"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629817865.4918172/events.out.tfevents.1629817865.c435e1c5ee04.920.161 +3 -0
- model-bin/finetune/base/log/1629818505.946728/events.out.tfevents.1629818505.c435e1c5ee04.920.163 +3 -0
- model-bin/finetune/base/log/1629819241.3187594/events.out.tfevents.1629819241.c435e1c5ee04.920.165 +3 -0
- model-bin/finetune/base/log/1629819885.031035/events.out.tfevents.1629819885.c435e1c5ee04.920.167 +3 -0
- model-bin/finetune/base/log/1629820516.7180705/events.out.tfevents.1629820516.c435e1c5ee04.920.169 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629817865.c435e1c5ee04.920.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629818505.c435e1c5ee04.920.162 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629819241.c435e1c5ee04.920.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629819885.c435e1c5ee04.920.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629820516.c435e1c5ee04.920.168 +3 -0
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84a39b5180d338455ed67d30094154fcf946d9e7d3501d922a007942dfe87cdc
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/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:cc1a73ce5dbe572e94f0c0b10d196b8fffdfbb252126396bda8c1009ccb833fc
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/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 sha256:07739d86de1eb12402159535e5b866ccb6a1445936e5827362789c7416e3c93a
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/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:f8026a5d49524a9f14745c2e29d3add8cbe3551d36a905ecf36a6d70478ad2c4
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/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 sha256:9165409afc414fbb8cbb563ef7519a5bc6fd7f59200f227203dc9a383fddcc46
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -180585,11 +180585,806 @@
|
|
| 180585 |
"eval_steps_per_second": 0.69,
|
| 180586 |
"eval_wer": 0.20920346062052506,
|
| 180587 |
"step": 63838
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 180588 |
}
|
| 180589 |
],
|
| 180590 |
-
"max_steps":
|
| 180591 |
"num_train_epochs": 5000,
|
| 180592 |
-
"total_flos": 1.
|
| 180593 |
"trial_name": null,
|
| 180594 |
"trial_params": null
|
| 180595 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 515.0,
|
| 5 |
+
"global_step": 64461,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 180585 |
"eval_steps_per_second": 0.69,
|
| 180586 |
"eval_wer": 0.20920346062052506,
|
| 180587 |
"step": 63838
|
| 180588 |
+
},
|
| 180589 |
+
{
|
| 180590 |
+
"epoch": 514.02,
|
| 180591 |
+
"learning_rate": 8.993333333333334e-06,
|
| 180592 |
+
"loss": 0.4305,
|
| 180593 |
+
"step": 63840
|
| 180594 |
+
},
|
| 180595 |
+
{
|
| 180596 |
+
"epoch": 514.06,
|
| 180597 |
+
"learning_rate": 8.993253205128206e-06,
|
| 180598 |
+
"loss": 0.3187,
|
| 180599 |
+
"step": 63845
|
| 180600 |
+
},
|
| 180601 |
+
{
|
| 180602 |
+
"epoch": 514.1,
|
| 180603 |
+
"learning_rate": 8.993173076923077e-06,
|
| 180604 |
+
"loss": 0.5928,
|
| 180605 |
+
"step": 63850
|
| 180606 |
+
},
|
| 180607 |
+
{
|
| 180608 |
+
"epoch": 514.14,
|
| 180609 |
+
"learning_rate": 8.99309294871795e-06,
|
| 180610 |
+
"loss": 0.4087,
|
| 180611 |
+
"step": 63855
|
| 180612 |
+
},
|
| 180613 |
+
{
|
| 180614 |
+
"epoch": 514.18,
|
| 180615 |
+
"learning_rate": 8.993012820512821e-06,
|
| 180616 |
+
"loss": 0.685,
|
| 180617 |
+
"step": 63860
|
| 180618 |
+
},
|
| 180619 |
+
{
|
| 180620 |
+
"epoch": 514.22,
|
| 180621 |
+
"learning_rate": 8.992932692307693e-06,
|
| 180622 |
+
"loss": 1.1175,
|
| 180623 |
+
"step": 63865
|
| 180624 |
+
},
|
| 180625 |
+
{
|
| 180626 |
+
"epoch": 514.26,
|
| 180627 |
