"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630210217.851614/events.out.tfevents.1630210217.cc93b136ebf5.1086.29 +3 -0
- model-bin/finetune/base/log/1630210657.964172/events.out.tfevents.1630210657.cc93b136ebf5.1086.31 +3 -0
- model-bin/finetune/base/log/1630211223.3432353/events.out.tfevents.1630211223.cc93b136ebf5.1086.33 +3 -0
- model-bin/finetune/base/log/1630211666.4765499/events.out.tfevents.1630211666.cc93b136ebf5.1086.35 +3 -0
- model-bin/finetune/base/log/1630212104.3499668/events.out.tfevents.1630212104.cc93b136ebf5.1086.37 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630210217.cc93b136ebf5.1086.28 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630210657.cc93b136ebf5.1086.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630211223.cc93b136ebf5.1086.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630211666.cc93b136ebf5.1086.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630212104.cc93b136ebf5.1086.36 +3 -0
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/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 sha256:7772525640a71b94879c52f0a8df49e5016ee6aa63e7d2a2617e69741a553f6a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/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:f2c46c95f08599e97c7b8af2fcb752dc29b3c24da837b1bf54096ea462fa1b1d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/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:ab856d5c1a50208d454e8f6bea1f46dc6469b0f803326d076dac2a148f7479be
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/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 sha256:24c524215320dbd91b7fda70e307ff61c64b39d6458057cd5b75b3d4be7f9c2c
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/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:eeb9204627a0bfb8dbb675ece7e93c026210f3a03315b8bc633c50705545ebfd
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -278604,11 +278604,800 @@
|
|
| 278604 |
"eval_steps_per_second": 0.616,
|
| 278605 |
"eval_wer": 0.1768541741311358,
|
| 278606 |
"step": 140875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 278607 |
}
|
| 278608 |
],
|
| 278609 |
-
"max_steps":
|
| 278610 |
"num_train_epochs": 5000,
|
| 278611 |
-
"total_flos": 3.
|
| 278612 |
"trial_name": null,
|
| 278613 |
"trial_params": null
|
| 278614 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1140.995983935743,
|
| 5 |
+
"global_step": 141498,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 278604 |
"eval_steps_per_second": 0.616,
|
| 278605 |
"eval_wer": 0.1768541741311358,
|
| 278606 |
"step": 140875
|
| 278607 |
+
},
|
| 278608 |
+
{
|
| 278609 |
+
"epoch": 1127.04,
|
| 278610 |
+
"learning_rate": 7.741114701130857e-06,
|
| 278611 |
+
"loss": 0.2934,
|
| 278612 |
+
"step": 140880
|
| 278613 |
+
},
|
| 278614 |
+
{
|
| 278615 |
+
"epoch": 1127.08,
|
| 278616 |
+
"learning_rate": 7.741033925686592e-06,
|
| 278617 |
+
"loss": 0.2902,
|
| 278618 |
+
"step": 140885
|
| 278619 |
+
},
|
| 278620 |
+
{
|
| 278621 |
+
"epoch": 1127.12,
|
| 278622 |
+
"learning_rate": 7.740953150242327e-06,
|
| 278623 |
+
"loss": 0.3104,
|
| 278624 |
+
"step": 140890
|
| 278625 |
+
},
|
| 278626 |
+
{
|
| 278627 |
+
"epoch": 1127.16,
|
| 278628 |
+
"learning_rate": 7.740872374798062e-06,
|
| 278629 |
+
"loss": 0.4304,
|
| 278630 |
+
"step": 140895
|
| 278631 |
+
},
|
| 278632 |
+
{
|
| 278633 |
+
"epoch": 1127.2,
|
| 278634 |
+
"learning_rate": 7.740791599353797e-06,
|
| 278635 |
+
"loss": 1.0707,
|
| 278636 |
+
"step": 140900
|
| 278637 |
+
},
|
| 278638 |
+
{
|
| 278639 |
+
"epoch": 1127.24,
|
| 278640 |
+
"learning_rate": 7.740710823909532e-06,
|
| 278641 |
+
"loss": 0.345,
|
| 278642 |
+
"step": 140905
|
| 278643 |
+
