Commit
•
53819cf
1
Parent(s):
6e056eb
Upload 11 files
Browse files- README (7).md +9 -0
- adapter_config.json +22 -0
- adapter_model.bin +3 -0
- optimizer.pt +3 -0
- rng_state.pth +3 -0
- scheduler.pt +3 -0
- special_tokens_map.json +7 -0
- tokenizer_config.json +5 -0
- trainer_state.json +4485 -0
- training_args.bin +3 -0
- vocab.txt +33 -0
README (7).md
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
---
|
4 |
+
## Training procedure
|
5 |
+
|
6 |
+
### Framework versions
|
7 |
+
|
8 |
+
|
9 |
+
- PEFT 0.5.0
|
adapter_config.json
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"auto_mapping": null,
|
3 |
+
"base_model_name_or_path": "facebook/esm2_t12_35M_UR50D",
|
4 |
+
"bias": "none",
|
5 |
+
"fan_in_fan_out": false,
|
6 |
+
"inference_mode": true,
|
7 |
+
"init_lora_weights": true,
|
8 |
+
"layers_pattern": null,
|
9 |
+
"layers_to_transform": null,
|
10 |
+
"lora_alpha": 1,
|
11 |
+
"lora_dropout": 0.5,
|
12 |
+
"modules_to_save": null,
|
13 |
+
"peft_type": "LORA",
|
14 |
+
"r": 2,
|
15 |
+
"revision": null,
|
16 |
+
"target_modules": [
|
17 |
+
"query",
|
18 |
+
"key",
|
19 |
+
"value"
|
20 |
+
],
|
21 |
+
"task_type": "TOKEN_CLS"
|
22 |
+
}
|
adapter_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:14ac0473d827fc9e20f5e5429f4e592a0bbcbdbeede91e7614d648d96c1e214d
|
3 |
+
size 307151
|
optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:53d4ca6fbd99b8984cfaaca6d6fddb61f061190bc046996f90f1397838f88802
|
3 |
+
size 622661
|
rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:344c6c5dd0cc20065fee40ed0bc217fbb0ed68ec8c4e3c30b20fb4da3da7a5a7
|
3 |
+
size 14575
|
scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7509c51376a8e9c0c6b6cd48f9a71e6a0832c6d090a55a7c7908bbb0bd73583e
|
3 |
+
size 627
|
special_tokens_map.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cls_token": "<cls>",
|
3 |
+
"eos_token": "<eos>",
|
4 |
+
"mask_token": "<mask>",
|
5 |
+
"pad_token": "<pad>",
|
6 |
+
"unk_token": "<unk>"
|
7 |
+
}
|
tokenizer_config.json
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"clean_up_tokenization_spaces": true,
|
3 |
+
"model_max_length": 1024,
|
4 |
+
"tokenizer_class": "EsmTokenizer"
|
5 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,4485 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.35228549223252686,
|
3 |
+
"best_model_checkpoint": "esm2_t12_35M_lora_binding_sites_2023-09-24_04-19-04/checkpoint-148440",
|
4 |
+
"epoch": 1.0,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 148440,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.0,
|
13 |
+
"learning_rate": 0.00037015670298819826,
|
14 |
+
"loss": 0.5221,
|
15 |
+
"step": 200
|
16 |
+
},
|
17 |
+
{
|
18 |
+
"epoch": 0.0,
|
19 |
+
"learning_rate": 0.0003701563931502581,
|
20 |
+
"loss": 0.3727,
|
21 |
+
"step": 400
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"epoch": 0.0,
|
25 |
+
"learning_rate": 0.00037015587606324446,
|
26 |
+
"loss": 0.3152,
|
27 |
+
"step": 600
|
28 |
+
},
|
29 |
+
{
|
30 |
+
"epoch": 0.01,
|
31 |
+
"learning_rate": 0.00037015515172773665,
|
32 |
+
"loss": 0.2911,
|
33 |
+
"step": 800
|
34 |
+
},
|
35 |
+
{
|
36 |
+
"epoch": 0.01,
|
37 |
+
"learning_rate": 0.0003701542201445457,
|
38 |
+
"loss": 0.2675,
|
39 |
+
"step": 1000
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.01,
|
43 |
+
"learning_rate": 0.0003701530813147147,
|
44 |
+
"loss": 0.2512,
|
45 |
+
"step": 1200
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"epoch": 0.01,
|
49 |
+
"learning_rate": 0.00037015173523951894,
|
50 |
+
"loss": 0.2466,
|
51 |
+
"step": 1400
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"epoch": 0.01,
|
55 |
+
"learning_rate": 0.00037015018192046586,
|
56 |
+
"loss": 0.2271,
|
57 |
+
"step": 1600
|
58 |
+
},
|
59 |
+
{
|
60 |
+
"epoch": 0.01,
|
61 |
+
"learning_rate": 0.0003701484213592947,
|
62 |
+
"loss": 0.2265,
|
63 |
+
"step": 1800
|
64 |
+
},
|
65 |
+
{
|
66 |
+
"epoch": 0.01,
|
67 |
+
"learning_rate": 0.000370146453557977,
|
68 |
+
"loss": 0.2299,
|
69 |
+
"step": 2000
|
70 |
+
},
|
71 |
+
{
|
72 |
+
"epoch": 0.01,
|
73 |
+
"learning_rate": 0.0003701442899094131,
|
74 |
+
"loss": 0.22,
|
75 |
+
"step": 2200
|
76 |
+
},
|
77 |
+
{
|
78 |
+
"epoch": 0.02,
|
79 |
+
"learning_rate": 0.0003701419086708161,
|
80 |
+
"loss": 0.2157,
|
81 |
+
"step": 2400
|
82 |
+
},
|
83 |
+
{
|
84 |
+
"epoch": 0.02,
|
85 |
+
"learning_rate": 0.00037013932019936536,
|
86 |
+
"loss": 0.215,
|
87 |
+
"step": 2600
|
88 |
+
},
|
89 |
+
{
|
90 |
+
"epoch": 0.02,
|
91 |
+
"learning_rate": 0.0003701365244979595,
|
92 |
+
"loss": 0.2036,
|
93 |
+
"step": 2800
|
94 |
+
},
|
95 |
+
{
|
96 |
+
"epoch": 0.02,
|
97 |
+
"learning_rate": 0.00037013352156972917,
|
98 |
+
"loss": 0.1992,
|
99 |
+
"step": 3000
|
100 |
+
},
|
101 |
+
{
|
102 |
+
"epoch": 0.02,
|
103 |
+
"learning_rate": 0.00037013031141803696,
|
104 |
+
"loss": 0.2077,
|
105 |
+
"step": 3200
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"epoch": 0.02,
|
109 |
+
"learning_rate": 0.00037012689404647756,
|
110 |
+
"loss": 0.1882,
|
111 |
+
"step": 3400
|
112 |
+
},
|
113 |
+
{
|
114 |
+
"epoch": 0.02,
|
115 |
+
"learning_rate": 0.0003701232694588778,
|
116 |
+
"loss": 0.1935,
|
117 |
+
"step": 3600
|
118 |
+
},
|
119 |
+
{
|
120 |
+
"epoch": 0.03,
|
121 |
+
"learning_rate": 0.00037011943765929636,
|
122 |
+
"loss": 0.1966,
|
123 |
+
"step": 3800
|
124 |
+
},
|
125 |
+
{
|
126 |
+
"epoch": 0.03,
|
127 |
+
"learning_rate": 0.0003701154193624823,
|
128 |
+
"loss": 0.1852,
|
129 |
+
"step": 4000
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"epoch": 0.03,
|
133 |
