"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/trainer_state.json +953 -5
- model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630223877.27181/events.out.tfevents.1630223877.cc93b136ebf5.1086.89 +3 -0
- model-bin/finetune/base/log/1630224307.777163/events.out.tfevents.1630224307.cc93b136ebf5.1086.91 +3 -0
- model-bin/finetune/base/log/1630224741.833629/events.out.tfevents.1630224741.cc93b136ebf5.1086.93 +3 -0
- model-bin/finetune/base/log/1630225180.2778423/events.out.tfevents.1630225180.cc93b136ebf5.1086.95 +3 -0
- model-bin/finetune/base/log/1630225606.0373123/events.out.tfevents.1630225606.cc93b136ebf5.1086.97 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630223877.cc93b136ebf5.1086.88 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630224307.cc93b136ebf5.1086.90 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630224741.cc93b136ebf5.1086.92 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630225180.cc93b136ebf5.1086.94 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630225606.cc93b136ebf5.1086.96 +3 -0
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/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:f244af29f882f04c01f58f5ca77480cdcb5dcbb73f0ef209191eccc6999e0a52
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/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 sha256:5cd7d37f4506f232619d1098d435d1703fd863ae76efe0903c1754396c8044ff
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/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 sha256:062f3ea956c2658f1d60fd363016c1ce7fcd6aa3edc3e881707a9c05a0819134
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/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 sha256:83acee9eaf83f2490bff1778f538e95bbe74b4277d68159d818d81dd1a84bdb7
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/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 sha256:2db674888116994c2e24ef5e8fbb6b9cf9bf43cfcdefb2459d7c42aa98d1dd80
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -283191,11 +283191,959 @@
|
|
| 283191 |
"eval_steps_per_second": 0.645,
|
| 283192 |
"eval_wer": 0.1758169934640523,
|
| 283193 |
"step": 144483
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 283194 |
}
|
| 283195 |
],
|
| 283196 |
"max_steps": 620000,
|
| 283197 |
"num_train_epochs": 5000,
|
| 283198 |
-
"total_flos": 4.
|
| 283199 |
"trial_name": null,
|
| 283200 |
"trial_params": null
|
| 283201 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.16790141568253503,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-144607",
|
| 4 |
+
"epoch": 1170.995983935743,
|
| 5 |
+
"global_step": 145228,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 283191 |
"eval_steps_per_second": 0.645,
|
| 283192 |
"eval_wer": 0.1758169934640523,
|
| 283193 |
"step": 144483
|
| 283194 |
+
},
|
| 283195 |
+
{
|
| 283196 |
+
"epoch": 1165.02,
|
| 283197 |
+
"learning_rate": 7.682907915993538e-06,
|
| 283198 |
+
"loss": 0.2721,
|
| 283199 |
+
"step": 144485
|
| 283200 |
+
},
|
| 283201 |
+
{
|
| 283202 |
+
"epoch": 1165.06,
|
| 283203 |
+
"learning_rate": 7.682827140549274e-06,
|
| 283204 |
+
"loss": 0.2887,
|
| 283205 |
+
"step": 144490
|
| 283206 |
+
},
|
| 283207 |
+
{
|
| 283208 |
+
"epoch": 1165.1,
|
| 283209 |
+
"learning_rate": 7.682746365105008e-06,
|
| 283210 |
+
"loss": 0.339,
|
| 283211 |
+
"step": 144495
|
| 283212 |
+
},
|
| 283213 |
+
{
|
| 283214 |
+
"epoch": 1165.14,
|
| 283215 |
+
"learning_rate": 7.682665589660744e-06,
|
| 283216 |
+
"loss": 0.2786,
|
| 283217 |
+
"step": 144500
|
| 283218 |
+
},
|
| 283219 |
+
{
|
| 283220 |
+
"epoch": 1165.18,
|
| 283221 |
+
"learning_rate": 7.682584814216478e-06,
|
| 283222 |
+
"loss": 0.5607,
|
| 283223 |
+
"step": 144505
|
| 283224 |
+
},
|
| 283225 |
+
{
|
| 283226 |
+
"epoch": 1165.22,
|
| 283227 |
+
"learning_rate": 7.682504038772214e-06,
|
| 283228 |
+
"loss": 0.9914,
|
| 283229 |
+
"step": 144510
|
| 283230 |
+
},
|
| 283231 |
+
{
|
| 283232 |
+
"epoch": 1165.26,
|
| 283233 |
+
"learning_rate": 7.68242326332795e-06,
|
| 283234 |
+
"loss": 0.2515,
|
| 283235 |
+
"step": 144515
|
| 283236 |
+
},
|
| 283237 |
+
{
|
| 283238 |
+
"epoch": 1165.3,
|
| 283239 |
+
"learning_rate": 7.682342487883684e-06,
|
| 283240 |
+
"loss": 0.3679,
|
| 283241 |
+
"step": 144520
|
| 283242 |
+
},
|
| 283243 |
+
{
|
| 283244 |
+
"epoch": 1165.34,
|
