"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/trainer_state.json +1511 -5
- model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629501847.633379/events.out.tfevents.1629501847.35c5ee475cfb.1049.1 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629501847.35c5ee475cfb.1049.0 +3 -0
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91f9bcba86de42b4153d909623f25e249d240c0ef350959b16f9c15de8ab54ce
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/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:dc48e24bdea4aa6e9567feadd468568aa6c0cf12ca5446336dc464e0f8e18a3a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/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:b369545a9503f4cc8000185be4bdd35fd2089613f6ee743e7f2721fc54ab603c
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/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:ce3810148dc64fc4ee35fde35dce8c39633c05854459451e2d7e9e6b8eeea1c9
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/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:bcbd0f5e8286f9c7dfd4aaef1552e076fd8458d4e8907633db212a3051d69fbb
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/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,
|
|
@@ -75546,11 +75546,1517 @@
|
|
| 75546 |
"eval_steps_per_second": 0.704,
|
| 75547 |
"eval_wer": 0.3300044716053063,
|
| 75548 |
"step": 12440
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75549 |
}
|
| 75550 |
],
|
| 75551 |
"max_steps": 620000,
|
| 75552 |
"num_train_epochs": 5000,
|
| 75553 |
-
"total_flos": 3.
|
| 75554 |
"trial_name": null,
|
| 75555 |
"trial_params": null
|
| 75556 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.3129266521423384,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-12564",
|
| 4 |
+
"epoch": 101.99598393574297,
|
| 5 |
+
"global_step": 12688,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 75546 |
"eval_steps_per_second": 0.704,
|
| 75547 |
"eval_wer": 0.3300044716053063,
|
| 75548 |
"step": 12440
|
| 75549 |
+
},
|
| 75550 |
+
{
|
| 75551 |
+
"epoch": 100.01,
|
| 75552 |
+
"learning_rate": 9.815266558966076e-06,
|
| 75553 |
+
"loss": 0.8036,
|
| 75554 |
+
"step": 12441
|
| 75555 |
+
},
|
| 75556 |
+
{
|
| 75557 |
+
"epoch": 100.02,
|
| 75558 |
+
"learning_rate": 9.815250403877223e-06,
|
| 75559 |
+
"loss": 0.6576,
|
| 75560 |
+
"step": 12442
|
| 75561 |
+
},
|
| 75562 |
+
{
|
| 75563 |
+
"epoch": 100.02,
|
| 75564 |
+
"learning_rate": 9.81523424878837e-06,
|
| 75565 |
+
"loss": 0.7578,
|
| 75566 |
+
"step": 12443
|
| 75567 |
+
},
|
| 75568 |
+
{
|
| 75569 |
+
"epoch": 100.03,
|
| 75570 |
+
"learning_rate": 9.815218093699516e-06,
|
| 75571 |
+
"loss": 0.6257,
|
| 75572 |
+
"step": 12444
|
| 75573 |
+
},
|
| 75574 |
+
{
|
| 75575 |
+
"epoch": 100.04,
|
| 75576 |
+
"learning_rate": 9.815201938610663e-06,
|
| 75577 |
+
"loss": 1.417,
|
| 75578 |
+
"step": 12445
|
| 75579 |
+
},
|
| 75580 |
+
{
|
| 75581 |
+
"epoch": 100.05,
|
| 75582 |
+
"learning_rate": 9.81518578352181e-06,
|
| 75583 |
+
"loss": 0.6262,
|
| 75584 |
+
"step": 12446
|
| 75585 |
+
},
|
| 75586 |
+
{
|
| 75587 |
+
"epoch": 100.06,
|
| 75588 |
+
"learning_rate": 9.815169628432957e-06,
|
| 75589 |
+
"loss": 1.4295,
|
| 75590 |
+
"step": 12447
|
| 75591 |
+
},
|
| 75592 |
+
{
|
| 75593 |
+
"epoch": 100.06,
|
| 75594 |
+
"learning_rate": 9.815153473344105e-06,
|
| 75595 |
+
"loss": 0.7108,
|
| 75596 |
+
"step": 12448
|
| 75597 |
+
},
|
| 75598 |
+
{
|
| 75599 |
+
"epoch": 100.07,
|
| 75600 |
+
"learning_rate": 9.815137318255252e-06,
|
| 75601 |
+
"loss": 0.5995,
|
| 75602 |
+
"step": 12449
|
| 75603 |
+
},
|
| 75604 |
+
{
|
| 75605 |
+
"epoch": 100.08,
|
| 75606 |
+
"learning_rate": 9.815121163166397e-06,
|
| 75607 |
+
"loss": 0.8812,
|
| 75608 |
+
"step": 12450
|
| 75609 |
+
},
|
| 75610 |
+
{
|
| 75611 |
+
"epoch": 100.09,
|
| 75612 |
+
"learning_rate": 9.815105008077546e-06,
|
| 75613 |
+
"loss": 0.7275,
|
| 75614 |
+
"step": 12451
|
| 75615 |
+
},
|
| 75616 |
+
{
|
| 75617 |
+
"epoch": 100.1,
|
| 75618 |
+
"learning_rate": 9.815088852988692e-06,
|
| 75619 |
+
"loss": 0.6411,
|
| 75620 |
+
"step": 12452
|
| 75621 |
+
},
|
| 75622 |
+
{
|
| 75623 |
+
"epoch": 100.1,
|
| 75624 |
+
"learning_rate": 9.81507269789984e-06,
|
| 75625 |
+
"loss": 0.7317,
|
| 75626 |
+
"step": 12453
|
| 75627 |
+
},
|
| 75628 |
+
{
|
| 75629 |
+
"epoch": 100.11,
|
| 75630 |
+
"learning_rate": 9.815056542810986e-06,
|
| 75631 |
+
"loss": 1.1441,
|
| 75632 |
+
"step": 12454
|
| 75633 |
+
},
|
| 75634 |
+
{
|
| 75635 |
+
"epoch": 100.12,
|
| 75636 |
+
"learning_rate": 9.815040387722133e-06,
|
| 75637 |
+
"loss": 0.7951,
|
| 75638 |
+
"step": 12455
|
| 75639 |
+
},
|
| 75640 |
+
{
|
| 75641 |
+
"epoch": 100.13,
|
| 75642 |
+
"learning_rate": 9.81502423263328e-06,
|
| 75643 |
+
"loss": 0.7124,
|
| 75644 |
+
"step": 12456
|
| 75645 |
+
},
|
| 75646 |
+
{
|
| 75647 |
+
"epoch": 100.14,
|
| 75648 |
+
"learning_rate": 9.815008077544427e-06,
|
| 75649 |
+
"loss": 0.7847,
|
| 75650 |
+
"step": 12457
|
| 75651 |
+
},
|
| 75652 |
+
{
|
| 75653 |
+
"epoch": 100.14,
|
| 75654 |
+
"learning_rate": 9.814991922455575e-06,
|
| 75655 |
+
"loss": 0.9751,
|
| 75656 |
+
"step": 12458
|
| 75657 |
+
},
|
| 75658 |
+
{
|
| 75659 |
+
"epoch": 100.15,
|
| 75660 |
+
"learning_rate": 9.814975767366722e-06,
|
| 75661 |
+
"loss": 0.9651,
|
| 75662 |
+
"step": 12459
|
| 75663 |
+
},
|
| 75664 |
+
{
|
| 75665 |
+
"epoch": 100.16,
|
| 75666 |
+
"learning_rate": 9.814959612277867e-06,
|
| 75667 |
+
"loss": 1.1636,
|
| 75668 |
+
"step": 12460
|
| 75669 |
+
},
|
| 75670 |
+
{
|
| 75671 |
+
"epoch": 100.17,
|
| 75672 |
+
"learning_rate": 9.814943457189016e-06,
|
| 75673 |
+
"loss": 1.6023,
|
| 75674 |
+
"step": 12461
|
| 75675 |
+
},
|
| 75676 |
+
{
|
| 75677 |
+
"epoch": 100.18,
|
| 75678 |
+
"learning_rate": 9.814927302100162e-06,
|
| 75679 |
+
"loss": 1.4563,
|
| 75680 |
+
"step": 12462
|
| 75681 |
+
