"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629892093.865816/events.out.tfevents.1629892093.7e498afd5545.7645.15 +3 -0
- model-bin/finetune/base/log/1629892567.0899503/events.out.tfevents.1629892567.7e498afd5545.7645.17 +3 -0
- model-bin/finetune/base/log/1629893045.2545307/events.out.tfevents.1629893045.7e498afd5545.7645.19 +3 -0
- model-bin/finetune/base/log/1629893510.1288207/events.out.tfevents.1629893510.7e498afd5545.7645.21 +3 -0
- model-bin/finetune/base/log/1629893973.3909373/events.out.tfevents.1629893973.7e498afd5545.7645.23 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629892093.7e498afd5545.7645.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629892566.7e498afd5545.7645.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629893045.7e498afd5545.7645.18 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629893510.7e498afd5545.7645.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629893973.7e498afd5545.7645.22 +3 -0
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/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 sha256:2f7011fd213a00609d063c7de52646f933c02885b6729b9df100cc88653c0770
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/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:e61021c5f6956a83f28794793bf05ffaa3847dfe43fd62b5cb1b5e70078ceae4
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/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:b43bab0cd2e069d88f2cb00182ba42d4d4428e663533bd0454a533cf361690e2
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/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 sha256:55268cedf33cbd9726c9ca2afafd5dc42f5d2a582af18496d5e36eaf0ef58f59
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/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 sha256:2f21781a404675bc1af4e71bb00027419ad339e8a09ed59aa2536f2b39085f87
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -197532,11 +197532,800 @@
|
|
| 197532 |
"eval_steps_per_second": 0.624,
|
| 197533 |
"eval_wer": 0.18589790672052883,
|
| 197534 |
"step": 77155
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197535 |
}
|
| 197536 |
],
|
| 197537 |
"max_steps": 620000,
|
| 197538 |
"num_train_epochs": 5000,
|
| 197539 |
-
"total_flos": 2.
|
| 197540 |
"trial_name": null,
|
| 197541 |
"trial_params": null
|
| 197542 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 626.995983935743,
|
| 5 |
+
"global_step": 77779,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 197532 |
"eval_steps_per_second": 0.624,
|
| 197533 |
"eval_wer": 0.18589790672052883,
|
| 197534 |
"step": 77155
|
| 197535 |
+
},
|
| 197536 |
+
{
|
| 197537 |
+
"epoch": 617.04,
|
| 197538 |
+
"learning_rate": 8.770096930533118e-06,
|
| 197539 |
+
"loss": 0.4089,
|
| 197540 |
+
"step": 77160
|
| 197541 |
+
},
|
| 197542 |
+
{
|
| 197543 |
+
"epoch": 617.08,
|
| 197544 |
+
"learning_rate": 8.770016155088854e-06,
|
| 197545 |
+
"loss": 0.2633,
|
| 197546 |
+
"step": 77165
|
| 197547 |
+
},
|
| 197548 |
+
{
|
| 197549 |
+
"epoch": 617.12,
|
| 197550 |
+
"learning_rate": 8.769935379644588e-06,
|
| 197551 |
+
"loss": 0.362,
|
| 197552 |
+
"step": 77170
|
| 197553 |
+
},
|
| 197554 |
+
{
|
| 197555 |
+
"epoch": 617.16,
|
| 197556 |
+
"learning_rate": 8.769854604200324e-06,
|
| 197557 |
+
"loss": 0.4616,
|
| 197558 |
+
"step": 77175
|
| 197559 |
+
},
|
| 197560 |
+
{
|
| 197561 |
+
"epoch": 617.2,
|
| 197562 |
+
"learning_rate": 8.769773828756058e-06,
|
| 197563 |
+
"loss": 1.1666,
|
| 197564 |
+
"step": 77180
|
| 197565 |
+
},
|
| 197566 |
+
{
|
| 197567 |
+
"epoch": 617.24,
|
| 197568 |
+
"learning_rate": 8.769693053311794e-06,
|
| 197569 |
+
"loss": 0.3367,
|