| 283245 |
+
"learning_rate": 7.68226171243942e-06,
|
| 283246 |
+
"loss": 0.3706,
|
| 283247 |
+
"step": 144525
|
| 283248 |
+
},
|
| 283249 |
+
{
|
| 283250 |
+
"epoch": 1165.38,
|
| 283251 |
+
"learning_rate": 7.682180936995154e-06,
|
| 283252 |
+
"loss": 0.5414,
|
| 283253 |
+
"step": 144530
|
| 283254 |
+
},
|
| 283255 |
+
{
|
| 283256 |
+
"epoch": 1165.42,
|
| 283257 |
+
"learning_rate": 7.68210016155089e-06,
|
| 283258 |
+
"loss": 1.0052,
|
| 283259 |
+
"step": 144535
|
| 283260 |
+
},
|
| 283261 |
+
{
|
| 283262 |
+
"epoch": 1165.46,
|
| 283263 |
+
"learning_rate": 7.682019386106624e-06,
|
| 283264 |
+
"loss": 0.2638,
|
| 283265 |
+
"step": 144540
|
| 283266 |
+
},
|
| 283267 |
+
{
|
| 283268 |
+
"epoch": 1165.5,
|
| 283269 |
+
"learning_rate": 7.68193861066236e-06,
|
| 283270 |
+
"loss": 0.2948,
|
| 283271 |
+
"step": 144545
|
| 283272 |
+
},
|
| 283273 |
+
{
|
| 283274 |
+
"epoch": 1165.54,
|
| 283275 |
+
"learning_rate": 7.681857835218094e-06,
|
| 283276 |
+
"loss": 0.4818,
|
| 283277 |
+
"step": 144550
|
| 283278 |
+
},
|
| 283279 |
+
{
|
| 283280 |
+
"epoch": 1165.58,
|
| 283281 |
+
"learning_rate": 7.68177705977383e-06,
|
| 283282 |
+
"loss": 0.6016,
|
| 283283 |
+
"step": 144555
|
| 283284 |
+
},
|
| 283285 |
+
{
|
| 283286 |
+
"epoch": 1165.62,
|
| 283287 |
+
"learning_rate": 7.681696284329564e-06,
|
| 283288 |
+
"loss": 0.8445,
|
| 283289 |
+
"step": 144560
|
| 283290 |
+
},
|
| 283291 |
+
{
|
| 283292 |
+
"epoch": 1165.66,
|
| 283293 |
+
"learning_rate": 7.6816155088853e-06,
|
| 283294 |
+
"loss": 0.246,
|
| 283295 |
+
"step": 144565
|
| 283296 |
+
},
|
| 283297 |
+
{
|
| 283298 |
+
"epoch": 1165.7,
|
| 283299 |
+
"learning_rate": 7.681534733441035e-06,
|
| 283300 |
+
"loss": 0.2843,
|
| 283301 |
+
"step": 144570
|
| 283302 |
+
},
|
| 283303 |
+
{
|
| 283304 |
+
"epoch": 1165.74,
|
| 283305 |
+
"learning_rate": 7.68145395799677e-06,
|
| 283306 |
+
"loss": 0.3326,
|
| 283307 |
+
"step": 144575
|
| 283308 |
+
},
|
| 283309 |
+
{
|
| 283310 |
+
"epoch": 1165.78,
|
| 283311 |
+
"learning_rate": 7.681373182552505e-06,
|
| 283312 |
+
"loss": 0.5409,
|
| 283313 |
+
"step": 144580
|
| 283314 |
+
},
|
| 283315 |
+
{
|
| 283316 |
+
"epoch": 1165.82,
|
| 283317 |
+
"learning_rate": 7.68129240710824e-06,
|
| 283318 |
+
"loss": 0.9051,
|
| 283319 |
+
"step": 144585
|
| 283320 |
+
},
|
| 283321 |
+
{
|
| 283322 |
+
"epoch": 1165.86,
|
| 283323 |
+
"learning_rate": 7.681211631663975e-06,
|
| 283324 |
+
"loss": 0.2603,
|
| 283325 |
+
"step": 144590
|
| 283326 |
+
},
|
| 283327 |
+
{
|
| 283328 |
+
"epoch": 1165.9,
|
| 283329 |
+
"learning_rate": 7.68113085621971e-06,
|
| 283330 |
+
"loss": 0.2964,
|
| 283331 |
+
"step": 144595
|
| 283332 |
+
},
|
| 283333 |
+
{
|
| 283334 |
+
"epoch": 1165.94,
|
| 283335 |
+
"learning_rate": 7.681050080775445e-06,
|
| 283336 |
+
"loss": 0.368,
|
| 283337 |
+
"step": 144600
|
| 283338 |
+
},
|
| 283339 |
+
{
|
| 283340 |
+
"epoch": 1165.98,
|
| 283341 |
+
"learning_rate": 7.68096930533118e-06,
|
| 283342 |
+
"loss": 0.6023,
|
| 283343 |
+
"step": 144605
|
| 283344 |
+
},
|
| 283345 |
+
{
|
| 283346 |
+
"epoch": 1166.0,
|
| 283347 |
+
"eval_loss": 0.3452383875846863,
|
| 283348 |
+
"eval_runtime": 39.7586,
|
| 283349 |
+
"eval_samples_per_second": 21.052,
|
| 283350 |
+
"eval_steps_per_second": 0.679,
|
| 283351 |
+
"eval_wer": 0.16790141568253503,
|
| 283352 |
+
"step": 144607
|
| 283353 |
+
},
|
| 283354 |
+
{
|
| 283355 |
+
"epoch": 1166.02,
|
| 283356 |
+
"learning_rate": 7.680888529886915e-06,
|
| 283357 |
+
"loss": 0.3216,
|
| 283358 |
+
"step": 144610
|
| 283359 |
+
},
|
| 283360 |
+
{
|
| 283361 |
+
"epoch": 1166.06,
|
| 283362 |
+
"learning_rate": 7.68080775444265e-06,
|
| 283363 |
+
"loss": 0.2238,
|
| 283364 |
+
"step": 144615
|
| 283365 |
+
},
|
| 283366 |
+
{
|
| 283367 |
+
"epoch": 1166.1,
|
| 283368 |
+
"learning_rate": 7.680726978998385e-06,
|
| 283369 |
+
"loss": 0.318,
|
| 283370 |
+
"step": 144620
|
| 283371 |
+
},
|
| 283372 |
+
{
|
| 283373 |
+
"epoch": 1166.14,
|
| 283374 |
+
"learning_rate": 7.680646203554121e-06,
|
| 283375 |
+
"loss": 0.357,
|
| 283376 |
+
"step": 144625
|
| 283377 |
+
},
|
| 283378 |
+
{
|
| 283379 |
+
"epoch": 1166.18,
|
| 283380 |
+
"learning_rate": 7.680565428109855e-06,
|
| 283381 |
+
"loss": 0.8028,
|
| 283382 |
+
"step": 144630
|
| 283383 |
+
},
|
| 283384 |
+
{
|
