"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/trainer_state.json +1111 -4
- model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630120105.1763353/events.out.tfevents.1630120105.86bb0ddabf9b.1042.61 +3 -0
- model-bin/finetune/base/log/1630120512.9155197/events.out.tfevents.1630120512.86bb0ddabf9b.1042.63 +3 -0
- model-bin/finetune/base/log/1630122518.2177079/events.out.tfevents.1630122518.86bb0ddabf9b.4092.1 +3 -0
- model-bin/finetune/base/log/1630122917.3648393/events.out.tfevents.1630122917.86bb0ddabf9b.4092.3 +3 -0
- model-bin/finetune/base/log/1630123316.325624/events.out.tfevents.1630123316.86bb0ddabf9b.4092.5 +3 -0
- model-bin/finetune/base/log/1630123712.2616467/events.out.tfevents.1630123712.86bb0ddabf9b.4092.7 +3 -0
- model-bin/finetune/base/log/1630124107.8066275/events.out.tfevents.1630124107.86bb0ddabf9b.4092.9 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630120105.86bb0ddabf9b.1042.60 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630120512.86bb0ddabf9b.1042.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630122518.86bb0ddabf9b.4092.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630122917.86bb0ddabf9b.4092.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630123316.86bb0ddabf9b.4092.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630123712.86bb0ddabf9b.4092.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630124107.86bb0ddabf9b.4092.8 +3 -0
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/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 sha256:824bdd08c7df579935f156e28209c69d29ed3abb148c9e232ef084a81064b9f6
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/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 sha256:4aaa5ecf7855c883590283b7ea54bd9eae633f0bbdccd988047d5691c3d47d2d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/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 sha256:86357397e153f05f9f45bd29caf9103edf34eadb7eb5627658dcefb2f331a9ca
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/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 sha256:33dab386f477ad09c561d8867f5af2329e81990e9d6cd66bae89bce2f60ded3e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/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:bc05e9bcb031fbc882d03868ce2a3c5093e01bdd14c38bd6f2151f7329b735e5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -249144,11 +249144,1118 @@
|
|
| 249144 |
"eval_steps_per_second": 0.701,
|
| 249145 |
"eval_wer": 0.18467364669191358,
|
| 249146 |
"step": 117722
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 249147 |
}
|
| 249148 |
],
|
| 249149 |
-
"max_steps":
|
| 249150 |
"num_train_epochs": 5000,
|
| 249151 |
-
"total_flos": 3.
|
| 249152 |
"trial_name": null,
|
| 249153 |
"trial_params": null
|
| 249154 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 956.0,
|
| 5 |
+
"global_step": 118592,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 249144 |
"eval_steps_per_second": 0.701,
|
| 249145 |
"eval_wer": 0.18467364669191358,
|
| 249146 |
"step": 117722
|
| 249147 |
+
},
|
| 249148 |
+
{
|
| 249149 |
+
"epoch": 941.02,
|
| 249150 |
+
"learning_rate": 8.130160256410257e-06,
|
| 249151 |
+
"loss": 0.3663,
|
| 249152 |
+
"step": 117725
|
| 249153 |
+
},
|
| 249154 |
+
{
|
| 249155 |
+
"epoch": 941.06,
|
| 249156 |
+
"learning_rate": 8.130080128205129e-06,
|
| 249157 |
+
"loss": 0.282,
|
| 249158 |
+
"step": 117730
|
| 249159 |
+
},
|
| 249160 |
+
{
|
| 249161 |
+
"epoch": 941.1,
|
| 249162 |
+
"learning_rate": 8.13e-06,
|
| 249163 |
+
"loss": 0.312,
|
| 249164 |
+
"step": 117735
|
| 249165 |
+
},
|
| 249166 |
+
{
|
| 249167 |
+
"epoch": 941.14,
|
| 249168 |
+
"learning_rate": 8.129919871794873e-06,
|
| 249169 |
+
"loss": 0.3393,
|
| 249170 |
+
"step": 117740
|
| 249171 |
+
},
|
| 249172 |
+
{
|
| 249173 |
+
"epoch": 941.18,
|
| 249174 |
+
"learning_rate": 8.129839743589744e-06,
|
| 249175 |
+
"loss": 0.7809,
|
| 249176 |
+
"step": 117745
|
| 249177 |
+
},
|
| 249178 |
+
{
|
| 249179 |
+
"epoch": 941.22,
|
| 249180 |
+
"learning_rate": 8.129759615384616e-06,
|
| 249181 |
+
"loss": 0.8867,
|
| 249182 |
+
"step": 117750
|
| 249183 |
+
},
|
| 249184 |
+
{
|
| 249185 |
+
"epoch": 941.26,
|
| 249186 |
+
"learning_rate": 8.129679487179489e-06,
|
| 249187 |
+
"loss": 0.2999,
|
| 249188 |
+
"step": 117755
|
| 249189 |
+
},
|
| 249190 |
+
{
|
| 249191 |
+
"epoch": 941.3,
|
| 249192 |
+
"learning_rate": 8.129599358974359e-06,
|
| 249193 |
+
"loss": 0.2899,
|
| 249194 |
+
"step": 117760
|
| 249195 |
+
},
|
| 249196 |
+
{
|
| 249197 |
+
"epoch": 941.34,
|
| 249198 |
+
"learning_rate": 8.129519230769232e-06,
|
| 249199 |
+
"loss": 0.3377,
|
| 249200 |
+
"step": 117765
|
| 249201 |
+
},
|
| 249202 |
+
{
|
| 249203 |
+
"epoch": 941.38,
|
| 249204 |
+
"learning_rate": 8.129439102564103e-06,
|
| 249205 |
+
"loss": 0.8005,
|
| 249206 |
+
"step": 117770
|
| 249207 |
+
},
|
| 249208 |
+
{
|
| 249209 |
+
"epoch": 941.42,
|
| 249210 |
+
"learning_rate": 8.129358974358974e-06,
|
| 249211 |
+
"loss": 0.8139,
|
| 249212 |
+
"step": 117775
|
| 249213 |
+
},
|