+
"learning_rate": 8.992852564102564e-06,
|
| 180628 |
+
"loss": 0.3149,
|
| 180629 |
+
"step": 63870
|
| 180630 |
+
},
|
| 180631 |
+
{
|
| 180632 |
+
"epoch": 514.3,
|
| 180633 |
+
"learning_rate": 8.992772435897437e-06,
|
| 180634 |
+
"loss": 0.3267,
|
| 180635 |
+
"step": 63875
|
| 180636 |
+
},
|
| 180637 |
+
{
|
| 180638 |
+
"epoch": 514.34,
|
| 180639 |
+
"learning_rate": 8.992692307692308e-06,
|
| 180640 |
+
"loss": 0.4176,
|
| 180641 |
+
"step": 63880
|
| 180642 |
+
},
|
| 180643 |
+
{
|
| 180644 |
+
"epoch": 514.38,
|
| 180645 |
+
"learning_rate": 8.99261217948718e-06,
|
| 180646 |
+
"loss": 0.9252,
|
| 180647 |
+
"step": 63885
|
| 180648 |
+
},
|
| 180649 |
+
{
|
| 180650 |
+
"epoch": 514.42,
|
| 180651 |
+
"learning_rate": 8.992532051282053e-06,
|
| 180652 |
+
"loss": 1.2078,
|
| 180653 |
+
"step": 63890
|
| 180654 |
+
},
|
| 180655 |
+
{
|
| 180656 |
+
"epoch": 514.46,
|
| 180657 |
+
"learning_rate": 8.992451923076924e-06,
|
| 180658 |
+
"loss": 0.3078,
|
| 180659 |
+
"step": 63895
|
| 180660 |
+
},
|
| 180661 |
+
{
|
| 180662 |
+
"epoch": 514.5,
|
| 180663 |
+
"learning_rate": 8.992371794871796e-06,
|
| 180664 |
+
"loss": 0.3263,
|
| 180665 |
+
"step": 63900
|
| 180666 |
+
},
|
| 180667 |
+
{
|
| 180668 |
+
"epoch": 514.54,
|
| 180669 |
+
"learning_rate": 8.992291666666667e-06,
|
| 180670 |
+
"loss": 0.4472,
|
| 180671 |
+
"step": 63905
|
| 180672 |
+
},
|
| 180673 |
+
{
|
| 180674 |
+
"epoch": 514.58,
|
| 180675 |
+
"learning_rate": 8.99221153846154e-06,
|
| 180676 |
+
"loss": 0.6959,
|
| 180677 |
+
"step": 63910
|
| 180678 |
+
},
|
| 180679 |
+
{
|
| 180680 |
+
"epoch": 514.62,
|
| 180681 |
+
"learning_rate": 8.992131410256411e-06,
|
| 180682 |
+
"loss": 1.0351,
|
| 180683 |
+
"step": 63915
|
| 180684 |
+
},
|
| 180685 |
+
{
|
| 180686 |
+
"epoch": 514.66,
|
| 180687 |
+
"learning_rate": 8.992051282051283e-06,
|
| 180688 |
+
"loss": 0.3149,
|
| 180689 |
+
"step": 63920
|
| 180690 |
+
},
|
| 180691 |
+
{
|
| 180692 |
+
"epoch": 514.7,
|
| 180693 |
+
"learning_rate": 8.991971153846154e-06,
|
| 180694 |
+
"loss": 0.3994,
|
| 180695 |
+
"step": 63925
|
| 180696 |
+
},
|
| 180697 |
+
{
|
| 180698 |
+
"epoch": 514.74,
|
| 180699 |
+
"learning_rate": 8.991891025641027e-06,
|
| 180700 |
+
"loss": 0.3566,
|
| 180701 |
+
"step": 63930
|
| 180702 |
+
},
|
| 180703 |
+
{
|
| 180704 |
+
"epoch": 514.78,
|
| 180705 |
+
"learning_rate": 8.991810897435898e-06,
|
| 180706 |
+
"loss": 0.7522,
|
| 180707 |
+
"step": 63935
|
| 180708 |
+
},
|
| 180709 |
+
{
|
| 180710 |
+
"epoch": 514.82,
|
| 180711 |
+
"learning_rate": 8.99173076923077e-06,
|
| 180712 |
+
"loss": 1.0754,
|
| 180713 |
+
"step": 63940
|
| 180714 |
+
},
|
| 180715 |
+
{
|
| 180716 |
+
"epoch": 514.86,
|
| 180717 |
+
"learning_rate": 8.991650641025643e-06,
|
| 180718 |
+
"loss": 0.3273,
|
| 180719 |
+
"step": 63945
|
| 180720 |
+
},
|
| 180721 |
+
{
|
| 180722 |
+
"epoch": 514.9,
|
| 180723 |
+
"learning_rate": 8.991570512820513e-06,
|
| 180724 |
+
"loss": 0.3926,
|
| 180725 |
+
"step": 63950
|
| 180726 |
+
},
|
| 180727 |
+
{
|
| 180728 |
+
"epoch": 514.94,
|
| 180729 |
+
"learning_rate": 8.991490384615386e-06,
|
| 180730 |
+
"loss": 0.4042,
|
| 180731 |
+
"step": 63955
|
| 180732 |
+
},
|
| 180733 |
+
{
|
| 180734 |
+
"epoch": 514.98,
|
| 180735 |
+
"learning_rate": 8.991410256410257e-06,
|
| 180736 |
+
"loss": 0.8227,
|
| 180737 |
+
"step": 63960
|
| 180738 |
+
},
|
| 180739 |
+
{
|
| 180740 |
+
"epoch": 515.0,
|
| 180741 |
+
"eval_loss": 0.38096028566360474,
|
| 180742 |
+
"eval_runtime": 40.0604,
|
| 180743 |
+
"eval_samples_per_second": 21.068,
|
| 180744 |
+
"eval_steps_per_second": 0.674,
|
| 180745 |
+
"eval_wer": 0.19324738229218605,
