"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629830879.2497265/events.out.tfevents.1629830879.c435e1c5ee04.920.201 +3 -0
- model-bin/finetune/base/log/1629831513.1270049/events.out.tfevents.1629831513.c435e1c5ee04.920.203 +3 -0
- model-bin/finetune/base/log/1629832142.3883011/events.out.tfevents.1629832142.c435e1c5ee04.920.205 +3 -0
- model-bin/finetune/base/log/1629832782.2191467/events.out.tfevents.1629832782.c435e1c5ee04.920.207 +3 -0
- model-bin/finetune/base/log/1629833409.5438683/events.out.tfevents.1629833409.c435e1c5ee04.920.209 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629830879.c435e1c5ee04.920.200 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629831513.c435e1c5ee04.920.202 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629832142.c435e1c5ee04.920.204 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629832782.c435e1c5ee04.920.206 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629833409.c435e1c5ee04.920.208 +3 -0
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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:a4305b79063344e8dcfa70fc08a440bf763d7a5555ef19d1c29a843c47a2bdd0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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 sha256:3b864d526ba5e96bbeec05edbe0fb74e2fce15748318e767dedff344b89f04bf
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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 sha256:4d3efe634557e655c20c363dfd2ec88502e52a1a2fe24613e53efd8f78e3b4d7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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 sha256:890e7963e056aec6747d41244e17966edf587e5f50edd32309235c4d785e577b
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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 sha256:84754f1abc95954e6ecb719c67cba56880d4f0682e71be6ded68f83ed242eea3
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/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,
|
|
@@ -183753,11 +183753,800 @@
|
|
| 183753 |
"eval_steps_per_second": 0.677,
|
| 183754 |
"eval_wer": 0.19705408178220724,
|
| 183755 |
"step": 66326
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 183756 |
}
|
| 183757 |
],
|
| 183758 |
"max_steps": 620000,
|
| 183759 |
"num_train_epochs": 5000,
|
| 183760 |
-
"total_flos": 1.
|
| 183761 |
"trial_name": null,
|
| 183762 |
"trial_params": null
|
| 183763 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18588425381903642,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-60729",
|
| 4 |
+
"epoch": 538.995983935743,
|
| 5 |
+
"global_step": 66949,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 183753 |
"eval_steps_per_second": 0.677,
|
| 183754 |
"eval_wer": 0.19705408178220724,
|
| 183755 |
"step": 66326
|
| 183756 |
+
},
|
| 183757 |
+
{
|
| 183758 |
+
"epoch": 530.03,
|
| 183759 |
+
"learning_rate": 8.953445512820514e-06,
|
| 183760 |
+
"loss": 0.396,
|
| 183761 |
+
"step": 66330
|
| 183762 |
+
},
|
| 183763 |
+
{
|
| 183764 |
+
"epoch": 530.07,
|
| 183765 |
+
"learning_rate": 8.953365384615385e-06,
|
| 183766 |
+
"loss": 0.3424,
|
| 183767 |
+
"step": 66335
|
| 183768 |
+
},
|
| 183769 |
+
{
|
| 183770 |
+
"epoch": 530.11,
|
| 183771 |
+
"learning_rate": 8.953285256410256e-06,
|
| 183772 |
+
"loss": 0.3231,
|
| 183773 |
+
"step": 66340
|
| 183774 |
+
},
|
| 183775 |
+
{
|
| 183776 |
+
"epoch": 530.15,
|
| 183777 |
+
"learning_rate": 8.95320512820513e-06,
|
| 183778 |
+
"loss": 0.4369,
|
| 183779 |
+
"step": 66345
|
| 183780 |
+
},
|
| 183781 |
+
{
|
| 183782 |
+
"epoch": 530.19,
|
| 183783 |
+
"learning_rate": 8.953125e-06,
|
| 183784 |
+
"loss": 0.9247,
|
| 183785 |
+
"step": 66350
|
| 183786 |
+
},