},
|
| 75682 |
+
{
|
| 75683 |
+
"epoch": 100.18,
|
| 75684 |
+
"learning_rate": 9.81491114701131e-06,
|
| 75685 |
+
"loss": 1.6947,
|
| 75686 |
+
"step": 12463
|
| 75687 |
+
},
|
| 75688 |
+
{
|
| 75689 |
+
"epoch": 100.19,
|
| 75690 |
+
"learning_rate": 9.814894991922456e-06,
|
| 75691 |
+
"loss": 1.8084,
|
| 75692 |
+
"step": 12464
|
| 75693 |
+
},
|
| 75694 |
+
{
|
| 75695 |
+
"epoch": 100.2,
|
| 75696 |
+
"learning_rate": 9.814878836833603e-06,
|
| 75697 |
+
"loss": 2.32,
|
| 75698 |
+
"step": 12465
|
| 75699 |
+
},
|
| 75700 |
+
{
|
| 75701 |
+
"epoch": 100.21,
|
| 75702 |
+
"learning_rate": 9.81486268174475e-06,
|
| 75703 |
+
"loss": 0.9217,
|
| 75704 |
+
"step": 12466
|
| 75705 |
+
},
|
| 75706 |
+
{
|
| 75707 |
+
"epoch": 100.22,
|
| 75708 |
+
"learning_rate": 9.814846526655898e-06,
|
| 75709 |
+
"loss": 0.7915,
|
| 75710 |
+
"step": 12467
|
| 75711 |
+
},
|
| 75712 |
+
{
|
| 75713 |
+
"epoch": 100.22,
|
| 75714 |
+
"learning_rate": 9.814830371567045e-06,
|
| 75715 |
+
"loss": 0.6502,
|
| 75716 |
+
"step": 12468
|
| 75717 |
+
},
|
| 75718 |
+
{
|
| 75719 |
+
"epoch": 100.23,
|
| 75720 |
+
"learning_rate": 9.814814216478192e-06,
|
| 75721 |
+
"loss": 0.7164,
|
| 75722 |
+
"step": 12469
|
| 75723 |
+
},
|
| 75724 |
+
{
|
| 75725 |
+
"epoch": 100.24,
|
| 75726 |
+
"learning_rate": 9.814798061389339e-06,
|
| 75727 |
+
"loss": 0.5594,
|
| 75728 |
+
"step": 12470
|
| 75729 |
+
},
|
| 75730 |
+
{
|
| 75731 |
+
"epoch": 100.25,
|
| 75732 |
+
"learning_rate": 9.814781906300486e-06,
|
| 75733 |
+
"loss": 0.9229,
|
| 75734 |
+
"step": 12471
|
| 75735 |
+
},
|
| 75736 |
+
{
|
| 75737 |
+
"epoch": 100.26,
|
| 75738 |
+
"learning_rate": 9.814765751211632e-06,
|
| 75739 |
+
"loss": 0.61,
|
| 75740 |
+
"step": 12472
|
| 75741 |
+
},
|
| 75742 |
+
{
|
| 75743 |
+
"epoch": 100.27,
|
| 75744 |
+
"learning_rate": 9.81474959612278e-06,
|
| 75745 |
+
"loss": 0.5965,
|
| 75746 |
+
"step": 12473
|
| 75747 |
+
},
|
| 75748 |
+
{
|
| 75749 |
+
"epoch": 100.27,
|
| 75750 |
+
"learning_rate": 9.814733441033926e-06,
|
| 75751 |
+
"loss": 0.7654,
|
| 75752 |
+
"step": 12474
|
| 75753 |
+
},
|
| 75754 |
+
{
|
| 75755 |
+
"epoch": 100.28,
|
| 75756 |
+
"learning_rate": 9.814717285945073e-06,
|
| 75757 |
+
"loss": 0.7178,
|
| 75758 |
+
"step": 12475
|
| 75759 |
+
},
|
| 75760 |
+
{
|
| 75761 |
+
"epoch": 100.29,
|
| 75762 |
+
"learning_rate": 9.81470113085622e-06,
|
| 75763 |
+
"loss": 0.5147,
|
| 75764 |
+
"step": 12476
|
| 75765 |
+
},
|
| 75766 |
+
{
|
| 75767 |
+
"epoch": 100.3,
|
| 75768 |
+
"learning_rate": 9.814684975767368e-06,
|
| 75769 |
+
"loss": 0.6269,
|
| 75770 |
+
"step": 12477
|
| 75771 |
+
},
|
| 75772 |
+
{
|
| 75773 |
+
"epoch": 100.31,
|
| 75774 |
+
"learning_rate": 9.814668820678515e-06,
|
| 75775 |
+
"loss": 0.7695,
|
| 75776 |
+
"step": 12478
|
| 75777 |
+
},
|
| 75778 |
+
{
|
| 75779 |
+
"epoch": 100.31,
|
| 75780 |
+
"learning_rate": 9.814652665589662e-06,
|
| 75781 |
+
"loss": 0.7091,
|
| 75782 |
+
"step": 12479
|
| 75783 |
+
},
|
| 75784 |
+
{
|
| 75785 |
+
"epoch": 100.32,
|
| 75786 |
+
"learning_rate": 9.814636510500809e-06,
|
| 75787 |
+
"loss": 0.7882,
|
| 75788 |
+
"step": 12480
|
| 75789 |
+
},
|
| 75790 |
+
{
|
| 75791 |
+
"epoch": 100.33,
|
| 75792 |
+
"learning_rate": 9.814620355411956e-06,
|
| 75793 |
+
"loss": 0.7148,
|
| 75794 |
+
"step": 12481
|
| 75795 |
+
},
|
| 75796 |
+
{
|
| 75797 |
+
"epoch": 100.34,
|
| 75798 |
+
"learning_rate": 9.814604200323102e-06,
|
| 75799 |
+
"loss": 0.8351,
|
| 75800 |
+
"step": 12482
|
| 75801 |
+
},
|
| 75802 |
+
{
|
| 75803 |
+
"epoch": 100.35,
|
| 75804 |
+
"learning_rate": 9.814588045234251e-06,
|
| 75805 |
+
"loss": 1.0891,
|
| 75806 |
+
"step": 12483
|
| 75807 |
+
},
|
| 75808 |
+
{
|
| 75809 |
+
"epoch": 100.35,
|
| 75810 |
+
"learning_rate": 9.814571890145396e-06,
|
| 75811 |
+
"loss": 1.3347,
|
| 75812 |
+
"step": 12484
|
| 75813 |
+
},
|
| 75814 |
+
{
|
| 75815 |
+
"epoch": 100.36,
|
| 75816 |
+
"learning_rate": 9.814555735056543e-06,
|
| 75817 |
+
"loss": 1.0697,
|
| 75818 |
+
"step": 12485
|
| 75819 |
+
},
|
| 75820 |
+
{
|
| 75821 |
+
"epoch": 100.37,
|
| 75822 |
+
"learning_rate": 9.81453957996769e-06,
|
| 75823 |
+
"loss": 1.052,
|
| 75824 |
+
"step": 12486
|
| 75825 |
+
},
|
| 75826 |
+
{
|
| 75827 |
+
"epoch": 100.38,
|
| 75828 |
+
"learning_rate": 9.814523424878838e-06,
|
| 75829 |
+
"loss": 1.3821,
|
| 75830 |
+
"step": 12487
|
| 75831 |
+
},
|
| 75832 |
+
{
|
| 75833 |
+
"epoch": 100.39,
|
| 75834 |
+
"learning_rate": 9.814507269789985e-06,
|
| 75835 |
+
"loss": 1.494,
|
| 75836 |
+
"step": 12488
|
| 75837 |
+
},
|
| 75838 |
+
{
|
| 75839 |
+
"epoch": 100.39,
|
| 75840 |
+
"learning_rate": 9.814491114701132e-06,
|
| 75841 |
+
"loss": 1.7524,
|
| 75842 |
+
"step": 12489
|
| 75843 |
+
},
|
| 75844 |
+
{
|
| 75845 |
+
"epoch": 100.4,
|
| 75846 |
+
"learning_rate": 9.814474959612279e-06,
|
| 75847 |
+
"loss": 2.5587,
|
| 75848 |
+
"step": 12490
|
| 75849 |
+
},
|
| 75850 |
+
{
|
| 75851 |
+
"epoch": 100.41,
|
| 75852 |
+
"learning_rate": 9.814458804523426e-06,
|
| 75853 |
+
"loss": 0.836,
|
| 75854 |
+
"step": 12491
|
| 75855 |
+
},
|
| 75856 |
+
{
|
| 75857 |
+
"epoch": 100.42,
|
| 75858 |
+
"learning_rate": 9.814442649434572e-06,
|
| 75859 |
+
"loss": 1.004,
|
| 75860 |
+
"step": 12492
|
| 75861 |
+
},
|
| 75862 |
+
{
|
| 75863 |
+
"epoch": 100.43,
|
| 75864 |
+
"learning_rate": 9.814426494345721e-06,
|
| 75865 |
+
"loss": 1.4414,
|
| 75866 |
+
"step": 12493
|
| 75867 |
+
},
|
| 75868 |
+
{
|
| 75869 |
+
"epoch": 100.43,
|
| 75870 |
+
"learning_rate": 9.814410339256866e-06,
|
| 75871 |
+
"loss": 0.6718,
|
| 75872 |
+
"step": 12494
|
| 75873 |
+
},
|
| 75874 |
+
{
|
| 75875 |
+
"epoch": 100.44,
|
| 75876 |
+
"learning_rate": 9.814394184168013e-06,
|
| 75877 |
+
"loss": 0.6228,
|
| 75878 |
+
"step": 12495
|
| 75879 |
+
},
|
| 75880 |
+
{
|
| 75881 |
+
"epoch": 100.45,
|
| 75882 |
+