| 249214 |
+
{
|
| 249215 |
+
"epoch": 941.46,
|
| 249216 |
+
"learning_rate": 8.129278846153846e-06,
|
| 249217 |
+
"loss": 0.2583,
|
| 249218 |
+
"step": 117780
|
| 249219 |
+
},
|
| 249220 |
+
{
|
| 249221 |
+
"epoch": 941.5,
|
| 249222 |
+
"learning_rate": 8.129198717948719e-06,
|
| 249223 |
+
"loss": 0.2831,
|
| 249224 |
+
"step": 117785
|
| 249225 |
+
},
|
| 249226 |
+
{
|
| 249227 |
+
"epoch": 941.54,
|
| 249228 |
+
"learning_rate": 8.12911858974359e-06,
|
| 249229 |
+
"loss": 0.3552,
|
| 249230 |
+
"step": 117790
|
| 249231 |
+
},
|
| 249232 |
+
{
|
| 249233 |
+
"epoch": 941.58,
|
| 249234 |
+
"learning_rate": 8.129038461538461e-06,
|
| 249235 |
+
"loss": 0.7863,
|
| 249236 |
+
"step": 117795
|
| 249237 |
+
},
|
| 249238 |
+
{
|
| 249239 |
+
"epoch": 941.62,
|
| 249240 |
+
"learning_rate": 8.128958333333335e-06,
|
| 249241 |
+
"loss": 0.8211,
|
| 249242 |
+
"step": 117800
|
| 249243 |
+
},
|
| 249244 |
+
{
|
| 249245 |
+
"epoch": 941.66,
|
| 249246 |
+
"learning_rate": 8.128878205128206e-06,
|
| 249247 |
+
"loss": 0.3191,
|
| 249248 |
+
"step": 117805
|
| 249249 |
+
},
|
| 249250 |
+
{
|
| 249251 |
+
"epoch": 941.7,
|
| 249252 |
+
"learning_rate": 8.128798076923077e-06,
|
| 249253 |
+
"loss": 0.3874,
|
| 249254 |
+
"step": 117810
|
| 249255 |
+
},
|
| 249256 |
+
{
|
| 249257 |
+
"epoch": 941.74,
|
| 249258 |
+
"learning_rate": 8.128717948717949e-06,
|
| 249259 |
+
"loss": 0.418,
|
| 249260 |
+
"step": 117815
|
| 249261 |
+
},
|
| 249262 |
+
{
|
| 249263 |
+
"epoch": 941.78,
|
| 249264 |
+
"learning_rate": 8.128637820512822e-06,
|
| 249265 |
+
"loss": 0.8736,
|
| 249266 |
+
"step": 117820
|
| 249267 |
+
},
|
| 249268 |
+
{
|
| 249269 |
+
"epoch": 941.82,
|
| 249270 |
+
"learning_rate": 8.128557692307693e-06,
|
| 249271 |
+
"loss": 0.9431,
|
| 249272 |
+
"step": 117825
|
| 249273 |
+
},
|
| 249274 |
+
{
|
| 249275 |
+
"epoch": 941.86,
|
| 249276 |
+
"learning_rate": 8.128477564102564e-06,
|
| 249277 |
+
"loss": 0.2709,
|
| 249278 |
+
"step": 117830
|
| 249279 |
+
},
|
| 249280 |
+
{
|
| 249281 |
+
"epoch": 941.9,
|
| 249282 |
+
"learning_rate": 8.128397435897436e-06,
|
| 249283 |
+
"loss": 0.2832,
|
| 249284 |
+
"step": 117835
|
| 249285 |
+
},
|
| 249286 |
+
{
|
| 249287 |
+
"epoch": 941.94,
|
| 249288 |
+
"learning_rate": 8.128317307692309e-06,
|
| 249289 |
+
"loss": 0.3677,
|
| 249290 |
+
"step": 117840
|
| 249291 |
+
},
|
| 249292 |
+
{
|
| 249293 |
+
"epoch": 941.98,
|
| 249294 |
+
"learning_rate": 8.12823717948718e-06,
|
| 249295 |
+
"loss": 0.8872,
|
| 249296 |
+
"step": 117845
|
| 249297 |
+
},
|
| 249298 |
+
{
|
| 249299 |
+
"epoch": 942.0,
|
| 249300 |
+
"eval_loss": 0.32367920875549316,
|
| 249301 |
+
"eval_runtime": 39.3812,
|
| 249302 |
+
"eval_samples_per_second": 21.305,
|
| 249303 |
+
"eval_steps_per_second": 0.686,
|
| 249304 |
+
"eval_wer": 0.1751803751803752,
|
| 249305 |
+
"step": 117847
|
| 249306 |
+
},
|
| 249307 |
+
{
|
| 249308 |
+
"epoch": 950.02,
|
| 249309 |
+
"learning_rate": 8.128157051282051e-06,
|
| 249310 |
+
"loss": 0.3156,
|
| 249311 |
+
"step": 117850
|
| 249312 |
+
},
|
| 249313 |
+
{
|
| 249314 |
+
"epoch": 950.06,
|
| 249315 |
+
"learning_rate": 8.128076923076925e-06,
|
| 249316 |
+
"loss": 0.2643,
|
| 249317 |
+
"step": 117855
|
| 249318 |
+
},
|
| 249319 |
+
{
|
| 249320 |
+
"epoch": 950.1,
|
| 249321 |
+
"learning_rate": 8.127996794871796e-06,
|
| 249322 |
+
"loss": 0.2823,
|
| 249323 |
+
"step": 117860
|
| 249324 |
+
},
|
| 249325 |
+
{
|
| 249326 |
+
"epoch": 950.15,
|
| 249327 |
+
"learning_rate": 8.127916666666667e-06,
|
| 249328 |
+
"loss": 0.3811,
|
| 249329 |
+
"step": 117865
|
| 249330 |
+
},
|
| 249331 |
+
{
|
| 249332 |
+
"epoch": 950.19,
|
| 249333 |
+
"learning_rate": 8.127836538461539e-06,
|
| 249334 |
+
"loss": 0.7131,
|
| 249335 |
+
"step": 117870
|
| 249336 |
+
},
|
| 249337 |
+
{
|
| 249338 |
+
"epoch": 950.23,
|
| 249339 |
+
"learning_rate": 8.127756410256412e-06,
|
| 249340 |
+
"loss": 0.9671,
|
| 249341 |
+
"step": 117875
|
| 249342 |
+
},
|
| 249343 |
+
{
|
| 249344 |
+
"epoch": 950.27,
|
| 249345 |
+
"learning_rate": 8.127676282051283e-06,
|
| 249346 |
+
"loss": 0.2583,
|
| 249347 |
+
"step": 117880
|
| 249348 |
+
},
|
| 249349 |
+
{
|
| 249350 |
+
"epoch": 950.31,
|
| 249351 |
+
"learning_rate": 8.127596153846154e-06,
|
| 249352 |
+
"loss": 0.2812,
|
| 249353 |
+
"step": 117885
|
| 249354 |
+
},
|
| 249355 |
+
{
|
| 249356 |
+
"epoch": 950.35,
|
| 249357 |
+
"learning_rate": 8.127516025641027e-06,
|
| 249358 |
+
"loss": 0.3449,
|
| 249359 |
+
"step": 117890
|
| 249360 |
+
},
|
| 249361 |
+
{
|
| 249362 |
+
"epoch": 950.39,
|
| 249363 |
+
"learning_rate": 8.127435897435899e-06,
|
| 249364 |
+
"loss": 0.7129,
|
| 249365 |
+
"step": 117895
|
| 249366 |
+
},
|
| 249367 |
+
{
|
| 249368 |
+
"epoch": 950.43,
|
| 249369 |
+
"learning_rate": 8.12735576923077e-06,
|
| 249370 |
+
"loss": 0.8821,
|
| 249371 |
+
"step": 117900
|
| 249372 |
+
},
|
| 249373 |
+
{
|
| 249374 |
+
"epoch": 950.47,
|
| 249375 |
+
"learning_rate": 8.127275641025642e-06,
|
| 249376 |
+
"loss": 0.2769,
|
| 249377 |
+
"step": 117905
|