| 197570 |
+
"step": 77185
|
| 197571 |
+
},
|
| 197572 |
+
{
|
| 197573 |
+
"epoch": 617.28,
|
| 197574 |
+
"learning_rate": 8.769612277867528e-06,
|
| 197575 |
+
"loss": 0.328,
|
| 197576 |
+
"step": 77190
|
| 197577 |
+
},
|
| 197578 |
+
{
|
| 197579 |
+
"epoch": 617.32,
|
| 197580 |
+
"learning_rate": 8.769531502423264e-06,
|
| 197581 |
+
"loss": 0.3628,
|
| 197582 |
+
"step": 77195
|
| 197583 |
+
},
|
| 197584 |
+
{
|
| 197585 |
+
"epoch": 617.36,
|
| 197586 |
+
"learning_rate": 8.769450726978998e-06,
|
| 197587 |
+
"loss": 0.514,
|
| 197588 |
+
"step": 77200
|
| 197589 |
+
},
|
| 197590 |
+
{
|
| 197591 |
+
"epoch": 617.4,
|
| 197592 |
+
"learning_rate": 8.769369951534734e-06,
|
| 197593 |
+
"loss": 1.3437,
|
| 197594 |
+
"step": 77205
|
| 197595 |
+
},
|
| 197596 |
+
{
|
| 197597 |
+
"epoch": 617.44,
|
| 197598 |
+
"learning_rate": 8.76928917609047e-06,
|
| 197599 |
+
"loss": 0.3451,
|
| 197600 |
+
"step": 77210
|
| 197601 |
+
},
|
| 197602 |
+
{
|
| 197603 |
+
"epoch": 617.48,
|
| 197604 |
+
"learning_rate": 8.769208400646204e-06,
|
| 197605 |
+
"loss": 0.2851,
|
| 197606 |
+
"step": 77215
|
| 197607 |
+
},
|
| 197608 |
+
{
|
| 197609 |
+
"epoch": 617.52,
|
| 197610 |
+
"learning_rate": 8.76912762520194e-06,
|
| 197611 |
+
"loss": 0.3446,
|
| 197612 |
+
"step": 77220
|
| 197613 |
+
},
|
| 197614 |
+
{
|
| 197615 |
+
"epoch": 617.56,
|
| 197616 |
+
"learning_rate": 8.769046849757674e-06,
|
| 197617 |
+
"loss": 0.4927,
|
| 197618 |
+
"step": 77225
|
| 197619 |
+
},
|
| 197620 |
+
{
|
| 197621 |
+
"epoch": 617.6,
|
| 197622 |
+
"learning_rate": 8.76896607431341e-06,
|
| 197623 |
+
"loss": 1.215,
|
| 197624 |
+
"step": 77230
|
| 197625 |
+
},
|
| 197626 |
+
{
|
| 197627 |
+
"epoch": 617.64,
|
| 197628 |
+
"learning_rate": 8.768885298869144e-06,
|
| 197629 |
+
"loss": 0.2961,
|
| 197630 |
+
"step": 77235
|
| 197631 |
+
},
|
| 197632 |
+
{
|
| 197633 |
+
"epoch": 617.68,
|
| 197634 |
+
"learning_rate": 8.76880452342488e-06,
|
| 197635 |
+
"loss": 0.2855,
|
| 197636 |
+
"step": 77240
|
| 197637 |
+
},
|
| 197638 |
+
{
|
| 197639 |
+
"epoch": 617.72,
|
| 197640 |
+
"learning_rate": 8.768723747980614e-06,
|
| 197641 |
+
"loss": 0.3748,
|
| 197642 |
+
"step": 77245
|
| 197643 |
+
},
|
| 197644 |
+
{
|
| 197645 |
+
"epoch": 617.76,
|
| 197646 |
+
"learning_rate": 8.76864297253635e-06,
|
| 197647 |
+
"loss": 0.5142,
|
| 197648 |
+
"step": 77250
|
| 197649 |
+
},
|
| 197650 |
+
{
|
| 197651 |
+
"epoch": 617.8,
|
| 197652 |
+
"learning_rate": 8.768562197092084e-06,
|
| 197653 |
+
"loss": 1.1865,
|
| 197654 |
+
"step": 77255
|
| 197655 |
+
},
|
| 197656 |
+
{
|
| 197657 |
+
"epoch": 617.84,
|
| 197658 |
+
"learning_rate": 8.76848142164782e-06,
|
| 197659 |
+
"loss": 0.3083,
|
| 197660 |
+
"step": 77260
|
| 197661 |
+
},
|
| 197662 |
+
{
|
| 197663 |
+
"epoch": 617.88,
|
| 197664 |
+
"learning_rate": 8.768400646203555e-06,
|
| 197665 |
+
"loss": 0.2774,
|
| 197666 |
+
"step": 77265
|
| 197667 |
+
},
|
| 197668 |
+
{
|
| 197669 |
+
"epoch": 617.92,
|
| 197670 |
+
"learning_rate": 8.76831987075929e-06,
|
| 197671 |
+
"loss": 0.3284,
|
| 197672 |
+
"step": 77270
|
| 197673 |
+
},
|
| 197674 |
+
{
|
| 197675 |
+
"epoch": 617.96,
|
| 197676 |
+
"learning_rate": 8.768239095315025e-06,
|
| 197677 |
+
"loss": 0.5053,
|
| 197678 |
+
"step": 77275
|
| 197679 |
+
},
|
| 197680 |
+
{
|
| 197681 |
+
"epoch": 618.0,
|
| 197682 |
+
"learning_rate": 8.76815831987076e-06,
|
| 197683 |
+
"loss": 1.2728,
|
| 197684 |
+
"step": 77280
|
| 197685 |
+
},
|
| 197686 |
+
{
|
| 197687 |
+
"epoch": 618.0,
|
| 197688 |
+
"eval_loss": 0.37362736463546753,
|
| 197689 |
+
"eval_runtime": 41.9714,
|
| 197690 |
+