"learning_rate": 9.814378029079161e-06,
|
| 75883 |
+
"loss": 0.5357,
|
| 75884 |
+
"step": 12496
|
| 75885 |
+
},
|
| 75886 |
+
{
|
| 75887 |
+
"epoch": 100.46,
|
| 75888 |
+
"learning_rate": 9.814361873990308e-06,
|
| 75889 |
+
"loss": 0.6663,
|
| 75890 |
+
"step": 12497
|
| 75891 |
+
},
|
| 75892 |
+
{
|
| 75893 |
+
"epoch": 100.47,
|
| 75894 |
+
"learning_rate": 9.814345718901455e-06,
|
| 75895 |
+
"loss": 0.6478,
|
| 75896 |
+
"step": 12498
|
| 75897 |
+
},
|
| 75898 |
+
{
|
| 75899 |
+
"epoch": 100.47,
|
| 75900 |
+
"learning_rate": 9.814329563812602e-06,
|
| 75901 |
+
"loss": 0.714,
|
| 75902 |
+
"step": 12499
|
| 75903 |
+
},
|
| 75904 |
+
{
|
| 75905 |
+
"epoch": 100.48,
|
| 75906 |
+
"learning_rate": 9.814313408723749e-06,
|
| 75907 |
+
"loss": 0.7126,
|
| 75908 |
+
"step": 12500
|
| 75909 |
+
},
|
| 75910 |
+
{
|
| 75911 |
+
"epoch": 100.49,
|
| 75912 |
+
"learning_rate": 9.814297253634896e-06,
|
| 75913 |
+
"loss": 0.8535,
|
| 75914 |
+
"step": 12501
|
| 75915 |
+
},
|
| 75916 |
+
{
|
| 75917 |
+
"epoch": 100.5,
|
| 75918 |
+
"learning_rate": 9.814281098546042e-06,
|
| 75919 |
+
"loss": 0.8269,
|
| 75920 |
+
"step": 12502
|
| 75921 |
+
},
|
| 75922 |
+
{
|
| 75923 |
+
"epoch": 100.51,
|
| 75924 |
+
"learning_rate": 9.814264943457191e-06,
|
| 75925 |
+
"loss": 0.8202,
|
| 75926 |
+
"step": 12503
|
| 75927 |
+
},
|
| 75928 |
+
{
|
| 75929 |
+
"epoch": 100.51,
|
| 75930 |
+
"learning_rate": 9.814248788368336e-06,
|
| 75931 |
+
"loss": 0.8583,
|
| 75932 |
+
"step": 12504
|
| 75933 |
+
},
|
| 75934 |
+
{
|
| 75935 |
+
"epoch": 100.52,
|
| 75936 |
+
"learning_rate": 9.814232633279483e-06,
|
| 75937 |
+
"loss": 0.888,
|
| 75938 |
+
"step": 12505
|
| 75939 |
+
},
|
| 75940 |
+
{
|
| 75941 |
+
"epoch": 100.53,
|
| 75942 |
+
"learning_rate": 9.814216478190631e-06,
|
| 75943 |
+
"loss": 0.8864,
|
| 75944 |
+
"step": 12506
|
| 75945 |
+
},
|
| 75946 |
+
{
|
| 75947 |
+
"epoch": 100.54,
|
| 75948 |
+
"learning_rate": 9.814200323101778e-06,
|
| 75949 |
+
"loss": 0.76,
|
| 75950 |
+
"step": 12507
|
| 75951 |
+
},
|
| 75952 |
+
{
|
| 75953 |
+
"epoch": 100.55,
|
| 75954 |
+
"learning_rate": 9.814184168012925e-06,
|
| 75955 |
+
"loss": 0.896,
|
| 75956 |
+
"step": 12508
|
| 75957 |
+
},
|
| 75958 |
+
{
|
| 75959 |
+
"epoch": 100.55,
|
| 75960 |
+
"learning_rate": 9.814168012924072e-06,
|
| 75961 |
+
"loss": 1.3523,
|
| 75962 |
+
"step": 12509
|
| 75963 |
+
},
|
| 75964 |
+
{
|
| 75965 |
+
"epoch": 100.56,
|
| 75966 |
+
"learning_rate": 9.814151857835219e-06,
|
| 75967 |
+
"loss": 1.3223,
|
| 75968 |
+
"step": 12510
|
| 75969 |
+
},
|
| 75970 |
+
{
|
| 75971 |
+
"epoch": 100.57,
|
| 75972 |
+
"learning_rate": 9.814135702746366e-06,
|
| 75973 |
+
"loss": 1.3627,
|
| 75974 |
+
"step": 12511
|
| 75975 |
+
},
|
| 75976 |
+
{
|
| 75977 |
+
"epoch": 100.58,
|
| 75978 |
+
"learning_rate": 9.814119547657512e-06,
|
| 75979 |
+
"loss": 1.3998,
|
| 75980 |
+
"step": 12512
|
| 75981 |
+
},
|
| 75982 |
+
{
|
| 75983 |
+
"epoch": 100.59,
|
| 75984 |
+
"learning_rate": 9.814103392568661e-06,
|
| 75985 |
+
"loss": 1.8094,
|
| 75986 |
+
"step": 12513
|
| 75987 |
+
},
|
| 75988 |
+
{
|
| 75989 |
+
"epoch": 100.59,
|
| 75990 |
+
"learning_rate": 9.814087237479808e-06,
|
| 75991 |
+
"loss": 1.9263,
|
| 75992 |
+
"step": 12514
|
| 75993 |
+
},
|
| 75994 |
+
{
|
| 75995 |
+
"epoch": 100.6,
|
| 75996 |
+
"learning_rate": 9.814071082390953e-06,
|
| 75997 |
+
"loss": 2.5968,
|
| 75998 |
+
"step": 12515
|
| 75999 |
+
},
|
| 76000 |
+
{
|
| 76001 |
+
"epoch": 100.61,
|
| 76002 |
+
"learning_rate": 9.814054927302101e-06,
|
| 76003 |
+
"loss": 0.8728,
|
| 76004 |
+
"step": 12516
|
| 76005 |
+
},
|
| 76006 |
+
{
|
| 76007 |
+
"epoch": 100.62,
|
| 76008 |
+
"learning_rate": 9.814038772213248e-06,
|
| 76009 |
+
"loss": 0.9975,
|
| 76010 |
+
"step": 12517
|
| 76011 |
+
},
|
| 76012 |
+
{
|
| 76013 |
+
"epoch": 100.63,
|
| 76014 |
+
"learning_rate": 9.814022617124395e-06,
|
| 76015 |
+
"loss": 0.8238,
|
| 76016 |
+
"step": 12518
|
| 76017 |
+
},
|
| 76018 |
+
{
|
| 76019 |
+
"epoch": 100.63,
|
| 76020 |
+
"learning_rate": 9.814006462035542e-06,
|
| 76021 |
+
"loss": 0.7486,
|
| 76022 |
+
"step": 12519
|
| 76023 |
+
},
|
| 76024 |
+
{
|
| 76025 |
+
"epoch": 100.64,
|
| 76026 |
+
"learning_rate": 9.813990306946689e-06,
|
| 76027 |
+
"loss": 0.6136,
|
| 76028 |
+
"step": 12520
|
| 76029 |
+
},
|
| 76030 |
+
{
|
| 76031 |
+
"epoch": 100.65,
|
| 76032 |
+
"learning_rate": 9.813974151857835e-06,
|
| 76033 |
+
"loss": 0.6641,
|
| 76034 |
+
"step": 12521
|
| 76035 |
+
},
|
| 76036 |
+
{
|
| 76037 |
+
"epoch": 100.66,
|
| 76038 |
+
"learning_rate": 9.813957996768984e-06,
|
| 76039 |
+
"loss": 0.6873,
|
| 76040 |
+
"step": 12522
|
| 76041 |
+
},
|
| 76042 |
+
{
|
| 76043 |
+
"epoch": 100.67,
|
| 76044 |
+
"learning_rate": 9.81394184168013e-06,
|
| 76045 |
+
"loss": 1.2726,
|
| 76046 |
+
"step": 12523
|
| 76047 |
+
},
|
| 76048 |
+
{
|
| 76049 |
+
"epoch": 100.67,
|
| 76050 |
+
"learning_rate": 9.813925686591278e-06,
|
| 76051 |
+
"loss": 0.5694,
|
| 76052 |
+
"step": 12524
|
| 76053 |
+
},
|
| 76054 |
+
{
|
| 76055 |
+
"epoch": 100.68,
|
| 76056 |
+
"learning_rate": 9.813909531502423e-06,
|
| 76057 |
+
"loss": 0.6132,
|
| 76058 |
+
"step": 12525
|
| 76059 |
+
},
|
| 76060 |
+
{
|
| 76061 |
+
"epoch": 100.69,
|
| 76062 |
+
"learning_rate": 9.813893376413571e-06,
|
| 76063 |
+
"loss": 0.7758,
|
| 76064 |
+
"step": 12526
|
| 76065 |
+
},
|
| 76066 |
+
{
|
| 76067 |
+
"epoch": 100.7,
|
| 76068 |
+
"learning_rate": 9.813877221324718e-06,
|
| 76069 |
+
"loss": 0.8513,
|
| 76070 |
+
"step": 12527
|
| 76071 |
+
},
|
| 76072 |
+
{
|
| 76073 |
+
"epoch": 100.71,
|
| 76074 |
+
"learning_rate": 9.813861066235865e-06,
|
| 76075 |
+
"loss": 0.5893,
|
| 76076 |
+
"step": 12528
|
| 76077 |
+
},
|
| 76078 |
+
{
|
| 76079 |
+
"epoch": 100.71,
|
| 76080 |
+
"learning_rate": 9.813844911147012e-06,
|
| 76081 |
+
"loss": 0.8308,
|