| 283385 |
+
"epoch": 1166.22,
|
| 283386 |
+
"learning_rate": 7.680484652665591e-06,
|
| 283387 |
+
"loss": 0.8776,
|
| 283388 |
+
"step": 144635
|
| 283389 |
+
},
|
| 283390 |
+
{
|
| 283391 |
+
"epoch": 1166.27,
|
| 283392 |
+
"learning_rate": 7.680403877221325e-06,
|
| 283393 |
+
"loss": 0.2889,
|
| 283394 |
+
"step": 144640
|
| 283395 |
+
},
|
| 283396 |
+
{
|
| 283397 |
+
"epoch": 1166.31,
|
| 283398 |
+
"learning_rate": 7.680323101777061e-06,
|
| 283399 |
+
"loss": 0.3034,
|
| 283400 |
+
"step": 144645
|
| 283401 |
+
},
|
| 283402 |
+
{
|
| 283403 |
+
"epoch": 1166.35,
|
| 283404 |
+
"learning_rate": 7.680242326332795e-06,
|
| 283405 |
+
"loss": 0.3402,
|
| 283406 |
+
"step": 144650
|
| 283407 |
+
},
|
| 283408 |
+
{
|
| 283409 |
+
"epoch": 1166.39,
|
| 283410 |
+
"learning_rate": 7.680161550888531e-06,
|
| 283411 |
+
"loss": 0.6687,
|
| 283412 |
+
"step": 144655
|
| 283413 |
+
},
|
| 283414 |
+
{
|
| 283415 |
+
"epoch": 1166.43,
|
| 283416 |
+
"learning_rate": 7.680080775444265e-06,
|
| 283417 |
+
"loss": 0.8415,
|
| 283418 |
+
"step": 144660
|
| 283419 |
+
},
|
| 283420 |
+
{
|
| 283421 |
+
"epoch": 1166.47,
|
| 283422 |
+
"learning_rate": 7.680000000000001e-06,
|
| 283423 |
+
"loss": 0.283,
|
| 283424 |
+
"step": 144665
|
| 283425 |
+
},
|
| 283426 |
+
{
|
| 283427 |
+
"epoch": 1166.51,
|
| 283428 |
+
"learning_rate": 7.679919224555735e-06,
|
| 283429 |
+
"loss": 0.2845,
|
| 283430 |
+
"step": 144670
|
| 283431 |
+
},
|
| 283432 |
+
{
|
| 283433 |
+
"epoch": 1166.55,
|
| 283434 |
+
"learning_rate": 7.679838449111471e-06,
|
| 283435 |
+
"loss": 0.3641,
|
| 283436 |
+
"step": 144675
|
| 283437 |
+
},
|
| 283438 |
+
{
|
| 283439 |
+
"epoch": 1166.59,
|
| 283440 |
+
"learning_rate": 7.679757673667205e-06,
|
| 283441 |
+
"loss": 0.6544,
|
| 283442 |
+
"step": 144680
|
| 283443 |
+
},
|
| 283444 |
+
{
|
| 283445 |
+
"epoch": 1166.63,
|
| 283446 |
+
"learning_rate": 7.679676898222941e-06,
|
| 283447 |
+
"loss": 0.8825,
|
| 283448 |
+
"step": 144685
|
| 283449 |
+
},
|
| 283450 |
+
{
|
| 283451 |
+
"epoch": 1166.67,
|
| 283452 |
+
"learning_rate": 7.679596122778677e-06,
|
| 283453 |
+
"loss": 0.308,
|
| 283454 |
+
"step": 144690
|
| 283455 |
+
},
|
| 283456 |
+
{
|
| 283457 |
+
"epoch": 1166.71,
|
| 283458 |
+
"learning_rate": 7.679515347334411e-06,
|
| 283459 |
+
"loss": 0.2938,
|
| 283460 |
+
"step": 144695
|
| 283461 |
+
},
|
| 283462 |
+
{
|
| 283463 |
+
"epoch": 1166.75,
|
| 283464 |
+
"learning_rate": 7.679434571890147e-06,
|
| 283465 |
+
"loss": 0.3421,
|
| 283466 |
+
"step": 144700
|
| 283467 |
+
},
|
| 283468 |
+
{
|
| 283469 |
+
"epoch": 1166.79,
|
| 283470 |
+
"learning_rate": 7.679353796445881e-06,
|
| 283471 |
+
"loss": 0.7238,
|
| 283472 |
+
"step": 144705
|
| 283473 |
+
},
|
| 283474 |
+
{
|
| 283475 |
+
"epoch": 1166.83,
|
| 283476 |
+
"learning_rate": 7.679273021001617e-06,
|
| 283477 |
+
"loss": 0.8722,
|
| 283478 |
+
"step": 144710
|
| 283479 |
+
},
|
| 283480 |
+
{
|
| 283481 |
+
"epoch": 1166.87,
|
| 283482 |
+
"learning_rate": 7.679192245557351e-06,
|
| 283483 |
+
"loss": 0.2458,
|
| 283484 |
+
"step": 144715
|
| 283485 |
+
},
|
| 283486 |
+
{
|
| 283487 |
+
"epoch": 1166.91,
|
| 283488 |
+
"learning_rate": 7.679111470113087e-06,
|
| 283489 |
+
"loss": 0.2701,
|
| 283490 |
+
"step": 144720
|
| 283491 |
+
},
|
| 283492 |
+
{
|
| 283493 |
+
"epoch": 1166.95,
|
| 283494 |
+
"learning_rate": 7.679030694668821e-06,
|
| 283495 |
+
"loss": 0.3769,
|
| 283496 |
+
"step": 144725
|
| 283497 |
+
},
|
| 283498 |
+
{
|
| 283499 |
+
"epoch": 1166.99,
|
| 283500 |
+
"learning_rate": 7.678949919224557e-06,
|
| 283501 |
+
"loss": 0.8216,
|
| 283502 |
+
"step": 144730
|
| 283503 |
+
},
|
| 283504 |
+
{
|
| 283505 |
+
"epoch": 1167.0,
|
| 283506 |
+
"eval_loss": 0.3796067237854004,
|
| 283507 |
+
"eval_runtime": 41.6728,
|
| 283508 |
+
"eval_samples_per_second": 20.085,
|
| 283509 |
+
"eval_steps_per_second": 0.648,
|
| 283510 |
+
"eval_wer": 0.18190386427898209,
|
| 283511 |
+
"step": 144731
|
| 283512 |
+
},
|
| 283513 |
+
{
|
| 283514 |
+
"epoch": 1167.03,
|
| 283515 |
+
"learning_rate": 7.67886914378029e-06,
|
| 283516 |
+
"loss": 0.2946,
|
| 283517 |
+
"step": 144735
|
| 283518 |
+
},
|
| 283519 |
+
{
|
| 283520 |
+
"epoch": 1167.07,
|
| 283521 |
+
"learning_rate": 7.678788368336027e-06,
|
| 283522 |
+
"loss": 0.2505,
|
| 283523 |
+
"step": 144740