+
"learning_rate": 0.0003701111741880463,
|
134 |
+
"loss": 0.1871,
|
135 |
+
"step": 4200
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"epoch": 0.03,
|
139 |
+
"learning_rate": 0.00037010672181517294,
|
140 |
+
"loss": 0.1653,
|
141 |
+
"step": 4400
|
142 |
+
},
|
143 |
+
{
|
144 |
+
"epoch": 0.03,
|
145 |
+
"learning_rate": 0.0003701020622488479,
|
146 |
+
"loss": 0.1865,
|
147 |
+
"step": 4600
|
148 |
+
},
|
149 |
+
{
|
150 |
+
"epoch": 0.03,
|
151 |
+
"learning_rate": 0.00037009719549428895,
|
152 |
+
"loss": 0.1746,
|
153 |
+
"step": 4800
|
154 |
+
},
|
155 |
+
{
|
156 |
+
"epoch": 0.03,
|
157 |
+
"learning_rate": 0.0003700921215569458,
|
158 |
+
"loss": 0.1787,
|
159 |
+
"step": 5000
|
160 |
+
},
|
161 |
+
{
|
162 |
+
"epoch": 0.04,
|
163 |
+
"learning_rate": 0.00037008684044250026,
|
164 |
+
"loss": 0.1873,
|
165 |
+
"step": 5200
|
166 |
+
},
|
167 |
+
{
|
168 |
+
"epoch": 0.04,
|
169 |
+
"learning_rate": 0.00037008135215686603,
|
170 |
+
"loss": 0.1742,
|
171 |
+
"step": 5400
|
172 |
+
},
|
173 |
+
{
|
174 |
+
"epoch": 0.04,
|
175 |
+
"learning_rate": 0.0003700756567061889,
|
176 |
+
"loss": 0.1777,
|
177 |
+
"step": 5600
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"epoch": 0.04,
|
181 |
+
"learning_rate": 0.0003700697540968466,
|
182 |
+
"loss": 0.1792,
|
183 |
+
"step": 5800
|
184 |
+
},
|
185 |
+
{
|
186 |
+
"epoch": 0.04,
|
187 |
+
"learning_rate": 0.00037006364433544883,
|
188 |
+
"loss": 0.1794,
|
189 |
+
"step": 6000
|
190 |
+
},
|
191 |
+
{
|
192 |
+
"epoch": 0.04,
|
193 |
+
"learning_rate": 0.0003700573274288372,
|
194 |
+
"loss": 0.1708,
|
195 |
+
"step": 6200
|
196 |
+
},
|
197 |
+
{
|
198 |
+
"epoch": 0.04,
|
199 |
+
"learning_rate": 0.00037005080338408546,
|
200 |
+
"loss": 0.161,
|
201 |
+
"step": 6400
|
202 |
+
},
|
203 |
+
{
|
204 |
+
"epoch": 0.04,
|
205 |
+
"learning_rate": 0.0003700440722084991,
|
206 |
+
"loss": 0.1661,
|
207 |
+
"step": 6600
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.05,
|
211 |
+
"learning_rate": 0.00037003720431783943,
|
212 |
+
"loss": 0.1833,
|
213 |
+
"step": 6800
|
214 |
+
},
|
215 |
+
{
|
216 |
+
"epoch": 0.05,
|
217 |
+
"learning_rate": 0.0003700300609745444,
|
218 |
+
"loss": 0.1802,
|
219 |
+
"step": 7000
|
220 |
+
},
|
221 |
+
{
|
222 |
+
"epoch": 0.05,
|
223 |
+
"learning_rate": 0.00037002271052364195,
|
224 |
+
"loss": 0.1557,
|
225 |
+
"step": 7200
|
226 |
+
},
|
227 |
+
{
|
228 |
+
"epoch": 0.05,
|
229 |
+
"learning_rate": 0.0003700151529733631,
|
230 |
+
"loss": 0.1763,
|
231 |
+
"step": 7400
|
232 |
+
},
|
233 |
+
{
|
234 |
+
"epoch": 0.05,
|
235 |
+
"learning_rate": 0.0003700073883321708,
|
236 |
+
"loss": 0.1661,
|
237 |
+
"step": 7600
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 0.05,
|
241 |
+
"learning_rate": 0.0003699994166087597,
|
242 |
+
"loss": 0.152,
|
243 |
+
"step": 7800
|
244 |
+
},
|
245 |
+
{
|
246 |
+
"epoch": 0.05,
|
247 |
+
"learning_rate": 0.00036999123781205663,
|
248 |
+
"loss": 0.1616,
|
249 |
+
"step": 8000
|
250 |
+
},
|
251 |
+
{
|
252 |
+
"epoch": 0.06,
|
253 |
+
"learning_rate": 0.00036998285195121995,
|
254 |
+
"loss": 0.1552,
|
255 |
+
"step": 8200
|
256 |
+
},
|
257 |
+
{
|
258 |
+
"epoch": 0.06,
|
259 |
+
"learning_rate": 0.00036997425903564034,
|
260 |
+
"loss": 0.1595,
|
261 |
+
"step": 8400
|
262 |
+
},
|
263 |
+
{
|
264 |
+
"epoch": 0.06,
|
265 |
+
"learning_rate": 0.00036996545907493994,
|
266 |
+
"loss": 0.1567,
|
267 |
+
"step": 8600
|
268 |
+
},
|
269 |
+
{
|
270 |
+
"epoch": 0.06,
|
271 |
+
"learning_rate": 0.0003699564520789729,
|
272 |
+
"loss": 0.1572,
|
273 |
+
"step": 8800
|
274 |
+
},
|
275 |
+
{
|
276 |
+
"epoch": 0.06,
|
277 |
+
"learning_rate": 0.00036994723805782516,
|
278 |
+
"loss": 0.1628,
|
279 |
+
"step": 9000
|
280 |
+
},
|
281 |
+
{
|
282 |
+
"epoch": 0.06,
|
283 |
+
"learning_rate": 0.00036993781702181466,
|
284 |
+
"loss": 0.1562,
|
285 |
+
"step": 9200
|
286 |
+
},
|
287 |
+
{
|
288 |
+
"epoch": 0.06,
|
289 |
+
"learning_rate": 0.00036992818898149083,
|
290 |
+
"loss": 0.1583,
|
291 |
+
"step": 9400
|
292 |
+
},
|
293 |
+
{
|
294 |
+
"epoch": 0.06,
|
295 |
+
"learning_rate": 0.00036991835394763526,
|
296 |
+
"loss": 0.1514,
|
297 |
+
"step": 9600
|
298 |
+
},
|
299 |
+
{
|
300 |
+
"epoch": 0.07,
|
301 |
+
"learning_rate": 0.000369908311931261,
|
302 |
+
"loss": 0.1651,
|
303 |
+
"step": 9800
|
304 |
+
},
|
305 |
+
{
|
306 |
+
"epoch": 0.07,
|
307 |
+
"learning_rate": 0.0003698980629436131,
|
308 |
+
"loss": 0.1488,
|
309 |
+
"step": 10000
|
310 |
+
},
|
311 |
+
{
|
312 |
+
"epoch": 0.07,
|
313 |
+
"learning_rate": 0.0003698876069961683,
|
314 |
+
"loss": 0.1471,
|
315 |
+
"step": 10200
|
316 |
+
},
|
317 |
+
{
|
318 |
+
"epoch": 0.07,
|
319 |
+
"learning_rate": 0.0003698769441006351,
|
320 |
+
"loss": 0.1516,
|
321 |
+
"step": 10400
|
322 |
+
},
|
323 |
+
{
|
324 |
+
"epoch": 0.07,
|
325 |
+
"learning_rate": 0.0003698660742689538,
|
326 |
+
"loss": 0.1411,
|
327 |
+
"step": 10600
|
328 |
+
},
|
329 |
+
{
|
330 |
+
"epoch": 0.07,
|
331 |
+
"learning_rate": 0.0003698549975132963,
|
332 |
+
"loss": 0.1573,
|
333 |
+
"step": 10800
|
334 |
+
},
|
335 |
+
{
|
336 |
+
"epoch": 0.07,
|
337 |
+
"learning_rate": 0.00036984377077907405,
|
338 |
+
"loss": 0.1472,
|
339 |
+
"step": 11000
|
340 |
+
},
|
341 |
+
{
|
342 |
+
"epoch": 0.08,
|
343 |
+
"learning_rate": 0.0003698322812473697,
|
344 |
+
"loss": 0.1445,
|
345 |
+
"step": 11200
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"epoch": 0.08,
|
349 |
+
"learning_rate": 0.0003698205848295303,
|
350 |
+
"loss": 0.1544,
|
351 |
+
"step": 11400
|
352 |
+
},