},
|
| 278644 |
+
{
|
| 278645 |
+
"epoch": 1127.28,
|
| 278646 |
+
"learning_rate": 7.740630048465267e-06,
|
| 278647 |
+
"loss": 0.2858,
|
| 278648 |
+
"step": 140910
|
| 278649 |
+
},
|
| 278650 |
+
{
|
| 278651 |
+
"epoch": 1127.32,
|
| 278652 |
+
"learning_rate": 7.740549273021002e-06,
|
| 278653 |
+
"loss": 0.27,
|
| 278654 |
+
"step": 140915
|
| 278655 |
+
},
|
| 278656 |
+
{
|
| 278657 |
+
"epoch": 1127.36,
|
| 278658 |
+
"learning_rate": 7.740468497576737e-06,
|
| 278659 |
+
"loss": 0.4203,
|
| 278660 |
+
"step": 140920
|
| 278661 |
+
},
|
| 278662 |
+
{
|
| 278663 |
+
"epoch": 1127.4,
|
| 278664 |
+
"learning_rate": 7.740387722132471e-06,
|
| 278665 |
+
"loss": 1.2038,
|
| 278666 |
+
"step": 140925
|
| 278667 |
+
},
|
| 278668 |
+
{
|
| 278669 |
+
"epoch": 1127.44,
|
| 278670 |
+
"learning_rate": 7.740306946688207e-06,
|
| 278671 |
+
"loss": 0.3352,
|
| 278672 |
+
"step": 140930
|
| 278673 |
+
},
|
| 278674 |
+
{
|
| 278675 |
+
"epoch": 1127.48,
|
| 278676 |
+
"learning_rate": 7.740226171243941e-06,
|
| 278677 |
+
"loss": 0.2844,
|
| 278678 |
+
"step": 140935
|
| 278679 |
+
},
|
| 278680 |
+
{
|
| 278681 |
+
"epoch": 1127.52,
|
| 278682 |
+
"learning_rate": 7.740145395799677e-06,
|
| 278683 |
+
"loss": 0.3658,
|
| 278684 |
+
"step": 140940
|
| 278685 |
+
},
|
| 278686 |
+
{
|
| 278687 |
+
"epoch": 1127.56,
|
| 278688 |
+
"learning_rate": 7.740064620355413e-06,
|
| 278689 |
+
"loss": 0.4165,
|
| 278690 |
+
"step": 140945
|
| 278691 |
+
},
|
| 278692 |
+
{
|
| 278693 |
+
"epoch": 1127.6,
|
| 278694 |
+
"learning_rate": 7.739983844911147e-06,
|
| 278695 |
+
"loss": 1.1142,
|
| 278696 |
+
"step": 140950
|
| 278697 |
+
},
|
| 278698 |
+
{
|
| 278699 |
+
"epoch": 1127.64,
|
| 278700 |
+
"learning_rate": 7.739903069466883e-06,
|
| 278701 |
+
"loss": 0.3433,
|
| 278702 |
+
"step": 140955
|
| 278703 |
+
},
|
| 278704 |
+
{
|
| 278705 |
+
"epoch": 1127.68,
|
| 278706 |
+
"learning_rate": 7.739822294022617e-06,
|
| 278707 |
+
"loss": 0.3123,
|
| 278708 |
+
"step": 140960
|
| 278709 |
+
},
|
| 278710 |
+
{
|
| 278711 |
+
"epoch": 1127.72,
|
| 278712 |
+
"learning_rate": 7.739741518578353e-06,
|
| 278713 |
+
"loss": 0.2778,
|
| 278714 |
+
"step": 140965
|
| 278715 |
+
},
|
| 278716 |
+
{
|
| 278717 |
+
"epoch": 1127.76,
|
| 278718 |
+
"learning_rate": 7.739660743134087e-06,
|
| 278719 |
+
"loss": 0.4403,
|
| 278720 |
+
"step": 140970
|
| 278721 |
+
},
|
| 278722 |
+
{
|
| 278723 |
+
"epoch": 1127.8,
|
| 278724 |
+
"learning_rate": 7.739579967689823e-06,
|
| 278725 |
+
"loss": 1.2477,
|
| 278726 |
+
"step": 140975
|
| 278727 |
+
},
|
| 278728 |
+
{
|
| 278729 |
+
"epoch": 1127.84,
|
| 278730 |
+
"learning_rate": 7.739499192245557e-06,
|
| 278731 |
+
"loss": 0.3184,
|
| 278732 |
+
"step": 140980
|
| 278733 |
+
},
|
| 278734 |
+
{
|
| 278735 |
+
"epoch": 1127.88,
|
| 278736 |
+
"learning_rate": 7.739418416801293e-06,
|
| 278737 |
+
"loss": 0.2304,
|
| 278738 |
+
"step": 140985
|
| 278739 |
+
},
|
| 278740 |
+
{
|
| 278741 |
+
"epoch": 1127.92,
|
| 278742 |
+
"learning_rate": 7.739337641357027e-06,
|
| 278743 |
+
"loss": 0.3165,
|
| 278744 |
+
"step": 140990
|
| 278745 |
+
},
|
| 278746 |
+
{
|
| 278747 |
+
"epoch": 1127.96,
|
| 278748 |
+
"learning_rate": 7.739256865912763e-06,
|
| 278749 |
+
"loss": 0.3891,
|
| 278750 |
+
"step": 140995
|
| 278751 |
+
},
|
| 278752 |
+
{
|
| 278753 |
+
"epoch": 1128.0,
|
| 278754 |
+
"learning_rate": 7.739176090468499e-06,
|
| 278755 |
+
"loss": 1.1212,
|
| 278756 |
+
"step": 141000
|
| 278757 |
+
},
|
| 278758 |
+
{
|
| 278759 |
+
"epoch": 1128.0,
|
| 278760 |
+
"eval_loss": 0.31474193930625916,
|
| 278761 |
+
"eval_runtime": 42.6982,
|
| 278762 |
+
"eval_samples_per_second": 19.579,
|