"eval_samples_per_second": 19.918,
|
| 197691 |
+
"eval_steps_per_second": 0.643,
|
| 197692 |
+
"eval_wer": 0.1899724677582959,
|
| 197693 |
+
"step": 77280
|
| 197694 |
+
},
|
| 197695 |
+
{
|
| 197696 |
+
"epoch": 618.04,
|
| 197697 |
+
"learning_rate": 8.768077544426495e-06,
|
| 197698 |
+
"loss": 0.4184,
|
| 197699 |
+
"step": 77285
|
| 197700 |
+
},
|
| 197701 |
+
{
|
| 197702 |
+
"epoch": 618.08,
|
| 197703 |
+
"learning_rate": 8.76799676898223e-06,
|
| 197704 |
+
"loss": 0.3041,
|
| 197705 |
+
"step": 77290
|
| 197706 |
+
},
|
| 197707 |
+
{
|
| 197708 |
+
"epoch": 618.12,
|
| 197709 |
+
"learning_rate": 8.767915993537965e-06,
|
| 197710 |
+
"loss": 0.371,
|
| 197711 |
+
"step": 77295
|
| 197712 |
+
},
|
| 197713 |
+
{
|
| 197714 |
+
"epoch": 618.16,
|
| 197715 |
+
"learning_rate": 8.7678352180937e-06,
|
| 197716 |
+
"loss": 0.516,
|
| 197717 |
+
"step": 77300
|
| 197718 |
+
},
|
| 197719 |
+
{
|
| 197720 |
+
"epoch": 618.2,
|
| 197721 |
+
"learning_rate": 8.767754442649435e-06,
|
| 197722 |
+
"loss": 1.1994,
|
| 197723 |
+
"step": 77305
|
| 197724 |
+
},
|
| 197725 |
+
{
|
| 197726 |
+
"epoch": 618.24,
|
| 197727 |
+
"learning_rate": 8.76767366720517e-06,
|
| 197728 |
+
"loss": 0.3946,
|
| 197729 |
+
"step": 77310
|
| 197730 |
+
},
|
| 197731 |
+
{
|
| 197732 |
+
"epoch": 618.28,
|
| 197733 |
+
"learning_rate": 8.767592891760905e-06,
|
| 197734 |
+
"loss": 0.32,
|
| 197735 |
+
"step": 77315
|
| 197736 |
+
},
|
| 197737 |
+
{
|
| 197738 |
+
"epoch": 618.32,
|
| 197739 |
+
"learning_rate": 8.767512116316641e-06,
|
| 197740 |
+
"loss": 0.318,
|
| 197741 |
+
"step": 77320
|
| 197742 |
+
},
|
| 197743 |
+
{
|
| 197744 |
+
"epoch": 618.36,
|
| 197745 |
+
"learning_rate": 8.767431340872375e-06,
|
| 197746 |
+
"loss": 0.4967,
|
| 197747 |
+
"step": 77325
|
| 197748 |
+
},
|
| 197749 |
+
{
|
| 197750 |
+
"epoch": 618.4,
|
| 197751 |
+
"learning_rate": 8.767350565428111e-06,
|
| 197752 |
+
"loss": 1.3815,
|
| 197753 |
+
"step": 77330
|
| 197754 |
+
},
|
| 197755 |
+
{
|
| 197756 |
+
"epoch": 618.44,
|
| 197757 |
+
"learning_rate": 8.767269789983845e-06,
|
| 197758 |
+
"loss": 0.3609,
|
| 197759 |
+
"step": 77335
|
| 197760 |
+
},
|
| 197761 |
+
{
|
| 197762 |
+
"epoch": 618.48,
|
| 197763 |
+
"learning_rate": 8.767189014539581e-06,
|
| 197764 |
+
"loss": 0.3027,
|
| 197765 |
+
"step": 77340
|
| 197766 |
+
},
|
| 197767 |
+
{
|
| 197768 |
+
"epoch": 618.52,
|
| 197769 |
+
"learning_rate": 8.767108239095315e-06,
|
| 197770 |
+
"loss": 0.3554,
|
| 197771 |
+
"step": 77345
|
| 197772 |
+
},
|
| 197773 |
+
{
|
| 197774 |
+
"epoch": 618.56,
|
| 197775 |
+
"learning_rate": 8.767027463651051e-06,
|
| 197776 |
+
"loss": 0.4742,
|
| 197777 |
+
"step": 77350
|
| 197778 |
+
},
|
| 197779 |
+
{
|
| 197780 |
+
"epoch": 618.6,
|
| 197781 |
+
"learning_rate": 8.766946688206785e-06,
|
| 197782 |
+
"loss": 1.1874,
|
| 197783 |
+
"step": 77355
|
| 197784 |
+
},
|
| 197785 |
+
{
|
| 197786 |
+
"epoch": 618.64,
|
| 197787 |
+
"learning_rate": 8.766865912762521e-06,
|
| 197788 |
+
"loss": 0.3542,
|
| 197789 |
+
"step": 77360
|
| 197790 |
+
},
|
| 197791 |
+
{
|
| 197792 |
+
"epoch": 618.68,
|
| 197793 |
+
"learning_rate": 8.766785137318255e-06,
|
| 197794 |
+
"loss": 0.2895,
|
| 197795 |
+
"step": 77365
|
| 197796 |
+
},
|
| 197797 |
+
{
|
| 197798 |
+
"epoch": 618.72,
|
| 197799 |
+
"learning_rate": 8.766704361873991e-06,
|
| 197800 |
+
"loss": 0.31,
|
| 197801 |
+
"step": 77370
|
| 197802 |
+
},
|
| 197803 |
+
{
|
| 197804 |
+
"epoch": 618.76,
|
| 197805 |
+
"learning_rate": 8.766623586429725e-06,
|
| 197806 |
+
"loss": 0.4415,
|
| 197807 |
+
"step": 77375
|
| 197808 |
+
},
|
| 197809 |
+
{
|
| 197810 |
+
"epoch": 618.8,