|
353 |
+
{
|
354 |
+
"epoch": 0.08,
|
355 |
+
"learning_rate": 0.0003698086815386536,
|
356 |
+
"loss": 0.1545,
|
357 |
+
"step": 11600
|
358 |
+
},
|
359 |
+
{
|
360 |
+
"epoch": 0.08,
|
361 |
+
"learning_rate": 0.00036979657138806857,
|
362 |
+
"loss": 0.1405,
|
363 |
+
"step": 11800
|
364 |
+
},
|
365 |
+
{
|
366 |
+
"epoch": 0.08,
|
367 |
+
"learning_rate": 0.00036978425439133624,
|
368 |
+
"loss": 0.1558,
|
369 |
+
"step": 12000
|
370 |
+
},
|
371 |
+
{
|
372 |
+
"epoch": 0.08,
|
373 |
+
"learning_rate": 0.00036977173056224906,
|
374 |
+
"loss": 0.154,
|
375 |
+
"step": 12200
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"epoch": 0.08,
|
379 |
+
"learning_rate": 0.0003697589999148311,
|
380 |
+
"loss": 0.1572,
|
381 |
+
"step": 12400
|
382 |
+
},
|
383 |
+
{
|
384 |
+
"epoch": 0.08,
|
385 |
+
"learning_rate": 0.0003697460624633381,
|
386 |
+
"loss": 0.1462,
|
387 |
+
"step": 12600
|
388 |
+
},
|
389 |
+
{
|
390 |
+
"epoch": 0.09,
|
391 |
+
"learning_rate": 0.0003697329182222573,
|
392 |
+
"loss": 0.1413,
|
393 |
+
"step": 12800
|
394 |
+
},
|
395 |
+
{
|
396 |
+
"epoch": 0.09,
|
397 |
+
"learning_rate": 0.0003697195672063075,
|
398 |
+
"loss": 0.1513,
|
399 |
+
"step": 13000
|
400 |
+
},
|
401 |
+
{
|
402 |
+
"epoch": 0.09,
|
403 |
+
"learning_rate": 0.00036970600943043915,
|
404 |
+
"loss": 0.1496,
|
405 |
+
"step": 13200
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.09,
|
409 |
+
"learning_rate": 0.00036969231424668924,
|
410 |
+
"loss": 0.1537,
|
411 |
+
"step": 13400
|
412 |
+
},
|
413 |
+
{
|
414 |
+
"epoch": 0.09,
|
415 |
+
"learning_rate": 0.0003696783440303689,
|
416 |
+
"loss": 0.147,
|
417 |
+
"step": 13600
|
418 |
+
},
|
419 |
+
{
|
420 |
+
"epoch": 0.09,
|
421 |
+
"learning_rate": 0.0003696641671002913,
|
422 |
+
"loss": 0.153,
|
423 |
+
"step": 13800
|
424 |
+
},
|
425 |
+
{
|
426 |
+
"epoch": 0.09,
|
427 |
+
"learning_rate": 0.0003696497834723318,
|
428 |
+
"loss": 0.1424,
|
429 |
+
"step": 14000
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 0.1,
|
433 |
+
"learning_rate": 0.0003696351931625971,
|
434 |
+
"loss": 0.1413,
|
435 |
+
"step": 14200
|
436 |
+
},
|
437 |
+
{
|
438 |
+
"epoch": 0.1,
|
439 |
+
"learning_rate": 0.00036962039618742523,
|
440 |
+
"loss": 0.1382,
|
441 |
+
"step": 14400
|
442 |
+
},
|
443 |
+
{
|
444 |
+
"epoch": 0.1,
|
445 |
+
"learning_rate": 0.00036960539256338583,
|
446 |
+
"loss": 0.1432,
|
447 |
+
"step": 14600
|
448 |
+
},
|
449 |
+
{
|
450 |
+
"epoch": 0.1,
|
451 |
+
"learning_rate": 0.0003695901823072799,
|
452 |
+
"loss": 0.1377,
|
453 |
+
"step": 14800
|
454 |
+
},
|
455 |
+
{
|
456 |
+
"epoch": 0.1,
|
457 |
+
"learning_rate": 0.0003695747654361396,
|
458 |
+
"loss": 0.1535,
|
459 |
+
"step": 15000
|
460 |
+
},
|
461 |
+
{
|
462 |
+
"epoch": 0.1,
|
463 |
+
"learning_rate": 0.0003695591419672288,
|
464 |
+
"loss": 0.1338,
|
465 |
+
"step": 15200
|
466 |
+
},
|
467 |
+
{
|
468 |
+
"epoch": 0.1,
|
469 |
+
"learning_rate": 0.0003695433915821276,
|
470 |
+
"loss": 0.1478,
|
471 |
+
"step": 15400
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"epoch": 0.11,
|
475 |
+
"learning_rate": 0.0003695273560031605,
|
476 |
+
"loss": 0.1427,
|
477 |
+
"step": 15600
|
478 |
+
},
|
479 |
+
{
|
480 |
+
"epoch": 0.11,
|
481 |
+
"learning_rate": 0.00036951111387951155,
|
482 |
+
"loss": 0.149,
|
483 |
+
"step": 15800
|
484 |
+
},
|
485 |
+
{
|
486 |
+
"epoch": 0.11,
|
487 |
+
"learning_rate": 0.00036949466522936866,
|
488 |
+
"loss": 0.1435,
|
489 |
+
"step": 16000
|
490 |
+
},
|
491 |
+
{
|
492 |
+
"epoch": 0.11,
|
493 |
+
"learning_rate": 0.0003694780100711509,
|
494 |
+
"loss": 0.1281,
|
495 |
+
"step": 16200
|
496 |
+
},
|
497 |
+
{
|
498 |
+
"epoch": 0.11,
|
499 |
+
"learning_rate": 0.00036946123324535805,
|
500 |
+
"loss": 0.156,
|
501 |
+
"step": 16400
|
502 |
+
},
|
503 |
+
{
|
504 |
+
"epoch": 0.11,
|
505 |
+
"learning_rate": 0.0003694441661594781,
|
506 |
+
"loss": 0.1463,
|
507 |
+
"step": 16600
|
508 |
+
},
|
509 |
+
{
|
510 |
+
"epoch": 0.11,
|
511 |
+
"learning_rate": 0.0003694268926220718,
|
512 |
+
"loss": 0.1403,
|
513 |
+
"step": 16800
|
514 |
+
},
|
515 |
+
{
|
516 |
+
"epoch": 0.11,
|
517 |
+
"learning_rate": 0.00036940941265248186,
|
518 |
+
"loss": 0.1346,
|
519 |
+
"step": 17000
|
520 |
+
},
|
521 |
+
{
|
522 |
+
"epoch": 0.12,
|
523 |
+
"learning_rate": 0.0003693917262702823,
|
524 |
+
"loss": 0.1271,
|
525 |
+
"step": 17200
|
526 |
+
},
|
527 |
+
{
|
528 |
+
"epoch": 0.12,
|
529 |
+
"learning_rate": 0.0003693738334952782,
|
530 |
+
"loss": 0.1397,
|
531 |
+
"step": 17400
|
532 |
+
},
|
533 |
+
{
|
534 |
+
"epoch": 0.12,
|
535 |
+
"learning_rate": 0.0003693557343475058,
|
536 |
+
"loss": 0.1281,
|
537 |
+
"step": 17600
|
538 |
+
},
|
539 |
+
{
|
540 |
+
"epoch": 0.12,
|
541 |
+
"learning_rate": 0.0003693374288472324,
|
542 |
+
"loss": 0.1448,
|
543 |
+
"step": 17800
|
544 |
+
},
|
545 |
+
{
|
546 |
+
"epoch": 0.12,
|
547 |
+
"learning_rate": 0.00036931891701495626,
|
548 |
+
"loss": 0.1354,
|
549 |
+
"step": 18000
|
550 |
+
},
|
551 |
+
{
|
552 |
+
"epoch": 0.12,
|
553 |
+
"learning_rate": 0.0003693001988714069,
|
554 |
+
"loss": 0.1408,
|
555 |
+
"step": 18200
|
556 |
+
},
|
557 |
+
{
|
558 |
+
"epoch": 0.12,
|
559 |
+
"learning_rate": 0.0003692812744375448,
|
560 |
+
"loss": 0.131,
|
561 |
+
"step": 18400
|
562 |
+
},
|
563 |
+
{
|
564 |
+
"epoch": 0.13,
|
565 |
+
"learning_rate": 0.0003692621437345614,
|
566 |
+
"loss": 0.1347,
|
567 |
+
"step": 18600
|
568 |
+
},
|
569 |
+
{
|
570 |
+
"epoch": 0.13,
|
571 |
+
"learning_rate": 0.00036924280678387906,
|
572 |
+
"loss": 0.1622,