|
| 180746 |
+
"step": 63962
|
| 180747 |
+
},
|
| 180748 |
+
{
|
| 180749 |
+
"epoch": 515.02,
|
| 180750 |
+
"learning_rate": 8.991330128205128e-06,
|
| 180751 |
+
"loss": 0.3509,
|
| 180752 |
+
"step": 63965
|
| 180753 |
+
},
|
| 180754 |
+
{
|
| 180755 |
+
"epoch": 515.06,
|
| 180756 |
+
"learning_rate": 8.99125e-06,
|
| 180757 |
+
"loss": 0.3487,
|
| 180758 |
+
"step": 63970
|
| 180759 |
+
},
|
| 180760 |
+
{
|
| 180761 |
+
"epoch": 515.1,
|
| 180762 |
+
"learning_rate": 8.991169871794873e-06,
|
| 180763 |
+
"loss": 0.3623,
|
| 180764 |
+
"step": 63975
|
| 180765 |
+
},
|
| 180766 |
+
{
|
| 180767 |
+
"epoch": 515.14,
|
| 180768 |
+
"learning_rate": 8.991089743589744e-06,
|
| 180769 |
+
"loss": 0.4096,
|
| 180770 |
+
"step": 63980
|
| 180771 |
+
},
|
| 180772 |
+
{
|
| 180773 |
+
"epoch": 515.18,
|
| 180774 |
+
"learning_rate": 8.991009615384615e-06,
|
| 180775 |
+
"loss": 0.8031,
|
| 180776 |
+
"step": 63985
|
| 180777 |
+
},
|
| 180778 |
+
{
|
| 180779 |
+
"epoch": 515.22,
|
| 180780 |
+
"learning_rate": 8.990929487179488e-06,
|
| 180781 |
+
"loss": 0.9627,
|
| 180782 |
+
"step": 63990
|
| 180783 |
+
},
|
| 180784 |
+
{
|
| 180785 |
+
"epoch": 515.27,
|
| 180786 |
+
"learning_rate": 8.99084935897436e-06,
|
| 180787 |
+
"loss": 0.3321,
|
| 180788 |
+
"step": 63995
|
| 180789 |
+
},
|
| 180790 |
+
{
|
| 180791 |
+
"epoch": 515.31,
|
| 180792 |
+
"learning_rate": 8.990769230769231e-06,
|
| 180793 |
+
"loss": 1.0533,
|
| 180794 |
+
"step": 64000
|
| 180795 |
+
},
|
| 180796 |
+
{
|
| 180797 |
+
"epoch": 515.35,
|
| 180798 |
+
"learning_rate": 8.990689102564103e-06,
|
| 180799 |
+
"loss": 0.349,
|
| 180800 |
+
"step": 64005
|
| 180801 |
+
},
|
| 180802 |
+
{
|
| 180803 |
+
"epoch": 515.39,
|
| 180804 |
+
"learning_rate": 8.990608974358976e-06,
|
| 180805 |
+
"loss": 0.7616,
|
| 180806 |
+
"step": 64010
|
| 180807 |
+
},
|
| 180808 |
+
{
|
| 180809 |
+
"epoch": 515.43,
|
| 180810 |
+
"learning_rate": 8.990528846153847e-06,
|
| 180811 |
+
"loss": 0.9069,
|
| 180812 |
+
"step": 64015
|
| 180813 |
+
},
|
| 180814 |
+
{
|
| 180815 |
+
"epoch": 515.47,
|
| 180816 |
+
"learning_rate": 8.990448717948718e-06,
|
| 180817 |
+
"loss": 0.3199,
|
| 180818 |
+
"step": 64020
|
| 180819 |
+
},
|
| 180820 |
+
{
|
| 180821 |
+
"epoch": 515.51,
|
| 180822 |
+
"learning_rate": 8.99036858974359e-06,
|
| 180823 |
+
"loss": 0.3591,
|
| 180824 |
+
"step": 64025
|
| 180825 |
+
},
|
| 180826 |
+
{
|
| 180827 |
+
"epoch": 515.55,
|
| 180828 |
+
"learning_rate": 8.990288461538463e-06,
|
| 180829 |
+
"loss": 0.3535,
|
| 180830 |
+
"step": 64030
|
| 180831 |
+
},
|
| 180832 |
+
{
|
| 180833 |
+
"epoch": 515.59,
|
| 180834 |
+
"learning_rate": 8.990208333333334e-06,
|
| 180835 |
+
"loss": 0.773,
|
| 180836 |
+
"step": 64035
|
| 180837 |
+
},
|
| 180838 |
+
{
|
| 180839 |
+
"epoch": 515.63,
|
| 180840 |
+
"learning_rate": 8.990128205128205e-06,
|
| 180841 |
+
"loss": 0.9016,
|
| 180842 |
+
"step": 64040
|
| 180843 |
+
},
|
| 180844 |
+
{
|
| 180845 |
+
"epoch": 515.67,
|
| 180846 |
+
"learning_rate": 8.990048076923079e-06,
|
| 180847 |
+
"loss": 0.3039,
|
| 180848 |
+
"step": 64045
|
| 180849 |
+
},
|
| 180850 |
+
{
|
| 180851 |
+
"epoch": 515.71,
|
| 180852 |
+
"learning_rate": 8.98996794871795e-06,
|
| 180853 |
+
"loss": 0.2948,
|
| 180854 |
+
"step": 64050
|
| 180855 |
+
},
|
| 180856 |
+
{
|
| 180857 |
+
"epoch": 515.75,
|
| 180858 |
+
"learning_rate": 8.989887820512821e-06,
|
| 180859 |
+
"loss": 0.4245,
|
| 180860 |
+
"step": 64055
|
| 180861 |
+
},
|
| 180862 |
+
{
|
| 180863 |
+
"epoch": 515.79,
|
| 180864 |
+
"learning_rate": 8.989807692307693e-06,
|
| 180865 |
+
"loss": 0.8484,
|
| 180866 |
+
"step": 64060
|
| 180867 |