|
| 197811 |
+
"learning_rate": 8.766542810985461e-06,
|
| 197812 |
+
"loss": 1.3175,
|
| 197813 |
+
"step": 77380
|
| 197814 |
+
},
|
| 197815 |
+
{
|
| 197816 |
+
"epoch": 618.84,
|
| 197817 |
+
"learning_rate": 8.766462035541197e-06,
|
| 197818 |
+
"loss": 0.3257,
|
| 197819 |
+
"step": 77385
|
| 197820 |
+
},
|
| 197821 |
+
{
|
| 197822 |
+
"epoch": 618.88,
|
| 197823 |
+
"learning_rate": 8.766381260096931e-06,
|
| 197824 |
+
"loss": 0.3193,
|
| 197825 |
+
"step": 77390
|
| 197826 |
+
},
|
| 197827 |
+
{
|
| 197828 |
+
"epoch": 618.92,
|
| 197829 |
+
"learning_rate": 8.766300484652667e-06,
|
| 197830 |
+
"loss": 0.3781,
|
| 197831 |
+
"step": 77395
|
| 197832 |
+
},
|
| 197833 |
+
{
|
| 197834 |
+
"epoch": 618.96,
|
| 197835 |
+
"learning_rate": 8.766219709208401e-06,
|
| 197836 |
+
"loss": 0.5195,
|
| 197837 |
+
"step": 77400
|
| 197838 |
+
},
|
| 197839 |
+
{
|
| 197840 |
+
"epoch": 619.0,
|
| 197841 |
+
"learning_rate": 8.766138933764137e-06,
|
| 197842 |
+
"loss": 1.2633,
|
| 197843 |
+
"step": 77405
|
| 197844 |
+
},
|
| 197845 |
+
{
|
| 197846 |
+
"epoch": 619.0,
|
| 197847 |
+
"eval_loss": 0.38753950595855713,
|
| 197848 |
+
"eval_runtime": 41.8381,
|
| 197849 |
+
"eval_samples_per_second": 19.982,
|
| 197850 |
+
"eval_steps_per_second": 0.645,
|
| 197851 |
+
"eval_wer": 0.1945414847161572,
|
| 197852 |
+
"step": 77405
|
| 197853 |
+
},
|
| 197854 |
+
{
|
| 197855 |
+
"epoch": 619.04,
|
| 197856 |
+
"learning_rate": 8.76605815831987e-06,
|
| 197857 |
+
"loss": 0.3856,
|
| 197858 |
+
"step": 77410
|
| 197859 |
+
},
|
| 197860 |
+
{
|
| 197861 |
+
"epoch": 619.08,
|
| 197862 |
+
"learning_rate": 8.765977382875607e-06,
|
| 197863 |
+
"loss": 0.3124,
|
| 197864 |
+
"step": 77415
|
| 197865 |
+
},
|
| 197866 |
+
{
|
| 197867 |
+
"epoch": 619.12,
|
| 197868 |
+
"learning_rate": 8.76589660743134e-06,
|
| 197869 |
+
"loss": 0.4006,
|
| 197870 |
+
"step": 77420
|
| 197871 |
+
},
|
| 197872 |
+
{
|
| 197873 |
+
"epoch": 619.16,
|
| 197874 |
+
"learning_rate": 8.765815831987077e-06,
|
| 197875 |
+
"loss": 0.4272,
|
| 197876 |
+
"step": 77425
|
| 197877 |
+
},
|
| 197878 |
+
{
|
| 197879 |
+
"epoch": 619.2,
|
| 197880 |
+
"learning_rate": 8.76573505654281e-06,
|
| 197881 |
+
"loss": 1.1045,
|
| 197882 |
+
"step": 77430
|
| 197883 |
+
},
|
| 197884 |
+
{
|
| 197885 |
+
"epoch": 619.24,
|
| 197886 |
+
"learning_rate": 8.765654281098547e-06,
|
| 197887 |
+
"loss": 0.3543,
|
| 197888 |
+
"step": 77435
|
| 197889 |
+
},
|
| 197890 |
+
{
|
| 197891 |
+
"epoch": 619.28,
|
| 197892 |
+
"learning_rate": 8.765573505654282e-06,
|
| 197893 |
+
"loss": 0.388,
|
| 197894 |
+
"step": 77440
|
| 197895 |
+
},
|
| 197896 |
+
{
|
| 197897 |
+
"epoch": 619.32,
|
| 197898 |
+
"learning_rate": 8.765492730210017e-06,
|
| 197899 |
+
"loss": 0.3724,
|
| 197900 |
+
"step": 77445
|
| 197901 |
+
},
|
| 197902 |
+
{
|
| 197903 |
+
"epoch": 619.36,
|
| 197904 |
+
"learning_rate": 8.765411954765752e-06,
|
| 197905 |
+
"loss": 0.3966,
|
| 197906 |
+
"step": 77450
|
| 197907 |
+
},
|
| 197908 |
+
{
|
| 197909 |
+
"epoch": 619.4,
|
| 197910 |
+
"learning_rate": 8.765331179321487e-06,
|
| 197911 |
+
"loss": 1.2215,
|
| 197912 |
+
"step": 77455
|
| 197913 |
+
},
|
| 197914 |
+
{
|
| 197915 |
+
"epoch": 619.44,
|
| 197916 |
+
"learning_rate": 8.765250403877222e-06,
|
| 197917 |
+
"loss": 0.3409,
|
| 197918 |
+
"step": 77460
|
| 197919 |
+
},
|
| 197920 |
+
{
|
| 197921 |
+
"epoch": 619.48,
|
| 197922 |
+
"learning_rate": 8.765169628432957e-06,
|
| 197923 |
+
"loss": 0.3197,
|
| 197924 |
+
"step": 77465
|
| 197925 |
+
},
|
| 197926 |
+
{
|
| 197927 |
+
"epoch": 619.52,
|
| 197928 |
+
"learning_rate": 8.765088852988692e-06,
|