|
| 183787 |
+
{
|
| 183788 |
+
"epoch": 530.23,
|
| 183789 |
+
"learning_rate": 8.953044871794872e-06,
|
| 183790 |
+
"loss": 0.8341,
|
| 183791 |
+
"step": 66355
|
| 183792 |
+
},
|
| 183793 |
+
{
|
| 183794 |
+
"epoch": 530.27,
|
| 183795 |
+
"learning_rate": 8.952964743589743e-06,
|
| 183796 |
+
"loss": 0.2963,
|
| 183797 |
+
"step": 66360
|
| 183798 |
+
},
|
| 183799 |
+
{
|
| 183800 |
+
"epoch": 530.31,
|
| 183801 |
+
"learning_rate": 8.952884615384617e-06,
|
| 183802 |
+
"loss": 0.3888,
|
| 183803 |
+
"step": 66365
|
| 183804 |
+
},
|
| 183805 |
+
{
|
| 183806 |
+
"epoch": 530.35,
|
| 183807 |
+
"learning_rate": 8.952804487179488e-06,
|
| 183808 |
+
"loss": 0.4379,
|
| 183809 |
+
"step": 66370
|
| 183810 |
+
},
|
| 183811 |
+
{
|
| 183812 |
+
"epoch": 530.39,
|
| 183813 |
+
"learning_rate": 8.95272435897436e-06,
|
| 183814 |
+
"loss": 0.9638,
|
| 183815 |
+
"step": 66375
|
| 183816 |
+
},
|
| 183817 |
+
{
|
| 183818 |
+
"epoch": 530.43,
|
| 183819 |
+
"learning_rate": 8.952644230769232e-06,
|
| 183820 |
+
"loss": 0.9032,
|
| 183821 |
+
"step": 66380
|
| 183822 |
+
},
|
| 183823 |
+
{
|
| 183824 |
+
"epoch": 530.47,
|
| 183825 |
+
"learning_rate": 8.952564102564104e-06,
|
| 183826 |
+
"loss": 0.299,
|
| 183827 |
+
"step": 66385
|
| 183828 |
+
},
|
| 183829 |
+
{
|
| 183830 |
+
"epoch": 530.51,
|
| 183831 |
+
"learning_rate": 8.952483974358975e-06,
|
| 183832 |
+
"loss": 0.4344,
|
| 183833 |
+
"step": 66390
|
| 183834 |
+
},
|
| 183835 |
+
{
|
| 183836 |
+
"epoch": 530.55,
|
| 183837 |
+
"learning_rate": 8.952403846153846e-06,
|
| 183838 |
+
"loss": 0.4741,
|
| 183839 |
+
"step": 66395
|
| 183840 |
+
},
|
| 183841 |
+
{
|
| 183842 |
+
"epoch": 530.59,
|
| 183843 |
+
"learning_rate": 8.95232371794872e-06,
|
| 183844 |
+
"loss": 0.968,
|
| 183845 |
+
"step": 66400
|
| 183846 |
+
},
|
| 183847 |
+
{
|
| 183848 |
+
"epoch": 530.63,
|
| 183849 |
+
"learning_rate": 8.95224358974359e-06,
|
| 183850 |
+
"loss": 0.6931,
|
| 183851 |
+
"step": 66405
|
| 183852 |
+
},
|
| 183853 |
+
{
|
| 183854 |
+
"epoch": 530.67,
|
| 183855 |
+
"learning_rate": 8.952163461538462e-06,
|
| 183856 |
+
"loss": 0.3257,
|
| 183857 |
+
"step": 66410
|
| 183858 |
+
},
|
| 183859 |
+
{
|
| 183860 |
+
"epoch": 530.71,
|
| 183861 |
+
"learning_rate": 8.952083333333335e-06,
|
| 183862 |
+
"loss": 0.3857,
|
| 183863 |
+
"step": 66415
|
| 183864 |
+
},
|
| 183865 |
+
{
|
| 183866 |
+
"epoch": 530.75,
|
| 183867 |
+
"learning_rate": 8.952003205128207e-06,
|
| 183868 |
+
"loss": 0.4299,
|
| 183869 |
+
"step": 66420
|
| 183870 |
+
},
|
| 183871 |
+
{
|
| 183872 |
+
"epoch": 530.79,
|
| 183873 |
+
"learning_rate": 8.951923076923078e-06,
|
| 183874 |
+
"loss": 1.1066,
|
| 183875 |
+
"step": 66425
|
| 183876 |
+
},
|
| 183877 |
+
{
|
| 183878 |
+
"epoch": 530.83,
|
| 183879 |
+
"learning_rate": 8.95184294871795e-06,
|
| 183880 |
+
"loss": 0.9173,
|
| 183881 |
+
"step": 66430
|
| 183882 |
+
},
|
| 183883 |
+
{
|
| 183884 |
+
"epoch": 530.87,
|
| 183885 |
+
"learning_rate": 8.951762820512822e-06,
|
| 183886 |
+
"loss": 0.2734,
|
| 183887 |
+
"step": 66435
|
| 183888 |
+
},
|
| 183889 |
+
{
|
| 183890 |
+
"epoch": 530.91,
|
| 183891 |
+
"learning_rate": 8.951682692307692e-06,
|
| 183892 |
+
"loss": 0.3241,
|
| 183893 |
+
"step": 66440
|
| 183894 |
+
},
|
| 183895 |
+
{
|
| 183896 |
+
"epoch": 530.95,
|
| 183897 |
+
"learning_rate": 8.951602564102565e-06,
|
| 183898 |
+
"loss": 0.4544,
|
| 183899 |
+
"step": 66445
|
| 183900 |
+
},
|
| 183901 |
+
{
|
| 183902 |
+
"epoch": 530.99,
|
| 183903 |
+
"learning_rate": 8.951522435897436e-06,
|
| 183904 |
+
"loss": 1.0191,
|
| 183905 |
+
"step": 66450
|
| 183906 |
+
},
|
| 183907 |
+
{
|
| 183908 |
+
"epoch": 531.0,