+
},
|
| 180868 |
+
{
|
| 180869 |
+
"epoch": 515.83,
|
| 180870 |
+
"learning_rate": 8.989727564102566e-06,
|
| 180871 |
+
"loss": 0.936,
|
| 180872 |
+
"step": 64065
|
| 180873 |
+
},
|
| 180874 |
+
{
|
| 180875 |
+
"epoch": 515.87,
|
| 180876 |
+
"learning_rate": 8.989647435897435e-06,
|
| 180877 |
+
"loss": 0.3889,
|
| 180878 |
+
"step": 64070
|
| 180879 |
+
},
|
| 180880 |
+
{
|
| 180881 |
+
"epoch": 515.91,
|
| 180882 |
+
"learning_rate": 8.989567307692308e-06,
|
| 180883 |
+
"loss": 0.3237,
|
| 180884 |
+
"step": 64075
|
| 180885 |
+
},
|
| 180886 |
+
{
|
| 180887 |
+
"epoch": 515.95,
|
| 180888 |
+
"learning_rate": 8.989487179487181e-06,
|
| 180889 |
+
"loss": 0.4583,
|
| 180890 |
+
"step": 64080
|
| 180891 |
+
},
|
| 180892 |
+
{
|
| 180893 |
+
"epoch": 515.99,
|
| 180894 |
+
"learning_rate": 8.989407051282051e-06,
|
| 180895 |
+
"loss": 0.7796,
|
| 180896 |
+
"step": 64085
|
| 180897 |
+
},
|
| 180898 |
+
{
|
| 180899 |
+
"epoch": 516.0,
|
| 180900 |
+
"eval_loss": 0.39013779163360596,
|
| 180901 |
+
"eval_runtime": 39.1915,
|
| 180902 |
+
"eval_samples_per_second": 21.561,
|
| 180903 |
+
"eval_steps_per_second": 0.689,
|
| 180904 |
+
"eval_wer": 0.19555424267040838,
|
| 180905 |
+
"step": 64086
|
| 180906 |
+
},
|
| 180907 |
+
{
|
| 180908 |
+
"epoch": 512.03,
|
| 180909 |
+
"learning_rate": 8.989326923076924e-06,
|
| 180910 |
+
"loss": 0.3987,
|
| 180911 |
+
"step": 64090
|
| 180912 |
+
},
|
| 180913 |
+
{
|
| 180914 |
+
"epoch": 512.07,
|
| 180915 |
+
"learning_rate": 8.989246794871795e-06,
|
| 180916 |
+
"loss": 0.2729,
|
| 180917 |
+
"step": 64095
|
| 180918 |
+
},
|
| 180919 |
+
{
|
| 180920 |
+
"epoch": 512.11,
|
| 180921 |
+
"learning_rate": 8.989166666666669e-06,
|
| 180922 |
+
"loss": 0.3189,
|
| 180923 |
+
"step": 64100
|
| 180924 |
+
},
|
| 180925 |
+
{
|
| 180926 |
+
"epoch": 512.15,
|
| 180927 |
+
"learning_rate": 8.989086538461538e-06,
|
| 180928 |
+
"loss": 0.4544,
|
| 180929 |
+
"step": 64105
|
| 180930 |
+
},
|
| 180931 |
+
{
|
| 180932 |
+
"epoch": 512.19,
|
| 180933 |
+
"learning_rate": 8.989006410256411e-06,
|
| 180934 |
+
"loss": 0.9482,
|
| 180935 |
+
"step": 64110
|
| 180936 |
+
},
|
| 180937 |
+
{
|
| 180938 |
+
"epoch": 512.23,
|
| 180939 |
+
"learning_rate": 8.988926282051283e-06,
|
| 180940 |
+
"loss": 0.6932,
|
| 180941 |
+
"step": 64115
|
| 180942 |
+
},
|
| 180943 |
+
{
|
| 180944 |
+
"epoch": 512.27,
|
| 180945 |
+
"learning_rate": 8.988846153846154e-06,
|
| 180946 |
+
"loss": 0.2904,
|
| 180947 |
+
"step": 64120
|
| 180948 |
+
},
|
| 180949 |
+
{
|
| 180950 |
+
"epoch": 512.31,
|
| 180951 |
+
"learning_rate": 8.988766025641025e-06,
|
| 180952 |
+
"loss": 0.2751,
|
| 180953 |
+
"step": 64125
|
| 180954 |
+
},
|
| 180955 |
+
{
|
| 180956 |
+
"epoch": 512.35,
|
| 180957 |
+
"learning_rate": 8.988685897435898e-06,
|
| 180958 |
+
"loss": 0.3816,
|
| 180959 |
+
"step": 64130
|
| 180960 |
+
},
|
| 180961 |
+
{
|
| 180962 |
+
"epoch": 512.39,
|
| 180963 |
+
"learning_rate": 8.98860576923077e-06,
|
| 180964 |
+
"loss": 0.7908,
|
| 180965 |
+
"step": 64135
|
| 180966 |
+
},
|
| 180967 |
+
{
|
| 180968 |
+
"epoch": 512.43,
|
| 180969 |
+
"learning_rate": 8.988541666666667e-06,
|
| 180970 |
+
"loss": 0.8226,
|
| 180971 |
+
"step": 64140
|
| 180972 |
+
},
|
| 180973 |
+
{
|
| 180974 |
+
"epoch": 512.47,
|
| 180975 |
+
"learning_rate": 8.988461538461539e-06,
|
| 180976 |
+
"loss": 0.2868,
|
| 180977 |
+
"step": 64145
|
| 180978 |
+
},
|
| 180979 |
+
{
|
| 180980 |
+
"epoch": 512.51,
|
| 180981 |
+
"learning_rate": 8.988381410256412e-06,
|
| 180982 |
+
"loss": 0.3305,
|
| 180983 |
+
"step": 64150
|
| 180984 |
+
},
|
| 180985 |
+
{
|
| 180986 |
+
"epoch": 512.55,
|
| 180987 |