| 249378 |
+
},
|
| 249379 |
+
{
|
| 249380 |
+
"epoch": 950.51,
|
| 249381 |
+
"learning_rate": 8.127195512820515e-06,
|
| 249382 |
+
"loss": 0.3113,
|
| 249383 |
+
"step": 117910
|
| 249384 |
+
},
|
| 249385 |
+
{
|
| 249386 |
+
"epoch": 950.55,
|
| 249387 |
+
"learning_rate": 8.127115384615384e-06,
|
| 249388 |
+
"loss": 0.3599,
|
| 249389 |
+
"step": 117915
|
| 249390 |
+
},
|
| 249391 |
+
{
|
| 249392 |
+
"epoch": 950.59,
|
| 249393 |
+
"learning_rate": 8.127035256410257e-06,
|
| 249394 |
+
"loss": 0.6893,
|
| 249395 |
+
"step": 117920
|
| 249396 |
+
},
|
| 249397 |
+
{
|
| 249398 |
+
"epoch": 950.63,
|
| 249399 |
+
"learning_rate": 8.126955128205129e-06,
|
| 249400 |
+
"loss": 0.8505,
|
| 249401 |
+
"step": 117925
|
| 249402 |
+
},
|
| 249403 |
+
{
|
| 249404 |
+
"epoch": 950.67,
|
| 249405 |
+
"learning_rate": 8.126875e-06,
|
| 249406 |
+
"loss": 0.2853,
|
| 249407 |
+
"step": 117930
|
| 249408 |
+
},
|
| 249409 |
+
{
|
| 249410 |
+
"epoch": 950.71,
|
| 249411 |
+
"learning_rate": 8.126794871794871e-06,
|
| 249412 |
+
"loss": 0.31,
|
| 249413 |
+
"step": 117935
|
| 249414 |
+
},
|
| 249415 |
+
{
|
| 249416 |
+
"epoch": 950.75,
|
| 249417 |
+
"learning_rate": 8.126714743589744e-06,
|
| 249418 |
+
"loss": 0.3777,
|
| 249419 |
+
"step": 117940
|
| 249420 |
+
},
|
| 249421 |
+
{
|
| 249422 |
+
"epoch": 950.79,
|
| 249423 |
+
"learning_rate": 8.126634615384616e-06,
|
| 249424 |
+
"loss": 0.7106,
|
| 249425 |
+
"step": 117945
|
| 249426 |
+
},
|
| 249427 |
+
{
|
| 249428 |
+
"epoch": 950.83,
|
| 249429 |
+
"learning_rate": 8.126554487179487e-06,
|
| 249430 |
+
"loss": 0.7565,
|
| 249431 |
+
"step": 117950
|
| 249432 |
+
},
|
| 249433 |
+
{
|
| 249434 |
+
"epoch": 950.87,
|
| 249435 |
+
"learning_rate": 8.12647435897436e-06,
|
| 249436 |
+
"loss": 0.3012,
|
| 249437 |
+
"step": 117955
|
| 249438 |
+
},
|
| 249439 |
+
{
|
| 249440 |
+
"epoch": 950.91,
|
| 249441 |
+
"learning_rate": 8.126394230769232e-06,
|
| 249442 |
+
"loss": 0.2882,
|
| 249443 |
+
"step": 117960
|
| 249444 |
+
},
|
| 249445 |
+
{
|
| 249446 |
+
"epoch": 950.95,
|
| 249447 |
+
"learning_rate": 8.126314102564103e-06,
|
| 249448 |
+
"loss": 0.4194,
|
| 249449 |
+
"step": 117965
|
| 249450 |
+
},
|
| 249451 |
+
{
|
| 249452 |
+
"epoch": 950.99,
|
| 249453 |
+
"learning_rate": 8.126233974358974e-06,
|
| 249454 |
+
"loss": 1.0347,
|
| 249455 |
+
"step": 117970
|
| 249456 |
+
},
|
| 249457 |
+
{
|
| 249458 |
+
"epoch": 951.0,
|
| 249459 |
+
"eval_loss": 0.3759602904319763,
|
| 249460 |
+
"eval_runtime": 38.3896,
|
| 249461 |
+
"eval_samples_per_second": 21.855,
|
| 249462 |
+
"eval_steps_per_second": 0.703,
|
| 249463 |
+
"eval_wer": 0.17543605703119347,
|
| 249464 |
+
"step": 117971
|
| 249465 |
+
},
|
| 249466 |
+
{
|
| 249467 |
+
"epoch": 943.03,
|
| 249468 |
+
"learning_rate": 8.126153846153847e-06,
|
| 249469 |
+
"loss": 0.3532,
|
| 249470 |
+
"step": 117975
|
| 249471 |
+
},
|
| 249472 |
+
{
|
| 249473 |
+
"epoch": 943.07,
|
| 249474 |
+
"learning_rate": 8.126073717948719e-06,
|
| 249475 |
+
"loss": 0.2588,
|
| 249476 |
+
"step": 117980
|
| 249477 |
+
},
|
| 249478 |
+
{
|
| 249479 |
+
"epoch": 943.11,
|
| 249480 |
+
"learning_rate": 8.12599358974359e-06,
|
| 249481 |
+
"loss": 0.3235,
|
| 249482 |
+
"step": 117985
|
| 249483 |
+
},
|
| 249484 |
+
{
|
| 249485 |
+
"epoch": 943.15,
|
| 249486 |
+
"learning_rate": 8.125913461538463e-06,
|
| 249487 |
+
"loss": 0.4366,
|
| 249488 |
+
"step": 117990
|
| 249489 |
+
},
|
| 249490 |
+
{
|
| 249491 |
+
"epoch": 943.19,
|
| 249492 |
+
"learning_rate": 8.125833333333334e-06,
|
| 249493 |
+
"loss": 0.8924,
|
| 249494 |
+
"step": 117995
|
| 249495 |
+
},
|
| 249496 |
+
{
|
| 249497 |
+
"epoch": 943.23,
|
| 249498 |
+
"learning_rate": 8.125753205128206e-06,
|
| 249499 |
+
"loss": 0.6004,
|
| 249500 |
+
"step": 118000
|
| 249501 |
+
},
|
| 249502 |
+
{
|
| 249503 |
+
"epoch": 943.27,
|
| 249504 |
+
"learning_rate": 8.125673076923077e-06,
|
| 249505 |
+
"loss": 0.2538,
|
| 249506 |
+
"step": 118005
|
| 249507 |
+
},
|
| 249508 |
+
{
|
| 249509 |
+
"epoch": 943.31,
|
| 249510 |
+
"learning_rate": 8.12559294871795e-06,
|
| 249511 |
+
"loss": 0.3154,
|
| 249512 |
+
"step": 118010
|
| 249513 |
+
},
|
| 249514 |
+
{
|
| 249515 |
+
"epoch": 943.35,
|
| 249516 |
+
"learning_rate": 8.125512820512822e-06,
|
| 249517 |
+
"loss": 0.358,
|
| 249518 |
+
"step": 118015
|
| 249519 |
+
},
|
| 249520 |
+
{
|
| 249521 |
+
"epoch": 943.39,
|
| 249522 |
+
"learning_rate": 8.125432692307693e-06,
|
| 249523 |
+
"loss": 0.9675,
|
| 249524 |
+
"step": 118020
|
| 249525 |
+
},
|
| 249526 |
+
{
|
| 249527 |
+
"epoch": 943.43,
|
| 249528 |
+
"learning_rate": 8.125352564102564e-06,
|
| 249529 |
+
"loss": 0.676,
|
| 249530 |
+
"step": 118025
|
| 249531 |
+
},
|
| 249532 |
+
{
|
| 249533 |
+
"epoch": 943.47,
|
| 249534 |
+
"learning_rate": 8.125272435897437e-06,
|
| 249535 |
+
"loss": 0.2939,
|
| 249536 |
+
"step": 118030
|
| 249537 |
+
},
|
| 249538 |
+
{
|
| 249539 |
+
"epoch": 943.51,
|
| 249540 |
+
"learning_rate": 8.125192307692307e-06,
|
| 249541 |
+
"loss": 0.3516,
|
| 249542 |
+