|
| 183909 |
+
"eval_loss": 0.406460165977478,
|
| 183910 |
+
"eval_runtime": 40.1796,
|
| 183911 |
+
"eval_samples_per_second": 20.831,
|
| 183912 |
+
"eval_steps_per_second": 0.672,
|
| 183913 |
+
"eval_wer": 0.19212436752951528,
|
| 183914 |
+
"step": 66451
|
| 183915 |
+
},
|
| 183916 |
+
{
|
| 183917 |
+
"epoch": 531.03,
|
| 183918 |
+
"learning_rate": 8.951442307692308e-06,
|
| 183919 |
+
"loss": 0.6888,
|
| 183920 |
+
"step": 66455
|
| 183921 |
+
},
|
| 183922 |
+
{
|
| 183923 |
+
"epoch": 531.07,
|
| 183924 |
+
"learning_rate": 8.951362179487179e-06,
|
| 183925 |
+
"loss": 0.3343,
|
| 183926 |
+
"step": 66460
|
| 183927 |
+
},
|
| 183928 |
+
{
|
| 183929 |
+
"epoch": 531.11,
|
| 183930 |
+
"learning_rate": 8.951282051282052e-06,
|
| 183931 |
+
"loss": 0.3268,
|
| 183932 |
+
"step": 66465
|
| 183933 |
+
},
|
| 183934 |
+
{
|
| 183935 |
+
"epoch": 531.15,
|
| 183936 |
+
"learning_rate": 8.951201923076924e-06,
|
| 183937 |
+
"loss": 0.3481,
|
| 183938 |
+
"step": 66470
|
| 183939 |
+
},
|
| 183940 |
+
{
|
| 183941 |
+
"epoch": 531.19,
|
| 183942 |
+
"learning_rate": 8.951121794871795e-06,
|
| 183943 |
+
"loss": 0.9568,
|
| 183944 |
+
"step": 66475
|
| 183945 |
+
},
|
| 183946 |
+
{
|
| 183947 |
+
"epoch": 531.23,
|
| 183948 |
+
"learning_rate": 8.951041666666668e-06,
|
| 183949 |
+
"loss": 0.6338,
|
| 183950 |
+
"step": 66480
|
| 183951 |
+
},
|
| 183952 |
+
{
|
| 183953 |
+
"epoch": 531.27,
|
| 183954 |
+
"learning_rate": 8.95096153846154e-06,
|
| 183955 |
+
"loss": 0.3239,
|
| 183956 |
+
"step": 66485
|
| 183957 |
+
},
|
| 183958 |
+
{
|
| 183959 |
+
"epoch": 531.31,
|
| 183960 |
+
"learning_rate": 8.95088141025641e-06,
|
| 183961 |
+
"loss": 0.3678,
|
| 183962 |
+
"step": 66490
|
| 183963 |
+
},
|
| 183964 |
+
{
|
| 183965 |
+
"epoch": 531.35,
|
| 183966 |
+
"learning_rate": 8.950801282051282e-06,
|
| 183967 |
+
"loss": 0.4088,
|
| 183968 |
+
"step": 66495
|
| 183969 |
+
},
|
| 183970 |
+
{
|
| 183971 |
+
"epoch": 531.39,
|
| 183972 |
+
"learning_rate": 8.950721153846155e-06,
|
| 183973 |
+
"loss": 0.9323,
|
| 183974 |
+
"step": 66500
|
| 183975 |
+
},
|
| 183976 |
+
{
|
| 183977 |
+
"epoch": 531.43,
|
| 183978 |
+
"learning_rate": 8.950641025641026e-06,
|
| 183979 |
+
"loss": 0.6159,
|
| 183980 |
+
"step": 66505
|
| 183981 |
+
},
|
| 183982 |
+
{
|
| 183983 |
+
"epoch": 531.47,
|
| 183984 |
+
"learning_rate": 8.950560897435898e-06,
|
| 183985 |
+
"loss": 0.3322,
|
| 183986 |
+
"step": 66510
|
| 183987 |
+
},
|
| 183988 |
+
{
|
| 183989 |
+
"epoch": 531.51,
|
| 183990 |
+
"learning_rate": 8.950480769230771e-06,
|
| 183991 |
+
"loss": 0.3244,
|
| 183992 |
+
"step": 66515
|
| 183993 |
+
},
|
| 183994 |
+
{
|
| 183995 |
+
"epoch": 531.55,
|
| 183996 |
+
"learning_rate": 8.950400641025642e-06,
|
| 183997 |
+
"loss": 0.4675,
|
| 183998 |
+
"step": 66520
|
| 183999 |
+
},
|
| 184000 |
+
{
|
| 184001 |
+
"epoch": 531.59,
|
| 184002 |
+
"learning_rate": 8.950320512820514e-06,
|
| 184003 |
+
"loss": 0.8916,
|
| 184004 |
+
"step": 66525
|
| 184005 |
+
},
|
| 184006 |
+
{
|
| 184007 |
+
"epoch": 531.63,
|
| 184008 |
+
"learning_rate": 8.950240384615385e-06,
|
| 184009 |
+
"loss": 0.7,
|
| 184010 |
+
"step": 66530
|
| 184011 |
+
},
|
| 184012 |
+
{
|
| 184013 |
+
"epoch": 531.67,
|
| 184014 |
+
"learning_rate": 8.950160256410258e-06,
|
| 184015 |
+
"loss": 0.2958,
|
| 184016 |
+
"step": 66535
|
| 184017 |
+
},
|
| 184018 |
+
{
|
| 184019 |
+
"epoch": 531.71,
|
| 184020 |
+
"learning_rate": 8.95008012820513e-06,
|
| 184021 |
+
"loss": 0.3249,
|
| 184022 |
+
"step": 66540
|
| 184023 |
+
},
|
| 184024 |
+
{
|
| 184025 |
+
"epoch": 531.75,
|
| 184026 |
+
"learning_rate": 8.95e-06,
|
| 184027 |
+
"loss": 0.4388,