|
| 283524 |
+
},
|
| 283525 |
+
{
|
| 283526 |
+
"epoch": 1167.11,
|
| 283527 |
+
"learning_rate": 7.678707592891763e-06,
|
| 283528 |
+
"loss": 0.2865,
|
| 283529 |
+
"step": 144745
|
| 283530 |
+
},
|
| 283531 |
+
{
|
| 283532 |
+
"epoch": 1167.15,
|
| 283533 |
+
"learning_rate": 7.678626817447497e-06,
|
| 283534 |
+
"loss": 0.3545,
|
| 283535 |
+
"step": 144750
|
| 283536 |
+
},
|
| 283537 |
+
{
|
| 283538 |
+
"epoch": 1167.19,
|
| 283539 |
+
"learning_rate": 7.678546042003232e-06,
|
| 283540 |
+
"loss": 0.9134,
|
| 283541 |
+
"step": 144755
|
| 283542 |
+
},
|
| 283543 |
+
{
|
| 283544 |
+
"epoch": 1167.23,
|
| 283545 |
+
"learning_rate": 7.678465266558967e-06,
|
| 283546 |
+
"loss": 0.655,
|
| 283547 |
+
"step": 144760
|
| 283548 |
+
},
|
| 283549 |
+
{
|
| 283550 |
+
"epoch": 1167.27,
|
| 283551 |
+
"learning_rate": 7.678384491114702e-06,
|
| 283552 |
+
"loss": 0.2705,
|
| 283553 |
+
"step": 144765
|
| 283554 |
+
},
|
| 283555 |
+
{
|
| 283556 |
+
"epoch": 1167.31,
|
| 283557 |
+
"learning_rate": 7.678303715670437e-06,
|
| 283558 |
+
"loss": 0.2562,
|
| 283559 |
+
"step": 144770
|
| 283560 |
+
},
|
| 283561 |
+
{
|
| 283562 |
+
"epoch": 1167.35,
|
| 283563 |
+
"learning_rate": 7.678222940226172e-06,
|
| 283564 |
+
"loss": 0.4007,
|
| 283565 |
+
"step": 144775
|
| 283566 |
+
},
|
| 283567 |
+
{
|
| 283568 |
+
"epoch": 1167.39,
|
| 283569 |
+
"learning_rate": 7.678142164781907e-06,
|
| 283570 |
+
"loss": 0.8377,
|
| 283571 |
+
"step": 144780
|
| 283572 |
+
},
|
| 283573 |
+
{
|
| 283574 |
+
"epoch": 1167.43,
|
| 283575 |
+
"learning_rate": 7.678061389337642e-06,
|
| 283576 |
+
"loss": 0.7035,
|
| 283577 |
+
"step": 144785
|
| 283578 |
+
},
|
| 283579 |
+
{
|
| 283580 |
+
"epoch": 1167.47,
|
| 283581 |
+
"learning_rate": 7.677980613893377e-06,
|
| 283582 |
+
"loss": 0.2795,
|
| 283583 |
+
"step": 144790
|
| 283584 |
+
},
|
| 283585 |
+
{
|
| 283586 |
+
"epoch": 1167.51,
|
| 283587 |
+
"learning_rate": 7.677899838449112e-06,
|
| 283588 |
+
"loss": 0.2829,
|
| 283589 |
+
"step": 144795
|
| 283590 |
+
},
|
| 283591 |
+
{
|
| 283592 |
+
"epoch": 1167.55,
|
| 283593 |
+
"learning_rate": 7.677819063004848e-06,
|
| 283594 |
+
"loss": 0.3805,
|
| 283595 |
+
"step": 144800
|
| 283596 |
+
},
|
| 283597 |
+
{
|
| 283598 |
+
"epoch": 1167.59,
|
| 283599 |
+
"learning_rate": 7.677738287560582e-06,
|
| 283600 |
+
"loss": 0.7594,
|
| 283601 |
+
"step": 144805
|
| 283602 |
+
},
|
| 283603 |
+
{
|
| 283604 |
+
"epoch": 1167.63,
|
| 283605 |
+
"learning_rate": 7.677657512116318e-06,
|
| 283606 |
+
"loss": 0.6312,
|
| 283607 |
+
"step": 144810
|
| 283608 |
+
},
|
| 283609 |
+
{
|
| 283610 |
+
"epoch": 1167.67,
|
| 283611 |
+
"learning_rate": 7.677576736672052e-06,
|
| 283612 |
+
"loss": 0.257,
|
| 283613 |
+
"step": 144815
|
| 283614 |
+
},
|
| 283615 |
+
{
|
| 283616 |
+
"epoch": 1167.71,
|
| 283617 |
+
"learning_rate": 7.677495961227788e-06,
|
| 283618 |
+
"loss": 0.2697,
|
| 283619 |
+
"step": 144820
|
| 283620 |
+
},
|
| 283621 |
+
{
|
| 283622 |
+
"epoch": 1167.76,
|
| 283623 |
+
"learning_rate": 7.677415185783522e-06,
|
| 283624 |
+
"loss": 0.3065,
|
| 283625 |
+
"step": 144825
|
| 283626 |
+
},
|
| 283627 |
+
{
|
| 283628 |
+
"epoch": 1167.8,
|
| 283629 |
+
"learning_rate": 7.677334410339258e-06,
|
| 283630 |
+
"loss": 0.8168,
|
| 283631 |
+
"step": 144830
|
| 283632 |
+
},
|
| 283633 |
+
{
|
| 283634 |
+
"epoch": 1167.84,
|
| 283635 |
+
"learning_rate": 7.677253634894992e-06,
|
| 283636 |
+
"loss": 0.6347,
|
| 283637 |
+
"step": 144835
|
| 283638 |
+
},
|
| 283639 |
+
{
|
| 283640 |
+
"epoch": 1167.88,
|
| 283641 |
+
"learning_rate": 7.677172859450728e-06,
|
| 283642 |
+
"loss": 0.2511,
|
| 283643 |
+
"step": 144840
|
| 283644 |
+
},
|
| 283645 |
+
{
|
| 283646 |
+
"epoch": 1167.92,
|
| 283647 |
+
"learning_rate": 7.677092084006462e-06,
|
| 283648 |
+
"loss": 0.2972,
|
| 283649 |
+
"step": 144845
|
| 283650 |
+
},
|
| 283651 |
+
{
|
| 283652 |
+
"epoch": 1167.96,
|
| 283653 |
+
"learning_rate": 7.677011308562198e-06,
|
| 283654 |
+
"loss": 0.4141,
|
| 283655 |
+
"step": 144850
|
| 283656 |
+
},
|
| 283657 |
+
{
|
| 283658 |
+
"epoch": 1168.0,
|
| 283659 |
+
"learning_rate": 7.676930533117932e-06,
|
| 283660 |
+
"loss": 1.1673,
|
| 283661 |
+
"step": 144855
|
| 283662 |
+
},
|
| 283663 |
+
{
|
| 283664 |
+
"epoch": 1168.0,
|
| 283665 |
+