|
573 |
+
"step": 18800
|
574 |
+
},
|
575 |
+
{
|
576 |
+
"epoch": 0.13,
|
577 |
+
"learning_rate": 0.00036922326360715115,
|
578 |
+
"loss": 0.1395,
|
579 |
+
"step": 19000
|
580 |
+
},
|
581 |
+
{
|
582 |
+
"epoch": 0.13,
|
583 |
+
"learning_rate": 0.0003692035142262621,
|
584 |
+
"loss": 0.1348,
|
585 |
+
"step": 19200
|
586 |
+
},
|
587 |
+
{
|
588 |
+
"epoch": 0.13,
|
589 |
+
"learning_rate": 0.00036918355866332706,
|
590 |
+
"loss": 0.1491,
|
591 |
+
"step": 19400
|
592 |
+
},
|
593 |
+
{
|
594 |
+
"epoch": 0.13,
|
595 |
+
"learning_rate": 0.0003691633969406922,
|
596 |
+
"loss": 0.1351,
|
597 |
+
"step": 19600
|
598 |
+
},
|
599 |
+
{
|
600 |
+
"epoch": 0.13,
|
601 |
+
"learning_rate": 0.00036914302908093444,
|
602 |
+
"loss": 0.1387,
|
603 |
+
"step": 19800
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 0.13,
|
607 |
+
"learning_rate": 0.0003691224551068615,
|
608 |
+
"loss": 0.1466,
|
609 |
+
"step": 20000
|
610 |
+
},
|
611 |
+
{
|
612 |
+
"epoch": 0.14,
|
613 |
+
"learning_rate": 0.0003691016750415121,
|
614 |
+
"loss": 0.1238,
|
615 |
+
"step": 20200
|
616 |
+
},
|
617 |
+
{
|
618 |
+
"epoch": 0.14,
|
619 |
+
"learning_rate": 0.00036908068890815544,
|
620 |
+
"loss": 0.1465,
|
621 |
+
"step": 20400
|
622 |
+
},
|
623 |
+
{
|
624 |
+
"epoch": 0.14,
|
625 |
+
"learning_rate": 0.00036905949673029186,
|
626 |
+
"loss": 0.1405,
|
627 |
+
"step": 20600
|
628 |
+
},
|
629 |
+
{
|
630 |
+
"epoch": 0.14,
|
631 |
+
"learning_rate": 0.0003690382060350824,
|
632 |
+
"loss": 0.1367,
|
633 |
+
"step": 20800
|
634 |
+
},
|
635 |
+
{
|
636 |
+
"epoch": 0.14,
|
637 |
+
"learning_rate": 0.0003690166028695521,
|
638 |
+
"loss": 0.138,
|
639 |
+
"step": 21000
|
640 |
+
},
|
641 |
+
{
|
642 |
+
"epoch": 0.14,
|
643 |
+
"learning_rate": 0.00036899479373127795,
|
644 |
+
"loss": 0.1339,
|
645 |
+
"step": 21200
|
646 |
+
},
|
647 |
+
{
|
648 |
+
"epoch": 0.14,
|
649 |
+
"learning_rate": 0.0003689727786446815,
|
650 |
+
"loss": 0.1398,
|
651 |
+
"step": 21400
|
652 |
+
},
|
653 |
+
{
|
654 |
+
"epoch": 0.15,
|
655 |
+
"learning_rate": 0.0003689505576344153,
|
656 |
+
"loss": 0.1278,
|
657 |
+
"step": 21600
|
658 |
+
},
|
659 |
+
{
|
660 |
+
"epoch": 0.15,
|
661 |
+
"learning_rate": 0.0003689281307253622,
|
662 |
+
"loss": 0.1472,
|
663 |
+
"step": 21800
|
664 |
+
},
|
665 |
+
{
|
666 |
+
"epoch": 0.15,
|
667 |
+
"learning_rate": 0.00036890549794263563,
|
668 |
+
"loss": 0.1371,
|
669 |
+
"step": 22000
|
670 |
+
},
|
671 |
+
{
|
672 |
+
"epoch": 0.15,
|
673 |
+
"learning_rate": 0.00036888265931157973,
|
674 |
+
"loss": 0.1401,
|
675 |
+
"step": 22200
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"epoch": 0.15,
|
679 |
+
"learning_rate": 0.000368859614857769,
|
680 |
+
"loss": 0.1349,
|
681 |
+
"step": 22400
|
682 |
+
},
|
683 |
+
{
|
684 |
+
"epoch": 0.15,
|
685 |
+
"learning_rate": 0.0003688363646070085,
|
686 |
+
"loss": 0.1389,
|
687 |
+
"step": 22600
|
688 |
+
},
|
689 |
+
{
|
690 |
+
"epoch": 0.15,
|
691 |
+
"learning_rate": 0.0003688130263772538,
|
692 |
+
"loss": 0.1367,
|
693 |
+
"step": 22800
|
694 |
+
},
|
695 |
+
{
|
696 |
+
"epoch": 0.15,
|
697 |
+
"learning_rate": 0.000368789365639588,
|
698 |
+
"loss": 0.138,
|
699 |
+
"step": 23000
|
700 |
+
},
|
701 |
+
{
|
702 |
+
"epoch": 0.16,
|
703 |
+
"learning_rate": 0.000368765499183637,
|
704 |
+
"loss": 0.1284,
|
705 |
+
"step": 23200
|
706 |
+
},
|
707 |
+
{
|
708 |
+
"epoch": 0.16,
|
709 |
+
"learning_rate": 0.00036874142703612636,
|
710 |
+
"loss": 0.1224,
|
711 |
+
"step": 23400
|
712 |
+
},
|
713 |
+
{
|
714 |
+
"epoch": 0.16,
|
715 |
+
"learning_rate": 0.00036871714922401174,
|
716 |
+
"loss": 0.1329,
|
717 |
+
"step": 23600
|
718 |
+
},
|
719 |
+
{
|
720 |
+
"epoch": 0.16,
|
721 |
+
"learning_rate": 0.00036869266577447944,
|
722 |
+
"loss": 0.1408,
|
723 |
+
"step": 23800
|
724 |
+
},
|
725 |
+
{
|
726 |
+
"epoch": 0.16,
|
727 |
+
"learning_rate": 0.0003686679767149457,
|
728 |
+
"loss": 0.1425,
|
729 |
+
"step": 24000
|
730 |
+
},
|
731 |
+
{
|
732 |
+
"epoch": 0.16,
|
733 |
+
"learning_rate": 0.0003686430820730573,
|
734 |
+
"loss": 0.1288,
|
735 |
+
"step": 24200
|
736 |
+
},
|
737 |
+
{
|
738 |
+
"epoch": 0.16,
|
739 |
+
"learning_rate": 0.0003686179818766911,
|
740 |
+
"loss": 0.1299,
|
741 |
+
"step": 24400
|
742 |
+
},
|
743 |
+
{
|
744 |
+
"epoch": 0.17,
|
745 |
+
"learning_rate": 0.00036859267615395395,
|
746 |
+
"loss": 0.1377,
|
747 |
+
"step": 24600
|
748 |
+
},
|
749 |
+
{
|
750 |
+
"epoch": 0.17,
|
751 |
+
"learning_rate": 0.00036856729300041623,
|
752 |
+
"loss": 0.1319,
|
753 |
+
"step": 24800
|
754 |
+
},
|
755 |
+
{
|
756 |
+
"epoch": 0.17,
|
757 |
+
"learning_rate": 0.00036854157733745503,
|
758 |
+
"loss": 0.1284,
|
759 |
+
"step": 25000
|
760 |
+
},
|
761 |
+
{
|
762 |
+
"epoch": 0.17,
|
763 |
+
"learning_rate": 0.00036851565623368017,
|
764 |
+
"loss": 0.1223,
|
765 |
+
"step": 25200
|
766 |
+
},
|
767 |
+
{
|
768 |
+
"epoch": 0.17,
|
769 |
+
"learning_rate": 0.0003684895297181181,
|
770 |
+
"loss": 0.146,
|
771 |
+
"step": 25400
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"epoch": 0.17,
|
775 |
+
"learning_rate": 0.0003684631978200251,
|
776 |
+
"loss": 0.1486,
|
777 |
+
"step": 25600
|
778 |
+
},
|
779 |
+
{
|
780 |
+
"epoch": 0.17,
|
781 |
+
"learning_rate": 0.00036843666056888733,
|
782 |
+
"loss": 0.1227,
|
783 |
+
"step": 25800
|
784 |
+
},
|
785 |
+
{
|
786 |
+
"epoch": 0.18,
|
787 |
+
"learning_rate": 0.0003684099179944211,
|
788 |
+
"loss": 0.1254,
|
789 |
+
"step": 26000
|
790 |
+
},
|
791 |
+
{
|
792 |
+
"epoch": 0.18,
|
793 |
+