| 278763 |
+
"eval_steps_per_second": 0.632,
|
| 278764 |
+
"eval_wer": 0.17202457493927703,
|
| 278765 |
+
"step": 141000
|
| 278766 |
+
},
|
| 278767 |
+
{
|
| 278768 |
+
"epoch": 1128.04,
|
| 278769 |
+
"learning_rate": 7.739095315024233e-06,
|
| 278770 |
+
"loss": 0.3107,
|
| 278771 |
+
"step": 141005
|
| 278772 |
+
},
|
| 278773 |
+
{
|
| 278774 |
+
"epoch": 1128.08,
|
| 278775 |
+
"learning_rate": 7.739014539579969e-06,
|
| 278776 |
+
"loss": 0.2824,
|
| 278777 |
+
"step": 141010
|
| 278778 |
+
},
|
| 278779 |
+
{
|
| 278780 |
+
"epoch": 1128.12,
|
| 278781 |
+
"learning_rate": 7.738933764135703e-06,
|
| 278782 |
+
"loss": 0.285,
|
| 278783 |
+
"step": 141015
|
| 278784 |
+
},
|
| 278785 |
+
{
|
| 278786 |
+
"epoch": 1128.16,
|
| 278787 |
+
"learning_rate": 7.738852988691439e-06,
|
| 278788 |
+
"loss": 0.4485,
|
| 278789 |
+
"step": 141020
|
| 278790 |
+
},
|
| 278791 |
+
{
|
| 278792 |
+
"epoch": 1128.2,
|
| 278793 |
+
"learning_rate": 7.738772213247173e-06,
|
| 278794 |
+
"loss": 1.1809,
|
| 278795 |
+
"step": 141025
|
| 278796 |
+
},
|
| 278797 |
+
{
|
| 278798 |
+
"epoch": 1128.24,
|
| 278799 |
+
"learning_rate": 7.738691437802909e-06,
|
| 278800 |
+
"loss": 0.3743,
|
| 278801 |
+
"step": 141030
|
| 278802 |
+
},
|
| 278803 |
+
{
|
| 278804 |
+
"epoch": 1128.28,
|
| 278805 |
+
"learning_rate": 7.738610662358643e-06,
|
| 278806 |
+
"loss": 0.2676,
|
| 278807 |
+
"step": 141035
|
| 278808 |
+
},
|
| 278809 |
+
{
|
| 278810 |
+
"epoch": 1128.32,
|
| 278811 |
+
"learning_rate": 7.738529886914379e-06,
|
| 278812 |
+
"loss": 0.295,
|
| 278813 |
+
"step": 141040
|
| 278814 |
+
},
|
| 278815 |
+
{
|
| 278816 |
+
"epoch": 1128.36,
|
| 278817 |
+
"learning_rate": 7.738449111470113e-06,
|
| 278818 |
+
"loss": 0.403,
|
| 278819 |
+
"step": 141045
|
| 278820 |
+
},
|
| 278821 |
+
{
|
| 278822 |
+
"epoch": 1128.4,
|
| 278823 |
+
"learning_rate": 7.738368336025849e-06,
|
| 278824 |
+
"loss": 1.1188,
|
| 278825 |
+
"step": 141050
|
| 278826 |
+
},
|
| 278827 |
+
{
|
| 278828 |
+
"epoch": 1128.44,
|
| 278829 |
+
"learning_rate": 7.738287560581585e-06,
|
| 278830 |
+
"loss": 0.2713,
|
| 278831 |
+
"step": 141055
|
| 278832 |
+
},
|
| 278833 |
+
{
|
| 278834 |
+
"epoch": 1128.48,
|
| 278835 |
+
"learning_rate": 7.738206785137319e-06,
|
| 278836 |
+
"loss": 0.2981,
|
| 278837 |
+
"step": 141060
|
| 278838 |
+
},
|
| 278839 |
+
{
|
| 278840 |
+
"epoch": 1128.52,
|
| 278841 |
+
"learning_rate": 7.738126009693055e-06,
|
| 278842 |
+
"loss": 0.2873,
|
| 278843 |
+
"step": 141065
|
| 278844 |
+
},
|
| 278845 |
+
{
|
| 278846 |
+
"epoch": 1128.56,
|
| 278847 |
+
"learning_rate": 7.738045234248789e-06,
|
| 278848 |
+
"loss": 0.4451,
|
| 278849 |
+
"step": 141070
|
| 278850 |
+
},
|
| 278851 |
+
{
|
| 278852 |
+
"epoch": 1128.6,
|
| 278853 |
+
"learning_rate": 7.737964458804524e-06,
|
| 278854 |
+
"loss": 0.9997,
|
| 278855 |
+
"step": 141075
|
| 278856 |
+
},
|
| 278857 |
+
{
|
| 278858 |
+
"epoch": 1128.64,
|
| 278859 |
+
"learning_rate": 7.737883683360259e-06,
|
| 278860 |
+
"loss": 0.3084,
|
| 278861 |
+
"step": 141080
|
| 278862 |
+
},
|
| 278863 |
+
{
|
| 278864 |
+
"epoch": 1128.68,
|
| 278865 |
+
"learning_rate": 7.737802907915994e-06,
|
| 278866 |
+
"loss": 0.2833,
|
| 278867 |
+
"step": 141085
|
| 278868 |
+
},
|
| 278869 |
+
{
|
| 278870 |
+
"epoch": 1128.72,
|
| 278871 |
+
"learning_rate": 7.737722132471729e-06,
|
| 278872 |
+
"loss": 0.3011,
|
| 278873 |
+
"step": 141090
|
| 278874 |
+
},
|
| 278875 |
+
{
|
| 278876 |
+
"epoch": 1128.76,
|
| 278877 |
+
"learning_rate": 7.737641357027464e-06,
|
| 278878 |
+
"loss": 0.4076,
|
| 278879 |
+
"step": 141095
|
| 278880 |
+
},
|
| 278881 |
+
{
|
| 278882 |
+
"epoch": 1128.8,
|
| 278883 |