| 76082 |
+
"step": 12529
|
| 76083 |
+
},
|
| 76084 |
+
{
|
| 76085 |
+
"epoch": 100.72,
|
| 76086 |
+
"learning_rate": 9.813828756058159e-06,
|
| 76087 |
+
"loss": 0.7328,
|
| 76088 |
+
"step": 12530
|
| 76089 |
+
},
|
| 76090 |
+
{
|
| 76091 |
+
"epoch": 100.73,
|
| 76092 |
+
"learning_rate": 9.813812600969305e-06,
|
| 76093 |
+
"loss": 0.7198,
|
| 76094 |
+
"step": 12531
|
| 76095 |
+
},
|
| 76096 |
+
{
|
| 76097 |
+
"epoch": 100.74,
|
| 76098 |
+
"learning_rate": 9.813796445880454e-06,
|
| 76099 |
+
"loss": 0.6577,
|
| 76100 |
+
"step": 12532
|
| 76101 |
+
},
|
| 76102 |
+
{
|
| 76103 |
+
"epoch": 100.75,
|
| 76104 |
+
"learning_rate": 9.8137802907916e-06,
|
| 76105 |
+
"loss": 1.4733,
|
| 76106 |
+
"step": 12533
|
| 76107 |
+
},
|
| 76108 |
+
{
|
| 76109 |
+
"epoch": 100.76,
|
| 76110 |
+
"learning_rate": 9.813764135702748e-06,
|
| 76111 |
+
"loss": 0.9624,
|
| 76112 |
+
"step": 12534
|
| 76113 |
+
},
|
| 76114 |
+
{
|
| 76115 |
+
"epoch": 100.76,
|
| 76116 |
+
"learning_rate": 9.813747980613894e-06,
|
| 76117 |
+
"loss": 1.0521,
|
| 76118 |
+
"step": 12535
|
| 76119 |
+
},
|
| 76120 |
+
{
|
| 76121 |
+
"epoch": 100.77,
|
| 76122 |
+
"learning_rate": 9.813731825525041e-06,
|
| 76123 |
+
"loss": 1.3083,
|
| 76124 |
+
"step": 12536
|
| 76125 |
+
},
|
| 76126 |
+
{
|
| 76127 |
+
"epoch": 100.78,
|
| 76128 |
+
"learning_rate": 9.813715670436188e-06,
|
| 76129 |
+
"loss": 1.4047,
|
| 76130 |
+
"step": 12537
|
| 76131 |
+
},
|
| 76132 |
+
{
|
| 76133 |
+
"epoch": 100.79,
|
| 76134 |
+
"learning_rate": 9.813699515347335e-06,
|
| 76135 |
+
"loss": 1.3001,
|
| 76136 |
+
"step": 12538
|
| 76137 |
+
},
|
| 76138 |
+
{
|
| 76139 |
+
"epoch": 100.8,
|
| 76140 |
+
"learning_rate": 9.813683360258482e-06,
|
| 76141 |
+
"loss": 2.0648,
|
| 76142 |
+
"step": 12539
|
| 76143 |
+
},
|
| 76144 |
+
{
|
| 76145 |
+
"epoch": 100.8,
|
| 76146 |
+
"learning_rate": 9.813667205169629e-06,
|
| 76147 |
+
"loss": 2.4255,
|
| 76148 |
+
"step": 12540
|
| 76149 |
+
},
|
| 76150 |
+
{
|
| 76151 |
+
"epoch": 100.81,
|
| 76152 |
+
"learning_rate": 9.813651050080775e-06,
|
| 76153 |
+
"loss": 0.7719,
|
| 76154 |
+
"step": 12541
|
| 76155 |
+
},
|
| 76156 |
+
{
|
| 76157 |
+
"epoch": 100.82,
|
| 76158 |
+
"learning_rate": 9.813634894991924e-06,
|
| 76159 |
+
"loss": 0.8144,
|
| 76160 |
+
"step": 12542
|
| 76161 |
+
},
|
| 76162 |
+
{
|
| 76163 |
+
"epoch": 100.83,
|
| 76164 |
+
"learning_rate": 9.81361873990307e-06,
|
| 76165 |
+
"loss": 0.6646,
|
| 76166 |
+
"step": 12543
|
| 76167 |
+
},
|
| 76168 |
+
{
|
| 76169 |
+
"epoch": 100.84,
|
| 76170 |
+
"learning_rate": 9.813602584814218e-06,
|
| 76171 |
+
"loss": 0.7451,
|
| 76172 |
+
"step": 12544
|
| 76173 |
+
},
|
| 76174 |
+
{
|
| 76175 |
+
"epoch": 100.84,
|
| 76176 |
+
"learning_rate": 9.813586429725364e-06,
|
| 76177 |
+
"loss": 0.8135,
|
| 76178 |
+
"step": 12545
|
| 76179 |
+
},
|
| 76180 |
+
{
|
| 76181 |
+
"epoch": 100.85,
|
| 76182 |
+
"learning_rate": 9.813570274636511e-06,
|
| 76183 |
+
"loss": 0.7016,
|
| 76184 |
+
"step": 12546
|
| 76185 |
+
},
|
| 76186 |
+
{
|
| 76187 |
+
"epoch": 100.86,
|
| 76188 |
+
"learning_rate": 9.813554119547658e-06,
|
| 76189 |
+
"loss": 0.6032,
|
| 76190 |
+
"step": 12547
|
| 76191 |
+
},
|
| 76192 |
+
{
|
| 76193 |
+
"epoch": 100.87,
|
| 76194 |
+
"learning_rate": 9.813537964458807e-06,
|
| 76195 |
+
"loss": 0.9547,
|
| 76196 |
+
"step": 12548
|
| 76197 |
+
},
|
| 76198 |
+
{
|
| 76199 |
+
"epoch": 100.88,
|
| 76200 |
+
"learning_rate": 9.813521809369952e-06,
|
| 76201 |
+
"loss": 0.9344,
|
| 76202 |
+
"step": 12549
|
| 76203 |
+
},
|
| 76204 |
+
{
|
| 76205 |
+
"epoch": 100.88,
|
| 76206 |
+
"learning_rate": 9.813505654281099e-06,
|
| 76207 |
+
"loss": 1.6009,
|
| 76208 |
+
"step": 12550
|
| 76209 |
+
},
|
| 76210 |
+
{
|
| 76211 |
+
"epoch": 100.89,
|
| 76212 |
+
"learning_rate": 9.813489499192245e-06,
|
| 76213 |
+
"loss": 0.6731,
|
| 76214 |
+
"step": 12551
|
| 76215 |
+
},
|
| 76216 |
+
{
|
| 76217 |
+
"epoch": 100.9,
|
| 76218 |
+
"learning_rate": 9.813473344103394e-06,
|
| 76219 |
+
"loss": 0.8096,
|
| 76220 |
+
"step": 12552
|
| 76221 |
+
},
|
| 76222 |
+
{
|
| 76223 |
+
"epoch": 100.91,
|
| 76224 |
+
"learning_rate": 9.81345718901454e-06,
|
| 76225 |
+
"loss": 0.7104,
|
| 76226 |
+
"step": 12553
|
| 76227 |
+
},
|
| 76228 |
+
{
|
| 76229 |
+
"epoch": 100.92,
|
| 76230 |
+
"learning_rate": 9.813441033925688e-06,
|
| 76231 |
+
"loss": 0.684,
|
| 76232 |
+
"step": 12554
|
| 76233 |
+
},
|
| 76234 |
+
{
|
| 76235 |
+
"epoch": 100.92,
|
| 76236 |
+
"learning_rate": 9.813424878836834e-06,
|
| 76237 |
+
"loss": 0.7053,
|
| 76238 |
+
"step": 12555
|
| 76239 |
+
},
|
| 76240 |
+
{
|
| 76241 |
+
"epoch": 100.93,
|
| 76242 |
+
"learning_rate": 9.813408723747981e-06,
|
| 76243 |
+
"loss": 0.5725,
|
| 76244 |
+
"step": 12556
|
| 76245 |
+
},
|
| 76246 |
+
{
|
| 76247 |
+
"epoch": 100.94,
|
| 76248 |
+
"learning_rate": 9.813392568659128e-06,
|
| 76249 |
+
"loss": 0.8067,
|
| 76250 |
+
"step": 12557
|
| 76251 |
+
},
|
| 76252 |
+
{
|
| 76253 |
+
"epoch": 100.95,
|
| 76254 |
+
"learning_rate": 9.813376413570277e-06,
|
| 76255 |
+
"loss": 0.8155,
|
| 76256 |
+
"step": 12558
|
| 76257 |
+
},
|
| 76258 |
+
{
|
| 76259 |
+
"epoch": 100.96,
|
| 76260 |
+
"learning_rate": 9.813360258481422e-06,
|
| 76261 |
+
"loss": 1.2523,
|
| 76262 |
+
"step": 12559
|
| 76263 |
+
},
|
| 76264 |
+
{
|
| 76265 |
+
"epoch": 100.96,
|
| 76266 |
+
"learning_rate": 9.813344103392569e-06,
|
| 76267 |
+
"loss": 1.0687,
|
| 76268 |
+
"step": 12560
|
| 76269 |
+
},
|
| 76270 |
+
{
|
| 76271 |
+
"epoch": 100.97,
|
| 76272 |
+
"learning_rate": 9.813327948303717e-06,
|
| 76273 |
+
"loss": 1.3879,
|
| 76274 |
+
"step": 12561
|
| 76275 |
+
},
|
| 76276 |
+
{
|
| 76277 |
+
"epoch": 100.98,
|
| 76278 |
+
"learning_rate": 9.813311793214864e-06,
|
| 76279 |
+
"loss": 1.837,
|
| 76280 |
+
"step": 12562
|
| 76281 |
+
},
|
| 76282 |
+
{
|
| 76283 |