"learning_rate": 0.0003683829701265725,
|
794 |
+
"loss": 0.1289,
|
795 |
+
"step": 26200
|
796 |
+
},
|
797 |
+
{
|
798 |
+
"epoch": 0.18,
|
799 |
+
"learning_rate": 0.0003683558169955177,
|
800 |
+
"loss": 0.129,
|
801 |
+
"step": 26400
|
802 |
+
},
|
803 |
+
{
|
804 |
+
"epoch": 0.18,
|
805 |
+
"learning_rate": 0.00036832845863166245,
|
806 |
+
"loss": 0.1296,
|
807 |
+
"step": 26600
|
808 |
+
},
|
809 |
+
{
|
810 |
+
"epoch": 0.18,
|
811 |
+
"learning_rate": 0.0003683008950656426,
|
812 |
+
"loss": 0.1338,
|
813 |
+
"step": 26800
|
814 |
+
},
|
815 |
+
{
|
816 |
+
"epoch": 0.18,
|
817 |
+
"learning_rate": 0.0003682731263283235,
|
818 |
+
"loss": 0.1383,
|
819 |
+
"step": 27000
|
820 |
+
},
|
821 |
+
{
|
822 |
+
"epoch": 0.18,
|
823 |
+
"learning_rate": 0.0003682452928304226,
|
824 |
+
"loss": 0.1334,
|
825 |
+
"step": 27200
|
826 |
+
},
|
827 |
+
{
|
828 |
+
"epoch": 0.18,
|
829 |
+
"learning_rate": 0.0003682171148694867,
|
830 |
+
"loss": 0.1387,
|
831 |
+
"step": 27400
|
832 |
+
},
|
833 |
+
{
|
834 |
+
"epoch": 0.19,
|
835 |
+
"learning_rate": 0.0003681887318310681,
|
836 |
+
"loss": 0.1242,
|
837 |
+
"step": 27600
|
838 |
+
},
|
839 |
+
{
|
840 |
+
"epoch": 0.19,
|
841 |
+
"learning_rate": 0.00036816014374694997,
|
842 |
+
"loss": 0.1278,
|
843 |
+
"step": 27800
|
844 |
+
},
|
845 |
+
{
|
846 |
+
"epoch": 0.19,
|
847 |
+
"learning_rate": 0.000368131350649145,
|
848 |
+
"loss": 0.133,
|
849 |
+
"step": 28000
|
850 |
+
},
|
851 |
+
{
|
852 |
+
"epoch": 0.19,
|
853 |
+
"learning_rate": 0.0003681023525698955,
|
854 |
+
"loss": 0.1261,
|
855 |
+
"step": 28200
|
856 |
+
},
|
857 |
+
{
|
858 |
+
"epoch": 0.19,
|
859 |
+
"learning_rate": 0.0003680731495416734,
|
860 |
+
"loss": 0.1358,
|
861 |
+
"step": 28400
|
862 |
+
},
|
863 |
+
{
|
864 |
+
"epoch": 0.19,
|
865 |
+
"learning_rate": 0.00036804374159717996,
|
866 |
+
"loss": 0.1362,
|
867 |
+
"step": 28600
|
868 |
+
},
|
869 |
+
{
|
870 |
+
"epoch": 0.19,
|
871 |
+
"learning_rate": 0.0003680141287693461,
|
872 |
+
"loss": 0.1291,
|
873 |
+
"step": 28800
|
874 |
+
},
|
875 |
+
{
|
876 |
+
"epoch": 0.2,
|
877 |
+
"learning_rate": 0.00036798431109133197,
|
878 |
+
"loss": 0.1252,
|
879 |
+
"step": 29000
|
880 |
+
},
|
881 |
+
{
|
882 |
+
"epoch": 0.2,
|
883 |
+
"learning_rate": 0.0003679542885965272,
|
884 |
+
"loss": 0.1286,
|
885 |
+
"step": 29200
|
886 |
+
},
|
887 |
+
{
|
888 |
+
"epoch": 0.2,
|
889 |
+
"learning_rate": 0.0003679240613185508,
|
890 |
+
"loss": 0.1281,
|
891 |
+
"step": 29400
|
892 |
+
},
|
893 |
+
{
|
894 |
+
"epoch": 0.2,
|
895 |
+
"learning_rate": 0.0003678937819606453,
|
896 |
+
"loss": 0.1355,
|
897 |
+
"step": 29600
|
898 |
+
},
|
899 |
+
{
|
900 |
+
"epoch": 0.2,
|
901 |
+
"learning_rate": 0.00036786314624159095,
|
902 |
+
"loss": 0.1357,
|
903 |
+
"step": 29800
|
904 |
+
},
|
905 |
+
{
|
906 |
+
"epoch": 0.2,
|
907 |
+
"learning_rate": 0.0003678323058414256,
|
908 |
+
"loss": 0.1293,
|
909 |
+
"step": 30000
|
910 |
+
},
|
911 |
+
{
|
912 |
+
"epoch": 0.2,
|
913 |
+
"learning_rate": 0.0003678012607946841,
|
914 |
+
"loss": 0.132,
|
915 |
+
"step": 30200
|
916 |
+
},
|
917 |
+
{
|
918 |
+
"epoch": 0.2,
|
919 |
+
"learning_rate": 0.0003677700111361304,
|
920 |
+
"loss": 0.1237,
|
921 |
+
"step": 30400
|
922 |
+
},
|
923 |
+
{
|
924 |
+
"epoch": 0.21,
|
925 |
+
"learning_rate": 0.0003677385569007577,
|
926 |
+
"loss": 0.1306,
|
927 |
+
"step": 30600
|
928 |
+
},
|
929 |
+
{
|
930 |
+
"epoch": 0.21,
|
931 |
+
"learning_rate": 0.0003677068981237883,
|
932 |
+
"loss": 0.1231,
|
933 |
+
"step": 30800
|
934 |
+
},
|
935 |
+
{
|
936 |
+
"epoch": 0.21,
|
937 |
+
"learning_rate": 0.00036767519466573924,
|
938 |
+
"loss": 0.1405,
|
939 |
+
"step": 31000
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 0.21,
|
943 |
+
"learning_rate": 0.0003676431279344224,
|
944 |
+
"loss": 0.127,
|
945 |
+
"step": 31200
|
946 |
+
},
|
947 |
+
{
|
948 |
+
"epoch": 0.21,
|
949 |
+
"learning_rate": 0.00036761085676836956,
|
950 |
+
"loss": 0.1336,
|
951 |
+
"step": 31400
|
952 |
+
},
|
953 |
+
{
|
954 |
+
"epoch": 0.21,
|
955 |
+
"learning_rate": 0.0003675783812037177,
|
956 |
+
"loss": 0.1346,
|
957 |
+
"step": 31600
|
958 |
+
},
|
959 |
+
{
|
960 |
+
"epoch": 0.21,
|
961 |
+
"learning_rate": 0.0003675457012768328,
|
962 |
+
"loss": 0.117,
|
963 |
+
"step": 31800
|
964 |
+
},
|
965 |
+
{
|
966 |
+
"epoch": 0.22,
|
967 |
+
"learning_rate": 0.00036751281702430955,
|
968 |
+
"loss": 0.1277,
|
969 |
+
"step": 32000
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 0.22,
|
973 |
+
"learning_rate": 0.0003674797284829716,
|
974 |
+
"loss": 0.1181,
|
975 |
+
"step": 32200
|
976 |
+
},
|
977 |
+
{
|
978 |
+
"epoch": 0.22,
|
979 |
+
"learning_rate": 0.0003674464356898712,
|
980 |
+
"loss": 0.124,
|
981 |
+
"step": 32400
|
982 |
+
},
|
983 |
+
{
|
984 |
+
"epoch": 0.22,
|
985 |
+
"learning_rate": 0.0003674129386822893,
|
986 |
+
"loss": 0.137,
|
987 |
+
"step": 32600
|
988 |
+
},
|
989 |
+
{
|
990 |
+
"epoch": 0.22,
|
991 |
+
"learning_rate": 0.0003673792374977358,
|
992 |
+
"loss": 0.1154,
|
993 |
+
"step": 32800
|
994 |
+
},
|
995 |
+
{
|
996 |
+
"epoch": 0.22,
|
997 |
+
"learning_rate": 0.000367345332173949,
|
998 |
+
"loss": 0.1548,
|
999 |
+
"step": 33000
|
1000 |
+
},
|
1001 |
+
{
|
1002 |
+
"epoch": 0.22,
|
1003 |
+
"learning_rate": 0.0003673112227488958,
|
1004 |
+
"loss": 0.1304,
|
1005 |
+
"step": 33200
|
1006 |
+
},
|
1007 |
+
{
|
1008 |
+
"epoch": 0.23,
|
1009 |
+
"learning_rate": 0.0003672769092607718,
|
1010 |
+
"loss": 0.113,
|
1011 |
+
"step": 33400
|
1012 |
+
},
|
1013 |
+
{
|
1014 |
+
"epoch": 0.23,
|
1015 |