"step": 118035
|
| 249543 |
+
},
|
| 249544 |
+
{
|
| 249545 |
+
"epoch": 943.55,
|
| 249546 |
+
"learning_rate": 8.12511217948718e-06,
|
| 249547 |
+
"loss": 0.3801,
|
| 249548 |
+
"step": 118040
|
| 249549 |
+
},
|
| 249550 |
+
{
|
| 249551 |
+
"epoch": 943.59,
|
| 249552 |
+
"learning_rate": 8.125032051282053e-06,
|
| 249553 |
+
"loss": 0.8063,
|
| 249554 |
+
"step": 118045
|
| 249555 |
+
},
|
| 249556 |
+
{
|
| 249557 |
+
"epoch": 943.63,
|
| 249558 |
+
"learning_rate": 8.124951923076923e-06,
|
| 249559 |
+
"loss": 0.8844,
|
| 249560 |
+
"step": 118050
|
| 249561 |
+
},
|
| 249562 |
+
{
|
| 249563 |
+
"epoch": 943.67,
|
| 249564 |
+
"learning_rate": 8.124871794871796e-06,
|
| 249565 |
+
"loss": 0.3227,
|
| 249566 |
+
"step": 118055
|
| 249567 |
+
},
|
| 249568 |
+
{
|
| 249569 |
+
"epoch": 943.71,
|
| 249570 |
+
"learning_rate": 8.124791666666667e-06,
|
| 249571 |
+
"loss": 0.3623,
|
| 249572 |
+
"step": 118060
|
| 249573 |
+
},
|
| 249574 |
+
{
|
| 249575 |
+
"epoch": 943.75,
|
| 249576 |
+
"learning_rate": 8.124711538461539e-06,
|
| 249577 |
+
"loss": 0.3823,
|
| 249578 |
+
"step": 118065
|
| 249579 |
+
},
|
| 249580 |
+
{
|
| 249581 |
+
"epoch": 943.79,
|
| 249582 |
+
"learning_rate": 8.12463141025641e-06,
|
| 249583 |
+
"loss": 1.0083,
|
| 249584 |
+
"step": 118070
|
| 249585 |
+
},
|
| 249586 |
+
{
|
| 249587 |
+
"epoch": 943.83,
|
| 249588 |
+
"learning_rate": 8.124551282051283e-06,
|
| 249589 |
+
"loss": 0.5787,
|
| 249590 |
+
"step": 118075
|
| 249591 |
+
},
|
| 249592 |
+
{
|
| 249593 |
+
"epoch": 943.87,
|
| 249594 |
+
"learning_rate": 8.124471153846154e-06,
|
| 249595 |
+
"loss": 0.3008,
|
| 249596 |
+
"step": 118080
|
| 249597 |
+
},
|
| 249598 |
+
{
|
| 249599 |
+
"epoch": 943.91,
|
| 249600 |
+
"learning_rate": 8.124391025641026e-06,
|
| 249601 |
+
"loss": 0.3161,
|
| 249602 |
+
"step": 118085
|
| 249603 |
+
},
|
| 249604 |
+
{
|
| 249605 |
+
"epoch": 943.95,
|
| 249606 |
+
"learning_rate": 8.124310897435899e-06,
|
| 249607 |
+
"loss": 0.3868,
|
| 249608 |
+
"step": 118090
|
| 249609 |
+
},
|
| 249610 |
+
{
|
| 249611 |
+
"epoch": 943.99,
|
| 249612 |
+
"learning_rate": 8.12423076923077e-06,
|
| 249613 |
+
"loss": 0.8324,
|
| 249614 |
+
"step": 118095
|
| 249615 |
+
},
|
| 249616 |
+
{
|
| 249617 |
+
"epoch": 944.0,
|
| 249618 |
+
"eval_loss": 0.47608083486557007,
|
| 249619 |
+
"eval_runtime": 36.6003,
|
| 249620 |
+
"eval_samples_per_second": 22.923,
|
| 249621 |
+
"eval_steps_per_second": 0.738,
|
| 249622 |
+
"eval_wer": 0.1929916710437458,
|
| 249623 |
+
"step": 118096
|
| 249624 |
+
},
|
| 249625 |
+
{
|
| 249626 |
+
"epoch": 952.03,
|
| 249627 |
+
"learning_rate": 8.124150641025641e-06,
|
| 249628 |
+
"loss": 0.3001,
|
| 249629 |
+
"step": 118100
|
| 249630 |
+
},
|
| 249631 |
+
{
|
| 249632 |
+
"epoch": 952.07,
|
| 249633 |
+
"learning_rate": 8.124070512820513e-06,
|
| 249634 |
+
"loss": 0.2851,
|
| 249635 |
+
"step": 118105
|
| 249636 |
+
},
|
| 249637 |
+
{
|
| 249638 |
+
"epoch": 952.11,
|
| 249639 |
+
"learning_rate": 8.123990384615386e-06,
|
| 249640 |
+
"loss": 0.2918,
|
| 249641 |
+
"step": 118110
|
| 249642 |
+
},
|
| 249643 |
+
{
|
| 249644 |
+
"epoch": 952.15,
|
| 249645 |
+
"learning_rate": 8.123910256410257e-06,
|
| 249646 |
+
"loss": 0.3754,
|
| 249647 |
+
"step": 118115
|
| 249648 |
+
},
|
| 249649 |
+
{
|
| 249650 |
+
"epoch": 952.19,
|
| 249651 |
+
"learning_rate": 8.123830128205129e-06,
|
| 249652 |
+
"loss": 0.8105,
|
| 249653 |
+
"step": 118120
|
| 249654 |
+
},
|
| 249655 |
+
{
|
| 249656 |
+
"epoch": 952.23,
|
| 249657 |
+
"learning_rate": 8.12375e-06,
|
| 249658 |
+
"loss": 0.6714,
|
| 249659 |
+
"step": 118125
|
| 249660 |
+
},
|
| 249661 |
+
{
|
| 249662 |
+
"epoch": 952.27,
|
| 249663 |
+
"learning_rate": 8.123669871794873e-06,
|
| 249664 |
+
"loss": 0.3082,
|
| 249665 |
+
"step": 118130
|
| 249666 |
+
},
|
| 249667 |
+
{
|
| 249668 |
+
"epoch": 952.31,
|
| 249669 |
+
"learning_rate": 8.123589743589744e-06,
|
| 249670 |
+
"loss": 0.3015,
|
| 249671 |
+
"step": 118135
|
| 249672 |
+
},
|
| 249673 |
+
{
|
| 249674 |
+
"epoch": 952.35,
|
| 249675 |
+
"learning_rate": 8.123509615384616e-06,
|
| 249676 |
+
"loss": 0.4073,
|
| 249677 |
+
"step": 118140
|
| 249678 |
+
},
|
| 249679 |
+
{
|
| 249680 |
+
"epoch": 952.39,
|
| 249681 |
+
"learning_rate": 8.123429487179489e-06,
|
| 249682 |
+
"loss": 1.0138,
|
| 249683 |
+
"step": 118145
|
| 249684 |
+
},
|
| 249685 |
+
{
|
| 249686 |
+
"epoch": 952.43,
|
| 249687 |
+
"learning_rate": 8.12334935897436e-06,
|
| 249688 |
+
"loss": 0.6503,
|
| 249689 |
+
"step": 118150
|
| 249690 |
+
},
|
| 249691 |
+
{
|
| 249692 |
+
"epoch": 952.47,
|
| 249693 |
+
"learning_rate": 8.123269230769231e-06,
|
| 249694 |
+
"loss": 0.283,
|
| 249695 |
+
"step": 118155
|
| 249696 |
+
},
|
| 249697 |
+
{
|
| 249698 |
+
"epoch": 952.51,
|
| 249699 |
+
"learning_rate": 8.123189102564103e-06,
|
| 249700 |
+
"loss": 0.2965,
|
| 249701 |
+
"step": 118160
|
| 249702 |
+
},
|
| 249703 |
+
{
|
| 249704 |
+
"epoch": 952.55,
|
| 249705 |
+
"learning_rate": 8.123108974358976e-06,
|
| 249706 |
+
"loss": 0.4055,
|