+
"epoch": 100.99,
|
| 76284 |
+
"learning_rate": 9.81329563812601e-06,
|
| 76285 |
+
"loss": 1.81,
|
| 76286 |
+
"step": 12563
|
| 76287 |
+
},
|
| 76288 |
+
{
|
| 76289 |
+
"epoch": 101.0,
|
| 76290 |
+
"learning_rate": 9.813279483037158e-06,
|
| 76291 |
+
"loss": 2.7796,
|
| 76292 |
+
"step": 12564
|
| 76293 |
+
},
|
| 76294 |
+
{
|
| 76295 |
+
"epoch": 101.0,
|
| 76296 |
+
"eval_loss": 0.7402641177177429,
|
| 76297 |
+
"eval_runtime": 39.7449,
|
| 76298 |
+
"eval_samples_per_second": 21.059,
|
| 76299 |
+
"eval_steps_per_second": 0.679,
|
| 76300 |
+
"eval_wer": 0.3129266521423384,
|
| 76301 |
+
"step": 12564
|
| 76302 |
+
},
|
| 76303 |
+
{
|
| 76304 |
+
"epoch": 101.01,
|
| 76305 |
+
"learning_rate": 9.813263327948304e-06,
|
| 76306 |
+
"loss": 0.8226,
|
| 76307 |
+
"step": 12565
|
| 76308 |
+
},
|
| 76309 |
+
{
|
| 76310 |
+
"epoch": 101.02,
|
| 76311 |
+
"learning_rate": 9.813247172859451e-06,
|
| 76312 |
+
"loss": 0.926,
|
| 76313 |
+
"step": 12566
|
| 76314 |
+
},
|
| 76315 |
+
{
|
| 76316 |
+
"epoch": 101.02,
|
| 76317 |
+
"learning_rate": 9.813231017770598e-06,
|
| 76318 |
+
"loss": 0.7772,
|
| 76319 |
+
"step": 12567
|
| 76320 |
+
},
|
| 76321 |
+
{
|
| 76322 |
+
"epoch": 101.03,
|
| 76323 |
+
"learning_rate": 9.813214862681747e-06,
|
| 76324 |
+
"loss": 0.8102,
|
| 76325 |
+
"step": 12568
|
| 76326 |
+
},
|
| 76327 |
+
{
|
| 76328 |
+
"epoch": 101.04,
|
| 76329 |
+
"learning_rate": 9.813198707592892e-06,
|
| 76330 |
+
"loss": 0.6223,
|
| 76331 |
+
"step": 12569
|
| 76332 |
+
},
|
| 76333 |
+
{
|
| 76334 |
+
"epoch": 101.05,
|
| 76335 |
+
"learning_rate": 9.813182552504039e-06,
|
| 76336 |
+
"loss": 0.5981,
|
| 76337 |
+
"step": 12570
|
| 76338 |
+
},
|
| 76339 |
+
{
|
| 76340 |
+
"epoch": 101.06,
|
| 76341 |
+
"learning_rate": 9.813166397415187e-06,
|
| 76342 |
+
"loss": 0.6781,
|
| 76343 |
+
"step": 12571
|
| 76344 |
+
},
|
| 76345 |
+
{
|
| 76346 |
+
"epoch": 101.06,
|
| 76347 |
+
"learning_rate": 9.813150242326334e-06,
|
| 76348 |
+
"loss": 0.6371,
|
| 76349 |
+
"step": 12572
|
| 76350 |
+
},
|
| 76351 |
+
{
|
| 76352 |
+
"epoch": 101.07,
|
| 76353 |
+
"learning_rate": 9.81313408723748e-06,
|
| 76354 |
+
"loss": 0.6215,
|
| 76355 |
+
"step": 12573
|
| 76356 |
+
},
|
| 76357 |
+
{
|
| 76358 |
+
"epoch": 101.08,
|
| 76359 |
+
"learning_rate": 9.813117932148628e-06,
|
| 76360 |
+
"loss": 0.824,
|
| 76361 |
+
"step": 12574
|
| 76362 |
+
},
|
| 76363 |
+
{
|
| 76364 |
+
"epoch": 101.09,
|
| 76365 |
+
"learning_rate": 9.813101777059774e-06,
|
| 76366 |
+
"loss": 0.679,
|
| 76367 |
+
"step": 12575
|
| 76368 |
+
},
|
| 76369 |
+
{
|
| 76370 |
+
"epoch": 101.1,
|
| 76371 |
+
"learning_rate": 9.813085621970921e-06,
|
| 76372 |
+
"loss": 0.8155,
|
| 76373 |
+
"step": 12576
|
| 76374 |
+
},
|
| 76375 |
+
{
|
| 76376 |
+
"epoch": 101.1,
|
| 76377 |
+
"learning_rate": 9.81306946688207e-06,
|
| 76378 |
+
"loss": 0.973,
|
| 76379 |
+
"step": 12577
|
| 76380 |
+
},
|
| 76381 |
+
{
|
| 76382 |
+
"epoch": 101.11,
|
| 76383 |
+
"learning_rate": 9.813053311793217e-06,
|
| 76384 |
+
"loss": 1.6699,
|
| 76385 |
+
"step": 12578
|
| 76386 |
+
},
|
| 76387 |
+
{
|
| 76388 |
+
"epoch": 101.12,
|
| 76389 |
+
"learning_rate": 9.813037156704362e-06,
|
| 76390 |
+
"loss": 0.9268,
|
| 76391 |
+
"step": 12579
|
| 76392 |
+
},
|
| 76393 |
+
{
|
| 76394 |
+
"epoch": 101.13,
|
| 76395 |
+
"learning_rate": 9.813021001615509e-06,
|
| 76396 |
+
"loss": 1.0844,
|
| 76397 |
+
"step": 12580
|
| 76398 |
+
},
|
| 76399 |
+
{
|
| 76400 |
+
"epoch": 101.14,
|
| 76401 |
+
"learning_rate": 9.813004846526657e-06,
|
| 76402 |
+
"loss": 0.7213,
|
| 76403 |
+
"step": 12581
|
| 76404 |
+
},
|
| 76405 |
+
{
|
| 76406 |
+
"epoch": 101.14,
|
| 76407 |
+
"learning_rate": 9.812988691437804e-06,
|
| 76408 |
+
"loss": 0.7768,
|
| 76409 |
+
"step": 12582
|
| 76410 |
+
},
|
| 76411 |
+
{
|
| 76412 |
+
"epoch": 101.15,
|
| 76413 |
+
"learning_rate": 9.81297253634895e-06,
|
| 76414 |
+
"loss": 1.2389,
|
| 76415 |
+
"step": 12583
|
| 76416 |
+
},
|
| 76417 |
+
{
|
| 76418 |
+
"epoch": 101.16,
|
| 76419 |
+
"learning_rate": 9.812956381260098e-06,
|
| 76420 |
+
"loss": 1.2811,
|
| 76421 |
+
"step": 12584
|
| 76422 |
+
},
|
| 76423 |
+
{
|
| 76424 |
+
"epoch": 101.17,
|
| 76425 |
+
"learning_rate": 9.812940226171244e-06,
|
| 76426 |
+
"loss": 1.1199,
|
| 76427 |
+
"step": 12585
|
| 76428 |
+
},
|
| 76429 |
+
{
|
| 76430 |
+
"epoch": 101.18,
|
| 76431 |
+
"learning_rate": 9.812924071082391e-06,
|
| 76432 |
+
"loss": 1.9411,
|
| 76433 |
+
"step": 12586
|
| 76434 |
+
},
|
| 76435 |
+
{
|
| 76436 |
+
"epoch": 101.18,
|
| 76437 |
+
"learning_rate": 9.81290791599354e-06,
|
| 76438 |
+
"loss": 1.4737,
|
| 76439 |
+
"step": 12587
|
| 76440 |
+
},
|
| 76441 |
+
{
|
| 76442 |
+
"epoch": 101.19,
|
| 76443 |
+
"learning_rate": 9.812891760904687e-06,
|
| 76444 |
+
"loss": 1.7156,
|
| 76445 |
+
"step": 12588
|
| 76446 |
+
},
|
| 76447 |
+
{
|
| 76448 |
+
"epoch": 101.2,
|
| 76449 |
+
"learning_rate": 9.812875605815833e-06,
|
| 76450 |
+
"loss": 2.5709,
|
| 76451 |
+
"step": 12589
|
| 76452 |
+
},
|
| 76453 |
+
{
|
| 76454 |
+
"epoch": 101.21,
|
| 76455 |
+
"learning_rate": 9.81285945072698e-06,
|
| 76456 |
+
"loss": 0.9945,
|
| 76457 |
+
"step": 12590
|
| 76458 |
+
},
|
| 76459 |
+
{
|
| 76460 |
+
"epoch": 101.22,
|
| 76461 |
+
"learning_rate": 9.812843295638127e-06,
|
| 76462 |
+
"loss": 0.7821,
|
| 76463 |
+
"step": 12591
|
| 76464 |
+
},
|
| 76465 |
+
{
|
| 76466 |
+
"epoch": 101.22,
|
| 76467 |
+
"learning_rate": 9.812827140549274e-06,
|
| 76468 |
+
"loss": 0.6336,
|
| 76469 |
+
"step": 12592
|
| 76470 |
+
},
|
| 76471 |
+
{
|
| 76472 |
+
"epoch": 101.23,
|
| 76473 |
+
"learning_rate": 9.81281098546042e-06,
|
| 76474 |
+
"loss": 0.616,
|
| 76475 |
+
"step": 12593
|
| 76476 |
+
},
|
| 76477 |
+
{
|
| 76478 |
+
"epoch": 101.24,
|
| 76479 |
+
"learning_rate": 9.812794830371567e-06,
|
| 76480 |
+
"loss": 0.6655,
|
| 76481 |