+
"learning_rate": 8.988301282051283e-06,
|
| 180988 |
+
"loss": 0.4952,
|
| 180989 |
+
"step": 64155
|
| 180990 |
+
},
|
| 180991 |
+
{
|
| 180992 |
+
"epoch": 512.59,
|
| 180993 |
+
"learning_rate": 8.988221153846154e-06,
|
| 180994 |
+
"loss": 1.1173,
|
| 180995 |
+
"step": 64160
|
| 180996 |
+
},
|
| 180997 |
+
{
|
| 180998 |
+
"epoch": 512.63,
|
| 180999 |
+
"learning_rate": 8.988141025641026e-06,
|
| 181000 |
+
"loss": 0.8068,
|
| 181001 |
+
"step": 64165
|
| 181002 |
+
},
|
| 181003 |
+
{
|
| 181004 |
+
"epoch": 512.67,
|
| 181005 |
+
"learning_rate": 8.988060897435899e-06,
|
| 181006 |
+
"loss": 0.3757,
|
| 181007 |
+
"step": 64170
|
| 181008 |
+
},
|
| 181009 |
+
{
|
| 181010 |
+
"epoch": 512.71,
|
| 181011 |
+
"learning_rate": 8.98798076923077e-06,
|
| 181012 |
+
"loss": 0.3248,
|
| 181013 |
+
"step": 64175
|
| 181014 |
+
},
|
| 181015 |
+
{
|
| 181016 |
+
"epoch": 512.75,
|
| 181017 |
+
"learning_rate": 8.987900641025641e-06,
|
| 181018 |
+
"loss": 0.4137,
|
| 181019 |
+
"step": 64180
|
| 181020 |
+
},
|
| 181021 |
+
{
|
| 181022 |
+
"epoch": 512.79,
|
| 181023 |
+
"learning_rate": 8.987820512820513e-06,
|
| 181024 |
+
"loss": 1.0292,
|
| 181025 |
+
"step": 64185
|
| 181026 |
+
},
|
| 181027 |
+
{
|
| 181028 |
+
"epoch": 512.83,
|
| 181029 |
+
"learning_rate": 8.987740384615386e-06,
|
| 181030 |
+
"loss": 0.7227,
|
| 181031 |
+
"step": 64190
|
| 181032 |
+
},
|
| 181033 |
+
{
|
| 181034 |
+
"epoch": 512.87,
|
| 181035 |
+
"learning_rate": 8.987660256410257e-06,
|
| 181036 |
+
"loss": 0.317,
|
| 181037 |
+
"step": 64195
|
| 181038 |
+
},
|
| 181039 |
+
{
|
| 181040 |
+
"epoch": 512.91,
|
| 181041 |
+
"learning_rate": 8.987580128205129e-06,
|
| 181042 |
+
"loss": 0.3131,
|
| 181043 |
+
"step": 64200
|
| 181044 |
+
},
|
| 181045 |
+
{
|
| 181046 |
+
"epoch": 512.95,
|
| 181047 |
+
"learning_rate": 8.987500000000002e-06,
|
| 181048 |
+
"loss": 0.4147,
|
| 181049 |
+
"step": 64205
|
| 181050 |
+
},
|
| 181051 |
+
{
|
| 181052 |
+
"epoch": 512.99,
|
| 181053 |
+
"learning_rate": 8.987419871794873e-06,
|
| 181054 |
+
"loss": 1.1387,
|
| 181055 |
+
"step": 64210
|
| 181056 |
+
},
|
| 181057 |
+
{
|
| 181058 |
+
"epoch": 513.0,
|
| 181059 |
+
"eval_loss": 0.4046097695827484,
|
| 181060 |
+
"eval_runtime": 37.7066,
|
| 181061 |
+
"eval_samples_per_second": 22.198,
|
| 181062 |
+
"eval_steps_per_second": 0.716,
|
| 181063 |
+
"eval_wer": 0.1909584617642742,
|
| 181064 |
+
"step": 64211
|
| 181065 |
+
},
|
| 181066 |
+
{
|
| 181067 |
+
"epoch": 513.03,
|
| 181068 |
+
"learning_rate": 8.987339743589744e-06,
|
| 181069 |
+
"loss": 0.3277,
|
| 181070 |
+
"step": 64215
|
| 181071 |
+
},
|
| 181072 |
+
{
|
| 181073 |
+
"epoch": 513.07,
|
| 181074 |
+
"learning_rate": 8.987259615384616e-06,
|
| 181075 |
+
"loss": 0.3194,
|
| 181076 |
+
"step": 64220
|
| 181077 |
+
},
|
| 181078 |
+
{
|
| 181079 |
+
"epoch": 513.11,
|
| 181080 |
+
"learning_rate": 8.987179487179489e-06,
|
| 181081 |
+
"loss": 0.3342,
|
| 181082 |
+
"step": 64225
|
| 181083 |
+
},
|
| 181084 |
+
{
|
| 181085 |
+
"epoch": 513.15,
|
| 181086 |
+
"learning_rate": 8.98709935897436e-06,
|
| 181087 |
+
"loss": 0.3926,
|
| 181088 |
+
"step": 64230
|
| 181089 |
+
},
|
| 181090 |
+
{
|
| 181091 |
+
"epoch": 513.19,
|
| 181092 |
+
"learning_rate": 8.987019230769231e-06,
|
| 181093 |
+
"loss": 0.9825,
|
| 181094 |
+
"step": 64235
|
| 181095 |
+
},
|
| 181096 |
+
{
|
| 181097 |
+
"epoch": 513.23,
|
| 181098 |
+
"learning_rate": 8.986939102564103e-06,
|
| 181099 |
+
"loss": 0.6865,
|
| 181100 |
+
"step": 64240
|
| 181101 |
+
},
|
| 181102 |
+
{
|
| 181103 |
+
"epoch": 513.27,
|
| 181104 |
+
"learning_rate": 8.986858974358976e-06,
|
| 181105 |
+