"eval_loss": 0.34798669815063477,
|
| 283666 |
+
"eval_runtime": 41.3687,
|
| 283667 |
+
"eval_samples_per_second": 20.233,
|
| 283668 |
+
"eval_steps_per_second": 0.653,
|
| 283669 |
+
"eval_wer": 0.17150972252567184,
|
| 283670 |
+
"step": 144855
|
| 283671 |
+
},
|
| 283672 |
+
{
|
| 283673 |
+
"epoch": 1158.04,
|
| 283674 |
+
"learning_rate": 7.676849757673668e-06,
|
| 283675 |
+
"loss": 0.3284,
|
| 283676 |
+
"step": 144860
|
| 283677 |
+
},
|
| 283678 |
+
{
|
| 283679 |
+
"epoch": 1158.08,
|
| 283680 |
+
"learning_rate": 7.676768982229404e-06,
|
| 283681 |
+
"loss": 0.2587,
|
| 283682 |
+
"step": 144865
|
| 283683 |
+
},
|
| 283684 |
+
{
|
| 283685 |
+
"epoch": 1158.12,
|
| 283686 |
+
"learning_rate": 7.676688206785138e-06,
|
| 283687 |
+
"loss": 0.3396,
|
| 283688 |
+
"step": 144870
|
| 283689 |
+
},
|
| 283690 |
+
{
|
| 283691 |
+
"epoch": 1158.16,
|
| 283692 |
+
"learning_rate": 7.676607431340874e-06,
|
| 283693 |
+
"loss": 0.4414,
|
| 283694 |
+
"step": 144875
|
| 283695 |
+
},
|
| 283696 |
+
{
|
| 283697 |
+
"epoch": 1158.2,
|
| 283698 |
+
"learning_rate": 7.676526655896608e-06,
|
| 283699 |
+
"loss": 1.0351,
|
| 283700 |
+
"step": 144880
|
| 283701 |
+
},
|
| 283702 |
+
{
|
| 283703 |
+
"epoch": 1158.24,
|
| 283704 |
+
"learning_rate": 7.676445880452344e-06,
|
| 283705 |
+
"loss": 0.2941,
|
| 283706 |
+
"step": 144885
|
| 283707 |
+
},
|
| 283708 |
+
{
|
| 283709 |
+
"epoch": 1158.28,
|
| 283710 |
+
"learning_rate": 7.676365105008078e-06,
|
| 283711 |
+
"loss": 0.2294,
|
| 283712 |
+
"step": 144890
|
| 283713 |
+
},
|
| 283714 |
+
{
|
| 283715 |
+
"epoch": 1158.32,
|
| 283716 |
+
"learning_rate": 7.676284329563814e-06,
|
| 283717 |
+
"loss": 0.326,
|
| 283718 |
+
"step": 144895
|
| 283719 |
+
},
|
| 283720 |
+
{
|
| 283721 |
+
"epoch": 1158.36,
|
| 283722 |
+
"learning_rate": 7.676203554119548e-06,
|
| 283723 |
+
"loss": 0.4232,
|
| 283724 |
+
"step": 144900
|
| 283725 |
+
},
|
| 283726 |
+
{
|
| 283727 |
+
"epoch": 1158.4,
|
| 283728 |
+
"learning_rate": 7.676122778675284e-06,
|
| 283729 |
+
"loss": 1.1964,
|
| 283730 |
+
"step": 144905
|
| 283731 |
+
},
|
| 283732 |
+
{
|
| 283733 |
+
"epoch": 1158.44,
|
| 283734 |
+
"learning_rate": 7.676042003231018e-06,
|
| 283735 |
+
"loss": 0.2844,
|
| 283736 |
+
"step": 144910
|
| 283737 |
+
},
|
| 283738 |
+
{
|
| 283739 |
+
"epoch": 1158.48,
|
| 283740 |
+
"learning_rate": 7.675961227786754e-06,
|
| 283741 |
+
"loss": 0.2528,
|
| 283742 |
+
"step": 144915
|
| 283743 |
+
},
|
| 283744 |
+
{
|
| 283745 |
+
"epoch": 1158.52,
|
| 283746 |
+
"learning_rate": 7.67588045234249e-06,
|
| 283747 |
+
"loss": 0.313,
|
| 283748 |
+
"step": 144920
|
| 283749 |
+
},
|
| 283750 |
+
{
|
| 283751 |
+
"epoch": 1158.56,
|
| 283752 |
+
"learning_rate": 7.675799676898224e-06,
|
| 283753 |
+
"loss": 0.4891,
|
| 283754 |
+
"step": 144925
|
| 283755 |
+
},
|
| 283756 |
+
{
|
| 283757 |
+
"epoch": 1158.6,
|
| 283758 |
+
"learning_rate": 7.67571890145396e-06,
|
| 283759 |
+
"loss": 1.3052,
|
| 283760 |
+
"step": 144930
|
| 283761 |
+
},
|
| 283762 |
+
{
|
| 283763 |
+
"epoch": 1158.64,
|
| 283764 |
+
"learning_rate": 7.675638126009694e-06,
|
| 283765 |
+
"loss": 0.3103,
|
| 283766 |
+
"step": 144935
|
| 283767 |
+
},
|
| 283768 |
+
{
|
| 283769 |
+
"epoch": 1158.68,
|
| 283770 |
+
"learning_rate": 7.67555735056543e-06,
|
| 283771 |
+
"loss": 0.3308,
|
| 283772 |
+
"step": 144940
|
| 283773 |
+
},
|
| 283774 |
+
{
|
| 283775 |
+
"epoch": 1158.72,
|
| 283776 |
+
"learning_rate": 7.675476575121164e-06,
|
| 283777 |
+
"loss": 0.3203,
|
| 283778 |
+
"step": 144945
|
| 283779 |
+
},
|
| 283780 |
+
{
|
| 283781 |
+
"epoch": 1158.76,
|
| 283782 |
+
"learning_rate": 7.6753957996769e-06,
|
| 283783 |
+
"loss": 0.3745,
|
| 283784 |
+
"step": 144950
|
| 283785 |
+
},
|
| 283786 |
+
{
|
| 283787 |
+
"epoch": 1158.8,
|
| 283788 |
+
"learning_rate": 7.675315024232634e-06,
|
| 283789 |
+
"loss": 1.1487,
|
| 283790 |
+
"step": 144955
|
| 283791 |
+
},
|
| 283792 |
+
{
|
| 283793 |
+
"epoch": 1158.84,
|
| 283794 |
+
"learning_rate": 7.67523424878837e-06,
|
| 283795 |
+
"loss": 0.4522,
|
| 283796 |
+
"step": 144960
|
| 283797 |
+
},
|
| 283798 |
+
{
|
| 283799 |
+
"epoch": 1158.88,
|
| 283800 |
+
"learning_rate": 7.675153473344104e-06,
|
| 283801 |
+
"loss": 0.3011,
|
| 283802 |
+
"step": 144965
|
| 283803 |
+
},
|
| 283804 |
+
{
|
| 283805 |
+
"epoch": 1158.92,