|
| 184028 |
+
"step": 66545
|
| 184029 |
+
},
|
| 184030 |
+
{
|
| 184031 |
+
"epoch": 531.79,
|
| 184032 |
+
"learning_rate": 8.949919871794872e-06,
|
| 184033 |
+
"loss": 0.877,
|
| 184034 |
+
"step": 66550
|
| 184035 |
+
},
|
| 184036 |
+
{
|
| 184037 |
+
"epoch": 531.83,
|
| 184038 |
+
"learning_rate": 8.949839743589745e-06,
|
| 184039 |
+
"loss": 0.6828,
|
| 184040 |
+
"step": 66555
|
| 184041 |
+
},
|
| 184042 |
+
{
|
| 184043 |
+
"epoch": 531.87,
|
| 184044 |
+
"learning_rate": 8.949759615384615e-06,
|
| 184045 |
+
"loss": 0.3371,
|
| 184046 |
+
"step": 66560
|
| 184047 |
+
},
|
| 184048 |
+
{
|
| 184049 |
+
"epoch": 531.91,
|
| 184050 |
+
"learning_rate": 8.949679487179488e-06,
|
| 184051 |
+
"loss": 0.344,
|
| 184052 |
+
"step": 66565
|
| 184053 |
+
},
|
| 184054 |
+
{
|
| 184055 |
+
"epoch": 531.95,
|
| 184056 |
+
"learning_rate": 8.949599358974361e-06,
|
| 184057 |
+
"loss": 0.4809,
|
| 184058 |
+
"step": 66570
|
| 184059 |
+
},
|
| 184060 |
+
{
|
| 184061 |
+
"epoch": 531.99,
|
| 184062 |
+
"learning_rate": 8.94951923076923e-06,
|
| 184063 |
+
"loss": 1.0281,
|
| 184064 |
+
"step": 66575
|
| 184065 |
+
},
|
| 184066 |
+
{
|
| 184067 |
+
"epoch": 532.0,
|
| 184068 |
+
"eval_loss": 0.4383997619152069,
|
| 184069 |
+
"eval_runtime": 39.6825,
|
| 184070 |
+
"eval_samples_per_second": 21.092,
|
| 184071 |
+
"eval_steps_per_second": 0.68,
|
| 184072 |
+
"eval_wer": 0.20101062643977113,
|
| 184073 |
+
"step": 66576
|
| 184074 |
+
},
|
| 184075 |
+
{
|
| 184076 |
+
"epoch": 532.03,
|
| 184077 |
+
"learning_rate": 8.949439102564104e-06,
|
| 184078 |
+
"loss": 0.3595,
|
| 184079 |
+
"step": 66580
|
| 184080 |
+
},
|
| 184081 |
+
{
|
| 184082 |
+
"epoch": 532.07,
|
| 184083 |
+
"learning_rate": 8.949358974358975e-06,
|
| 184084 |
+
"loss": 0.3205,
|
| 184085 |
+
"step": 66585
|
| 184086 |
+
},
|
| 184087 |
+
{
|
| 184088 |
+
"epoch": 532.11,
|
| 184089 |
+
"learning_rate": 8.949278846153846e-06,
|
| 184090 |
+
"loss": 0.2612,
|
| 184091 |
+
"step": 66590
|
| 184092 |
+
},
|
| 184093 |
+
{
|
| 184094 |
+
"epoch": 532.15,
|
| 184095 |
+
"learning_rate": 8.949198717948718e-06,
|
| 184096 |
+
"loss": 0.4,
|
| 184097 |
+
"step": 66595
|
| 184098 |
+
},
|
| 184099 |
+
{
|
| 184100 |
+
"epoch": 532.19,
|
| 184101 |
+
"learning_rate": 8.94911858974359e-06,
|
| 184102 |
+
"loss": 0.9481,
|
| 184103 |
+
"step": 66600
|
| 184104 |
+
},
|
| 184105 |
+
{
|
| 184106 |
+
"epoch": 532.23,
|
| 184107 |
+
"learning_rate": 8.949038461538462e-06,
|
| 184108 |
+
"loss": 0.7302,
|
| 184109 |
+
"step": 66605
|
| 184110 |
+
},
|
| 184111 |
+
{
|
| 184112 |
+
"epoch": 532.27,
|
| 184113 |
+
"learning_rate": 8.948958333333333e-06,
|
| 184114 |
+
"loss": 0.3677,
|
| 184115 |
+
"step": 66610
|
| 184116 |
+
},
|
| 184117 |
+
{
|
| 184118 |
+
"epoch": 532.31,
|
| 184119 |
+
"learning_rate": 8.948878205128207e-06,
|
| 184120 |
+
"loss": 0.3763,
|
| 184121 |
+
"step": 66615
|
| 184122 |
+
},
|
| 184123 |
+
{
|
| 184124 |
+
"epoch": 532.35,
|
| 184125 |
+
"learning_rate": 8.948798076923078e-06,
|
| 184126 |
+
"loss": 0.4758,
|
| 184127 |
+
"step": 66620
|
| 184128 |
+
},
|
| 184129 |
+
{
|
| 184130 |
+
"epoch": 532.39,
|
| 184131 |
+
"learning_rate": 8.94871794871795e-06,
|
| 184132 |
+
"loss": 0.9979,
|
| 184133 |
+
"step": 66625
|
| 184134 |
+
},
|
| 184135 |
+
{
|
| 184136 |
+
"epoch": 532.43,
|
| 184137 |
+
"learning_rate": 8.94863782051282e-06,
|
| 184138 |
+
"loss": 0.7855,
|
| 184139 |
+
"step": 66630
|
| 184140 |
+
},
|
| 184141 |
+
{
|
| 184142 |
+
"epoch": 532.47,
|
| 184143 |
+
"learning_rate": 8.948557692307694e-06,
|
| 184144 |
+
"loss": 0.318,
|
| 184145 |
+
"step": 66635
|
| 184146 |
+
},
|
| 184147 |
+
{
|
| 184148 |
+