"loss": 0.3306,
|
| 181106 |
+
"step": 64245
|
| 181107 |
+
},
|
| 181108 |
+
{
|
| 181109 |
+
"epoch": 513.31,
|
| 181110 |
+
"learning_rate": 8.986778846153847e-06,
|
| 181111 |
+
"loss": 0.375,
|
| 181112 |
+
"step": 64250
|
| 181113 |
+
},
|
| 181114 |
+
{
|
| 181115 |
+
"epoch": 513.35,
|
| 181116 |
+
"learning_rate": 8.986698717948719e-06,
|
| 181117 |
+
"loss": 0.4763,
|
| 181118 |
+
"step": 64255
|
| 181119 |
+
},
|
| 181120 |
+
{
|
| 181121 |
+
"epoch": 513.39,
|
| 181122 |
+
"learning_rate": 8.986618589743592e-06,
|
| 181123 |
+
"loss": 1.0595,
|
| 181124 |
+
"step": 64260
|
| 181125 |
+
},
|
| 181126 |
+
{
|
| 181127 |
+
"epoch": 513.43,
|
| 181128 |
+
"learning_rate": 8.986538461538461e-06,
|
| 181129 |
+
"loss": 0.7274,
|
| 181130 |
+
"step": 64265
|
| 181131 |
+
},
|
| 181132 |
+
{
|
| 181133 |
+
"epoch": 513.47,
|
| 181134 |
+
"learning_rate": 8.986458333333334e-06,
|
| 181135 |
+
"loss": 0.3112,
|
| 181136 |
+
"step": 64270
|
| 181137 |
+
},
|
| 181138 |
+
{
|
| 181139 |
+
"epoch": 513.51,
|
| 181140 |
+
"learning_rate": 8.986378205128206e-06,
|
| 181141 |
+
"loss": 0.3159,
|
| 181142 |
+
"step": 64275
|
| 181143 |
+
},
|
| 181144 |
+
{
|
| 181145 |
+
"epoch": 513.55,
|
| 181146 |
+
"learning_rate": 8.986298076923077e-06,
|
| 181147 |
+
"loss": 0.4657,
|
| 181148 |
+
"step": 64280
|
| 181149 |
+
},
|
| 181150 |
+
{
|
| 181151 |
+
"epoch": 513.59,
|
| 181152 |
+
"learning_rate": 8.986217948717948e-06,
|
| 181153 |
+
"loss": 0.9238,
|
| 181154 |
+
"step": 64285
|
| 181155 |
+
},
|
| 181156 |
+
{
|
| 181157 |
+
"epoch": 513.63,
|
| 181158 |
+
"learning_rate": 8.986137820512821e-06,
|
| 181159 |
+
"loss": 0.7649,
|
| 181160 |
+
"step": 64290
|
| 181161 |
+
},
|
| 181162 |
+
{
|
| 181163 |
+
"epoch": 513.67,
|
| 181164 |
+
"learning_rate": 8.986057692307693e-06,
|
| 181165 |
+
"loss": 0.3587,
|
| 181166 |
+
"step": 64295
|
| 181167 |
+
},
|
| 181168 |
+
{
|
| 181169 |
+
"epoch": 513.71,
|
| 181170 |
+
"learning_rate": 8.985977564102564e-06,
|
| 181171 |
+
"loss": 0.3466,
|
| 181172 |
+
"step": 64300
|
| 181173 |
+
},
|
| 181174 |
+
{
|
| 181175 |
+
"epoch": 513.75,
|
| 181176 |
+
"learning_rate": 8.985897435897437e-06,
|
| 181177 |
+
"loss": 0.4172,
|
| 181178 |
+
"step": 64305
|
| 181179 |
+
},
|
| 181180 |
+
{
|
| 181181 |
+
"epoch": 513.79,
|
| 181182 |
+
"learning_rate": 8.985817307692309e-06,
|
| 181183 |
+
"loss": 0.9061,
|
| 181184 |
+
"step": 64310
|
| 181185 |
+
},
|
| 181186 |
+
{
|
| 181187 |
+
"epoch": 513.83,
|
| 181188 |
+
"learning_rate": 8.98573717948718e-06,
|
| 181189 |
+
"loss": 0.7512,
|
| 181190 |
+
"step": 64315
|
| 181191 |
+
},
|
| 181192 |
+
{
|
| 181193 |
+
"epoch": 513.87,
|
| 181194 |
+
"learning_rate": 8.985657051282051e-06,
|
| 181195 |
+
"loss": 0.3268,
|
| 181196 |
+
"step": 64320
|
| 181197 |
+
},
|
| 181198 |
+
{
|
| 181199 |
+
"epoch": 513.91,
|
| 181200 |
+
"learning_rate": 8.985576923076924e-06,
|
| 181201 |
+
"loss": 0.2828,
|
| 181202 |
+
"step": 64325
|
| 181203 |
+
},
|
| 181204 |
+
{
|
| 181205 |
+
"epoch": 513.95,
|
| 181206 |
+
"learning_rate": 8.985496794871796e-06,
|
| 181207 |
+
"loss": 0.4606,
|
| 181208 |
+
"step": 64330
|
| 181209 |
+
},
|
| 181210 |
+
{
|
| 181211 |
+
"epoch": 513.99,
|
| 181212 |
+
"learning_rate": 8.985416666666667e-06,
|
| 181213 |
+
"loss": 1.0801,
|
| 181214 |
+
"step": 64335
|
| 181215 |
+
},
|
| 181216 |
+
{
|
| 181217 |
+
"epoch": 514.0,
|
| 181218 |
+
"eval_loss": 0.39479246735572815,
|
| 181219 |
+
"eval_runtime": 39.9341,
|
| 181220 |
+
"eval_samples_per_second": 20.96,
|
| 181221 |
+
"eval_steps_per_second": 0.676,
|
| 181222 |
+
"eval_wer": 0.1925839073721237,
|
| 181223 |
+
"step": 64336
|
| 181224 |
+
},
|
| 181225 |