|
| 283806 |
+
"learning_rate": 7.67507269789984e-06,
|
| 283807 |
+
"loss": 0.2663,
|
| 283808 |
+
"step": 144970
|
| 283809 |
+
},
|
| 283810 |
+
{
|
| 283811 |
+
"epoch": 1158.96,
|
| 283812 |
+
"learning_rate": 7.674991922455575e-06,
|
| 283813 |
+
"loss": 0.4301,
|
| 283814 |
+
"step": 144975
|
| 283815 |
+
},
|
| 283816 |
+
{
|
| 283817 |
+
"epoch": 1159.0,
|
| 283818 |
+
"learning_rate": 7.67491114701131e-06,
|
| 283819 |
+
"loss": 1.3414,
|
| 283820 |
+
"step": 144980
|
| 283821 |
+
},
|
| 283822 |
+
{
|
| 283823 |
+
"epoch": 1159.0,
|
| 283824 |
+
"eval_loss": 0.3729027807712555,
|
| 283825 |
+
"eval_runtime": 40.7509,
|
| 283826 |
+
"eval_samples_per_second": 20.539,
|
| 283827 |
+
"eval_steps_per_second": 0.663,
|
| 283828 |
+
"eval_wer": 0.17167350527549824,
|
| 283829 |
+
"step": 144980
|
| 283830 |
+
},
|
| 283831 |
+
{
|
| 283832 |
+
"epoch": 1169.04,
|
| 283833 |
+
"learning_rate": 7.674830371567045e-06,
|
| 283834 |
+
"loss": 0.3148,
|
| 283835 |
+
"step": 144985
|
| 283836 |
+
},
|
| 283837 |
+
{
|
| 283838 |
+
"epoch": 1169.08,
|
| 283839 |
+
"learning_rate": 7.67474959612278e-06,
|
| 283840 |
+
"loss": 0.2913,
|
| 283841 |
+
"step": 144990
|
| 283842 |
+
},
|
| 283843 |
+
{
|
| 283844 |
+
"epoch": 1169.12,
|
| 283845 |
+
"learning_rate": 7.674668820678515e-06,
|
| 283846 |
+
"loss": 0.2694,
|
| 283847 |
+
"step": 144995
|
| 283848 |
+
},
|
| 283849 |
+
{
|
| 283850 |
+
"epoch": 1169.16,
|
| 283851 |
+
"learning_rate": 7.67458804523425e-06,
|
| 283852 |
+
"loss": 0.478,
|
| 283853 |
+
"step": 145000
|
| 283854 |
+
},
|
| 283855 |
+
{
|
| 283856 |
+
"epoch": 1169.2,
|
| 283857 |
+
"learning_rate": 7.674507269789985e-06,
|
| 283858 |
+
"loss": 1.1648,
|
| 283859 |
+
"step": 145005
|
| 283860 |
+
},
|
| 283861 |
+
{
|
| 283862 |
+
"epoch": 1169.24,
|
| 283863 |
+
"learning_rate": 7.67442649434572e-06,
|
| 283864 |
+
"loss": 0.3068,
|
| 283865 |
+
"step": 145010
|
| 283866 |
+
},
|
| 283867 |
+
{
|
| 283868 |
+
"epoch": 1169.28,
|
| 283869 |
+
"learning_rate": 7.674345718901455e-06,
|
| 283870 |
+
"loss": 0.2923,
|
| 283871 |
+
"step": 145015
|
| 283872 |
+
},
|
| 283873 |
+
{
|
| 283874 |
+
"epoch": 1169.32,
|
| 283875 |
+
"learning_rate": 7.67426494345719e-06,
|
| 283876 |
+
"loss": 0.277,
|
| 283877 |
+
"step": 145020
|
| 283878 |
+
},
|
| 283879 |
+
{
|
| 283880 |
+
"epoch": 1169.36,
|
| 283881 |
+
"learning_rate": 7.674184168012925e-06,
|
| 283882 |
+
"loss": 0.5259,
|
| 283883 |
+
"step": 145025
|
| 283884 |
+
},
|
| 283885 |
+
{
|
| 283886 |
+
"epoch": 1169.4,
|
| 283887 |
+
"learning_rate": 7.67410339256866e-06,
|
| 283888 |
+
"loss": 1.3239,
|
| 283889 |
+
"step": 145030
|
| 283890 |
+
},
|
| 283891 |
+
{
|
| 283892 |
+
"epoch": 1169.44,
|
| 283893 |
+
"learning_rate": 7.674022617124395e-06,
|
| 283894 |
+
"loss": 0.2994,
|
| 283895 |
+
"step": 145035
|
| 283896 |
+
},
|
| 283897 |
+
{
|
| 283898 |
+
"epoch": 1169.48,
|
| 283899 |
+
"learning_rate": 7.673941841680131e-06,
|
| 283900 |
+
"loss": 0.2812,
|
| 283901 |
+
"step": 145040
|
| 283902 |
+
},
|
| 283903 |
+
{
|
| 283904 |
+
"epoch": 1169.52,
|
| 283905 |
+
"learning_rate": 7.673861066235865e-06,
|
| 283906 |
+
"loss": 0.2935,
|
| 283907 |
+
"step": 145045
|
| 283908 |
+
},
|
| 283909 |
+
{
|
| 283910 |
+
"epoch": 1169.56,
|
| 283911 |
+
"learning_rate": 7.673780290791601e-06,
|
| 283912 |
+
"loss": 0.4962,
|
| 283913 |
+
"step": 145050
|
| 283914 |
+
},
|
| 283915 |
+
{
|
| 283916 |
+
"epoch": 1169.6,
|
| 283917 |
+
"learning_rate": 7.673699515347335e-06,
|
| 283918 |
+
"loss": 1.2089,
|
| 283919 |
+
"step": 145055
|
| 283920 |
+
},
|
| 283921 |
+
{
|
| 283922 |
+
"epoch": 1169.64,
|
| 283923 |
+
"learning_rate": 7.673618739903071e-06,
|
| 283924 |
+
"loss": 0.2642,
|
| 283925 |
+
"step": 145060
|
| 283926 |
+
},
|
| 283927 |
+
{
|
| 283928 |
+
"epoch": 1169.68,
|
| 283929 |
+
"learning_rate": 7.673537964458805e-06,
|
| 283930 |
+
"loss": 0.2744,
|
| 283931 |
+
"step": 145065
|
| 283932 |
+
},
|
| 283933 |
+
{
|
| 283934 |
+
"epoch": 1169.72,
|
| 283935 |
+
"learning_rate": 7.673457189014541e-06,
|
| 283936 |
+
"loss": 0.278,
|
| 283937 |
+
"step": 145070
|
| 283938 |
+
},
|
| 283939 |
+
{
|
| 283940 |
+
"epoch": 1169.76,
|
| 283941 |
+
"learning_rate": 7.673376413570275e-06,
|
| 283942 |
+
"loss": 0.4382,
|
| 283943 |
+
"step": 145075
|
| 283944 |
+
},
|
| 283945 |