"epoch": 532.51,
|
| 184149 |
+
"learning_rate": 8.948477564102565e-06,
|
| 184150 |
+
"loss": 0.3541,
|
| 184151 |
+
"step": 66640
|
| 184152 |
+
},
|
| 184153 |
+
{
|
| 184154 |
+
"epoch": 532.55,
|
| 184155 |
+
"learning_rate": 8.948397435897436e-06,
|
| 184156 |
+
"loss": 0.406,
|
| 184157 |
+
"step": 66645
|
| 184158 |
+
},
|
| 184159 |
+
{
|
| 184160 |
+
"epoch": 532.59,
|
| 184161 |
+
"learning_rate": 8.948317307692308e-06,
|
| 184162 |
+
"loss": 1.0155,
|
| 184163 |
+
"step": 66650
|
| 184164 |
+
},
|
| 184165 |
+
{
|
| 184166 |
+
"epoch": 532.63,
|
| 184167 |
+
"learning_rate": 8.94823717948718e-06,
|
| 184168 |
+
"loss": 0.8103,
|
| 184169 |
+
"step": 66655
|
| 184170 |
+
},
|
| 184171 |
+
{
|
| 184172 |
+
"epoch": 532.67,
|
| 184173 |
+
"learning_rate": 8.948157051282052e-06,
|
| 184174 |
+
"loss": 0.3239,
|
| 184175 |
+
"step": 66660
|
| 184176 |
+
},
|
| 184177 |
+
{
|
| 184178 |
+
"epoch": 532.71,
|
| 184179 |
+
"learning_rate": 8.948076923076923e-06,
|
| 184180 |
+
"loss": 0.3431,
|
| 184181 |
+
"step": 66665
|
| 184182 |
+
},
|
| 184183 |
+
{
|
| 184184 |
+
"epoch": 532.75,
|
| 184185 |
+
"learning_rate": 8.947996794871797e-06,
|
| 184186 |
+
"loss": 0.4256,
|
| 184187 |
+
"step": 66670
|
| 184188 |
+
},
|
| 184189 |
+
{
|
| 184190 |
+
"epoch": 532.79,
|
| 184191 |
+
"learning_rate": 8.947916666666668e-06,
|
| 184192 |
+
"loss": 0.8556,
|
| 184193 |
+
"step": 66675
|
| 184194 |
+
},
|
| 184195 |
+
{
|
| 184196 |
+
"epoch": 532.83,
|
| 184197 |
+
"learning_rate": 8.94783653846154e-06,
|
| 184198 |
+
"loss": 0.6975,
|
| 184199 |
+
"step": 66680
|
| 184200 |
+
},
|
| 184201 |
+
{
|
| 184202 |
+
"epoch": 532.87,
|
| 184203 |
+
"learning_rate": 8.94775641025641e-06,
|
| 184204 |
+
"loss": 0.2859,
|
| 184205 |
+
"step": 66685
|
| 184206 |
+
},
|
| 184207 |
+
{
|
| 184208 |
+
"epoch": 532.91,
|
| 184209 |
+
"learning_rate": 8.947676282051284e-06,
|
| 184210 |
+
"loss": 0.3586,
|
| 184211 |
+
"step": 66690
|
| 184212 |
+
},
|
| 184213 |
+
{
|
| 184214 |
+
"epoch": 532.95,
|
| 184215 |
+
"learning_rate": 8.947596153846155e-06,
|
| 184216 |
+
"loss": 0.4585,
|
| 184217 |
+
"step": 66695
|
| 184218 |
+
},
|
| 184219 |
+
{
|
| 184220 |
+
"epoch": 532.99,
|
| 184221 |
+
"learning_rate": 8.947516025641026e-06,
|
| 184222 |
+
"loss": 0.9764,
|
| 184223 |
+
"step": 66700
|
| 184224 |
+
},
|
| 184225 |
+
{
|
| 184226 |
+
"epoch": 533.0,
|
| 184227 |
+
"eval_loss": 0.3940429985523224,
|
| 184228 |
+
"eval_runtime": 40.3665,
|
| 184229 |
+
"eval_samples_per_second": 20.71,
|
| 184230 |
+
"eval_steps_per_second": 0.669,
|
| 184231 |
+
"eval_wer": 0.18733183041233364,
|
| 184232 |
+
"step": 66701
|
| 184233 |
+
},
|
| 184234 |
+
{
|
| 184235 |
+
"epoch": 537.03,
|
| 184236 |
+
"learning_rate": 8.947435897435898e-06,
|
| 184237 |
+
"loss": 0.3689,
|
| 184238 |
+
"step": 66705
|
| 184239 |
+
},
|
| 184240 |
+
{
|
| 184241 |
+
"epoch": 537.07,
|
| 184242 |
+
"learning_rate": 8.94735576923077e-06,
|
| 184243 |
+
"loss": 0.2814,
|
| 184244 |
+
"step": 66710
|
| 184245 |
+
},
|
| 184246 |
+
{
|
| 184247 |
+
"epoch": 537.11,
|
| 184248 |
+
"learning_rate": 8.947275641025642e-06,
|
| 184249 |
+
"loss": 0.3363,
|
| 184250 |
+
"step": 66715
|
| 184251 |
+
},
|
| 184252 |
+
{
|
| 184253 |
+
"epoch": 537.15,
|
| 184254 |
+
"learning_rate": 8.947195512820514e-06,
|
| 184255 |
+
"loss": 0.4299,
|
| 184256 |
+
"step": 66720
|
| 184257 |
+
},
|
| 184258 |
+
{
|
| 184259 |
+
"epoch": 537.19,
|
| 184260 |
+
"learning_rate": 8.947115384615387e-06,
|
| 184261 |
+
"loss": 1.1418,
|
| 184262 |
+
"step": 66725
|
| 184263 |
+
},
|
| 184264 |
+
{
|
| 184265 |
+
"epoch": 537.23,
|
| 184266 |
+
"learning_rate": 8.947035256410256e-06,
|
| 184267 |