| 249707 |
+
"step": 118165
|
| 249708 |
+
},
|
| 249709 |
+
{
|
| 249710 |
+
"epoch": 952.59,
|
| 249711 |
+
"learning_rate": 8.123028846153847e-06,
|
| 249712 |
+
"loss": 0.856,
|
| 249713 |
+
"step": 118170
|
| 249714 |
+
},
|
| 249715 |
+
{
|
| 249716 |
+
"epoch": 952.63,
|
| 249717 |
+
"learning_rate": 8.122948717948719e-06,
|
| 249718 |
+
"loss": 0.6568,
|
| 249719 |
+
"step": 118175
|
| 249720 |
+
},
|
| 249721 |
+
{
|
| 249722 |
+
"epoch": 952.67,
|
| 249723 |
+
"learning_rate": 8.12286858974359e-06,
|
| 249724 |
+
"loss": 0.3454,
|
| 249725 |
+
"step": 118180
|
| 249726 |
+
},
|
| 249727 |
+
{
|
| 249728 |
+
"epoch": 952.71,
|
| 249729 |
+
"learning_rate": 8.122788461538463e-06,
|
| 249730 |
+
"loss": 0.3327,
|
| 249731 |
+
"step": 118185
|
| 249732 |
+
},
|
| 249733 |
+
{
|
| 249734 |
+
"epoch": 952.76,
|
| 249735 |
+
"learning_rate": 8.122708333333334e-06,
|
| 249736 |
+
"loss": 0.3899,
|
| 249737 |
+
"step": 118190
|
| 249738 |
+
},
|
| 249739 |
+
{
|
| 249740 |
+
"epoch": 952.8,
|
| 249741 |
+
"learning_rate": 8.122628205128206e-06,
|
| 249742 |
+
"loss": 0.9436,
|
| 249743 |
+
"step": 118195
|
| 249744 |
+
},
|
| 249745 |
+
{
|
| 249746 |
+
"epoch": 952.84,
|
| 249747 |
+
"learning_rate": 8.122548076923079e-06,
|
| 249748 |
+
"loss": 0.5313,
|
| 249749 |
+
"step": 118200
|
| 249750 |
+
},
|
| 249751 |
+
{
|
| 249752 |
+
"epoch": 952.88,
|
| 249753 |
+
"learning_rate": 8.122467948717948e-06,
|
| 249754 |
+
"loss": 0.2523,
|
| 249755 |
+
"step": 118205
|
| 249756 |
+
},
|
| 249757 |
+
{
|
| 249758 |
+
"epoch": 952.92,
|
| 249759 |
+
"learning_rate": 8.122387820512822e-06,
|
| 249760 |
+
"loss": 0.3375,
|
| 249761 |
+
"step": 118210
|
| 249762 |
+
},
|
| 249763 |
+
{
|
| 249764 |
+
"epoch": 952.96,
|
| 249765 |
+
"learning_rate": 8.122307692307693e-06,
|
| 249766 |
+
"loss": 0.4248,
|
| 249767 |
+
"step": 118215
|
| 249768 |
+
},
|
| 249769 |
+
{
|
| 249770 |
+
"epoch": 953.0,
|
| 249771 |
+
"learning_rate": 8.122227564102564e-06,
|
| 249772 |
+
"loss": 0.9152,
|
| 249773 |
+
"step": 118220
|
| 249774 |
+
},
|
| 249775 |
+
{
|
| 249776 |
+
"epoch": 953.0,
|
| 249777 |
+
"eval_loss": 0.3905293643474579,
|
| 249778 |
+
"eval_runtime": 36.1162,
|
| 249779 |
+
"eval_samples_per_second": 23.258,
|
| 249780 |
+
"eval_steps_per_second": 0.748,
|
| 249781 |
+
"eval_wer": 0.18685816341717518,
|
| 249782 |
+
"step": 118220
|
| 249783 |
+
},
|
| 249784 |
+
{
|
| 249785 |
+
"epoch": 953.04,
|
| 249786 |
+
"learning_rate": 8.122147435897436e-06,
|
| 249787 |
+
"loss": 0.3212,
|
| 249788 |
+
"step": 118225
|
| 249789 |
+
},
|
| 249790 |
+
{
|
| 249791 |
+
"epoch": 953.08,
|
| 249792 |
+
"learning_rate": 8.122067307692309e-06,
|
| 249793 |
+
"loss": 0.3003,
|
| 249794 |
+
"step": 118230
|
| 249795 |
+
},
|
| 249796 |
+
{
|
| 249797 |
+
"epoch": 953.12,
|
| 249798 |
+
"learning_rate": 8.12198717948718e-06,
|
| 249799 |
+
"loss": 0.3147,
|
| 249800 |
+
"step": 118235
|
| 249801 |
+
},
|
| 249802 |
+
{
|
| 249803 |
+
"epoch": 953.16,
|
| 249804 |
+
"learning_rate": 8.121907051282051e-06,
|
| 249805 |
+
"loss": 0.4301,
|
| 249806 |
+
"step": 118240
|
| 249807 |
+
},
|
| 249808 |
+
{
|
| 249809 |
+
"epoch": 953.2,
|
| 249810 |
+
"learning_rate": 8.121826923076924e-06,
|
| 249811 |
+
"loss": 1.0591,
|
| 249812 |
+
"step": 118245
|
| 249813 |
+
},
|
| 249814 |
+
{
|
| 249815 |
+
"epoch": 953.24,
|
| 249816 |
+
"learning_rate": 8.121746794871796e-06,
|
| 249817 |
+
"loss": 0.3709,
|
| 249818 |
+
"step": 118250
|
| 249819 |
+
},
|
| 249820 |
+
{
|
| 249821 |
+
"epoch": 953.28,
|
| 249822 |
+
"learning_rate": 8.121666666666667e-06,
|
| 249823 |
+
"loss": 0.3037,
|
| 249824 |
+
"step": 118255
|
| 249825 |
+
},
|
| 249826 |
+
{
|
| 249827 |
+
"epoch": 953.32,
|
| 249828 |
+
"learning_rate": 8.121586538461538e-06,
|
| 249829 |
+
"loss": 0.3362,
|
| 249830 |
+
"step": 118260
|
| 249831 |
+
},
|
| 249832 |
+
{
|
| 249833 |
+
"epoch": 953.36,
|
| 249834 |
+
"learning_rate": 8.121506410256412e-06,
|
| 249835 |
+
"loss": 0.5091,
|
| 249836 |
+
"step": 118265
|
| 249837 |
+
},
|
| 249838 |
+
{
|
| 249839 |
+
"epoch": 953.4,
|
| 249840 |
+
"learning_rate": 8.121426282051283e-06,
|
| 249841 |
+
"loss": 1.252,
|
| 249842 |
+
"step": 118270
|
| 249843 |
+
},
|
| 249844 |
+
{
|
| 249845 |
+
"epoch": 953.44,
|
| 249846 |
+
"learning_rate": 8.121346153846154e-06,
|
| 249847 |
+
"loss": 0.3504,
|
| 249848 |
+
"step": 118275
|
| 249849 |
+
},
|
| 249850 |
+
{
|
| 249851 |
+
"epoch": 953.48,
|
| 249852 |
+
"learning_rate": 8.121266025641026e-06,
|
| 249853 |
+
"loss": 0.2573,
|
| 249854 |
+
"step": 118280
|
| 249855 |
+
},
|
| 249856 |
+
{
|
| 249857 |
+
"epoch": 953.52,
|
| 249858 |
+
"learning_rate": 8.121185897435899e-06,
|
| 249859 |
+
"loss": 0.3457,
|
| 249860 |
+
"step": 118285
|
| 249861 |
+
},
|
| 249862 |
+
{
|
| 249863 |
+
"epoch": 953.56,
|
| 249864 |
+
"learning_rate": 8.12110576923077e-06,
|
| 249865 |
+
"loss": 0.5181,
|
| 249866 |
+
"step": 118290
|
| 249867 |
+
},
|
| 249868 |
+
{
|
| 249869 |
+
"epoch": 953.6,
|
| 249870 |
+
"learning_rate": 8.121025641025641e-06,
|
| 249871 |
+