+
{
|
| 283946 |
+
"epoch": 1169.8,
|
| 283947 |
+
"learning_rate": 7.673295638126011e-06,
|
| 283948 |
+
"loss": 1.0989,
|
| 283949 |
+
"step": 145080
|
| 283950 |
+
},
|
| 283951 |
+
{
|
| 283952 |
+
"epoch": 1169.84,
|
| 283953 |
+
"learning_rate": 7.673214862681745e-06,
|
| 283954 |
+
"loss": 0.3134,
|
| 283955 |
+
"step": 145085
|
| 283956 |
+
},
|
| 283957 |
+
{
|
| 283958 |
+
"epoch": 1169.88,
|
| 283959 |
+
"learning_rate": 7.673134087237481e-06,
|
| 283960 |
+
"loss": 0.3014,
|
| 283961 |
+
"step": 145090
|
| 283962 |
+
},
|
| 283963 |
+
{
|
| 283964 |
+
"epoch": 1169.92,
|
| 283965 |
+
"learning_rate": 7.673053311793217e-06,
|
| 283966 |
+
"loss": 0.2862,
|
| 283967 |
+
"step": 145095
|
| 283968 |
+
},
|
| 283969 |
+
{
|
| 283970 |
+
"epoch": 1169.96,
|
| 283971 |
+
"learning_rate": 7.672972536348951e-06,
|
| 283972 |
+
"loss": 0.5695,
|
| 283973 |
+
"step": 145100
|
| 283974 |
+
},
|
| 283975 |
+
{
|
| 283976 |
+
"epoch": 1170.0,
|
| 283977 |
+
"eval_loss": 0.39969587326049805,
|
| 283978 |
+
"eval_runtime": 40.575,
|
| 283979 |
+
"eval_samples_per_second": 20.628,
|
| 283980 |
+
"eval_steps_per_second": 0.665,
|
| 283981 |
+
"eval_wer": 0.1783857926653191,
|
| 283982 |
+
"step": 145104
|
| 283983 |
+
},
|
| 283984 |
+
{
|
| 283985 |
+
"epoch": 1170.01,
|
| 283986 |
+
"learning_rate": 7.672891760904687e-06,
|
| 283987 |
+
"loss": 0.3477,
|
| 283988 |
+
"step": 145105
|
| 283989 |
+
},
|
| 283990 |
+
{
|
| 283991 |
+
"epoch": 1170.05,
|
| 283992 |
+
"learning_rate": 7.67281098546042e-06,
|
| 283993 |
+
"loss": 0.3213,
|
| 283994 |
+
"step": 145110
|
| 283995 |
+
},
|
| 283996 |
+
{
|
| 283997 |
+
"epoch": 1170.09,
|
| 283998 |
+
"learning_rate": 7.672730210016157e-06,
|
| 283999 |
+
"loss": 0.275,
|
| 284000 |
+
"step": 145115
|
| 284001 |
+
},
|
| 284002 |
+
{
|
| 284003 |
+
"epoch": 1170.13,
|
| 284004 |
+
"learning_rate": 7.67264943457189e-06,
|
| 284005 |
+
"loss": 0.3189,
|
| 284006 |
+
"step": 145120
|
| 284007 |
+
},
|
| 284008 |
+
{
|
| 284009 |
+
"epoch": 1170.17,
|
| 284010 |
+
"learning_rate": 7.672568659127627e-06,
|
| 284011 |
+
"loss": 0.5509,
|
| 284012 |
+
"step": 145125
|
| 284013 |
+
},
|
| 284014 |
+
{
|
| 284015 |
+
"epoch": 1170.21,
|
| 284016 |
+
"learning_rate": 7.67248788368336e-06,
|
| 284017 |
+
"loss": 1.0719,
|
| 284018 |
+
"step": 145130
|
| 284019 |
+
},
|
| 284020 |
+
{
|
| 284021 |
+
"epoch": 1170.25,
|
| 284022 |
+
"learning_rate": 7.672407108239097e-06,
|
| 284023 |
+
"loss": 0.2919,
|
| 284024 |
+
"step": 145135
|
| 284025 |
+
},
|
| 284026 |
+
{
|
| 284027 |
+
"epoch": 1170.29,
|
| 284028 |
+
"learning_rate": 7.67232633279483e-06,
|
| 284029 |
+
"loss": 0.2532,
|
| 284030 |
+
"step": 145140
|
| 284031 |
+
},
|
| 284032 |
+
{
|
| 284033 |
+
"epoch": 1170.33,
|
| 284034 |
+
"learning_rate": 7.672245557350567e-06,
|
| 284035 |
+
"loss": 0.3189,
|
| 284036 |
+
"step": 145145
|
| 284037 |
+
},
|
| 284038 |
+
{
|
| 284039 |
+
"epoch": 1170.37,
|
| 284040 |
+
"learning_rate": 7.672164781906302e-06,
|
| 284041 |
+
"loss": 0.4942,
|
| 284042 |
+
"step": 145150
|
| 284043 |
+
},
|
| 284044 |
+
{
|
| 284045 |
+
"epoch": 1170.41,
|
| 284046 |
+
"learning_rate": 7.672084006462037e-06,
|
| 284047 |
+
"loss": 1.1852,
|
| 284048 |
+
"step": 145155
|
| 284049 |
+
},
|
| 284050 |
+
{
|
| 284051 |
+
"epoch": 1170.45,
|
| 284052 |
+
"learning_rate": 7.672003231017772e-06,
|
| 284053 |
+
"loss": 0.3175,
|
| 284054 |
+
"step": 145160
|
| 284055 |
+
},
|
| 284056 |
+
{
|
| 284057 |
+
"epoch": 1170.49,
|
| 284058 |
+
"learning_rate": 7.671922455573507e-06,
|
| 284059 |
+
"loss": 0.3137,
|
| 284060 |
+
"step": 145165
|
| 284061 |
+
},
|
| 284062 |
+
{
|
| 284063 |
+
"epoch": 1170.53,
|
| 284064 |
+
"learning_rate": 7.671841680129242e-06,
|
| 284065 |
+
"loss": 0.2959,
|
| 284066 |
+
"step": 145170
|
| 284067 |
+
},
|
| 284068 |
+
{
|
| 284069 |
+
"epoch": 1170.57,
|
| 284070 |
+
"learning_rate": 7.671760904684977e-06,
|
| 284071 |
+
"loss": 0.4471,
|
| 284072 |
+
"step": 145175
|
| 284073 |
+
},
|
| 284074 |
+
{
|
| 284075 |
+
"epoch": 1170.61,
|
| 284076 |
+
"learning_rate": 7.671680129240712e-06,
|
| 284077 |
+
"loss": 1.0092,
|
| 284078 |
+
"step": 145180
|
| 284079 |
+
},
|
| 284080 |
+
{
|
| 284081 |
+
"epoch": 1170.65,
|
| 284082 |
+
"learning_rate": 7.671599353796446e-06,
|
| 284083 |
+
"loss": 0.3617,
|
| 284084 |
+