| 197929 |
+
"loss": 0.3963,
|
| 197930 |
+
"step": 77470
|
| 197931 |
+
},
|
| 197932 |
+
{
|
| 197933 |
+
"epoch": 619.56,
|
| 197934 |
+
"learning_rate": 8.765008077544427e-06,
|
| 197935 |
+
"loss": 0.4343,
|
| 197936 |
+
"step": 77475
|
| 197937 |
+
},
|
| 197938 |
+
{
|
| 197939 |
+
"epoch": 619.6,
|
| 197940 |
+
"learning_rate": 8.764927302100162e-06,
|
| 197941 |
+
"loss": 1.3709,
|
| 197942 |
+
"step": 77480
|
| 197943 |
+
},
|
| 197944 |
+
{
|
| 197945 |
+
"epoch": 619.64,
|
| 197946 |
+
"learning_rate": 8.764846526655896e-06,
|
| 197947 |
+
"loss": 0.2994,
|
| 197948 |
+
"step": 77485
|
| 197949 |
+
},
|
| 197950 |
+
{
|
| 197951 |
+
"epoch": 619.68,
|
| 197952 |
+
"learning_rate": 8.764765751211632e-06,
|
| 197953 |
+
"loss": 0.346,
|
| 197954 |
+
"step": 77490
|
| 197955 |
+
},
|
| 197956 |
+
{
|
| 197957 |
+
"epoch": 619.72,
|
| 197958 |
+
"learning_rate": 8.764684975767368e-06,
|
| 197959 |
+
"loss": 0.3852,
|
| 197960 |
+
"step": 77495
|
| 197961 |
+
},
|
| 197962 |
+
{
|
| 197963 |
+
"epoch": 619.76,
|
| 197964 |
+
"learning_rate": 8.764604200323102e-06,
|
| 197965 |
+
"loss": 0.5026,
|
| 197966 |
+
"step": 77500
|
| 197967 |
+
},
|
| 197968 |
+
{
|
| 197969 |
+
"epoch": 619.8,
|
| 197970 |
+
"learning_rate": 8.764523424878838e-06,
|
| 197971 |
+
"loss": 1.3511,
|
| 197972 |
+
"step": 77505
|
| 197973 |
+
},
|
| 197974 |
+
{
|
| 197975 |
+
"epoch": 619.84,
|
| 197976 |
+
"learning_rate": 8.764442649434572e-06,
|
| 197977 |
+
"loss": 0.3118,
|
| 197978 |
+
"step": 77510
|
| 197979 |
+
},
|
| 197980 |
+
{
|
| 197981 |
+
"epoch": 619.88,
|
| 197982 |
+
"learning_rate": 8.764361873990308e-06,
|
| 197983 |
+
"loss": 0.3242,
|
| 197984 |
+
"step": 77515
|
| 197985 |
+
},
|
| 197986 |
+
{
|
| 197987 |
+
"epoch": 619.92,
|
| 197988 |
+
"learning_rate": 8.764281098546042e-06,
|
| 197989 |
+
"loss": 0.3453,
|
| 197990 |
+
"step": 77520
|
| 197991 |
+
},
|
| 197992 |
+
{
|
| 197993 |
+
"epoch": 619.96,
|
| 197994 |
+
"learning_rate": 8.764200323101778e-06,
|
| 197995 |
+
"loss": 0.4682,
|
| 197996 |
+
"step": 77525
|
| 197997 |
+
},
|
| 197998 |
+
{
|
| 197999 |
+
"epoch": 620.0,
|
| 198000 |
+
"learning_rate": 8.764119547657512e-06,
|
| 198001 |
+
"loss": 1.1805,
|
| 198002 |
+
"step": 77530
|
| 198003 |
+
},
|
| 198004 |
+
{
|
| 198005 |
+
"epoch": 620.0,
|
| 198006 |
+
"eval_loss": 0.4079735279083252,
|
| 198007 |
+
"eval_runtime": 43.0502,
|
| 198008 |
+
"eval_samples_per_second": 19.419,
|
| 198009 |
+
"eval_steps_per_second": 0.627,
|
| 198010 |
+
"eval_wer": 0.18670113456843315,
|
| 198011 |
+
"step": 77530
|
| 198012 |
+
},
|
| 198013 |
+
{
|
| 198014 |
+
"epoch": 620.04,
|
| 198015 |
+
"learning_rate": 8.764038772213248e-06,
|
| 198016 |
+
"loss": 0.3656,
|
| 198017 |
+
"step": 77535
|
| 198018 |
+
},
|
| 198019 |
+
{
|
| 198020 |
+
"epoch": 620.08,
|
| 198021 |
+
"learning_rate": 8.763957996768982e-06,
|
| 198022 |
+
"loss": 0.2756,
|
| 198023 |
+
"step": 77540
|
| 198024 |
+
},
|
| 198025 |
+
{
|
| 198026 |
+
"epoch": 620.12,
|
| 198027 |
+
"learning_rate": 8.763877221324718e-06,
|
| 198028 |
+
"loss": 0.4158,
|
| 198029 |
+
"step": 77545
|
| 198030 |
+
},
|
| 198031 |
+
{
|
| 198032 |
+
"epoch": 620.16,
|
| 198033 |
+
"learning_rate": 8.763796445880452e-06,
|
| 198034 |
+
"loss": 0.4729,
|
| 198035 |
+
"step": 77550
|
| 198036 |
+
},
|
| 198037 |
+
{
|
| 198038 |
+
"epoch": 620.2,
|
| 198039 |
+
"learning_rate": 8.763715670436188e-06,
|
| 198040 |
+
"loss": 1.4564,
|
| 198041 |
+
"step": 77555
|
| 198042 |
+
},
|
| 198043 |
+
{
|
| 198044 |
+
"epoch": 620.24,
|
| 198045 |
+
"learning_rate": 8.763634894991924e-06,
|
| 198046 |
+
"loss": 0.3279,
|
| 198047 |
+
"step": 77560