"loss": 1.2063,
|
| 249872 |
+
"step": 118295
|
| 249873 |
+
},
|
| 249874 |
+
{
|
| 249875 |
+
"epoch": 953.64,
|
| 249876 |
+
"learning_rate": 8.120945512820514e-06,
|
| 249877 |
+
"loss": 0.2867,
|
| 249878 |
+
"step": 118300
|
| 249879 |
+
},
|
| 249880 |
+
{
|
| 249881 |
+
"epoch": 953.68,
|
| 249882 |
+
"learning_rate": 8.120865384615386e-06,
|
| 249883 |
+
"loss": 0.3127,
|
| 249884 |
+
"step": 118305
|
| 249885 |
+
},
|
| 249886 |
+
{
|
| 249887 |
+
"epoch": 953.72,
|
| 249888 |
+
"learning_rate": 8.120785256410257e-06,
|
| 249889 |
+
"loss": 0.3992,
|
| 249890 |
+
"step": 118310
|
| 249891 |
+
},
|
| 249892 |
+
{
|
| 249893 |
+
"epoch": 953.76,
|
| 249894 |
+
"learning_rate": 8.120705128205129e-06,
|
| 249895 |
+
"loss": 0.4856,
|
| 249896 |
+
"step": 118315
|
| 249897 |
+
},
|
| 249898 |
+
{
|
| 249899 |
+
"epoch": 953.8,
|
| 249900 |
+
"learning_rate": 8.120625000000002e-06,
|
| 249901 |
+
"loss": 1.1835,
|
| 249902 |
+
"step": 118320
|
| 249903 |
+
},
|
| 249904 |
+
{
|
| 249905 |
+
"epoch": 953.84,
|
| 249906 |
+
"learning_rate": 8.120544871794871e-06,
|
| 249907 |
+
"loss": 0.2638,
|
| 249908 |
+
"step": 118325
|
| 249909 |
+
},
|
| 249910 |
+
{
|
| 249911 |
+
"epoch": 953.88,
|
| 249912 |
+
"learning_rate": 8.120464743589744e-06,
|
| 249913 |
+
"loss": 0.2925,
|
| 249914 |
+
"step": 118330
|
| 249915 |
+
},
|
| 249916 |
+
{
|
| 249917 |
+
"epoch": 953.92,
|
| 249918 |
+
"learning_rate": 8.120384615384616e-06,
|
| 249919 |
+
"loss": 0.2602,
|
| 249920 |
+
"step": 118335
|
| 249921 |
+
},
|
| 249922 |
+
{
|
| 249923 |
+
"epoch": 953.96,
|
| 249924 |
+
"learning_rate": 8.120304487179487e-06,
|
| 249925 |
+
"loss": 0.4767,
|
| 249926 |
+
"step": 118340
|
| 249927 |
+
},
|
| 249928 |
+
{
|
| 249929 |
+
"epoch": 954.0,
|
| 249930 |
+
"eval_loss": 0.5237769484519958,
|
| 249931 |
+
"eval_runtime": 37.4896,
|
| 249932 |
+
"eval_samples_per_second": 22.406,
|
| 249933 |
+
"eval_steps_per_second": 0.72,
|
| 249934 |
+
"eval_wer": 0.18330558169840006,
|
| 249935 |
+
"step": 118344
|
| 249936 |
+
},
|
| 249937 |
+
{
|
| 249938 |
+
"epoch": 954.01,
|
| 249939 |
+
"learning_rate": 8.12022435897436e-06,
|
| 249940 |
+
"loss": 0.3711,
|
| 249941 |
+
"step": 118345
|
| 249942 |
+
},
|
| 249943 |
+
{
|
| 249944 |
+
"epoch": 954.05,
|
| 249945 |
+
"learning_rate": 8.120144230769231e-06,
|
| 249946 |
+
"loss": 0.3008,
|
| 249947 |
+
"step": 118350
|
| 249948 |
+
},
|
| 249949 |
+
{
|
| 249950 |
+
"epoch": 954.09,
|
| 249951 |
+
"learning_rate": 8.120064102564103e-06,
|
| 249952 |
+
"loss": 0.2683,
|
| 249953 |
+
"step": 118355
|
| 249954 |
+
},
|
| 249955 |
+
{
|
| 249956 |
+
"epoch": 954.13,
|
| 249957 |
+
"learning_rate": 8.119983974358974e-06,
|
| 249958 |
+
"loss": 0.2991,
|
| 249959 |
+
"step": 118360
|
| 249960 |
+
},
|
| 249961 |
+
{
|
| 249962 |
+
"epoch": 954.17,
|
| 249963 |
+
"learning_rate": 8.119903846153847e-06,
|
| 249964 |
+
"loss": 0.4376,
|
| 249965 |
+
"step": 118365
|
| 249966 |
+
},
|
| 249967 |
+
{
|
| 249968 |
+
"epoch": 954.21,
|
| 249969 |
+
"learning_rate": 8.119823717948719e-06,
|
| 249970 |
+
"loss": 1.0462,
|
| 249971 |
+
"step": 118370
|
| 249972 |
+
},
|
| 249973 |
+
{
|
| 249974 |
+
"epoch": 954.25,
|
| 249975 |
+
"learning_rate": 8.11974358974359e-06,
|
| 249976 |
+
"loss": 0.2917,
|
| 249977 |
+
"step": 118375
|
| 249978 |
+
},
|
| 249979 |
+
{
|
| 249980 |
+
"epoch": 954.29,
|
| 249981 |
+
"learning_rate": 8.119663461538461e-06,
|
| 249982 |
+
"loss": 0.2586,
|
| 249983 |
+
"step": 118380
|
| 249984 |
+
},
|
| 249985 |
+
{
|
| 249986 |
+
"epoch": 954.33,
|
| 249987 |
+
"learning_rate": 8.119583333333334e-06,
|
| 249988 |
+
"loss": 0.32,
|
| 249989 |
+
"step": 118385
|
| 249990 |
+
},
|
| 249991 |
+
{
|
| 249992 |
+
"epoch": 954.37,
|
| 249993 |
+
"learning_rate": 8.119503205128206e-06,
|
| 249994 |
+
"loss": 0.5745,
|
| 249995 |
+
"step": 118390
|
| 249996 |
+
},
|
| 249997 |
+
{
|
| 249998 |
+
"epoch": 954.41,
|
| 249999 |
+
"learning_rate": 8.119423076923077e-06,
|
| 250000 |
+
"loss": 1.1579,
|
| 250001 |
+
"step": 118395
|
| 250002 |
+
},
|
| 250003 |
+
{
|
| 250004 |
+
"epoch": 954.45,
|
| 250005 |
+
"learning_rate": 8.11934294871795e-06,
|
| 250006 |
+
"loss": 0.2653,
|
| 250007 |
+
"step": 118400
|
| 250008 |
+
},
|
| 250009 |
+
{
|
| 250010 |
+
"epoch": 954.49,
|
| 250011 |
+
"learning_rate": 8.119262820512821e-06,
|
| 250012 |
+
"loss": 0.2799,
|
| 250013 |
+
"step": 118405
|
| 250014 |
+
},
|
| 250015 |
+
{
|
| 250016 |
+
"epoch": 954.53,
|
| 250017 |
+
"learning_rate": 8.119182692307693e-06,
|
| 250018 |
+
"loss": 0.317,
|
| 250019 |
+
"step": 118410
|
| 250020 |
+
},
|
| 250021 |
+
{
|
| 250022 |
+
"epoch": 954.57,
|
| 250023 |
+
"learning_rate": 8.119102564102564e-06,
|
| 250024 |
+
"loss": 0.524,
|
| 250025 |
+
"step": 118415
|
| 250026 |
+
},
|
| 250027 |
+
{
|
| 250028 |
+
"epoch": 954.61,
|
| 250029 |
+
"learning_rate": 8.119022435897437e-06,
|
| 250030 |
+
"loss": 1.0891,
|
| 250031 |
+
"step": 118420
|
| 250032 |
+
},
|
| 250033 |
+
{
|
| 250034 |
+
"epoch": 954.65,
|
| 250035 |
+
"learning_rate": 8.118942307692309e-06,