"step": 145185
|
| 284085 |
+
},
|
| 284086 |
+
{
|
| 284087 |
+
"epoch": 1170.69,
|
| 284088 |
+
"learning_rate": 7.671518578352182e-06,
|
| 284089 |
+
"loss": 0.3141,
|
| 284090 |
+
"step": 145190
|
| 284091 |
+
},
|
| 284092 |
+
{
|
| 284093 |
+
"epoch": 1170.73,
|
| 284094 |
+
"learning_rate": 7.671437802907916e-06,
|
| 284095 |
+
"loss": 0.2982,
|
| 284096 |
+
"step": 145195
|
| 284097 |
+
},
|
| 284098 |
+
{
|
| 284099 |
+
"epoch": 1170.77,
|
| 284100 |
+
"learning_rate": 7.671357027463652e-06,
|
| 284101 |
+
"loss": 0.4728,
|
| 284102 |
+
"step": 145200
|
| 284103 |
+
},
|
| 284104 |
+
{
|
| 284105 |
+
"epoch": 1170.81,
|
| 284106 |
+
"learning_rate": 7.671276252019386e-06,
|
| 284107 |
+
"loss": 1.086,
|
| 284108 |
+
"step": 145205
|
| 284109 |
+
},
|
| 284110 |
+
{
|
| 284111 |
+
"epoch": 1170.85,
|
| 284112 |
+
"learning_rate": 7.671195476575122e-06,
|
| 284113 |
+
"loss": 0.3326,
|
| 284114 |
+
"step": 145210
|
| 284115 |
+
},
|
| 284116 |
+
{
|
| 284117 |
+
"epoch": 1170.89,
|
| 284118 |
+
"learning_rate": 7.671114701130858e-06,
|
| 284119 |
+
"loss": 0.2725,
|
| 284120 |
+
"step": 145215
|
| 284121 |
+
},
|
| 284122 |
+
{
|
| 284123 |
+
"epoch": 1170.93,
|
| 284124 |
+
"learning_rate": 7.671033925686592e-06,
|
| 284125 |
+
"loss": 0.3219,
|
| 284126 |
+
"step": 145220
|
| 284127 |
+
},
|
| 284128 |
+
{
|
| 284129 |
+
"epoch": 1170.97,
|
| 284130 |
+
"learning_rate": 7.670953150242328e-06,
|
| 284131 |
+
"loss": 0.5993,
|
| 284132 |
+
"step": 145225
|
| 284133 |
+
},
|
| 284134 |
+
{
|
| 284135 |
+
"epoch": 1171.0,
|
| 284136 |
+
"eval_loss": 0.3494146466255188,
|
| 284137 |
+
"eval_runtime": 40.4075,
|
| 284138 |
+
"eval_samples_per_second": 20.714,
|
| 284139 |
+
"eval_steps_per_second": 0.668,
|
| 284140 |
+
"eval_wer": 0.17565191913272782,
|
| 284141 |
+
"step": 145228
|
| 284142 |
}
|
| 284143 |
],
|
| 284144 |
"max_steps": 620000,
|
| 284145 |
"num_train_epochs": 5000,
|
| 284146 |
+
"total_flos": 4.0867175962962303e+20,
|
| 284147 |
"trial_name": null,
|
| 284148 |
"trial_params": null
|
| 284149 |
}
|
model-bin/finetune/base/{checkpoint-144483 → checkpoint-145228}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630223877.27181/events.out.tfevents.1630223877.cc93b136ebf5.1086.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:541a4403eff3fe5f96205cb5a76127e31eb9fa9b7a413c1843896e9f31a4544b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630224307.777163/events.out.tfevents.1630224307.cc93b136ebf5.1086.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e00d1be20e613894766ead4a3fc8dc19982433c5c3a7bea445d138e7a5fbd0dd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630224741.833629/events.out.tfevents.1630224741.cc93b136ebf5.1086.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b3b258fd88108704b0743959decd53428659573f1d9db8c28d0048e8b978026
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630225180.2778423/events.out.tfevents.1630225180.cc93b136ebf5.1086.95
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:20d0f0e9819294b7d986c2c05f1336440882cd6232ab586b367c5935408ede28
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630225606.0373123/events.out.tfevents.1630225606.cc93b136ebf5.1086.97
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7242827e78eded1c4c5247d2bf2e15c428b2c488d6443016c26e40b13aeb9b3e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630223877.cc93b136ebf5.1086.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed03fcdd0817189279a78610c7be52bcb116e163722189d8594201c6e788e95d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630224307.cc93b136ebf5.1086.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae84d3859a5df226803db7b8d1fb0ceaf72afa015f211d04c36e8eccd39adc41
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630224741.cc93b136ebf5.1086.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af0e9c293650400c9f8281ef6d70a52bf5fbf3aa7f6abc719288141dce8223df
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630225180.cc93b136ebf5.1086.94
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a95c35378f294017d1c950c25ebd29e223cbf9f7fec59102b0616d06586ca30f
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630225606.cc93b136ebf5.1086.96
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ce068135a96c05ef8feceb8b6e91cccc6fb29605b6f857c55266fa1930df460
|
| 3 |
+
size 8622
|