+
"loss": 0.6762,
|
| 184268 |
+
"step": 66730
|
| 184269 |
+
},
|
| 184270 |
+
{
|
| 184271 |
+
"epoch": 537.27,
|
| 184272 |
+
"learning_rate": 8.94695512820513e-06,
|
| 184273 |
+
"loss": 0.2862,
|
| 184274 |
+
"step": 66735
|
| 184275 |
+
},
|
| 184276 |
+
{
|
| 184277 |
+
"epoch": 537.31,
|
| 184278 |
+
"learning_rate": 8.946875e-06,
|
| 184279 |
+
"loss": 0.3495,
|
| 184280 |
+
"step": 66740
|
| 184281 |
+
},
|
| 184282 |
+
{
|
| 184283 |
+
"epoch": 537.35,
|
| 184284 |
+
"learning_rate": 8.946794871794872e-06,
|
| 184285 |
+
"loss": 0.4347,
|
| 184286 |
+
"step": 66745
|
| 184287 |
+
},
|
| 184288 |
+
{
|
| 184289 |
+
"epoch": 537.39,
|
| 184290 |
+
"learning_rate": 8.946714743589743e-06,
|
| 184291 |
+
"loss": 0.9885,
|
| 184292 |
+
"step": 66750
|
| 184293 |
+
},
|
| 184294 |
+
{
|
| 184295 |
+
"epoch": 537.43,
|
| 184296 |
+
"learning_rate": 8.946634615384616e-06,
|
| 184297 |
+
"loss": 0.8602,
|
| 184298 |
+
"step": 66755
|
| 184299 |
+
},
|
| 184300 |
+
{
|
| 184301 |
+
"epoch": 537.47,
|
| 184302 |
+
"learning_rate": 8.946554487179488e-06,
|
| 184303 |
+
"loss": 0.3301,
|
| 184304 |
+
"step": 66760
|
| 184305 |
+
},
|
| 184306 |
+
{
|
| 184307 |
+
"epoch": 537.51,
|
| 184308 |
+
"learning_rate": 8.946474358974359e-06,
|
| 184309 |
+
"loss": 0.3302,
|
| 184310 |
+
"step": 66765
|
| 184311 |
+
},
|
| 184312 |
+
{
|
| 184313 |
+
"epoch": 537.55,
|
| 184314 |
+
"learning_rate": 8.946394230769232e-06,
|
| 184315 |
+
"loss": 0.4187,
|
| 184316 |
+
"step": 66770
|
| 184317 |
+
},
|
| 184318 |
+
{
|
| 184319 |
+
"epoch": 537.59,
|
| 184320 |
+
"learning_rate": 8.946314102564104e-06,
|
| 184321 |
+
"loss": 1.0443,
|
| 184322 |
+
"step": 66775
|
| 184323 |
+
},
|
| 184324 |
+
{
|
| 184325 |
+
"epoch": 537.63,
|
| 184326 |
+
"learning_rate": 8.946233974358975e-06,
|
| 184327 |
+
"loss": 0.6987,
|
| 184328 |
+
"step": 66780
|
| 184329 |
+
},
|
| 184330 |
+
{
|
| 184331 |
+
"epoch": 537.67,
|
| 184332 |
+
"learning_rate": 8.946153846153846e-06,
|
| 184333 |
+
"loss": 0.2443,
|
| 184334 |
+
"step": 66785
|
| 184335 |
+
},
|
| 184336 |
+
{
|
| 184337 |
+
"epoch": 537.71,
|
| 184338 |
+
"learning_rate": 8.94607371794872e-06,
|
| 184339 |
+
"loss": 0.3445,
|
| 184340 |
+
"step": 66790
|
| 184341 |
+
},
|
| 184342 |
+
{
|
| 184343 |
+
"epoch": 537.76,
|
| 184344 |
+
"learning_rate": 8.94599358974359e-06,
|
| 184345 |
+
"loss": 0.4564,
|
| 184346 |
+
"step": 66795
|
| 184347 |
+
},
|
| 184348 |
+
{
|
| 184349 |
+
"epoch": 537.8,
|
| 184350 |
+
"learning_rate": 8.945913461538462e-06,
|
| 184351 |
+
"loss": 0.9662,
|
| 184352 |
+
"step": 66800
|
| 184353 |
+
},
|
| 184354 |
+
{
|
| 184355 |
+
"epoch": 537.84,
|
| 184356 |
+
"learning_rate": 8.945833333333333e-06,
|
| 184357 |
+
"loss": 0.7671,
|
| 184358 |
+
"step": 66805
|
| 184359 |
+
},
|
| 184360 |
+
{
|
| 184361 |
+
"epoch": 537.88,
|
| 184362 |
+
"learning_rate": 8.945753205128206e-06,
|
| 184363 |
+
"loss": 0.3258,
|
| 184364 |
+
"step": 66810
|
| 184365 |
+
},
|
| 184366 |
+
{
|
| 184367 |
+
"epoch": 537.92,
|
| 184368 |
+
"learning_rate": 8.945673076923078e-06,
|
| 184369 |
+
"loss": 0.3372,
|
| 184370 |
+
"step": 66815
|
| 184371 |
+
},
|
| 184372 |
+
{
|
| 184373 |
+
"epoch": 537.96,
|
| 184374 |
+
"learning_rate": 8.945592948717949e-06,
|
| 184375 |
+
"loss": 0.4195,
|
| 184376 |
+
"step": 66820
|
| 184377 |
+
},
|
| 184378 |
+
{
|
| 184379 |
+
"epoch": 538.0,
|
| 184380 |
+
"learning_rate": 8.945512820512822e-06,
|
| 184381 |
+
"loss": 1.161,
|
| 184382 |
+
"step": 66825
|
| 184383 |
+
},
|
| 184384 |
+
{
|
| 184385 |
+
"epoch": 538.0,
|
| 184386 |
+
"eval_loss": 0.3947887122631073,
|
| 184387 |
+
"eval_runtime": 39.1445,
|
| 184388 |
+