|
| 250036 |
+
"loss": 0.2812,
|
| 250037 |
+
"step": 118425
|
| 250038 |
+
},
|
| 250039 |
+
{
|
| 250040 |
+
"epoch": 954.69,
|
| 250041 |
+
"learning_rate": 8.11886217948718e-06,
|
| 250042 |
+
"loss": 0.3314,
|
| 250043 |
+
"step": 118430
|
| 250044 |
+
},
|
| 250045 |
+
{
|
| 250046 |
+
"epoch": 954.73,
|
| 250047 |
+
"learning_rate": 8.118782051282051e-06,
|
| 250048 |
+
"loss": 0.3237,
|
| 250049 |
+
"step": 118435
|
| 250050 |
+
},
|
| 250051 |
+
{
|
| 250052 |
+
"epoch": 954.77,
|
| 250053 |
+
"learning_rate": 8.118701923076924e-06,
|
| 250054 |
+
"loss": 0.6018,
|
| 250055 |
+
"step": 118440
|
| 250056 |
+
},
|
| 250057 |
+
{
|
| 250058 |
+
"epoch": 954.81,
|
| 250059 |
+
"learning_rate": 8.118621794871796e-06,
|
| 250060 |
+
"loss": 1.1353,
|
| 250061 |
+
"step": 118445
|
| 250062 |
+
},
|
| 250063 |
+
{
|
| 250064 |
+
"epoch": 954.85,
|
| 250065 |
+
"learning_rate": 8.118541666666667e-06,
|
| 250066 |
+
"loss": 0.2756,
|
| 250067 |
+
"step": 118450
|
| 250068 |
+
},
|
| 250069 |
+
{
|
| 250070 |
+
"epoch": 954.89,
|
| 250071 |
+
"learning_rate": 8.11846153846154e-06,
|
| 250072 |
+
"loss": 0.3134,
|
| 250073 |
+
"step": 118455
|
| 250074 |
+
},
|
| 250075 |
+
{
|
| 250076 |
+
"epoch": 954.93,
|
| 250077 |
+
"learning_rate": 8.118381410256411e-06,
|
| 250078 |
+
"loss": 0.2983,
|
| 250079 |
+
"step": 118460
|
| 250080 |
+
},
|
| 250081 |
+
{
|
| 250082 |
+
"epoch": 954.97,
|
| 250083 |
+
"learning_rate": 8.118301282051283e-06,
|
| 250084 |
+
"loss": 0.5291,
|
| 250085 |
+
"step": 118465
|
| 250086 |
+
},
|
| 250087 |
+
{
|
| 250088 |
+
"epoch": 955.0,
|
| 250089 |
+
"eval_loss": 0.39194896817207336,
|
| 250090 |
+
"eval_runtime": 36.8334,
|
| 250091 |
+
"eval_samples_per_second": 22.805,
|
| 250092 |
+
"eval_steps_per_second": 0.733,
|
| 250093 |
+
"eval_wer": 0.17496909315686132,
|
| 250094 |
+
"step": 118468
|
| 250095 |
+
},
|
| 250096 |
+
{
|
| 250097 |
+
"epoch": 955.02,
|
| 250098 |
+
"learning_rate": 8.118221153846154e-06,
|
| 250099 |
+
"loss": 0.3078,
|
| 250100 |
+
"step": 118470
|
| 250101 |
+
},
|
| 250102 |
+
{
|
| 250103 |
+
"epoch": 955.06,
|
| 250104 |
+
"learning_rate": 8.118141025641027e-06,
|
| 250105 |
+
"loss": 0.3107,
|
| 250106 |
+
"step": 118475
|
| 250107 |
+
},
|
| 250108 |
+
{
|
| 250109 |
+
"epoch": 955.1,
|
| 250110 |
+
"learning_rate": 8.118060897435897e-06,
|
| 250111 |
+
"loss": 0.2864,
|
| 250112 |
+
"step": 118480
|
| 250113 |
+
},
|
| 250114 |
+
{
|
| 250115 |
+
"epoch": 955.14,
|
| 250116 |
+
"learning_rate": 8.11798076923077e-06,
|
| 250117 |
+
"loss": 0.3148,
|
| 250118 |
+
"step": 118485
|
| 250119 |
+
},
|
| 250120 |
+
{
|
| 250121 |
+
"epoch": 955.18,
|
| 250122 |
+
"learning_rate": 8.117900641025643e-06,
|
| 250123 |
+
"loss": 0.5148,
|
| 250124 |
+
"step": 118490
|
| 250125 |
+
},
|
| 250126 |
+
{
|
| 250127 |
+
"epoch": 955.22,
|
| 250128 |
+
"learning_rate": 8.117820512820513e-06,
|
| 250129 |
+
"loss": 0.8883,
|
| 250130 |
+
"step": 118495
|
| 250131 |
+
},
|
| 250132 |
+
{
|
| 250133 |
+
"epoch": 955.26,
|
| 250134 |
+
"learning_rate": 8.117740384615386e-06,
|
| 250135 |
+
"loss": 0.3215,
|
| 250136 |
+
"step": 118500
|
| 250137 |
+
},
|
| 250138 |
+
{
|
| 250139 |
+
"epoch": 955.3,
|
| 250140 |
+
"learning_rate": 8.117660256410257e-06,
|
| 250141 |
+
"loss": 0.2788,
|
| 250142 |
+
"step": 118505
|
| 250143 |
+
},
|
| 250144 |
+
{
|
| 250145 |
+
"epoch": 955.34,
|
| 250146 |
+
"learning_rate": 8.117580128205128e-06,
|
| 250147 |
+
"loss": 0.4172,
|
| 250148 |
+
"step": 118510
|
| 250149 |
+
},
|
| 250150 |
+
{
|
| 250151 |
+
"epoch": 955.38,
|
| 250152 |
+
"learning_rate": 8.1175e-06,
|
| 250153 |
+
"loss": 0.7161,
|
| 250154 |
+
"step": 118515
|
| 250155 |
+
},
|
| 250156 |
+
{
|
| 250157 |
+
"epoch": 955.42,
|
| 250158 |
+
"learning_rate": 8.117419871794873e-06,
|
| 250159 |
+
"loss": 1.0977,
|
| 250160 |
+
"step": 118520
|
| 250161 |
+
},
|
| 250162 |
+
{
|
| 250163 |
+
"epoch": 955.46,
|
| 250164 |
+
"learning_rate": 8.117339743589744e-06,
|
| 250165 |
+
"loss": 0.2489,
|
| 250166 |
+
"step": 118525
|
| 250167 |
+
},
|
| 250168 |
+
{
|
| 250169 |
+
"epoch": 955.5,
|
| 250170 |
+
"learning_rate": 8.117259615384616e-06,
|
| 250171 |
+
"loss": 0.2679,
|
| 250172 |
+
"step": 118530
|
| 250173 |
+
},
|
| 250174 |
+
{
|
| 250175 |
+
"epoch": 955.54,
|
| 250176 |
+
"learning_rate": 8.117179487179487e-06,
|
| 250177 |
+
"loss": 0.311,
|
| 250178 |
+
"step": 118535
|
| 250179 |
+
},
|
| 250180 |
+
{
|
| 250181 |
+
"epoch": 955.58,
|
| 250182 |
+
"learning_rate": 8.11709935897436e-06,
|
| 250183 |
+
"loss": 0.6199,
|
| 250184 |
+
"step": 118540
|
| 250185 |
+
},
|
| 250186 |
+
{
|
| 250187 |
+
"epoch": 955.62,
|
| 250188 |
+
"learning_rate": 8.117019230769231e-06,
|
| 250189 |
+
"loss": 1.0061,
|
| 250190 |
+
"step": 118545
|
| 250191 |
+
},
|
| 250192 |
+
{
|
| 250193 |
+
"epoch": 955.66,
|
| 250194 |
+
"learning_rate": 8.116939102564103e-06,
|
| 250195 |
+
"loss": 0.2723,
|
| 250196 |
+
"step": 118550
|
| 250197 |
+
},
|
| 250198 |
+
{
|
| 250199 |
+
"epoch": 955.7,
|
| 250200 |
+
"learning_rate": 8.116858974358976e-06,