|
| 198048 |
+
},
|
| 198049 |
+
{
|
| 198050 |
+
"epoch": 620.28,
|
| 198051 |
+
"learning_rate": 8.763554119547658e-06,
|
| 198052 |
+
"loss": 0.3054,
|
| 198053 |
+
"step": 77565
|
| 198054 |
+
},
|
| 198055 |
+
{
|
| 198056 |
+
"epoch": 620.32,
|
| 198057 |
+
"learning_rate": 8.763473344103394e-06,
|
| 198058 |
+
"loss": 0.3148,
|
| 198059 |
+
"step": 77570
|
| 198060 |
+
},
|
| 198061 |
+
{
|
| 198062 |
+
"epoch": 620.36,
|
| 198063 |
+
"learning_rate": 8.763392568659128e-06,
|
| 198064 |
+
"loss": 0.4941,
|
| 198065 |
+
"step": 77575
|
| 198066 |
+
},
|
| 198067 |
+
{
|
| 198068 |
+
"epoch": 620.4,
|
| 198069 |
+
"learning_rate": 8.763311793214864e-06,
|
| 198070 |
+
"loss": 1.3589,
|
| 198071 |
+
"step": 77580
|
| 198072 |
+
},
|
| 198073 |
+
{
|
| 198074 |
+
"epoch": 620.44,
|
| 198075 |
+
"learning_rate": 8.763231017770598e-06,
|
| 198076 |
+
"loss": 0.3719,
|
| 198077 |
+
"step": 77585
|
| 198078 |
+
},
|
| 198079 |
+
{
|
| 198080 |
+
"epoch": 620.48,
|
| 198081 |
+
"learning_rate": 8.763150242326334e-06,
|
| 198082 |
+
"loss": 0.2943,
|
| 198083 |
+
"step": 77590
|
| 198084 |
+
},
|
| 198085 |
+
{
|
| 198086 |
+
"epoch": 620.52,
|
| 198087 |
+
"learning_rate": 8.763069466882068e-06,
|
| 198088 |
+
"loss": 0.414,
|
| 198089 |
+
"step": 77595
|
| 198090 |
+
},
|
| 198091 |
+
{
|
| 198092 |
+
"epoch": 620.56,
|
| 198093 |
+
"learning_rate": 8.762988691437804e-06,
|
| 198094 |
+
"loss": 0.3896,
|
| 198095 |
+
"step": 77600
|
| 198096 |
+
},
|
| 198097 |
+
{
|
| 198098 |
+
"epoch": 620.6,
|
| 198099 |
+
"learning_rate": 8.762907915993538e-06,
|
| 198100 |
+
"loss": 1.2367,
|
| 198101 |
+
"step": 77605
|
| 198102 |
+
},
|
| 198103 |
+
{
|
| 198104 |
+
"epoch": 620.64,
|
| 198105 |
+
"learning_rate": 8.762827140549274e-06,
|
| 198106 |
+
"loss": 0.3736,
|
| 198107 |
+
"step": 77610
|
| 198108 |
+
},
|
| 198109 |
+
{
|
| 198110 |
+
"epoch": 620.68,
|
| 198111 |
+
"learning_rate": 8.76274636510501e-06,
|
| 198112 |
+
"loss": 0.3219,
|
| 198113 |
+
"step": 77615
|
| 198114 |
+
},
|
| 198115 |
+
{
|
| 198116 |
+
"epoch": 620.72,
|
| 198117 |
+
"learning_rate": 8.762665589660744e-06,
|
| 198118 |
+
"loss": 0.4256,
|
| 198119 |
+
"step": 77620
|
| 198120 |
+
},
|
| 198121 |
+
{
|
| 198122 |
+
"epoch": 620.76,
|
| 198123 |
+
"learning_rate": 8.76258481421648e-06,
|
| 198124 |
+
"loss": 0.5015,
|
| 198125 |
+
"step": 77625
|
| 198126 |
+
},
|
| 198127 |
+
{
|
| 198128 |
+
"epoch": 620.8,
|
| 198129 |
+
"learning_rate": 8.762504038772214e-06,
|
| 198130 |
+
"loss": 1.3466,
|
| 198131 |
+
"step": 77630
|
| 198132 |
+
},
|
| 198133 |
+
{
|
| 198134 |
+
"epoch": 620.84,
|
| 198135 |
+
"learning_rate": 8.76242326332795e-06,
|
| 198136 |
+
"loss": 0.3469,
|
| 198137 |
+
"step": 77635
|
| 198138 |
+
},
|
| 198139 |
+
{
|
| 198140 |
+
"epoch": 620.88,
|
| 198141 |
+
"learning_rate": 8.762342487883684e-06,
|
| 198142 |
+
"loss": 0.2818,
|
| 198143 |
+
"step": 77640
|
| 198144 |
+
},
|
| 198145 |
+
{
|
| 198146 |
+
"epoch": 620.92,
|
| 198147 |
+
"learning_rate": 8.76226171243942e-06,
|
| 198148 |
+
"loss": 0.3715,
|
| 198149 |
+
"step": 77645
|
| 198150 |
+
},
|
| 198151 |
+
{
|
| 198152 |
+
"epoch": 620.96,
|
| 198153 |
+
"learning_rate": 8.762180936995154e-06,
|
| 198154 |
+
"loss": 0.5095,
|
| 198155 |
+
"step": 77650
|
| 198156 |
+
},
|
| 198157 |
+
{
|
| 198158 |
+
"epoch": 621.0,
|
| 198159 |
+
"learning_rate": 8.76210016155089e-06,
|
| 198160 |
+
"loss": 1.3371,
|
| 198161 |
+
"step": 77655
|
| 198162 |
+
},
|
| 198163 |
+
{
|
| 198164 |
+
"epoch": 621.0,
|
| 198165 |
+
"eval_loss": 0.3949979543685913,
|
| 198166 |
+
"eval_runtime": 41.8082,
|
| 198167 |
+
"eval_samples_per_second": 20.02,