+
"step": 12594
|
| 76482 |
+
},
|
| 76483 |
+
{
|
| 76484 |
+
"epoch": 101.25,
|
| 76485 |
+
"learning_rate": 9.812778675282714e-06,
|
| 76486 |
+
"loss": 0.5965,
|
| 76487 |
+
"step": 12595
|
| 76488 |
+
},
|
| 76489 |
+
{
|
| 76490 |
+
"epoch": 101.26,
|
| 76491 |
+
"learning_rate": 9.812762520193861e-06,
|
| 76492 |
+
"loss": 0.6148,
|
| 76493 |
+
"step": 12596
|
| 76494 |
+
},
|
| 76495 |
+
{
|
| 76496 |
+
"epoch": 101.27,
|
| 76497 |
+
"learning_rate": 9.81274636510501e-06,
|
| 76498 |
+
"loss": 0.5628,
|
| 76499 |
+
"step": 12597
|
| 76500 |
+
},
|
| 76501 |
+
{
|
| 76502 |
+
"epoch": 101.27,
|
| 76503 |
+
"learning_rate": 9.812730210016157e-06,
|
| 76504 |
+
"loss": 0.5296,
|
| 76505 |
+
"step": 12598
|
| 76506 |
+
},
|
| 76507 |
+
{
|
| 76508 |
+
"epoch": 101.28,
|
| 76509 |
+
"learning_rate": 9.812714054927303e-06,
|
| 76510 |
+
"loss": 0.5968,
|
| 76511 |
+
"step": 12599
|
| 76512 |
+
},
|
| 76513 |
+
{
|
| 76514 |
+
"epoch": 101.29,
|
| 76515 |
+
"learning_rate": 9.81269789983845e-06,
|
| 76516 |
+
"loss": 0.8654,
|
| 76517 |
+
"step": 12600
|
| 76518 |
+
},
|
| 76519 |
+
{
|
| 76520 |
+
"epoch": 101.3,
|
| 76521 |
+
"learning_rate": 9.812681744749597e-06,
|
| 76522 |
+
"loss": 0.6842,
|
| 76523 |
+
"step": 12601
|
| 76524 |
+
},
|
| 76525 |
+
{
|
| 76526 |
+
"epoch": 101.31,
|
| 76527 |
+
"learning_rate": 9.812665589660744e-06,
|
| 76528 |
+
"loss": 1.2858,
|
| 76529 |
+
"step": 12602
|
| 76530 |
+
},
|
| 76531 |
+
{
|
| 76532 |
+
"epoch": 101.31,
|
| 76533 |
+
"learning_rate": 9.81264943457189e-06,
|
| 76534 |
+
"loss": 0.6153,
|
| 76535 |
+
"step": 12603
|
| 76536 |
+
},
|
| 76537 |
+
{
|
| 76538 |
+
"epoch": 101.32,
|
| 76539 |
+
"learning_rate": 9.812633279483037e-06,
|
| 76540 |
+
"loss": 0.7943,
|
| 76541 |
+
"step": 12604
|
| 76542 |
+
},
|
| 76543 |
+
{
|
| 76544 |
+
"epoch": 101.33,
|
| 76545 |
+
"learning_rate": 9.812617124394184e-06,
|
| 76546 |
+
"loss": 0.8027,
|
| 76547 |
+
"step": 12605
|
| 76548 |
+
},
|
| 76549 |
+
{
|
| 76550 |
+
"epoch": 101.34,
|
| 76551 |
+
"learning_rate": 9.812600969305331e-06,
|
| 76552 |
+
"loss": 0.7428,
|
| 76553 |
+
"step": 12606
|
| 76554 |
+
},
|
| 76555 |
+
{
|
| 76556 |
+
"epoch": 101.35,
|
| 76557 |
+
"learning_rate": 9.81258481421648e-06,
|
| 76558 |
+
"loss": 0.9211,
|
| 76559 |
+
"step": 12607
|
| 76560 |
+
},
|
| 76561 |
+
{
|
| 76562 |
+
"epoch": 101.35,
|
| 76563 |
+
"learning_rate": 9.812568659127626e-06,
|
| 76564 |
+
"loss": 0.9136,
|
| 76565 |
+
"step": 12608
|
| 76566 |
+
},
|
| 76567 |
+
{
|
| 76568 |
+
"epoch": 101.36,
|
| 76569 |
+
"learning_rate": 9.812552504038773e-06,
|
| 76570 |
+
"loss": 1.1029,
|
| 76571 |
+
"step": 12609
|
| 76572 |
+
},
|
| 76573 |
+
{
|
| 76574 |
+
"epoch": 101.37,
|
| 76575 |
+
"learning_rate": 9.81253634894992e-06,
|
| 76576 |
+
"loss": 0.9693,
|
| 76577 |
+
"step": 12610
|
| 76578 |
+
},
|
| 76579 |
+
{
|
| 76580 |
+
"epoch": 101.38,
|
| 76581 |
+
"learning_rate": 9.812520193861067e-06,
|
| 76582 |
+
"loss": 1.3601,
|
| 76583 |
+
"step": 12611
|
| 76584 |
+
},
|
| 76585 |
+
{
|
| 76586 |
+
"epoch": 101.39,
|
| 76587 |
+
"learning_rate": 9.812504038772214e-06,
|
| 76588 |
+
"loss": 1.6806,
|
| 76589 |
+
"step": 12612
|
| 76590 |
+
},
|
| 76591 |
+
{
|
| 76592 |
+
"epoch": 101.39,
|
| 76593 |
+
"learning_rate": 9.812487883683362e-06,
|
| 76594 |
+
"loss": 1.9042,
|
| 76595 |
+
"step": 12613
|
| 76596 |
+
},
|
| 76597 |
+
{
|
| 76598 |
+
"epoch": 101.4,
|
| 76599 |
+
"learning_rate": 9.812471728594507e-06,
|
| 76600 |
+
"loss": 2.6731,
|
| 76601 |
+
"step": 12614
|
| 76602 |
+
},
|
| 76603 |
+
{
|
| 76604 |
+
"epoch": 101.41,
|
| 76605 |
+
"learning_rate": 9.812455573505654e-06,
|
| 76606 |
+
"loss": 0.7313,
|
| 76607 |
+
"step": 12615
|
| 76608 |
+
},
|
| 76609 |
+
{
|
| 76610 |
+
"epoch": 101.42,
|
| 76611 |
+
"learning_rate": 9.812439418416803e-06,
|
| 76612 |
+
"loss": 0.8018,
|
| 76613 |
+
"step": 12616
|
| 76614 |
+
},
|
| 76615 |
+
{
|
| 76616 |
+
"epoch": 101.43,
|
| 76617 |
+
"learning_rate": 9.81242326332795e-06,
|
| 76618 |
+
"loss": 0.8298,
|
| 76619 |
+
"step": 12617
|
| 76620 |
+
},
|
| 76621 |
+
{
|
| 76622 |
+
"epoch": 101.43,
|
| 76623 |
+
"learning_rate": 9.812407108239096e-06,
|
| 76624 |
+
"loss": 0.9246,
|
| 76625 |
+
"step": 12618
|
| 76626 |
+
},
|
| 76627 |
+
{
|
| 76628 |
+
"epoch": 101.44,
|
| 76629 |
+
"learning_rate": 9.812390953150243e-06,
|
| 76630 |
+
"loss": 1.429,
|
| 76631 |
+
"step": 12619
|
| 76632 |
+
},
|
| 76633 |
+
{
|
| 76634 |
+
"epoch": 101.45,
|
| 76635 |
+
"learning_rate": 9.81237479806139e-06,
|
| 76636 |
+
"loss": 0.67,
|
| 76637 |
+
"step": 12620
|
| 76638 |
+
},
|
| 76639 |
+
{
|
| 76640 |
+
"epoch": 101.46,
|
| 76641 |
+
"learning_rate": 9.812358642972537e-06,
|
| 76642 |
+
"loss": 0.5775,
|
| 76643 |
+
"step": 12621
|
| 76644 |
+
},
|
| 76645 |
+
{
|
| 76646 |
+
"epoch": 101.47,
|
| 76647 |
+
"learning_rate": 9.812342487883684e-06,
|
| 76648 |
+
"loss": 0.6864,
|
| 76649 |
+
"step": 12622
|
| 76650 |
+
},
|
| 76651 |
+
{
|
| 76652 |
+
"epoch": 101.47,
|
| 76653 |
+
"learning_rate": 9.812326332794832e-06,
|
| 76654 |
+
"loss": 0.5996,
|
| 76655 |
+
"step": 12623
|
| 76656 |
+
},
|
| 76657 |
+
{
|
| 76658 |
+
"epoch": 101.48,
|
| 76659 |
+
"learning_rate": 9.812310177705977e-06,
|
| 76660 |
+
"loss": 0.7286,
|
| 76661 |
+
"step": 12624
|
| 76662 |
+
},
|
| 76663 |
+
{
|
| 76664 |
+
"epoch": 101.49,
|
| 76665 |
+
"learning_rate": 9.812294022617124e-06,
|
| 76666 |
+
"loss": 0.6334,
|
| 76667 |
+
"step": 12625
|
| 76668 |
+
},
|
| 76669 |
+
{
|
| 76670 |
+
"epoch": 101.5,
|
| 76671 |
+
"learning_rate": 9.812277867528273e-06,
|
| 76672 |
+
"loss": 0.6695,
|
| 76673 |
+
"step": 12626
|
| 76674 |
+
},
|
| 76675 |
+
{
|
| 76676 |
+
"epoch": 101.51,
|
| 76677 |
+
"learning_rate": 9.81226171243942e-06,
|
| 76678 |
+
"loss": 0.6042,
|
| 76679 |
+
"step": 12627
|
| 76680 |
+
},
|
| 76681 |
+
{
|
| 76682 |
+