+
"learning_rate": 0.0003672423917480009,
|
1016 |
+
"loss": 0.1383,
|
1017 |
+
"step": 33600
|
1018 |
+
},
|
1019 |
+
{
|
1020 |
+
"epoch": 0.23,
|
1021 |
+
"learning_rate": 0.0003672076702492356,
|
1022 |
+
"loss": 0.1286,
|
1023 |
+
"step": 33800
|
1024 |
+
},
|
1025 |
+
{
|
1026 |
+
"epoch": 0.23,
|
1027 |
+
"learning_rate": 0.0003671727448033568,
|
1028 |
+
"loss": 0.1233,
|
1029 |
+
"step": 34000
|
1030 |
+
},
|
1031 |
+
{
|
1032 |
+
"epoch": 0.23,
|
1033 |
+
"learning_rate": 0.0003671376154494739,
|
1034 |
+
"loss": 0.1311,
|
1035 |
+
"step": 34200
|
1036 |
+
},
|
1037 |
+
{
|
1038 |
+
"epoch": 0.23,
|
1039 |
+
"learning_rate": 0.00036710228222692435,
|
1040 |
+
"loss": 0.1342,
|
1041 |
+
"step": 34400
|
1042 |
+
},
|
1043 |
+
{
|
1044 |
+
"epoch": 0.23,
|
1045 |
+
"learning_rate": 0.0003670669233674915,
|
1046 |
+
"loss": 0.1317,
|
1047 |
+
"step": 34600
|
1048 |
+
},
|
1049 |
+
{
|
1050 |
+
"epoch": 0.23,
|
1051 |
+
"learning_rate": 0.0003670311835453819,
|
1052 |
+
"loss": 0.1343,
|
1053 |
+
"step": 34800
|
1054 |
+
},
|
1055 |
+
{
|
1056 |
+
"epoch": 0.24,
|
1057 |
+
"learning_rate": 0.00036699523997378743,
|
1058 |
+
"loss": 0.1177,
|
1059 |
+
"step": 35000
|
1060 |
+
},
|
1061 |
+
{
|
1062 |
+
"epoch": 0.24,
|
1063 |
+
"learning_rate": 0.00036695909269295746,
|
1064 |
+
"loss": 0.1269,
|
1065 |
+
"step": 35200
|
1066 |
+
},
|
1067 |
+
{
|
1068 |
+
"epoch": 0.24,
|
1069 |
+
"learning_rate": 0.0003669227417433694,
|
1070 |
+
"loss": 0.1374,
|
1071 |
+
"step": 35400
|
1072 |
+
},
|
1073 |
+
{
|
1074 |
+
"epoch": 0.24,
|
1075 |
+
"learning_rate": 0.0003668861871657289,
|
1076 |
+
"loss": 0.1216,
|
1077 |
+
"step": 35600
|
1078 |
+
},
|
1079 |
+
{
|
1080 |
+
"epoch": 0.24,
|
1081 |
+
"learning_rate": 0.00036684942900096955,
|
1082 |
+
"loss": 0.1234,
|
1083 |
+
"step": 35800
|
1084 |
+
},
|
1085 |
+
{
|
1086 |
+
"epoch": 0.24,
|
1087 |
+
"learning_rate": 0.00036681246729025277,
|
1088 |
+
"loss": 0.1312,
|
1089 |
+
"step": 36000
|
1090 |
+
},
|
1091 |
+
{
|
1092 |
+
"epoch": 0.24,
|
1093 |
+
"learning_rate": 0.0003667753020749681,
|
1094 |
+
"loss": 0.1219,
|
1095 |
+
"step": 36200
|
1096 |
+
},
|
1097 |
+
{
|
1098 |
+
"epoch": 0.25,
|
1099 |
+
"learning_rate": 0.0003667379333967329,
|
1100 |
+
"loss": 0.1231,
|
1101 |
+
"step": 36400
|
1102 |
+
},
|
1103 |
+
{
|
1104 |
+
"epoch": 0.25,
|
1105 |
+
"learning_rate": 0.00036670036129739236,
|
1106 |
+
"loss": 0.1236,
|
1107 |
+
"step": 36600
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 0.25,
|
1111 |
+
"learning_rate": 0.0003666625858190194,
|
1112 |
+
"loss": 0.1299,
|
1113 |
+
"step": 36800
|
1114 |
+
},
|
1115 |
+
{
|
1116 |
+
"epoch": 0.25,
|
1117 |
+
"learning_rate": 0.00036662460700391487,
|
1118 |
+
"loss": 0.1284,
|
1119 |
+
"step": 37000
|
1120 |
+
},
|
1121 |
+
{
|
1122 |
+
"epoch": 0.25,
|
1123 |
+
"learning_rate": 0.0003665864248946071,
|
1124 |
+
"loss": 0.1276,
|
1125 |
+
"step": 37200
|
1126 |
+
},
|
1127 |
+
{
|
1128 |
+
"epoch": 0.25,
|
1129 |
+
"learning_rate": 0.00036654803953385206,
|
1130 |
+
"loss": 0.1227,
|
1131 |
+
"step": 37400
|
1132 |
+
},
|
1133 |
+
{
|
1134 |
+
"epoch": 0.25,
|
1135 |
+
"learning_rate": 0.00036650945096463363,
|
1136 |
+
"loss": 0.1275,
|
1137 |
+
"step": 37600
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 0.25,
|
1141 |
+
"learning_rate": 0.0003664706592301629,
|
1142 |
+
"loss": 0.1338,
|
1143 |
+
"step": 37800
|
1144 |
+
},
|
1145 |
+
{
|
1146 |
+
"epoch": 0.26,
|
1147 |
+
"learning_rate": 0.00036643166437387857,
|
1148 |
+
"loss": 0.1248,
|
1149 |
+
"step": 38000
|
1150 |
+
},
|
1151 |
+
{
|
1152 |
+
"epoch": 0.26,
|
1153 |
+
"learning_rate": 0.00036639246643944696,
|
1154 |
+
"loss": 0.1234,
|
1155 |
+
"step": 38200
|
1156 |
+
},
|
1157 |
+
{
|
1158 |
+
"epoch": 0.26,
|
1159 |
+
"learning_rate": 0.00036635306547076144,
|
1160 |
+
"loss": 0.1274,
|
1161 |
+
"step": 38400
|
1162 |
+
},
|
1163 |
+
{
|
1164 |
+
"epoch": 0.26,
|
1165 |
+
"learning_rate": 0.0003663134615119431,
|
1166 |
+
"loss": 0.1211,
|
1167 |
+
"step": 38600
|
1168 |
+
},
|
1169 |
+
{
|
1170 |
+
"epoch": 0.26,
|
1171 |
+
"learning_rate": 0.00036627365460734014,
|
1172 |
+
"loss": 0.1162,
|
1173 |
+
"step": 38800
|
1174 |
+
},
|
1175 |
+
{
|
1176 |
+
"epoch": 0.26,
|
1177 |
+
"learning_rate": 0.0003662336448015281,
|
1178 |
+
"loss": 0.1093,
|
1179 |
+
"step": 39000
|
1180 |
+
},
|
1181 |
+
{
|
1182 |
+
"epoch": 0.26,
|
1183 |
+
"learning_rate": 0.00036619363370715154,
|
1184 |
+
"loss": 0.1281,
|
1185 |
+
"step": 39200
|
1186 |
+
},
|
1187 |
+
{
|
1188 |
+
"epoch": 0.27,
|
1189 |
+
"learning_rate": 0.000366153219247501,
|
1190 |
+
"loss": 0.1143,
|
1191 |
+
"step": 39400
|
1192 |
+
},
|
1193 |
+
{
|
1194 |
+
"epoch": 0.27,
|
1195 |
+
"learning_rate": 0.00036611260202150397,
|
1196 |
+
"loss": 0.1301,
|
1197 |
+
"step": 39600
|
1198 |
+
},
|
1199 |
+
{
|
1200 |
+
"epoch": 0.27,
|
1201 |
+
"learning_rate": 0.0003660717820746435,
|
1202 |
+
"loss": 0.1312,
|
1203 |
+
"step": 39800
|
1204 |
+
},
|
1205 |
+
{
|
1206 |
+
"epoch": 0.27,
|
1207 |
+
"learning_rate": 0.0003660307594526293,
|
1208 |
+
"loss": 0.1249,
|
1209 |
+
"step": 40000
|
1210 |
+
},
|
1211 |
+
{
|
1212 |
+
"epoch": 0.27,
|
1213 |
+
"learning_rate": 0.00036598953420139834,
|
1214 |
+
"loss": 0.13,
|
1215 |
+
"step": 40200
|
1216 |
+
},
|
1217 |
+
{
|
1218 |
+
"epoch": 0.27,
|
1219 |
+
"learning_rate": 0.00036594810636711434,
|
1220 |
+
"loss": 0.1215,
|
1221 |
+
"step": 40400
|
1222 |
+
},
|
1223 |
+
{
|
1224 |
+
"epoch": 0.27,
|
1225 |
+
"learning_rate": 0.000365906475996168,
|
1226 |
+
"loss": 0.1347,
|