+
"learning_rate": 7.737560581583199e-06,
|
| 278884 |
+
"loss": 1.1358,
|
| 278885 |
+
"step": 141100
|
| 278886 |
+
},
|
| 278887 |
+
{
|
| 278888 |
+
"epoch": 1128.84,
|
| 278889 |
+
"learning_rate": 7.737479806138934e-06,
|
| 278890 |
+
"loss": 0.3086,
|
| 278891 |
+
"step": 141105
|
| 278892 |
+
},
|
| 278893 |
+
{
|
| 278894 |
+
"epoch": 1128.88,
|
| 278895 |
+
"learning_rate": 7.737399030694669e-06,
|
| 278896 |
+
"loss": 0.2608,
|
| 278897 |
+
"step": 141110
|
| 278898 |
+
},
|
| 278899 |
+
{
|
| 278900 |
+
"epoch": 1128.92,
|
| 278901 |
+
"learning_rate": 7.737318255250404e-06,
|
| 278902 |
+
"loss": 0.3053,
|
| 278903 |
+
"step": 141115
|
| 278904 |
+
},
|
| 278905 |
+
{
|
| 278906 |
+
"epoch": 1128.96,
|
| 278907 |
+
"learning_rate": 7.73723747980614e-06,
|
| 278908 |
+
"loss": 0.4529,
|
| 278909 |
+
"step": 141120
|
| 278910 |
+
},
|
| 278911 |
+
{
|
| 278912 |
+
"epoch": 1129.0,
|
| 278913 |
+
"learning_rate": 7.737156704361874e-06,
|
| 278914 |
+
"loss": 0.9868,
|
| 278915 |
+
"step": 141125
|
| 278916 |
+
},
|
| 278917 |
+
{
|
| 278918 |
+
"epoch": 1129.0,
|
| 278919 |
+
"eval_loss": 0.3641243577003479,
|
| 278920 |
+
"eval_runtime": 42.3242,
|
| 278921 |
+
"eval_samples_per_second": 19.752,
|
| 278922 |
+
"eval_steps_per_second": 0.638,
|
| 278923 |
+
"eval_wer": 0.17473699917928823,
|
| 278924 |
+
"step": 141125
|
| 278925 |
+
},
|
| 278926 |
+
{
|
| 278927 |
+
"epoch": 1129.04,
|
| 278928 |
+
"learning_rate": 7.73707592891761e-06,
|
| 278929 |
+
"loss": 0.2973,
|
| 278930 |
+
"step": 141130
|
| 278931 |
+
},
|
| 278932 |
+
{
|
| 278933 |
+
"epoch": 1129.08,
|
| 278934 |
+
"learning_rate": 7.736995153473344e-06,
|
| 278935 |
+
"loss": 0.2701,
|
| 278936 |
+
"step": 141135
|
| 278937 |
+
},
|
| 278938 |
+
{
|
| 278939 |
+
"epoch": 1129.12,
|
| 278940 |
+
"learning_rate": 7.73691437802908e-06,
|
| 278941 |
+
"loss": 0.2608,
|
| 278942 |
+
"step": 141140
|
| 278943 |
+
},
|
| 278944 |
+
{
|
| 278945 |
+
"epoch": 1129.16,
|
| 278946 |
+
"learning_rate": 7.736833602584814e-06,
|
| 278947 |
+
"loss": 0.5024,
|
| 278948 |
+
"step": 141145
|
| 278949 |
+
},
|
| 278950 |
+
{
|
| 278951 |
+
"epoch": 1129.2,
|
| 278952 |
+
"learning_rate": 7.73675282714055e-06,
|
| 278953 |
+
"loss": 1.3191,
|
| 278954 |
+
"step": 141150
|
| 278955 |
+
},
|
| 278956 |
+
{
|
| 278957 |
+
"epoch": 1129.24,
|
| 278958 |
+
"learning_rate": 7.736672051696284e-06,
|
| 278959 |
+
"loss": 0.2705,
|
| 278960 |
+
"step": 141155
|
| 278961 |
+
},
|
| 278962 |
+
{
|
| 278963 |
+
"epoch": 1129.28,
|
| 278964 |
+
"learning_rate": 7.73659127625202e-06,
|
| 278965 |
+
"loss": 0.26,
|
| 278966 |
+
"step": 141160
|
| 278967 |
+
},
|
| 278968 |
+
{
|
| 278969 |
+
"epoch": 1129.32,
|
| 278970 |
+
"learning_rate": 7.736510500807754e-06,
|
| 278971 |
+
"loss": 0.3361,
|
| 278972 |
+
"step": 141165
|
| 278973 |
+
},
|
| 278974 |
+
{
|
| 278975 |
+
"epoch": 1129.36,
|
| 278976 |
+
"learning_rate": 7.73642972536349e-06,
|
| 278977 |
+
"loss": 0.3538,
|
| 278978 |
+
"step": 141170
|
| 278979 |
+
},
|
| 278980 |
+
{
|
| 278981 |
+
"epoch": 1129.4,
|
| 278982 |
+
"learning_rate": 7.736348949919226e-06,
|
| 278983 |
+
"loss": 1.2496,
|
| 278984 |
+
"step": 141175
|
| 278985 |
+
},
|
| 278986 |
+
{
|
| 278987 |
+
"epoch": 1129.44,
|
| 278988 |
+
"learning_rate": 7.73626817447496e-06,
|
| 278989 |
+
"loss": 0.3176,
|
| 278990 |
+
"step": 141180
|
| 278991 |
+
},
|
| 278992 |
+
{
|
| 278993 |
+
"epoch": 1129.48,
|
| 278994 |
+
"learning_rate": 7.736187399030696e-06,
|
| 278995 |
+
"loss": 0.249,
|
| 278996 |
+
"step": 141185
|
| 278997 |
+
},
|
| 278998 |
+
{
|
| 278999 |
+
"epoch": 1129.52,
|
| 279000 |
+
"learning_rate": 7.73610662358643e-06,
|
| 279001 |
+