"epoch": 101.51,
|
| 76683 |
+
"learning_rate": 9.812245557350566e-06,
|
| 76684 |
+
"loss": 0.7287,
|
| 76685 |
+
"step": 12628
|
| 76686 |
+
},
|
| 76687 |
+
{
|
| 76688 |
+
"epoch": 101.52,
|
| 76689 |
+
"learning_rate": 9.812229402261713e-06,
|
| 76690 |
+
"loss": 0.5906,
|
| 76691 |
+
"step": 12629
|
| 76692 |
+
},
|
| 76693 |
+
{
|
| 76694 |
+
"epoch": 101.53,
|
| 76695 |
+
"learning_rate": 9.81221324717286e-06,
|
| 76696 |
+
"loss": 0.7655,
|
| 76697 |
+
"step": 12630
|
| 76698 |
+
},
|
| 76699 |
+
{
|
| 76700 |
+
"epoch": 101.54,
|
| 76701 |
+
"learning_rate": 9.812197092084007e-06,
|
| 76702 |
+
"loss": 0.8739,
|
| 76703 |
+
"step": 12631
|
| 76704 |
+
},
|
| 76705 |
+
{
|
| 76706 |
+
"epoch": 101.55,
|
| 76707 |
+
"learning_rate": 9.812180936995154e-06,
|
| 76708 |
+
"loss": 0.8849,
|
| 76709 |
+
"step": 12632
|
| 76710 |
+
},
|
| 76711 |
+
{
|
| 76712 |
+
"epoch": 101.55,
|
| 76713 |
+
"learning_rate": 9.812164781906302e-06,
|
| 76714 |
+
"loss": 0.9914,
|
| 76715 |
+
"step": 12633
|
| 76716 |
+
},
|
| 76717 |
+
{
|
| 76718 |
+
"epoch": 101.56,
|
| 76719 |
+
"learning_rate": 9.812148626817447e-06,
|
| 76720 |
+
"loss": 1.0731,
|
| 76721 |
+
"step": 12634
|
| 76722 |
+
},
|
| 76723 |
+
{
|
| 76724 |
+
"epoch": 101.57,
|
| 76725 |
+
"learning_rate": 9.812132471728594e-06,
|
| 76726 |
+
"loss": 1.3329,
|
| 76727 |
+
"step": 12635
|
| 76728 |
+
},
|
| 76729 |
+
{
|
| 76730 |
+
"epoch": 101.58,
|
| 76731 |
+
"learning_rate": 9.812116316639743e-06,
|
| 76732 |
+
"loss": 1.3646,
|
| 76733 |
+
"step": 12636
|
| 76734 |
+
},
|
| 76735 |
+
{
|
| 76736 |
+
"epoch": 101.59,
|
| 76737 |
+
"learning_rate": 9.81210016155089e-06,
|
| 76738 |
+
"loss": 1.503,
|
| 76739 |
+
"step": 12637
|
| 76740 |
+
},
|
| 76741 |
+
{
|
| 76742 |
+
"epoch": 101.59,
|
| 76743 |
+
"learning_rate": 9.812084006462036e-06,
|
| 76744 |
+
"loss": 2.0364,
|
| 76745 |
+
"step": 12638
|
| 76746 |
+
},
|
| 76747 |
+
{
|
| 76748 |
+
"epoch": 101.6,
|
| 76749 |
+
"learning_rate": 9.812067851373183e-06,
|
| 76750 |
+
"loss": 6.7083,
|
| 76751 |
+
"step": 12639
|
| 76752 |
+
},
|
| 76753 |
+
{
|
| 76754 |
+
"epoch": 101.61,
|
| 76755 |
+
"learning_rate": 9.81205169628433e-06,
|
| 76756 |
+
"loss": 0.7836,
|
| 76757 |
+
"step": 12640
|
| 76758 |
+
},
|
| 76759 |
+
{
|
| 76760 |
+
"epoch": 101.62,
|
| 76761 |
+
"learning_rate": 9.812035541195477e-06,
|
| 76762 |
+
"loss": 0.7141,
|
| 76763 |
+
"step": 12641
|
| 76764 |
+
},
|
| 76765 |
+
{
|
| 76766 |
+
"epoch": 101.63,
|
| 76767 |
+
"learning_rate": 9.812019386106625e-06,
|
| 76768 |
+
"loss": 0.6748,
|
| 76769 |
+
"step": 12642
|
| 76770 |
+
},
|
| 76771 |
+
{
|
| 76772 |
+
"epoch": 101.63,
|
| 76773 |
+
"learning_rate": 9.812003231017772e-06,
|
| 76774 |
+
"loss": 0.6543,
|
| 76775 |
+
"step": 12643
|
| 76776 |
+
},
|
| 76777 |
+
{
|
| 76778 |
+
"epoch": 101.64,
|
| 76779 |
+
"learning_rate": 9.811987075928917e-06,
|
| 76780 |
+
"loss": 0.6779,
|
| 76781 |
+
"step": 12644
|
| 76782 |
+
},
|
| 76783 |
+
{
|
| 76784 |
+
"epoch": 101.65,
|
| 76785 |
+
"learning_rate": 9.811970920840066e-06,
|
| 76786 |
+
"loss": 0.9245,
|
| 76787 |
+
"step": 12645
|
| 76788 |
+
},
|
| 76789 |
+
{
|
| 76790 |
+
"epoch": 101.66,
|
| 76791 |
+
"learning_rate": 9.811954765751213e-06,
|
| 76792 |
+
"loss": 0.6281,
|
| 76793 |
+
"step": 12646
|
| 76794 |
+
},
|
| 76795 |
+
{
|
| 76796 |
+
"epoch": 101.67,
|
| 76797 |
+
"learning_rate": 9.81193861066236e-06,
|
| 76798 |
+
"loss": 0.6212,
|
| 76799 |
+
"step": 12647
|
| 76800 |
+
},
|
| 76801 |
+
{
|
| 76802 |
+
"epoch": 101.67,
|
| 76803 |
+
"learning_rate": 9.811922455573506e-06,
|
| 76804 |
+
"loss": 1.0585,
|
| 76805 |
+
"step": 12648
|
| 76806 |
+
},
|
| 76807 |
+
{
|
| 76808 |
+
"epoch": 101.68,
|
| 76809 |
+
"learning_rate": 9.811906300484653e-06,
|
| 76810 |
+
"loss": 0.5934,
|
| 76811 |
+
"step": 12649
|
| 76812 |
+
},
|
| 76813 |
+
{
|
| 76814 |
+
"epoch": 101.69,
|
| 76815 |
+
"learning_rate": 9.8118901453958e-06,
|
| 76816 |
+
"loss": 0.8031,
|
| 76817 |
+
"step": 12650
|
| 76818 |
+
},
|
| 76819 |
+
{
|
| 76820 |
+
"epoch": 101.7,
|
| 76821 |
+
"learning_rate": 9.811873990306947e-06,
|
| 76822 |
+
"loss": 0.7886,
|
| 76823 |
+
"step": 12651
|
| 76824 |
+
},
|
| 76825 |
+
{
|
| 76826 |
+
"epoch": 101.71,
|
| 76827 |
+
"learning_rate": 9.811857835218095e-06,
|
| 76828 |
+
"loss": 0.5487,
|
| 76829 |
+
"step": 12652
|
| 76830 |
+
},
|
| 76831 |
+
{
|
| 76832 |
+
"epoch": 101.71,
|
| 76833 |
+
"learning_rate": 9.811841680129242e-06,
|
| 76834 |
+
"loss": 0.7441,
|
| 76835 |
+
"step": 12653
|
| 76836 |
+
},
|
| 76837 |
+
{
|
| 76838 |
+
"epoch": 101.72,
|
| 76839 |
+
"learning_rate": 9.811825525040389e-06,
|
| 76840 |
+
"loss": 0.6405,
|
| 76841 |
+
"step": 12654
|
| 76842 |
+
},
|
| 76843 |
+
{
|
| 76844 |
+
"epoch": 101.73,
|
| 76845 |
+
"learning_rate": 9.811809369951536e-06,
|
| 76846 |
+
"loss": 0.8121,
|
| 76847 |
+
"step": 12655
|
| 76848 |
+
},
|
| 76849 |
+
{
|
| 76850 |
+
"epoch": 101.74,
|
| 76851 |
+
"learning_rate": 9.811793214862683e-06,
|
| 76852 |
+
"loss": 0.7839,
|
| 76853 |
+
"step": 12656
|
| 76854 |
+
},
|
| 76855 |
+
{
|
| 76856 |
+
"epoch": 101.75,
|
| 76857 |
+
"learning_rate": 9.81177705977383e-06,
|
| 76858 |
+
"loss": 0.9364,
|
| 76859 |
+
"step": 12657
|
| 76860 |
+
},
|
| 76861 |
+
{
|
| 76862 |
+
"epoch": 101.76,
|
| 76863 |
+
"learning_rate": 9.811760904684976e-06,
|
| 76864 |
+
"loss": 1.1826,
|
| 76865 |
+
"step": 12658
|
| 76866 |
+
},
|
| 76867 |
+
{
|
| 76868 |
+
"epoch": 101.76,
|
| 76869 |
+
"learning_rate": 9.811744749596123e-06,
|
| 76870 |
+
"loss": 1.0468,
|
| 76871 |
+
"step": 12659
|
| 76872 |
+
},
|
| 76873 |
+
{
|
| 76874 |
+
"epoch": 101.77,
|
| 76875 |
+
"learning_rate": 9.81172859450727e-06,
|
| 76876 |
+
"loss": 1.2483,
|
| 76877 |
+
"step": 12660
|
| 76878 |
+
},
|
| 76879 |
+
{
|
| 76880 |
+
"epoch": 101.78,
|
| 76881 |
+
"learning_rate": 9.811712439418417e-06,
|
