |
+
},
|
| 189316 |
+
{
|
| 189317 |
+
"epoch": 570.12,
|
| 189318 |
+
"learning_rate": 8.883429487179488e-06,
|
| 189319 |
+
"loss": 0.304,
|
| 189320 |
+
"step": 70700
|
| 189321 |
+
},
|
| 189322 |
+
{
|
| 189323 |
+
"epoch": 570.16,
|
| 189324 |
+
"learning_rate": 8.88334935897436e-06,
|
| 189325 |
+
"loss": 0.4448,
|
| 189326 |
+
"step": 70705
|
| 189327 |
+
},
|
| 189328 |
+
{
|
| 189329 |
+
"epoch": 570.2,
|
| 189330 |
+
"learning_rate": 8.883269230769231e-06,
|
| 189331 |
+
"loss": 1.4043,
|
| 189332 |
+
"step": 70710
|
| 189333 |
+
},
|
| 189334 |
+
{
|
| 189335 |
+
"epoch": 570.24,
|
| 189336 |
+
"learning_rate": 8.883189102564104e-06,
|
| 189337 |
+
"loss": 0.3952,
|
| 189338 |
+
"step": 70715
|
| 189339 |
+
},
|
| 189340 |
+
{
|
| 189341 |
+
"epoch": 570.28,
|
| 189342 |
+
"learning_rate": 8.883108974358975e-06,
|
| 189343 |
+
"loss": 0.3139,
|
| 189344 |
+
"step": 70720
|
| 189345 |
+
},
|
| 189346 |
+
{
|
| 189347 |
+
"epoch": 570.32,
|
| 189348 |
+
"learning_rate": 8.883028846153847e-06,
|
| 189349 |
+
"loss": 0.3671,
|
| 189350 |
+
"step": 70725
|
| 189351 |
+
},
|
| 189352 |
+
{
|
| 189353 |
+
"epoch": 570.36,
|
| 189354 |
+
"learning_rate": 8.882948717948718e-06,
|
| 189355 |
+
"loss": 0.4336,
|
| 189356 |
+
"step": 70730
|
| 189357 |
+
},
|
| 189358 |
+
{
|
| 189359 |
+
"epoch": 570.4,
|
| 189360 |
+
"learning_rate": 8.882868589743591e-06,
|
| 189361 |
+
"loss": 1.2766,
|
| 189362 |
+
"step": 70735
|
| 189363 |
+
},
|
| 189364 |
+
{
|
| 189365 |
+
"epoch": 570.44,
|
| 189366 |
+
"learning_rate": 8.882788461538463e-06,
|
| 189367 |
+
"loss": 0.4221,
|
| 189368 |
+
"step": 70740
|
| 189369 |
+
},
|
| 189370 |
+
{
|
| 189371 |
+
"epoch": 570.48,
|
| 189372 |
+
"learning_rate": 8.882708333333334e-06,
|
| 189373 |
+
"loss": 0.3531,
|
| 189374 |
+
"step": 70745
|
| 189375 |
+
},
|
| 189376 |
+
{
|
| 189377 |
+
"epoch": 570.52,
|
| 189378 |
+
"learning_rate": 8.882628205128205e-06,
|
| 189379 |
+
"loss": 0.3561,
|
| 189380 |
+
"step": 70750
|
| 189381 |
+
},
|
| 189382 |
+
{
|
| 189383 |
+
"epoch": 570.56,
|
| 189384 |
+
"learning_rate": 8.882548076923078e-06,
|
| 189385 |
+
"loss": 0.4438,
|
| 189386 |
+
"step": 70755
|
| 189387 |
+
},
|
| 189388 |
+
{
|
| 189389 |
+
"epoch": 570.6,
|
| 189390 |
+
"learning_rate": 8.88246794871795e-06,
|
| 189391 |
+
"loss": 1.1606,
|
| 189392 |
+
"step": 70760
|
| 189393 |
+
},
|
| 189394 |
+
{
|
| 189395 |
+
"epoch": 570.64,
|
| 189396 |
+
"learning_rate": 8.882387820512821e-06,
|
| 189397 |
+
"loss": 0.3717,
|
| 189398 |
+
"step": 70765
|
| 189399 |
+
},
|
| 189400 |
+
{
|
| 189401 |
+
"epoch": 570.68,
|
| 189402 |
+
"learning_rate": 8.882307692307694e-06,
|
| 189403 |
+
"loss": 0.3317,
|
| 189404 |
+
"step": 70770
|
| 189405 |
+
},
|
| 189406 |
+
{
|
| 189407 |
+
"epoch": 570.72,
|
| 189408 |
+
"learning_rate": 8.882227564102565e-06,
|
| 189409 |
+
"loss": 0.4233,
|
| 189410 |
+
"step": 70775
|
| 189411 |
+
},
|
| 189412 |
+
{
|
| 189413 |
+
"epoch": 570.76,
|
| 189414 |
+
"learning_rate": 8.882147435897437e-06,
|
| 189415 |
+
"loss": 0.4874,
|
| 189416 |
+
"step": 70780
|
| 189417 |
+
},
|
| 189418 |
+
{
|
| 189419 |
+
"epoch": 570.8,
|
| 189420 |
+
"learning_rate": 8.882067307692308e-06,
|
| 189421 |
+
"loss": 1.2514,
|
| 189422 |
+
"step": 70785
|
| 189423 |
+
},
|
| 189424 |
+
{
|
| 189425 |
+
"epoch": 570.84,
|
| 189426 |
+
"learning_rate": 8.881987179487181e-06,
|
| 189427 |
+