"loss": 0.3081,
|
| 279002 |
+
"step": 141190
|
| 279003 |
+
},
|
| 279004 |
+
{
|
| 279005 |
+
"epoch": 1129.56,
|
| 279006 |
+
"learning_rate": 7.736025848142166e-06,
|
| 279007 |
+
"loss": 0.4101,
|
| 279008 |
+
"step": 141195
|
| 279009 |
+
},
|
| 279010 |
+
{
|
| 279011 |
+
"epoch": 1129.6,
|
| 279012 |
+
"learning_rate": 7.7359450726979e-06,
|
| 279013 |
+
"loss": 1.1433,
|
| 279014 |
+
"step": 141200
|
| 279015 |
+
},
|
| 279016 |
+
{
|
| 279017 |
+
"epoch": 1129.64,
|
| 279018 |
+
"learning_rate": 7.735864297253636e-06,
|
| 279019 |
+
"loss": 0.3121,
|
| 279020 |
+
"step": 141205
|
| 279021 |
+
},
|
| 279022 |
+
{
|
| 279023 |
+
"epoch": 1129.68,
|
| 279024 |
+
"learning_rate": 7.73578352180937e-06,
|
| 279025 |
+
"loss": 0.2492,
|
| 279026 |
+
"step": 141210
|
| 279027 |
+
},
|
| 279028 |
+
{
|
| 279029 |
+
"epoch": 1129.72,
|
| 279030 |
+
"learning_rate": 7.735702746365106e-06,
|
| 279031 |
+
"loss": 0.2631,
|
| 279032 |
+
"step": 141215
|
| 279033 |
+
},
|
| 279034 |
+
{
|
| 279035 |
+
"epoch": 1129.76,
|
| 279036 |
+
"learning_rate": 7.73562197092084e-06,
|
| 279037 |
+
"loss": 0.4334,
|
| 279038 |
+
"step": 141220
|
| 279039 |
+
},
|
| 279040 |
+
{
|
| 279041 |
+
"epoch": 1129.8,
|
| 279042 |
+
"learning_rate": 7.735541195476576e-06,
|
| 279043 |
+
"loss": 1.1294,
|
| 279044 |
+
"step": 141225
|
| 279045 |
+
},
|
| 279046 |
+
{
|
| 279047 |
+
"epoch": 1129.84,
|
| 279048 |
+
"learning_rate": 7.735460420032312e-06,
|
| 279049 |
+
"loss": 0.3479,
|
| 279050 |
+
"step": 141230
|
| 279051 |
+
},
|
| 279052 |
+
{
|
| 279053 |
+
"epoch": 1129.88,
|
| 279054 |
+
"learning_rate": 7.735379644588046e-06,
|
| 279055 |
+
"loss": 0.2833,
|
| 279056 |
+
"step": 141235
|
| 279057 |
+
},
|
| 279058 |
+
{
|
| 279059 |
+
"epoch": 1129.92,
|
| 279060 |
+
"learning_rate": 7.735298869143782e-06,
|
| 279061 |
+
"loss": 0.3542,
|
| 279062 |
+
"step": 141240
|
| 279063 |
+
},
|
| 279064 |
+
{
|
| 279065 |
+
"epoch": 1129.96,
|
| 279066 |
+
"learning_rate": 7.735218093699516e-06,
|
| 279067 |
+
"loss": 0.4441,
|
| 279068 |
+
"step": 141245
|
| 279069 |
+
},
|
| 279070 |
+
{
|
| 279071 |
+
"epoch": 1130.0,
|
| 279072 |
+
"learning_rate": 7.735137318255252e-06,
|
| 279073 |
+
"loss": 1.3531,
|
| 279074 |
+
"step": 141250
|
| 279075 |
+
},
|
| 279076 |
+
{
|
| 279077 |
+
"epoch": 1130.0,
|
| 279078 |
+
"eval_loss": 0.35339123010635376,
|
| 279079 |
+
"eval_runtime": 42.2845,
|
| 279080 |
+
"eval_samples_per_second": 19.818,
|
| 279081 |
+
"eval_steps_per_second": 0.639,
|
| 279082 |
+
"eval_wer": 0.16927854408897233,
|
| 279083 |
+
"step": 141250
|
| 279084 |
+
},
|
| 279085 |
+
{
|
| 279086 |
+
"epoch": 1139.04,
|
| 279087 |
+
"learning_rate": 7.735056542810986e-06,
|
| 279088 |
+
"loss": 0.3019,
|
| 279089 |
+
"step": 141255
|
| 279090 |
+
},
|
| 279091 |
+
{
|
| 279092 |
+
"epoch": 1139.08,
|
| 279093 |
+
"learning_rate": 7.734975767366722e-06,
|
| 279094 |
+
"loss": 0.2567,
|
| 279095 |
+
"step": 141260
|
| 279096 |
+
},
|
| 279097 |
+
{
|
| 279098 |
+
"epoch": 1139.12,
|
| 279099 |
+
"learning_rate": 7.734894991922456e-06,
|
| 279100 |
+
"loss": 0.3121,
|
| 279101 |
+
"step": 141265
|
| 279102 |
+
},
|
| 279103 |
+
{
|
| 279104 |
+
"epoch": 1139.16,
|
| 279105 |
+
"learning_rate": 7.734814216478192e-06,
|
| 279106 |
+
"loss": 0.3943,
|
| 279107 |
+
"step": 141270
|
| 279108 |
+
},
|
| 279109 |
+
{
|
| 279110 |
+
"epoch": 1139.2,
|
| 279111 |
+
"learning_rate": 7.734733441033926e-06,
|
| 279112 |
+
"loss": 1.1946,
|
| 279113 |
+
"step": 141275
|
| 279114 |
+
},
|
| 279115 |
+
{
|
| 279116 |
+
"epoch": 1139.24,
|
| 279117 |
+
"learning_rate": 7.734652665589662e-06,
|
| 279118 |
+
"loss": 0.3344,
|
| 279119 |
+
"step": 141280
|
| 279120 |