1227 |
+
"step": 40600
|
1228 |
+
},
|
1229 |
+
{
|
1230 |
+
"epoch": 0.27,
|
1231 |
+
"learning_rate": 0.00036586464313517655,
|
1232 |
+
"loss": 0.1219,
|
1233 |
+
"step": 40800
|
1234 |
+
},
|
1235 |
+
{
|
1236 |
+
"epoch": 0.28,
|
1237 |
+
"learning_rate": 0.00036582260783098424,
|
1238 |
+
"loss": 0.1228,
|
1239 |
+
"step": 41000
|
1240 |
+
},
|
1241 |
+
{
|
1242 |
+
"epoch": 0.28,
|
1243 |
+
"learning_rate": 0.000365780370130662,
|
1244 |
+
"loss": 0.1263,
|
1245 |
+
"step": 41200
|
1246 |
+
},
|
1247 |
+
{
|
1248 |
+
"epoch": 0.28,
|
1249 |
+
"learning_rate": 0.0003657379300815072,
|
1250 |
+
"loss": 0.1278,
|
1251 |
+
"step": 41400
|
1252 |
+
},
|
1253 |
+
{
|
1254 |
+
"epoch": 0.28,
|
1255 |
+
"learning_rate": 0.00036569528773104393,
|
1256 |
+
"loss": 0.1191,
|
1257 |
+
"step": 41600
|
1258 |
+
},
|
1259 |
+
{
|
1260 |
+
"epoch": 0.28,
|
1261 |
+
"learning_rate": 0.0003656526578530694,
|
1262 |
+
"loss": 0.1304,
|
1263 |
+
"step": 41800
|
1264 |
+
},
|
1265 |
+
{
|
1266 |
+
"epoch": 0.28,
|
1267 |
+
"learning_rate": 0.00036560961205437574,
|
1268 |
+
"loss": 0.1327,
|
1269 |
+
"step": 42000
|
1270 |
+
},
|
1271 |
+
{
|
1272 |
+
"epoch": 0.28,
|
1273 |
+
"learning_rate": 0.0003655663640980633,
|
1274 |
+
"loss": 0.1281,
|
1275 |
+
"step": 42200
|
1276 |
+
},
|
1277 |
+
{
|
1278 |
+
"epoch": 0.29,
|
1279 |
+
"learning_rate": 0.0003655229140325608,
|
1280 |
+
"loss": 0.1266,
|
1281 |
+
"step": 42400
|
1282 |
+
},
|
1283 |
+
{
|
1284 |
+
"epoch": 0.29,
|
1285 |
+
"learning_rate": 0.0003654792619065234,
|
1286 |
+
"loss": 0.1142,
|
1287 |
+
"step": 42600
|
1288 |
+
},
|
1289 |
+
{
|
1290 |
+
"epoch": 0.29,
|
1291 |
+
"learning_rate": 0.0003654354077688324,
|
1292 |
+
"loss": 0.1087,
|
1293 |
+
"step": 42800
|
1294 |
+
},
|
1295 |
+
{
|
1296 |
+
"epoch": 0.29,
|
1297 |
+
"learning_rate": 0.0003653913516685954,
|
1298 |
+
"loss": 0.1197,
|
1299 |
+
"step": 43000
|
1300 |
+
},
|
1301 |
+
{
|
1302 |
+
"epoch": 0.29,
|
1303 |
+
"learning_rate": 0.00036534709365514605,
|
1304 |
+
"loss": 0.1249,
|
1305 |
+
"step": 43200
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 0.29,
|
1309 |
+
"learning_rate": 0.0003653026337780443,
|
1310 |
+
"loss": 0.1217,
|
1311 |
+
"step": 43400
|
1312 |
+
},
|
1313 |
+
{
|
1314 |
+
"epoch": 0.29,
|
1315 |
+
"learning_rate": 0.00036525797208707584,
|
1316 |
+
"loss": 0.1167,
|
1317 |
+
"step": 43600
|
1318 |
+
},
|
1319 |
+
{
|
1320 |
+
"epoch": 0.3,
|
1321 |
+
"learning_rate": 0.0003652133334513314,
|
1322 |
+
"loss": 0.1265,
|
1323 |
+
"step": 43800
|
1324 |
+
},
|
1325 |
+
{
|
1326 |
+
"epoch": 0.3,
|
1327 |
+
"learning_rate": 0.00036516826929133405,
|
1328 |
+
"loss": 0.1287,
|
1329 |
+
"step": 44000
|
1330 |
+
},
|
1331 |
+
{
|
1332 |
+
"epoch": 0.3,
|
1333 |
+
"learning_rate": 0.00036512300346793045,
|
1334 |
+
"loss": 0.1254,
|
1335 |
+
"step": 44200
|
1336 |
+
},
|
1337 |
+
{
|
1338 |
+
"epoch": 0.3,
|
1339 |
+
"learning_rate": 0.00036507753603180906,
|
1340 |
+
"loss": 0.1234,
|
1341 |
+
"step": 44400
|
1342 |
+
},
|
1343 |
+
{
|
1344 |
+
"epoch": 0.3,
|
1345 |
+
"learning_rate": 0.000365031867033884,
|
1346 |
+
"loss": 0.1256,
|
1347 |
+
"step": 44600
|
1348 |
+
},
|
1349 |
+
{
|
1350 |
+
"epoch": 0.3,
|
1351 |
+
"learning_rate": 0.00036498599652529513,
|
1352 |
+
"loss": 0.1295,
|
1353 |
+
"step": 44800
|
1354 |
+
},
|
1355 |
+
{
|
1356 |
+
"epoch": 0.3,
|
1357 |
+
"learning_rate": 0.00036493992455740784,
|
1358 |
+
"loss": 0.129,
|
1359 |
+
"step": 45000
|
1360 |
+
},
|
1361 |
+
{
|
1362 |
+
"epoch": 0.3,
|
1363 |
+
"learning_rate": 0.00036489365118181326,
|
1364 |
+
"loss": 0.1252,
|
1365 |
+
"step": 45200
|
1366 |
+
},
|
1367 |
+
{
|
1368 |
+
"epoch": 0.31,
|
1369 |
+
"learning_rate": 0.000364847176450328,
|
1370 |
+
"loss": 0.1276,
|
1371 |
+
"step": 45400
|
1372 |
+
},
|
1373 |
+
{
|
1374 |
+
"epoch": 0.31,
|
1375 |
+
"learning_rate": 0.00036480050041499414,
|
1376 |
+
"loss": 0.1102,
|
1377 |
+
"step": 45600
|
1378 |
+
},
|
1379 |
+
{
|
1380 |
+
"epoch": 0.31,
|
1381 |
+
"learning_rate": 0.0003647536231280792,
|
1382 |
+
"loss": 0.125,
|
1383 |
+
"step": 45800
|
1384 |
+
},
|
1385 |
+
{
|
1386 |
+
"epoch": 0.31,
|
1387 |
+
"learning_rate": 0.00036470678053490165,
|
1388 |
+
"loss": 0.1421,
|
1389 |
+
"step": 46000
|
1390 |
+
},
|
1391 |
+
{
|
1392 |
+
"epoch": 0.31,
|
1393 |
+
"learning_rate": 0.00036465950190812875,
|
1394 |
+
"loss": 0.1281,
|
1395 |
+
"step": 46200
|
1396 |
+
},
|
1397 |
+
{
|
1398 |
+
"epoch": 0.31,
|
1399 |
+
"learning_rate": 0.000364612022187664,
|
1400 |
+
"loss": 0.1147,
|
1401 |
+
"step": 46400
|
1402 |
+
},
|
1403 |
+
{
|
1404 |
+
"epoch": 0.31,
|
1405 |
+
"learning_rate": 0.0003645643414266747,
|
1406 |
+
"loss": 0.1263,
|
1407 |
+
"step": 46600
|
1408 |
+
},
|
1409 |
+
{
|
1410 |
+
"epoch": 0.32,
|
1411 |
+
"learning_rate": 0.0003645164596785535,
|
1412 |
+
"loss": 0.1255,
|
1413 |
+
"step": 46800
|
1414 |
+
},
|
1415 |
+
{
|
1416 |
+
"epoch": 0.32,
|
1417 |
+
"learning_rate": 0.00036446837699691823,
|
1418 |
+
"loss": 0.1227,
|
1419 |
+
"step": 47000
|
1420 |
+
},
|
1421 |
+
{
|
1422 |
+
"epoch": 0.32,
|
1423 |
+
"learning_rate": 0.00036442009343561136,
|
1424 |
+
"loss": 0.1171,
|
1425 |
+
"step": 47200
|
1426 |
+
},
|
1427 |
+
{
|
1428 |
+
"epoch": 0.32,
|
1429 |
+
"learning_rate": 0.00036437160904870047,
|
1430 |
+
"loss": 0.1179,
|
1431 |
+
"step": 47400
|
1432 |
+
},
|
1433 |
+
{
|
1434 |
+
"epoch": 0.32,
|
1435 |
+
"learning_rate": 0.00036432292389047827,
|
1436 |
+
"loss": 0.1258,
|
1437 |
+
"step": 47600
|
1438 |
+
},
|
1439 |
+
{
|
1440 |
+
"epoch": 0.32,
|
1441 |
+