"loss": 0.308,
|
| 189428 |
+
"step": 70790
|
| 189429 |
+
},
|
| 189430 |
+
{
|
| 189431 |
+
"epoch": 570.88,
|
| 189432 |
+
"learning_rate": 8.88190705128205e-06,
|
| 189433 |
+
"loss": 0.2971,
|
| 189434 |
+
"step": 70795
|
| 189435 |
+
},
|
| 189436 |
+
{
|
| 189437 |
+
"epoch": 570.92,
|
| 189438 |
+
"learning_rate": 8.881826923076924e-06,
|
| 189439 |
+
"loss": 0.3294,
|
| 189440 |
+
"step": 70800
|
| 189441 |
+
},
|
| 189442 |
+
{
|
| 189443 |
+
"epoch": 570.96,
|
| 189444 |
+
"learning_rate": 8.881746794871797e-06,
|
| 189445 |
+
"loss": 0.5041,
|
| 189446 |
+
"step": 70805
|
| 189447 |
+
},
|
| 189448 |
+
{
|
| 189449 |
+
"epoch": 571.0,
|
| 189450 |
+
"eval_loss": 0.3579244613647461,
|
| 189451 |
+
"eval_runtime": 40.1364,
|
| 189452 |
+
"eval_samples_per_second": 20.929,
|
| 189453 |
+
"eval_steps_per_second": 0.673,
|
| 189454 |
+
"eval_wer": 0.18833597008485545,
|
| 189455 |
+
"step": 70809
|
| 189456 |
}
|
| 189457 |
],
|
| 189458 |
"max_steps": 620000,
|
| 189459 |
"num_train_epochs": 5000,
|
| 189460 |
+
"total_flos": 1.992660946951832e+20,
|
| 189461 |
"trial_name": null,
|
| 189462 |
"trial_params": null
|
| 189463 |
}
|
model-bin/finetune/base/{checkpoint-70187 → checkpoint-70809}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629855529.4768627/events.out.tfevents.1629855529.7e498afd5545.905.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a691631c88f84295fd9ab41c5956717343a6f9dec943932d6a3ad8f46030e3b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629856180.0808132/events.out.tfevents.1629856180.7e498afd5545.905.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d7a9169e28b1bce5e031d1aa0d74d07354c7bd91a9be4845b40d52cb4bbdf1d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629856939.1696703/events.out.tfevents.1629856939.7e498afd5545.905.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8adcac9c4c255d936902703e352e0e045101ded6951f4111567f54ce342c2a35
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629857599.5043178/events.out.tfevents.1629857599.7e498afd5545.905.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:145fddd07a3f4491f7a746dace0ef93d228c1fec0640cb33769649f55c299987
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629858276.3230329/events.out.tfevents.1629858276.7e498afd5545.905.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ef04fe4fbfefdcc0b6de0a1a16d7a2ca942190d8f97b43d3d147ad41760d4c44
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629855529.7e498afd5545.905.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:34507e83b75bf4bbc54a8bd105bd143753b51cf27592bb1d69b985c9f4189e64
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629856180.7e498afd5545.905.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c92cf6f8487beed64bbf54ef7187b87e4064683edf6121913f4cbb80ea2744c7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629856939.7e498afd5545.905.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0b5035b69e2011893044a066b819d4a21bdb64d19d3e1b29bdbac313a76640b5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629857599.7e498afd5545.905.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:89e235409744eeaf21e0dc20dead8539999f0032993c80fb08d09a71a54e26ff
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629858276.7e498afd5545.905.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8490d6025e42e5a0dbdd012564ebe0f57e0d403757bc257d5d1c103539ebcc5
|
| 3 |
+
size 8462
|