+
{
|
| 181226 |
+
"epoch": 514.03,
|
| 181227 |
+
"learning_rate": 8.985336538461538e-06,
|
| 181228 |
+
"loss": 0.3468,
|
| 181229 |
+
"step": 64340
|
| 181230 |
+
},
|
| 181231 |
+
{
|
| 181232 |
+
"epoch": 514.07,
|
| 181233 |
+
"learning_rate": 8.985256410256412e-06,
|
| 181234 |
+
"loss": 0.3658,
|
| 181235 |
+
"step": 64345
|
| 181236 |
+
},
|
| 181237 |
+
{
|
| 181238 |
+
"epoch": 514.11,
|
| 181239 |
+
"learning_rate": 8.985176282051283e-06,
|
| 181240 |
+
"loss": 0.3534,
|
| 181241 |
+
"step": 64350
|
| 181242 |
+
},
|
| 181243 |
+
{
|
| 181244 |
+
"epoch": 514.15,
|
| 181245 |
+
"learning_rate": 8.985096153846154e-06,
|
| 181246 |
+
"loss": 0.5208,
|
| 181247 |
+
"step": 64355
|
| 181248 |
+
},
|
| 181249 |
+
{
|
| 181250 |
+
"epoch": 514.19,
|
| 181251 |
+
"learning_rate": 8.985016025641027e-06,
|
| 181252 |
+
"loss": 0.9566,
|
| 181253 |
+
"step": 64360
|
| 181254 |
+
},
|
| 181255 |
+
{
|
| 181256 |
+
"epoch": 514.23,
|
| 181257 |
+
"learning_rate": 8.984935897435899e-06,
|
| 181258 |
+
"loss": 0.7008,
|
| 181259 |
+
"step": 64365
|
| 181260 |
+
},
|
| 181261 |
+
{
|
| 181262 |
+
"epoch": 514.27,
|
| 181263 |
+
"learning_rate": 8.98485576923077e-06,
|
| 181264 |
+
"loss": 0.3148,
|
| 181265 |
+
"step": 64370
|
| 181266 |
+
},
|
| 181267 |
+
{
|
| 181268 |
+
"epoch": 514.31,
|
| 181269 |
+
"learning_rate": 8.984775641025641e-06,
|
| 181270 |
+
"loss": 0.3496,
|
| 181271 |
+
"step": 64375
|
| 181272 |
+
},
|
| 181273 |
+
{
|
| 181274 |
+
"epoch": 514.35,
|
| 181275 |
+
"learning_rate": 8.984695512820514e-06,
|
| 181276 |
+
"loss": 0.4292,
|
| 181277 |
+
"step": 64380
|
| 181278 |
+
},
|
| 181279 |
+
{
|
| 181280 |
+
"epoch": 514.39,
|
| 181281 |
+
"learning_rate": 8.984615384615386e-06,
|
| 181282 |
+
"loss": 0.8532,
|
| 181283 |
+
"step": 64385
|
| 181284 |
+
},
|
| 181285 |
+
{
|
| 181286 |
+
"epoch": 514.43,
|
| 181287 |
+
"learning_rate": 8.984535256410257e-06,
|
| 181288 |
+
"loss": 0.7066,
|
| 181289 |
+
"step": 64390
|
| 181290 |
+
},
|
| 181291 |
+
{
|
| 181292 |
+
"epoch": 514.47,
|
| 181293 |
+
"learning_rate": 8.984455128205128e-06,
|
| 181294 |
+
"loss": 0.3225,
|
| 181295 |
+
"step": 64395
|
| 181296 |
+
},
|
| 181297 |
+
{
|
| 181298 |
+
"epoch": 514.51,
|
| 181299 |
+
"learning_rate": 8.984375000000002e-06,
|
| 181300 |
+
"loss": 0.407,
|
| 181301 |
+
"step": 64400
|
| 181302 |
+
},
|
| 181303 |
+
{
|
| 181304 |
+
"epoch": 514.55,
|
| 181305 |
+
"learning_rate": 8.984294871794873e-06,
|
| 181306 |
+
"loss": 0.4698,
|
| 181307 |
+
"step": 64405
|
| 181308 |
+
},
|
| 181309 |
+
{
|
| 181310 |
+
"epoch": 514.59,
|
| 181311 |
+
"learning_rate": 8.984214743589744e-06,
|
| 181312 |
+
"loss": 1.0325,
|
| 181313 |
+
"step": 64410
|
| 181314 |
+
},
|
| 181315 |
+
{
|
| 181316 |
+
"epoch": 514.63,
|
| 181317 |
+
"learning_rate": 8.984134615384617e-06,
|
| 181318 |
+
"loss": 0.6856,
|
| 181319 |
+
"step": 64415
|
| 181320 |
+
},
|
| 181321 |
+
{
|
| 181322 |
+
"epoch": 514.67,
|
| 181323 |
+
"learning_rate": 8.984054487179487e-06,
|
| 181324 |
+
"loss": 0.3071,
|
| 181325 |
+
"step": 64420
|
| 181326 |
+
},
|
| 181327 |
+
{
|
| 181328 |
+
"epoch": 514.71,
|
| 181329 |
+
"learning_rate": 8.98397435897436e-06,
|
| 181330 |
+
"loss": 0.3533,
|
| 181331 |
+
"step": 64425
|
| 181332 |
+
},
|
| 181333 |
+
{
|
| 181334 |
+
"epoch": 514.75,
|
| 181335 |
+
"learning_rate": 8.983894230769231e-06,
|
| 181336 |
+
"loss": 0.4645,
|
| 181337 |
+
"step": 64430
|
| 181338 |
+
},
|
| 181339 |
+
{
|
| 181340 |
+
"epoch": 514.79,
|
| 181341 |
+
"learning_rate": 8.983814102564103e-06,
|
| 181342 |
+
"loss": 0.8507,
|
| 181343 |
+
"step": 64435
|
| 181344 |
+
},
|
| 181345 |
+
{
|
| 181346 |
+
"epoch": 514.83,