+
},
|
| 279121 |
+
{
|
| 279122 |
+
"epoch": 1139.28,
|
| 279123 |
+
"learning_rate": 7.734571890145396e-06,
|
| 279124 |
+
"loss": 0.2405,
|
| 279125 |
+
"step": 141285
|
| 279126 |
+
},
|
| 279127 |
+
{
|
| 279128 |
+
"epoch": 1139.32,
|
| 279129 |
+
"learning_rate": 7.734491114701131e-06,
|
| 279130 |
+
"loss": 0.3154,
|
| 279131 |
+
"step": 141290
|
| 279132 |
+
},
|
| 279133 |
+
{
|
| 279134 |
+
"epoch": 1139.36,
|
| 279135 |
+
"learning_rate": 7.734410339256867e-06,
|
| 279136 |
+
"loss": 0.5155,
|
| 279137 |
+
"step": 141295
|
| 279138 |
+
},
|
| 279139 |
+
{
|
| 279140 |
+
"epoch": 1139.4,
|
| 279141 |
+
"learning_rate": 7.734329563812601e-06,
|
| 279142 |
+
"loss": 1.2843,
|
| 279143 |
+
"step": 141300
|
| 279144 |
+
},
|
| 279145 |
+
{
|
| 279146 |
+
"epoch": 1139.44,
|
| 279147 |
+
"learning_rate": 7.734248788368337e-06,
|
| 279148 |
+
"loss": 0.301,
|
| 279149 |
+
"step": 141305
|
| 279150 |
+
},
|
| 279151 |
+
{
|
| 279152 |
+
"epoch": 1139.48,
|
| 279153 |
+
"learning_rate": 7.734168012924071e-06,
|
| 279154 |
+
"loss": 0.2791,
|
| 279155 |
+
"step": 141310
|
| 279156 |
+
},
|
| 279157 |
+
{
|
| 279158 |
+
"epoch": 1139.52,
|
| 279159 |
+
"learning_rate": 7.734087237479807e-06,
|
| 279160 |
+
"loss": 0.2982,
|
| 279161 |
+
"step": 141315
|
| 279162 |
+
},
|
| 279163 |
+
{
|
| 279164 |
+
"epoch": 1139.56,
|
| 279165 |
+
"learning_rate": 7.734006462035541e-06,
|
| 279166 |
+
"loss": 0.4278,
|
| 279167 |
+
"step": 141320
|
| 279168 |
+
},
|
| 279169 |
+
{
|
| 279170 |
+
"epoch": 1139.6,
|
| 279171 |
+
"learning_rate": 7.733925686591277e-06,
|
| 279172 |
+
"loss": 1.299,
|
| 279173 |
+
"step": 141325
|
| 279174 |
+
},
|
| 279175 |
+
{
|
| 279176 |
+
"epoch": 1139.64,
|
| 279177 |
+
"learning_rate": 7.733844911147011e-06,
|
| 279178 |
+
"loss": 0.3192,
|
| 279179 |
+
"step": 141330
|
| 279180 |
+
},
|
| 279181 |
+
{
|
| 279182 |
+
"epoch": 1139.68,
|
| 279183 |
+
"learning_rate": 7.733764135702747e-06,
|
| 279184 |
+
"loss": 0.2698,
|
| 279185 |
+
"step": 141335
|
| 279186 |
+
},
|
| 279187 |
+
{
|
| 279188 |
+
"epoch": 1139.72,
|
| 279189 |
+
"learning_rate": 7.733683360258481e-06,
|
| 279190 |
+
"loss": 0.4436,
|
| 279191 |
+
"step": 141340
|
| 279192 |
+
},
|
| 279193 |
+
{
|
| 279194 |
+
"epoch": 1139.76,
|
| 279195 |
+
"learning_rate": 7.733602584814217e-06,
|
| 279196 |
+
"loss": 0.3944,
|
| 279197 |
+
"step": 141345
|
| 279198 |
+
},
|
| 279199 |
+
{
|
| 279200 |
+
"epoch": 1139.8,
|
| 279201 |
+
"learning_rate": 7.733521809369953e-06,
|
| 279202 |
+
"loss": 1.1616,
|
| 279203 |
+
"step": 141350
|
| 279204 |
+
},
|
| 279205 |
+
{
|
| 279206 |
+
"epoch": 1139.84,
|
| 279207 |
+
"learning_rate": 7.733441033925687e-06,
|
| 279208 |
+
"loss": 0.3037,
|
| 279209 |
+
"step": 141355
|
| 279210 |
+
},
|
| 279211 |
+
{
|
| 279212 |
+
"epoch": 1139.88,
|
| 279213 |
+
"learning_rate": 7.733360258481423e-06,
|
| 279214 |
+
"loss": 0.3252,
|
| 279215 |
+
"step": 141360
|
| 279216 |
+
},
|
| 279217 |
+
{
|
| 279218 |
+
"epoch": 1139.92,
|
| 279219 |
+
"learning_rate": 7.733279483037157e-06,
|
| 279220 |
+
"loss": 0.3114,
|
| 279221 |
+
"step": 141365
|
| 279222 |
+
},
|
| 279223 |
+
{
|
| 279224 |
+
"epoch": 1139.96,
|
| 279225 |
+
"learning_rate": 7.733198707592893e-06,
|
| 279226 |
+
"loss": 0.4506,
|
| 279227 |
+
"step": 141370
|
| 279228 |
+
},
|
| 279229 |
+
{
|
| 279230 |
+
"epoch": 1140.0,
|
| 279231 |
+
"eval_loss": 0.31875544786453247,
|
| 279232 |
+
"eval_runtime": 41.1552,
|
| 279233 |
+
"eval_samples_per_second": 20.362,
|
| 279234 |
+
"eval_steps_per_second": 0.656,
|
| 279235 |
+
"eval_wer": 0.18156947444204463,
|
| 279236 |
+
"step": 141374
|
| 279237 |
+
},
|
| 279238 |
+
{
|
| 279239 |
+
"epoch": 1140.01,