|
| 198168 |
+
"eval_steps_per_second": 0.646,
|
| 198169 |
+
"eval_wer": 0.19575146813601102,
|
| 198170 |
+
"step": 77655
|
| 198171 |
+
},
|
| 198172 |
+
{
|
| 198173 |
+
"epoch": 626.04,
|
| 198174 |
+
"learning_rate": 8.762019386106624e-06,
|
| 198175 |
+
"loss": 0.3413,
|
| 198176 |
+
"step": 77660
|
| 198177 |
+
},
|
| 198178 |
+
{
|
| 198179 |
+
"epoch": 626.08,
|
| 198180 |
+
"learning_rate": 8.76193861066236e-06,
|
| 198181 |
+
"loss": 0.3734,
|
| 198182 |
+
"step": 77665
|
| 198183 |
+
},
|
| 198184 |
+
{
|
| 198185 |
+
"epoch": 626.12,
|
| 198186 |
+
"learning_rate": 8.761857835218095e-06,
|
| 198187 |
+
"loss": 0.3673,
|
| 198188 |
+
"step": 77670
|
| 198189 |
+
},
|
| 198190 |
+
{
|
| 198191 |
+
"epoch": 626.16,
|
| 198192 |
+
"learning_rate": 8.76177705977383e-06,
|
| 198193 |
+
"loss": 0.4349,
|
| 198194 |
+
"step": 77675
|
| 198195 |
+
},
|
| 198196 |
+
{
|
| 198197 |
+
"epoch": 626.2,
|
| 198198 |
+
"learning_rate": 8.761696284329565e-06,
|
| 198199 |
+
"loss": 1.4509,
|
| 198200 |
+
"step": 77680
|
| 198201 |
+
},
|
| 198202 |
+
{
|
| 198203 |
+
"epoch": 626.24,
|
| 198204 |
+
"learning_rate": 8.7616155088853e-06,
|
| 198205 |
+
"loss": 0.3588,
|
| 198206 |
+
"step": 77685
|
| 198207 |
+
},
|
| 198208 |
+
{
|
| 198209 |
+
"epoch": 626.28,
|
| 198210 |
+
"learning_rate": 8.761534733441035e-06,
|
| 198211 |
+
"loss": 0.3109,
|
| 198212 |
+
"step": 77690
|
| 198213 |
+
},
|
| 198214 |
+
{
|
| 198215 |
+
"epoch": 626.32,
|
| 198216 |
+
"learning_rate": 8.76145395799677e-06,
|
| 198217 |
+
"loss": 0.3368,
|
| 198218 |
+
"step": 77695
|
| 198219 |
+
},
|
| 198220 |
+
{
|
| 198221 |
+
"epoch": 626.36,
|
| 198222 |
+
"learning_rate": 8.761373182552505e-06,
|
| 198223 |
+
"loss": 0.5113,
|
| 198224 |
+
"step": 77700
|
| 198225 |
+
},
|
| 198226 |
+
{
|
| 198227 |
+
"epoch": 626.4,
|
| 198228 |
+
"learning_rate": 8.76129240710824e-06,
|
| 198229 |
+
"loss": 1.1518,
|
| 198230 |
+
"step": 77705
|
| 198231 |
+
},
|
| 198232 |
+
{
|
| 198233 |
+
"epoch": 626.44,
|
| 198234 |
+
"learning_rate": 8.761211631663975e-06,
|
| 198235 |
+
"loss": 0.4165,
|
| 198236 |
+
"step": 77710
|
| 198237 |
+
},
|
| 198238 |
+
{
|
| 198239 |
+
"epoch": 626.48,
|
| 198240 |
+
"learning_rate": 8.76113085621971e-06,
|
| 198241 |
+
"loss": 0.3262,
|
| 198242 |
+
"step": 77715
|
| 198243 |
+
},
|
| 198244 |
+
{
|
| 198245 |
+
"epoch": 626.52,
|
| 198246 |
+
"learning_rate": 8.761050080775445e-06,
|
| 198247 |
+
"loss": 0.3053,
|
| 198248 |
+
"step": 77720
|
| 198249 |
+
},
|
| 198250 |
+
{
|
| 198251 |
+
"epoch": 626.56,
|
| 198252 |
+
"learning_rate": 8.76096930533118e-06,
|
| 198253 |
+
"loss": 0.4874,
|
| 198254 |
+
"step": 77725
|
| 198255 |
+
},
|
| 198256 |
+
{
|
| 198257 |
+
"epoch": 626.6,
|
| 198258 |
+
"learning_rate": 8.760888529886915e-06,
|
| 198259 |
+
"loss": 1.3094,
|
| 198260 |
+
"step": 77730
|
| 198261 |
+
},
|
| 198262 |
+
{
|
| 198263 |
+
"epoch": 626.64,
|
| 198264 |
+
"learning_rate": 8.760807754442651e-06,
|
| 198265 |
+
"loss": 0.3594,
|
| 198266 |
+
"step": 77735
|
| 198267 |
+
},
|
| 198268 |
+
{
|
| 198269 |
+
"epoch": 626.68,
|
| 198270 |
+
"learning_rate": 8.760726978998385e-06,
|
| 198271 |
+
"loss": 0.3254,
|
| 198272 |
+
"step": 77740
|
| 198273 |
+
},
|
| 198274 |
+
{
|
| 198275 |
+
"epoch": 626.72,
|
| 198276 |
+
"learning_rate": 8.760646203554121e-06,
|
| 198277 |
+
"loss": 0.3081,
|
| 198278 |
+
"step": 77745
|
| 198279 |
+
},
|
| 198280 |
+
{
|
| 198281 |
+
"epoch": 626.76,
|
| 198282 |
+
"learning_rate": 8.760565428109855e-06,
|
| 198283 |
+
"loss": 0.4125,
|
| 198284 |
+
"step": 77750
|
| 198285 |
+
},
|
| 198286 |
+
{
|
| 198287 |
+
"epoch": 626.8,
|
| 198288 |