|
| 250201 |
+
"loss": 0.3127,
|
| 250202 |
+
"step": 118555
|
| 250203 |
+
},
|
| 250204 |
+
{
|
| 250205 |
+
"epoch": 955.74,
|
| 250206 |
+
"learning_rate": 8.116778846153847e-06,
|
| 250207 |
+
"loss": 0.3329,
|
| 250208 |
+
"step": 118560
|
| 250209 |
+
},
|
| 250210 |
+
{
|
| 250211 |
+
"epoch": 955.78,
|
| 250212 |
+
"learning_rate": 8.116698717948718e-06,
|
| 250213 |
+
"loss": 0.6651,
|
| 250214 |
+
"step": 118565
|
| 250215 |
+
},
|
| 250216 |
+
{
|
| 250217 |
+
"epoch": 955.82,
|
| 250218 |
+
"learning_rate": 8.11661858974359e-06,
|
| 250219 |
+
"loss": 0.9592,
|
| 250220 |
+
"step": 118570
|
| 250221 |
+
},
|
| 250222 |
+
{
|
| 250223 |
+
"epoch": 955.86,
|
| 250224 |
+
"learning_rate": 8.116538461538463e-06,
|
| 250225 |
+
"loss": 0.3461,
|
| 250226 |
+
"step": 118575
|
| 250227 |
+
},
|
| 250228 |
+
{
|
| 250229 |
+
"epoch": 955.9,
|
| 250230 |
+
"learning_rate": 8.116458333333334e-06,
|
| 250231 |
+
"loss": 0.3026,
|
| 250232 |
+
"step": 118580
|
| 250233 |
+
},
|
| 250234 |
+
{
|
| 250235 |
+
"epoch": 955.94,
|
| 250236 |
+
"learning_rate": 8.116378205128206e-06,
|
| 250237 |
+
"loss": 0.3226,
|
| 250238 |
+
"step": 118585
|
| 250239 |
+
},
|
| 250240 |
+
{
|
| 250241 |
+
"epoch": 955.98,
|
| 250242 |
+
"learning_rate": 8.116298076923079e-06,
|
| 250243 |
+
"loss": 0.7119,
|
| 250244 |
+
"step": 118590
|
| 250245 |
+
},
|
| 250246 |
+
{
|
| 250247 |
+
"epoch": 956.0,
|
| 250248 |
+
"eval_loss": 0.450756698846817,
|
| 250249 |
+
"eval_runtime": 36.2725,
|
| 250250 |
+
"eval_samples_per_second": 23.13,
|
| 250251 |
+
"eval_steps_per_second": 0.744,
|
| 250252 |
+
"eval_wer": 0.19291367167646845,
|
| 250253 |
+
"step": 118592
|
| 250254 |
}
|
| 250255 |
],
|
| 250256 |
+
"max_steps": 620000,
|
| 250257 |
"num_train_epochs": 5000,
|
| 250258 |
+
"total_flos": 3.337482219007906e+20,
|
| 250259 |
"trial_name": null,
|
| 250260 |
"trial_params": null
|
| 250261 |
}
|
model-bin/finetune/base/{checkpoint-117722 → checkpoint-118592}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630120105.1763353/events.out.tfevents.1630120105.86bb0ddabf9b.1042.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1fd411888b605a4bce813c5a78a4aa71068cafffeb22509913b5b7e1b4008d09
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630120512.9155197/events.out.tfevents.1630120512.86bb0ddabf9b.1042.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d42dc9fe90f56f7df13a3a3d032acddc5d00dbac73340c9ac699ba0f09707501
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630122518.2177079/events.out.tfevents.1630122518.86bb0ddabf9b.4092.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6588e586958501de11ff0888ba2564d38e6d1733a041463d675164f2379ce878
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630122917.3648393/events.out.tfevents.1630122917.86bb0ddabf9b.4092.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8403ff0f34403c9492ffc6d3f681dfc73a9b39812bf0269b7e04e2700de8b658
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630123316.325624/events.out.tfevents.1630123316.86bb0ddabf9b.4092.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bd80980e3f96a26ecb60c46c327402515f8e7be256c342189b987105729d217
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630123712.2616467/events.out.tfevents.1630123712.86bb0ddabf9b.4092.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a0c6e6a4214dd36c6164af3529a289edf05223f9f1ecb89d7c006b4931c4ccea
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630124107.8066275/events.out.tfevents.1630124107.86bb0ddabf9b.4092.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:466ee521d8c1ee5da1c2626835d25c15032cd2021e61e89ea92f4e81ea163c8e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630120105.86bb0ddabf9b.1042.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ac3a79215d1dbb0818235f26b5a4a7423c498e1eda10a620a47200579974ea7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630120512.86bb0ddabf9b.1042.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f179f8e97a3037a75d368e022645f69f8a027815c8ff6a57cacaf8918135d44
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630122518.86bb0ddabf9b.4092.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c5ebe7fe919e08c99e09e92f23f1a35efd76edd491d3531a76aaa88dcbf8a5a5
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1630122917.86bb0ddabf9b.4092.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e7df4eaed912c7fdeaac75d6f5a4059e2d33f9879077b9d4f5b4c3e94fa22aaa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630123316.86bb0ddabf9b.4092.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68c8705b0f637ddc66a4231d385f6c0f252fd31ea76d254e961b4097ecab43a7
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630123712.86bb0ddabf9b.4092.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:473c4f96f82c1568c0b7f38b870f6e27f7cf39a1db2b97720035669fdce0ad81
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630124107.86bb0ddabf9b.4092.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f58b9291dc819badcbac31ec6c4430e209a9890f0468938fd443cb67aaae2474
|
| 3 |
+
size 8622
|