|
| 279240 |
+
"learning_rate": 7.733117932148627e-06,
|
| 279241 |
+
"loss": 0.3174,
|
| 279242 |
+
"step": 141375
|
| 279243 |
+
},
|
| 279244 |
+
{
|
| 279245 |
+
"epoch": 1140.05,
|
| 279246 |
+
"learning_rate": 7.733037156704363e-06,
|
| 279247 |
+
"loss": 0.3126,
|
| 279248 |
+
"step": 141380
|
| 279249 |
+
},
|
| 279250 |
+
{
|
| 279251 |
+
"epoch": 1140.09,
|
| 279252 |
+
"learning_rate": 7.732956381260097e-06,
|
| 279253 |
+
"loss": 0.3189,
|
| 279254 |
+
"step": 141385
|
| 279255 |
+
},
|
| 279256 |
+
{
|
| 279257 |
+
"epoch": 1140.13,
|
| 279258 |
+
"learning_rate": 7.732875605815833e-06,
|
| 279259 |
+
"loss": 0.299,
|
| 279260 |
+
"step": 141390
|
| 279261 |
+
},
|
| 279262 |
+
{
|
| 279263 |
+
"epoch": 1140.17,
|
| 279264 |
+
"learning_rate": 7.732794830371567e-06,
|
| 279265 |
+
"loss": 0.4947,
|
| 279266 |
+
"step": 141395
|
| 279267 |
+
},
|
| 279268 |
+
{
|
| 279269 |
+
"epoch": 1140.21,
|
| 279270 |
+
"learning_rate": 7.732714054927303e-06,
|
| 279271 |
+
"loss": 1.0491,
|
| 279272 |
+
"step": 141400
|
| 279273 |
+
},
|
| 279274 |
+
{
|
| 279275 |
+
"epoch": 1140.25,
|
| 279276 |
+
"learning_rate": 7.732633279483039e-06,
|
| 279277 |
+
"loss": 0.2623,
|
| 279278 |
+
"step": 141405
|
| 279279 |
+
},
|
| 279280 |
+
{
|
| 279281 |
+
"epoch": 1140.29,
|
| 279282 |
+
"learning_rate": 7.732552504038773e-06,
|
| 279283 |
+
"loss": 0.2973,
|
| 279284 |
+
"step": 141410
|
| 279285 |
+
},
|
| 279286 |
+
{
|
| 279287 |
+
"epoch": 1140.33,
|
| 279288 |
+
"learning_rate": 7.732471728594509e-06,
|
| 279289 |
+
"loss": 0.2994,
|
| 279290 |
+
"step": 141415
|
| 279291 |
+
},
|
| 279292 |
+
{
|
| 279293 |
+
"epoch": 1140.37,
|
| 279294 |
+
"learning_rate": 7.732390953150243e-06,
|
| 279295 |
+
"loss": 0.4631,
|
| 279296 |
+
"step": 141420
|
| 279297 |
+
},
|
| 279298 |
+
{
|
| 279299 |
+
"epoch": 1140.41,
|
| 279300 |
+
"learning_rate": 7.732310177705979e-06,
|
| 279301 |
+
"loss": 0.8482,
|
| 279302 |
+
"step": 141425
|
| 279303 |
+
},
|
| 279304 |
+
{
|
| 279305 |
+
"epoch": 1140.45,
|
| 279306 |
+
"learning_rate": 7.732229402261713e-06,
|
| 279307 |
+
"loss": 0.2911,
|
| 279308 |
+
"step": 141430
|
| 279309 |
+
},
|
| 279310 |
+
{
|
| 279311 |
+
"epoch": 1140.49,
|
| 279312 |
+
"learning_rate": 7.732148626817449e-06,
|
| 279313 |
+
"loss": 0.3115,
|
| 279314 |
+
"step": 141435
|
| 279315 |
+
},
|
| 279316 |
+
{
|
| 279317 |
+
"epoch": 1140.53,
|
| 279318 |
+
"learning_rate": 7.732067851373183e-06,
|
| 279319 |
+
"loss": 0.2556,
|
| 279320 |
+
"step": 141440
|
| 279321 |
+
},
|
| 279322 |
+
{
|
| 279323 |
+
"epoch": 1140.57,
|
| 279324 |
+
"learning_rate": 7.731987075928919e-06,
|
| 279325 |
+
"loss": 0.5297,
|
| 279326 |
+
"step": 141445
|
| 279327 |
+
},
|
| 279328 |
+
{
|
| 279329 |
+
"epoch": 1140.61,
|
| 279330 |
+
"learning_rate": 7.731906300484653e-06,
|
| 279331 |
+
"loss": 1.0871,
|
| 279332 |
+
"step": 141450
|
| 279333 |
+
},
|
| 279334 |
+
{
|
| 279335 |
+
"epoch": 1140.65,
|
| 279336 |
+
"learning_rate": 7.731825525040389e-06,
|
| 279337 |
+
"loss": 0.2652,
|
| 279338 |
+
"step": 141455
|
| 279339 |
+
},
|
| 279340 |
+
{
|
| 279341 |
+
"epoch": 1140.69,
|
| 279342 |
+
"learning_rate": 7.731744749596123e-06,
|
| 279343 |
+
"loss": 0.2827,
|
| 279344 |
+
"step": 141460
|
| 279345 |
+
},
|
| 279346 |
+
{
|
| 279347 |
+
"epoch": 1140.73,
|
| 279348 |
+
"learning_rate": 7.731663974151859e-06,
|
| 279349 |
+
"loss": 0.355,
|
| 279350 |
+
"step": 141465
|
| 279351 |
+
},
|
| 279352 |
+
{
|
| 279353 |
+
"epoch": 1140.77,
|
| 279354 |
+
"learning_rate": 7.731583198707594e-06,
|
| 279355 |
+
"loss": 0.4689,
|
| 279356 |
+
"step": 141470
|
| 279357 |
+
},
|
| 279358 |
+
{
|
| 279359 |
+
"epoch": 1140.81,
|