|
| 181347 |
+
"learning_rate": 8.983733974358974e-06,
|
| 181348 |
+
"loss": 0.6507,
|
| 181349 |
+
"step": 64440
|
| 181350 |
+
},
|
| 181351 |
+
{
|
| 181352 |
+
"epoch": 514.87,
|
| 181353 |
+
"learning_rate": 8.983653846153847e-06,
|
| 181354 |
+
"loss": 0.2712,
|
| 181355 |
+
"step": 64445
|
| 181356 |
+
},
|
| 181357 |
+
{
|
| 181358 |
+
"epoch": 514.91,
|
| 181359 |
+
"learning_rate": 8.983573717948719e-06,
|
| 181360 |
+
"loss": 0.3556,
|
| 181361 |
+
"step": 64450
|
| 181362 |
+
},
|
| 181363 |
+
{
|
| 181364 |
+
"epoch": 514.95,
|
| 181365 |
+
"learning_rate": 8.98349358974359e-06,
|
| 181366 |
+
"loss": 0.3635,
|
| 181367 |
+
"step": 64455
|
| 181368 |
+
},
|
| 181369 |
+
{
|
| 181370 |
+
"epoch": 514.99,
|
| 181371 |
+
"learning_rate": 8.983413461538463e-06,
|
| 181372 |
+
"loss": 1.0489,
|
| 181373 |
+
"step": 64460
|
| 181374 |
+
},
|
| 181375 |
+
{
|
| 181376 |
+
"epoch": 515.0,
|
| 181377 |
+
"eval_loss": 0.3951534926891327,
|
| 181378 |
+
"eval_runtime": 40.2216,
|
| 181379 |
+
"eval_samples_per_second": 20.81,
|
| 181380 |
+
"eval_steps_per_second": 0.671,
|
| 181381 |
+
"eval_wer": 0.19014445992520349,
|
| 181382 |
+
"step": 64461
|
| 181383 |
}
|
| 181384 |
],
|
| 181385 |
+
"max_steps": 625000,
|
| 181386 |
"num_train_epochs": 5000,
|
| 181387 |
+
"total_flos": 1.8139542363224536e+20,
|
| 181388 |
"trial_name": null,
|
| 181389 |
"trial_params": null
|
| 181390 |
}
|
model-bin/finetune/base/{checkpoint-63838 → checkpoint-64461}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629817865.4918172/events.out.tfevents.1629817865.c435e1c5ee04.920.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfc99d2e12bb5f2f9385da87c530a3106f3f9ce14dd97305ea719c8a08918a52
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629818505.946728/events.out.tfevents.1629818505.c435e1c5ee04.920.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a5d13745bc030fbc74019749474a110d7b7e8d86af79ab0ac84b5fabf7fcf8ea
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629819241.3187594/events.out.tfevents.1629819241.c435e1c5ee04.920.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4cc3e73dc9cd4a55f46e208f117a44a9cf6c20f5c061828f1db4cbfd93f2bddc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629819885.031035/events.out.tfevents.1629819885.c435e1c5ee04.920.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a05122c7c685705132692186300c3057b14aaeea2c0694fbfb9422bf80e5a672
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629820516.7180705/events.out.tfevents.1629820516.c435e1c5ee04.920.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:659c0cff7ca255954b7a667fee94310f45f2b579f3d5760557dd83328dde966c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629817865.c435e1c5ee04.920.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10f3d49c2cbcfa9ad662ed9c17e21ccb5536c13b4d262b9c287a4bbe6db25605
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629818505.c435e1c5ee04.920.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2772c35c003f5e856b69081e4d87ed0fb0388c2bf01ba72c01f009c7b22798ea
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629819241.c435e1c5ee04.920.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77d003bd73767b3550d266e9a783b97cdf79bd1637e96ae3269af40f4ebe6cb6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629819885.c435e1c5ee04.920.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f0b7eccff42426f79e087454180863cb5788e0ed4284b059fb0489e8d3fac1ee
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629820516.c435e1c5ee04.920.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ba75c3a56b4d3e2f2e7ef0100ce80474fabfe421047298d53777d9db9cb2158
|
| 3 |
+
size 8622
|