"eval_samples_per_second": 21.357,
|
| 184389 |
+
"eval_steps_per_second": 0.69,
|
| 184390 |
+
"eval_wer": 0.19173909813535397,
|
| 184391 |
+
"step": 66825
|
| 184392 |
+
},
|
| 184393 |
+
{
|
| 184394 |
+
"epoch": 538.04,
|
| 184395 |
+
"learning_rate": 8.945432692307694e-06,
|
| 184396 |
+
"loss": 0.3732,
|
| 184397 |
+
"step": 66830
|
| 184398 |
+
},
|
| 184399 |
+
{
|
| 184400 |
+
"epoch": 538.08,
|
| 184401 |
+
"learning_rate": 8.945352564102565e-06,
|
| 184402 |
+
"loss": 0.3088,
|
| 184403 |
+
"step": 66835
|
| 184404 |
+
},
|
| 184405 |
+
{
|
| 184406 |
+
"epoch": 538.12,
|
| 184407 |
+
"learning_rate": 8.945272435897436e-06,
|
| 184408 |
+
"loss": 0.3213,
|
| 184409 |
+
"step": 66840
|
| 184410 |
+
},
|
| 184411 |
+
{
|
| 184412 |
+
"epoch": 538.16,
|
| 184413 |
+
"learning_rate": 8.94519230769231e-06,
|
| 184414 |
+
"loss": 0.5747,
|
| 184415 |
+
"step": 66845
|
| 184416 |
+
},
|
| 184417 |
+
{
|
| 184418 |
+
"epoch": 538.2,
|
| 184419 |
+
"learning_rate": 8.945112179487179e-06,
|
| 184420 |
+
"loss": 1.2296,
|
| 184421 |
+
"step": 66850
|
| 184422 |
+
},
|
| 184423 |
+
{
|
| 184424 |
+
"epoch": 538.24,
|
| 184425 |
+
"learning_rate": 8.945032051282052e-06,
|
| 184426 |
+
"loss": 0.3106,
|
| 184427 |
+
"step": 66855
|
| 184428 |
+
},
|
| 184429 |
+
{
|
| 184430 |
+
"epoch": 538.28,
|
| 184431 |
+
"learning_rate": 8.944951923076923e-06,
|
| 184432 |
+
"loss": 0.2994,
|
| 184433 |
+
"step": 66860
|
| 184434 |
+
},
|
| 184435 |
+
{
|
| 184436 |
+
"epoch": 538.32,
|
| 184437 |
+
"learning_rate": 8.944871794871795e-06,
|
| 184438 |
+
"loss": 0.3399,
|
| 184439 |
+
"step": 66865
|
| 184440 |
+
},
|
| 184441 |
+
{
|
| 184442 |
+
"epoch": 538.36,
|
| 184443 |
+
"learning_rate": 8.944791666666668e-06,
|
| 184444 |
+
"loss": 0.4754,
|
| 184445 |
+
"step": 66870
|
| 184446 |
+
},
|
| 184447 |
+
{
|
| 184448 |
+
"epoch": 538.4,
|
| 184449 |
+
"learning_rate": 8.94471153846154e-06,
|
| 184450 |
+
"loss": 1.396,
|
| 184451 |
+
"step": 66875
|
| 184452 |
+
},
|
| 184453 |
+
{
|
| 184454 |
+
"epoch": 538.44,
|
| 184455 |
+
"learning_rate": 8.94463141025641e-06,
|
| 184456 |
+
"loss": 0.3324,
|
| 184457 |
+
"step": 66880
|
| 184458 |
+
},
|
| 184459 |
+
{
|
| 184460 |
+
"epoch": 538.48,
|
| 184461 |
+
"learning_rate": 8.944551282051282e-06,
|
| 184462 |
+
"loss": 0.2814,
|
| 184463 |
+
"step": 66885
|
| 184464 |
+
},
|
| 184465 |
+
{
|
| 184466 |
+
"epoch": 538.52,
|
| 184467 |
+
"learning_rate": 8.944471153846155e-06,
|
| 184468 |
+
"loss": 0.3601,
|
| 184469 |
+
"step": 66890
|
| 184470 |
+
},
|
| 184471 |
+
{
|
| 184472 |
+
"epoch": 538.56,
|
| 184473 |
+
"learning_rate": 8.944391025641026e-06,
|
| 184474 |
+
"loss": 0.5061,
|
| 184475 |
+
"step": 66895
|
| 184476 |
+
},
|
| 184477 |
+
{
|
| 184478 |
+
"epoch": 538.6,
|
| 184479 |
+
"learning_rate": 8.944310897435898e-06,
|
| 184480 |
+
"loss": 1.241,
|
| 184481 |
+
"step": 66900
|
| 184482 |
+
},
|
| 184483 |
+
{
|
| 184484 |
+
"epoch": 538.64,
|
| 184485 |
+
"learning_rate": 8.944230769230769e-06,
|
| 184486 |
+
"loss": 0.3642,
|
| 184487 |
+
"step": 66905
|
| 184488 |
+
},
|
| 184489 |
+
{
|
| 184490 |
+
"epoch": 538.68,
|
| 184491 |
+
"learning_rate": 8.944150641025642e-06,
|
| 184492 |
+
"loss": 0.3311,
|
| 184493 |
+
"step": 66910
|
| 184494 |
+
},
|
| 184495 |
+
{
|
| 184496 |
+
"epoch": 538.72,
|
| 184497 |
+
"learning_rate": 8.944070512820513e-06,
|
| 184498 |
+
"loss": 0.3155,
|
| 184499 |
+
"step": 66915
|
| 184500 |
+
},
|
| 184501 |
+
{
|
| 184502 |
+
"epoch": 538.76,
|
| 184503 |
+
"learning_rate": 8.943990384615385e-06,
|
| 184504 |
+
"loss": 0.4952,
|
| 184505 |
+
"step": 66920
|
| 184506 |
+
},
|
| 184507 |
+
{
|
| 184508 |
+
"epoch": 538.8,
|
| 184509 |
+