| 279360 |
+
"learning_rate": 7.731502423263329e-06,
|
| 279361 |
+
"loss": 1.1235,
|
| 279362 |
+
"step": 141475
|
| 279363 |
+
},
|
| 279364 |
+
{
|
| 279365 |
+
"epoch": 1140.85,
|
| 279366 |
+
"learning_rate": 7.731421647819064e-06,
|
| 279367 |
+
"loss": 0.2817,
|
| 279368 |
+
"step": 141480
|
| 279369 |
+
},
|
| 279370 |
+
{
|
| 279371 |
+
"epoch": 1140.89,
|
| 279372 |
+
"learning_rate": 7.731340872374799e-06,
|
| 279373 |
+
"loss": 0.2921,
|
| 279374 |
+
"step": 141485
|
| 279375 |
+
},
|
| 279376 |
+
{
|
| 279377 |
+
"epoch": 1140.93,
|
| 279378 |
+
"learning_rate": 7.731260096930534e-06,
|
| 279379 |
+
"loss": 0.3773,
|
| 279380 |
+
"step": 141490
|
| 279381 |
+
},
|
| 279382 |
+
{
|
| 279383 |
+
"epoch": 1140.97,
|
| 279384 |
+
"learning_rate": 7.731179321486269e-06,
|
| 279385 |
+
"loss": 0.5879,
|
| 279386 |
+
"step": 141495
|
| 279387 |
+
},
|
| 279388 |
+
{
|
| 279389 |
+
"epoch": 1141.0,
|
| 279390 |
+
"eval_loss": 0.35892051458358765,
|
| 279391 |
+
"eval_runtime": 42.4638,
|
| 279392 |
+
"eval_samples_per_second": 19.711,
|
| 279393 |
+
"eval_steps_per_second": 0.636,
|
| 279394 |
+
"eval_wer": 0.17664390945023356,
|
| 279395 |
+
"step": 141498
|
| 279396 |
}
|
| 279397 |
],
|
| 279398 |
+
"max_steps": 620000,
|
| 279399 |
"num_train_epochs": 5000,
|
| 279400 |
+
"total_flos": 3.9816733700847796e+20,
|
| 279401 |
"trial_name": null,
|
| 279402 |
"trial_params": null
|
| 279403 |
}
|
model-bin/finetune/base/{checkpoint-140875 → checkpoint-141498}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630210217.851614/events.out.tfevents.1630210217.cc93b136ebf5.1086.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:118b38257e05a24d2a91d0748cdc3d3589b462d43979c1591f92d417c6d29556
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630210657.964172/events.out.tfevents.1630210657.cc93b136ebf5.1086.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:09b068c8b675783c60467edaea3f1eeb52b4792cca9f28071c9f7393b54cc0ba
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630211223.3432353/events.out.tfevents.1630211223.cc93b136ebf5.1086.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:623fa061ebf8ef8da2656ace8984f1e08887c4ba34c1ad24792f43514939bc7f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630211666.4765499/events.out.tfevents.1630211666.cc93b136ebf5.1086.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40b508ee62df8c72803cb509d55d84b75e50ab061fe9a658cd9867bd15d228d2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630212104.3499668/events.out.tfevents.1630212104.cc93b136ebf5.1086.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d828a780babb72fb4e09959ea367573e2985e6bc37f17781219ac6b882fb00b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630210217.cc93b136ebf5.1086.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b066d7213be0c6ae99af6fd84a0b2016f307fdfaa6e2990510ac1724724b15a9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630210657.cc93b136ebf5.1086.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2df5faf6fdc93ac0f65debb1eac2526472cf83b335c73075df1c950c1071144f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630211223.cc93b136ebf5.1086.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f919acaf70ecdf3b475afd4d539579f9249de8571b0ecfc999dd86d804cd95ca
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630211666.cc93b136ebf5.1086.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e143304602907ed0ad03ff98a9f8eacf1b5221886cfc39bcc749f8ac9ee90de3
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630212104.cc93b136ebf5.1086.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:87117962305670ce0d44c9156f44a9ce22b508378c667d048479db4d410ebe31
|
| 3 |
+
size 8622
|