| 76882 |
+
"loss": 1.3404,
|
| 76883 |
+
"step": 12661
|
| 76884 |
+
},
|
| 76885 |
+
{
|
| 76886 |
+
"epoch": 101.79,
|
| 76887 |
+
"learning_rate": 9.811696284329565e-06,
|
| 76888 |
+
"loss": 1.5523,
|
| 76889 |
+
"step": 12662
|
| 76890 |
+
},
|
| 76891 |
+
{
|
| 76892 |
+
"epoch": 101.8,
|
| 76893 |
+
"learning_rate": 9.811680129240712e-06,
|
| 76894 |
+
"loss": 1.9791,
|
| 76895 |
+
"step": 12663
|
| 76896 |
+
},
|
| 76897 |
+
{
|
| 76898 |
+
"epoch": 101.8,
|
| 76899 |
+
"learning_rate": 9.811663974151859e-06,
|
| 76900 |
+
"loss": 2.5813,
|
| 76901 |
+
"step": 12664
|
| 76902 |
+
},
|
| 76903 |
+
{
|
| 76904 |
+
"epoch": 101.81,
|
| 76905 |
+
"learning_rate": 9.811647819063006e-06,
|
| 76906 |
+
"loss": 0.9987,
|
| 76907 |
+
"step": 12665
|
| 76908 |
+
},
|
| 76909 |
+
{
|
| 76910 |
+
"epoch": 101.82,
|
| 76911 |
+
"learning_rate": 9.811631663974153e-06,
|
| 76912 |
+
"loss": 0.5521,
|
| 76913 |
+
"step": 12666
|
| 76914 |
+
},
|
| 76915 |
+
{
|
| 76916 |
+
"epoch": 101.83,
|
| 76917 |
+
"learning_rate": 9.8116155088853e-06,
|
| 76918 |
+
"loss": 0.5785,
|
| 76919 |
+
"step": 12667
|
| 76920 |
+
},
|
| 76921 |
+
{
|
| 76922 |
+
"epoch": 101.84,
|
| 76923 |
+
"learning_rate": 9.811599353796446e-06,
|
| 76924 |
+
"loss": 0.5724,
|
| 76925 |
+
"step": 12668
|
| 76926 |
+
},
|
| 76927 |
+
{
|
| 76928 |
+
"epoch": 101.84,
|
| 76929 |
+
"learning_rate": 9.811583198707593e-06,
|
| 76930 |
+
"loss": 0.6612,
|
| 76931 |
+
"step": 12669
|
| 76932 |
+
},
|
| 76933 |
+
{
|
| 76934 |
+
"epoch": 101.85,
|
| 76935 |
+
"learning_rate": 9.81156704361874e-06,
|
| 76936 |
+
"loss": 0.575,
|
| 76937 |
+
"step": 12670
|
| 76938 |
+
},
|
| 76939 |
+
{
|
| 76940 |
+
"epoch": 101.86,
|
| 76941 |
+
"learning_rate": 9.811550888529889e-06,
|
| 76942 |
+
"loss": 0.9643,
|
| 76943 |
+
"step": 12671
|
| 76944 |
+
},
|
| 76945 |
+
{
|
| 76946 |
+
"epoch": 101.87,
|
| 76947 |
+
"learning_rate": 9.811534733441035e-06,
|
| 76948 |
+
"loss": 0.6386,
|
| 76949 |
+
"step": 12672
|
| 76950 |
+
},
|
| 76951 |
+
{
|
| 76952 |
+
"epoch": 101.88,
|
| 76953 |
+
"learning_rate": 9.811518578352182e-06,
|
| 76954 |
+
"loss": 0.8729,
|
| 76955 |
+
"step": 12673
|
| 76956 |
+
},
|
| 76957 |
+
{
|
| 76958 |
+
"epoch": 101.88,
|
| 76959 |
+
"learning_rate": 9.811502423263329e-06,
|
| 76960 |
+
"loss": 0.6118,
|
| 76961 |
+
"step": 12674
|
| 76962 |
+
},
|
| 76963 |
+
{
|
| 76964 |
+
"epoch": 101.89,
|
| 76965 |
+
"learning_rate": 9.811486268174476e-06,
|
| 76966 |
+
"loss": 0.5661,
|
| 76967 |
+
"step": 12675
|
| 76968 |
+
},
|
| 76969 |
+
{
|
| 76970 |
+
"epoch": 101.9,
|
| 76971 |
+
"learning_rate": 9.811470113085623e-06,
|
| 76972 |
+
"loss": 0.7745,
|
| 76973 |
+
"step": 12676
|
| 76974 |
+
},
|
| 76975 |
+
{
|
| 76976 |
+
"epoch": 101.91,
|
| 76977 |
+
"learning_rate": 9.81145395799677e-06,
|
| 76978 |
+
"loss": 0.7163,
|
| 76979 |
+
"step": 12677
|
| 76980 |
+
},
|
| 76981 |
+
{
|
| 76982 |
+
"epoch": 101.92,
|
| 76983 |
+
"learning_rate": 9.811437802907916e-06,
|
| 76984 |
+
"loss": 0.7856,
|
| 76985 |
+
"step": 12678
|
| 76986 |
+
},
|
| 76987 |
+
{
|
| 76988 |
+
"epoch": 101.92,
|
| 76989 |
+
"learning_rate": 9.811421647819063e-06,
|
| 76990 |
+
"loss": 0.7679,
|
| 76991 |
+
"step": 12679
|
| 76992 |
+
},
|
| 76993 |
+
{
|
| 76994 |
+
"epoch": 101.93,
|
| 76995 |
+
"learning_rate": 9.81140549273021e-06,
|
| 76996 |
+
"loss": 0.8701,
|
| 76997 |
+
"step": 12680
|
| 76998 |
+
},
|
| 76999 |
+
{
|
| 77000 |
+
"epoch": 101.94,
|
| 77001 |
+
"learning_rate": 9.811389337641358e-06,
|
| 77002 |
+
"loss": 0.7172,
|
| 77003 |
+
"step": 12681
|
| 77004 |
+
},
|
| 77005 |
+
{
|
| 77006 |
+
"epoch": 101.95,
|
| 77007 |
+
"learning_rate": 9.811373182552505e-06,
|
| 77008 |
+
"loss": 1.0521,
|
| 77009 |
+
"step": 12682
|
| 77010 |
+
},
|
| 77011 |
+
{
|
| 77012 |
+
"epoch": 101.96,
|
| 77013 |
+
"learning_rate": 9.811357027463652e-06,
|
| 77014 |
+
"loss": 0.886,
|
| 77015 |
+
"step": 12683
|
| 77016 |
+
},
|
| 77017 |
+
{
|
| 77018 |
+
"epoch": 101.96,
|
| 77019 |
+
"learning_rate": 9.811340872374799e-06,
|
| 77020 |
+
"loss": 1.2457,
|
| 77021 |
+
"step": 12684
|
| 77022 |
+
},
|
| 77023 |
+
{
|
| 77024 |
+
"epoch": 101.97,
|
| 77025 |
+
"learning_rate": 9.811324717285946e-06,
|
| 77026 |
+
"loss": 1.3873,
|
| 77027 |
+
"step": 12685
|
| 77028 |
+
},
|
| 77029 |
+
{
|
| 77030 |
+
"epoch": 101.98,
|
| 77031 |
+
"learning_rate": 9.811308562197093e-06,
|
| 77032 |
+
"loss": 1.3486,
|
| 77033 |
+
"step": 12686
|
| 77034 |
+
},
|
| 77035 |
+
{
|
| 77036 |
+
"epoch": 101.99,
|
| 77037 |
+
"learning_rate": 9.81129240710824e-06,
|
| 77038 |
+
"loss": 1.709,
|
| 77039 |
+
"step": 12687
|
| 77040 |
+
},
|
| 77041 |
+
{
|
| 77042 |
+
"epoch": 102.0,
|
| 77043 |
+
"learning_rate": 9.811276252019388e-06,
|
| 77044 |
+
"loss": 2.2157,
|
| 77045 |
+
"step": 12688
|
| 77046 |
+
},
|
| 77047 |
+
{
|
| 77048 |
+
"epoch": 102.0,
|
| 77049 |
+
"eval_loss": 0.6487979292869568,
|
| 77050 |
+
"eval_runtime": 37.9002,
|
| 77051 |
+
"eval_samples_per_second": 22.084,
|
| 77052 |
+
"eval_steps_per_second": 0.712,
|
| 77053 |
+
"eval_wer": 0.31620553359683795,
|
| 77054 |
+
"step": 12688
|
| 77055 |
}
|
| 77056 |
],
|
| 77057 |
"max_steps": 620000,
|
| 77058 |
"num_train_epochs": 5000,
|
| 77059 |
+
"total_flos": 3.5690837380972356e+19,
|
| 77060 |
"trial_name": null,
|
| 77061 |
"trial_params": null
|
| 77062 |
}
|
model-bin/finetune/base/{checkpoint-12440 → checkpoint-12688}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629501847.633379/events.out.tfevents.1629501847.35c5ee475cfb.1049.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:09be66cbc06885ee47bdb3e375e01c96b53b3a8e750c1417a4ba851d11ccb093
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629501847.35c5ee475cfb.1049.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:60dbb9363d6c6f1ccfe76e54d9f397c2b41d5a5f9bf9817e4c43a8ab30b9a267
|
| 3 |
+
size 24086
|