+
"learning_rate": 8.760484652665591e-06,
|
| 198289 |
+
"loss": 1.2727,
|
| 198290 |
+
"step": 77755
|
| 198291 |
+
},
|
| 198292 |
+
{
|
| 198293 |
+
"epoch": 626.84,
|
| 198294 |
+
"learning_rate": 8.760403877221325e-06,
|
| 198295 |
+
"loss": 0.3183,
|
| 198296 |
+
"step": 77760
|
| 198297 |
+
},
|
| 198298 |
+
{
|
| 198299 |
+
"epoch": 626.88,
|
| 198300 |
+
"learning_rate": 8.760323101777061e-06,
|
| 198301 |
+
"loss": 0.3672,
|
| 198302 |
+
"step": 77765
|
| 198303 |
+
},
|
| 198304 |
+
{
|
| 198305 |
+
"epoch": 626.92,
|
| 198306 |
+
"learning_rate": 8.760242326332795e-06,
|
| 198307 |
+
"loss": 0.3245,
|
| 198308 |
+
"step": 77770
|
| 198309 |
+
},
|
| 198310 |
+
{
|
| 198311 |
+
"epoch": 626.96,
|
| 198312 |
+
"learning_rate": 8.760161550888531e-06,
|
| 198313 |
+
"loss": 0.4834,
|
| 198314 |
+
"step": 77775
|
| 198315 |
+
},
|
| 198316 |
+
{
|
| 198317 |
+
"epoch": 627.0,
|
| 198318 |
+
"eval_loss": 0.4380192756652832,
|
| 198319 |
+
"eval_runtime": 42.5503,
|
| 198320 |
+
"eval_samples_per_second": 19.671,
|
| 198321 |
+
"eval_steps_per_second": 0.635,
|
| 198322 |
+
"eval_wer": 0.19297325642370214,
|
| 198323 |
+
"step": 77779
|
| 198324 |
}
|
| 198325 |
],
|
| 198326 |
"max_steps": 620000,
|
| 198327 |
"num_train_epochs": 5000,
|
| 198328 |
+
"total_flos": 2.1888125289934658e+20,
|
| 198329 |
"trial_name": null,
|
| 198330 |
"trial_params": null
|
| 198331 |
}
|
model-bin/finetune/base/{checkpoint-77155 → checkpoint-77779}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629892093.865816/events.out.tfevents.1629892093.7e498afd5545.7645.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:93b6706032dcaef909be43b5fe7a4c558845d351dc9c196ee7a47b6fbe16f68c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629892567.0899503/events.out.tfevents.1629892567.7e498afd5545.7645.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b3101719cd5afdbb2f7914ee12a139bd70492eed84bf82bdb1dcf857d2527a2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629893045.2545307/events.out.tfevents.1629893045.7e498afd5545.7645.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0355fbcef8893ca75f858f1841846c42bc4c2b5d1d3b408071860d74efda4f7e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629893510.1288207/events.out.tfevents.1629893510.7e498afd5545.7645.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca5317449e2f180e69d2ef819442715847aeccfca0f66584ed91a0f8059f7839
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629893973.3909373/events.out.tfevents.1629893973.7e498afd5545.7645.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d13be364e6cd4fcd39d562b3f0a9d11e0f53aa5162c3cfcc76a07ad5dcc44f0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629892093.7e498afd5545.7645.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c4ad90a8d0f68de4a48de6f09259d2c3f3da8cc2f22563aee5b7e4836c14eb4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629892566.7e498afd5545.7645.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a5adf725e71f1292675eddf7ef8d38ef14fe9c7b57c7e462c565c2c82a571eca
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629893045.7e498afd5545.7645.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c1771ab9166258a17533cdc266bb1be41a00d8a695c19f0ee7b17c1bb6559aa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629893510.7e498afd5545.7645.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4b988b11d6a85d25d2a05c50594be9524ebf1c8e4daa4030ae5837283db84ec
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629893973.7e498afd5545.7645.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ed23b1847a5e37f605602b624d41aa91c18bb2180f34333cc53c8aad8696a34
|
| 3 |
+
size 8462
|