"learning_rate": 8.943910256410258e-06,
|
| 184510 |
+
"loss": 1.324,
|
| 184511 |
+
"step": 66925
|
| 184512 |
+
},
|
| 184513 |
+
{
|
| 184514 |
+
"epoch": 538.84,
|
| 184515 |
+
"learning_rate": 8.94383012820513e-06,
|
| 184516 |
+
"loss": 0.3858,
|
| 184517 |
+
"step": 66930
|
| 184518 |
+
},
|
| 184519 |
+
{
|
| 184520 |
+
"epoch": 538.88,
|
| 184521 |
+
"learning_rate": 8.94375e-06,
|
| 184522 |
+
"loss": 0.3126,
|
| 184523 |
+
"step": 66935
|
| 184524 |
+
},
|
| 184525 |
+
{
|
| 184526 |
+
"epoch": 538.92,
|
| 184527 |
+
"learning_rate": 8.943669871794872e-06,
|
| 184528 |
+
"loss": 0.4153,
|
| 184529 |
+
"step": 66940
|
| 184530 |
+
},
|
| 184531 |
+
{
|
| 184532 |
+
"epoch": 538.96,
|
| 184533 |
+
"learning_rate": 8.943589743589745e-06,
|
| 184534 |
+
"loss": 0.4908,
|
| 184535 |
+
"step": 66945
|
| 184536 |
+
},
|
| 184537 |
+
{
|
| 184538 |
+
"epoch": 539.0,
|
| 184539 |
+
"eval_loss": 0.4666649401187897,
|
| 184540 |
+
"eval_runtime": 39.7494,
|
| 184541 |
+
"eval_samples_per_second": 21.032,
|
| 184542 |
+
"eval_steps_per_second": 0.679,
|
| 184543 |
+
"eval_wer": 0.20303119051105578,
|
| 184544 |
+
"step": 66949
|
| 184545 |
}
|
| 184546 |
],
|
| 184547 |
"max_steps": 620000,
|
| 184548 |
"num_train_epochs": 5000,
|
| 184549 |
+
"total_flos": 1.8839674795561802e+20,
|
| 184550 |
"trial_name": null,
|
| 184551 |
"trial_params": null
|
| 184552 |
}
|
model-bin/finetune/base/{checkpoint-66326 → checkpoint-66949}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629830879.2497265/events.out.tfevents.1629830879.c435e1c5ee04.920.201
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6b75676a0a25dc896e7207461989b0632eafc614c13a359f7aaef1cf96c72192
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629831513.1270049/events.out.tfevents.1629831513.c435e1c5ee04.920.203
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95578367bfe0c98bf978e589caeb48f0ebcf8a828aad2c575fd43f3fd2b30c20
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629832142.3883011/events.out.tfevents.1629832142.c435e1c5ee04.920.205
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e86a97f637df3a6ec1ece5aafd69d2db481f19c1cd9058e34b9dc464299cc6d4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629832782.2191467/events.out.tfevents.1629832782.c435e1c5ee04.920.207
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c93212efcd7f1a4db36f67a8d30d53dbac379210090f1a86635a30adef0a2d56
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629833409.5438683/events.out.tfevents.1629833409.c435e1c5ee04.920.209
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e6ef65c3871d02573b75fa0b9d0d3435d29ce67c10cfbce24f39e9a0b091ff1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629830879.c435e1c5ee04.920.200
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:25291100f3c75456226310622eb0631232e9a2e31acc0b8637445bcc4e620426
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629831513.c435e1c5ee04.920.202
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01f1bbbe487a27e4ced6f72dd38a0e369926931e3eedf287f6df34593805cf2d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629832142.c435e1c5ee04.920.204
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6a824752ad2615bbf723f157a7bd782eb94bbdd1ad3c483b26b63c499a87e58
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629832782.c435e1c5ee04.920.206
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8371499d74a7d558a9e71f461c29889769a702010dc01b552c2da272d678e078
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629833409.c435e1c5ee04.920.208
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d32682a66d7ceb182068ad415fb4e5a85eea63b4994c01e34fb4be2a1fc4fc6e
|
| 3 |
+
size 8462
|