"learning_rate": 0.0003642740380154618,
|
1442 |
+
"loss": 0.1297,
|
1443 |
+
"step": 47800
|
1444 |
+
},
|
1445 |
+
{
|
1446 |
+
"epoch": 0.32,
|
1447 |
+
"learning_rate": 0.00036422495147839327,
|
1448 |
+
"loss": 0.1272,
|
1449 |
+
"step": 48000
|
1450 |
+
},
|
1451 |
+
{
|
1452 |
+
"epoch": 0.32,
|
1453 |
+
"learning_rate": 0.0003641756643342394,
|
1454 |
+
"loss": 0.1152,
|
1455 |
+
"step": 48200
|
1456 |
+
},
|
1457 |
+
{
|
1458 |
+
"epoch": 0.33,
|
1459 |
+
"learning_rate": 0.0003641264245754532,
|
1460 |
+
"loss": 0.1162,
|
1461 |
+
"step": 48400
|
1462 |
+
},
|
1463 |
+
{
|
1464 |
+
"epoch": 0.33,
|
1465 |
+
"learning_rate": 0.00036407673738527173,
|
1466 |
+
"loss": 0.1195,
|
1467 |
+
"step": 48600
|
1468 |
+
},
|
1469 |
+
{
|
1470 |
+
"epoch": 0.33,
|
1471 |
+
"learning_rate": 0.00036402684975397404,
|
1472 |
+
"loss": 0.1252,
|
1473 |
+
"step": 48800
|
1474 |
+
},
|
1475 |
+
{
|
1476 |
+
"epoch": 0.33,
|
1477 |
+
"learning_rate": 0.0003639767617374241,
|
1478 |
+
"loss": 0.1175,
|
1479 |
+
"step": 49000
|
1480 |
+
},
|
1481 |
+
{
|
1482 |
+
"epoch": 0.33,
|
1483 |
+
"learning_rate": 0.00036392647339170993,
|
1484 |
+
"loss": 0.1249,
|
1485 |
+
"step": 49200
|
1486 |
+
},
|
1487 |
+
{
|
1488 |
+
"epoch": 0.33,
|
1489 |
+
"learning_rate": 0.0003638759847731442,
|
1490 |
+
"loss": 0.1199,
|
1491 |
+
"step": 49400
|
1492 |
+
},
|
1493 |
+
{
|
1494 |
+
"epoch": 0.33,
|
1495 |
+
"learning_rate": 0.0003638252959382637,
|
1496 |
+
"loss": 0.122,
|
1497 |
+
"step": 49600
|
1498 |
+
},
|
1499 |
+
{
|
1500 |
+
"epoch": 0.34,
|
1501 |
+
"learning_rate": 0.00036377440694382947,
|
1502 |
+
"loss": 0.1109,
|
1503 |
+
"step": 49800
|
1504 |
+
},
|
1505 |
+
{
|
1506 |
+
"epoch": 0.34,
|
1507 |
+
"learning_rate": 0.0003637233178468266,
|
1508 |
+
"loss": 0.1171,
|
1509 |
+
"step": 50000
|
1510 |
+
},
|
1511 |
+
{
|
1512 |
+
"epoch": 0.34,
|
1513 |
+
"learning_rate": 0.00036367202870446437,
|
1514 |
+
"loss": 0.1213,
|
1515 |
+
"step": 50200
|
1516 |
+
},
|
1517 |
+
{
|
1518 |
+
"epoch": 0.34,
|
1519 |
+
"learning_rate": 0.0003636207975172021,
|
1520 |
+
"loss": 0.1156,
|
1521 |
+
"step": 50400
|
1522 |
+
},
|
1523 |
+
{
|
1524 |
+
"epoch": 0.34,
|
1525 |
+
"learning_rate": 0.0003635691094561523,
|
1526 |
+
"loss": 0.1257,
|
1527 |
+
"step": 50600
|
1528 |
+
},
|
1529 |
+
{
|
1530 |
+
"epoch": 0.34,
|
1531 |
+
"learning_rate": 0.00036351722152242464,
|
1532 |
+
"loss": 0.1179,
|
1533 |
+
"step": 50800
|
1534 |
+
},
|
1535 |
+
{
|
1536 |
+
"epoch": 0.34,
|
1537 |
+
"learning_rate": 0.0003634651337741228,
|
1538 |
+
"loss": 0.1287,
|
1539 |
+
"step": 51000
|
1540 |
+
},
|
1541 |
+
{
|
1542 |
+
"epoch": 0.34,
|
1543 |
+
"learning_rate": 0.0003634128462695743,
|
1544 |
+
"loss": 0.1182,
|
1545 |
+
"step": 51200
|
1546 |
+
},
|
1547 |
+
{
|
1548 |
+
"epoch": 0.35,
|
1549 |
+
"learning_rate": 0.0003633606219999926,
|
1550 |
+
"loss": 0.1371,
|
1551 |
+
"step": 51400
|
1552 |
+
},
|
1553 |
+
{
|
1554 |
+
"epoch": 0.35,
|
1555 |
+
"learning_rate": 0.00036330793615687614,
|
1556 |
+
"loss": 0.1055,
|
1557 |
+
"step": 51600
|
1558 |
+
},
|
1559 |
+
{
|
1560 |
+
"epoch": 0.35,
|
1561 |
+
"learning_rate": 0.00036325505073354187,
|
1562 |
+
"loss": 0.1216,
|
1563 |
+
"step": 51800
|
1564 |
+
},
|
1565 |
+
{
|
1566 |
+
"epoch": 0.35,
|
1567 |
+
"learning_rate": 0.00036320196578921055,
|
1568 |
+
"loss": 0.1207,
|
1569 |
+
"step": 52000
|
1570 |
+
},
|
1571 |
+
{
|
1572 |
+
"epoch": 0.35,
|
1573 |
+
"learning_rate": 0.0003631486813833263,
|
1574 |
+
"loss": 0.131,
|
1575 |
+
"step": 52200
|
1576 |
+
},
|
1577 |
+
{
|
1578 |
+
"epoch": 0.35,
|
1579 |
+
"learning_rate": 0.00036309519757555674,
|
1580 |
+
"loss": 0.12,
|
1581 |
+
"step": 52400
|
1582 |
+
},
|
1583 |
+
{
|
1584 |
+
"epoch": 0.35,
|
1585 |
+
"learning_rate": 0.00036304151442579253,
|
1586 |
+
"loss": 0.1199,
|
1587 |
+
"step": 52600
|
1588 |
+
},
|
1589 |
+
{
|
1590 |
+
"epoch": 0.36,
|
1591 |
+
"learning_rate": 0.0003629876319941478,
|
1592 |
+
"loss": 0.116,
|
1593 |
+
"step": 52800
|
1594 |
+
},
|
1595 |
+
{
|
1596 |
+
"epoch": 0.36,
|
1597 |
+
"learning_rate": 0.00036293355034095967,
|
1598 |
+
"loss": 0.1134,
|
1599 |
+
"step": 53000
|
1600 |
+
},
|
1601 |
+
{
|
1602 |
+
"epoch": 0.36,
|
1603 |
+
"learning_rate": 0.00036287926952678845,
|
1604 |
+
"loss": 0.1133,
|
1605 |
+
"step": 53200
|
1606 |
+
},
|
1607 |
+
{
|
1608 |
+
"epoch": 0.36,
|
1609 |
+
"learning_rate": 0.00036282478961241737,
|
1610 |
+
"loss": 0.1254,
|
1611 |
+
"step": 53400
|
1612 |
+
},
|
1613 |
+
{
|
1614 |
+
"epoch": 0.36,
|
1615 |
+
"learning_rate": 0.00036277011065885274,
|
1616 |
+
"loss": 0.1207,
|
1617 |
+
"step": 53600
|
1618 |
+
},
|
1619 |
+
{
|
1620 |
+
"epoch": 0.36,
|
1621 |
+
"learning_rate": 0.00036271523272732363,
|
1622 |
+
"loss": 0.1222,
|
1623 |
+
"step": 53800
|
1624 |
+
},
|
1625 |
+
{
|
1626 |
+
"epoch": 0.36,
|
1627 |
+
"learning_rate": 0.00036266015587928203,
|
1628 |
+
"loss": 0.1161,
|
1629 |
+
"step": 54000
|
1630 |
+
},
|
1631 |
+
{
|
1632 |
+
"epoch": 0.37,
|
1633 |
+
"learning_rate": 0.0003626048801764025,
|
1634 |
+
"loss": 0.1216,
|
1635 |
+
"step": 54200
|
1636 |
+
},
|
1637 |
+
{
|
1638 |
+
"epoch": 0.37,
|
1639 |
+
"learning_rate": 0.00036254940568058254,
|
1640 |
+
"loss": 0.1253,
|
1641 |
+
"step": 54400
|
1642 |
+
},
|
1643 |
+
{
|
1644 |
+
"epoch": 0.37,
|
1645 |
+
"learning_rate": 0.000362493732453942,
|
1646 |
+
"loss": 0.1154,
|
1647 |
+
"step": 54600
|
1648 |
+
},
|
1649 |
+
{
|
1650 |
+
"epoch": 0.37,
|
1651 |
+
"learning_rate": 0.00036243786055882343,
|
1652 |
+
"loss": 0.1163,
|
1653 |
+
"step": 54800
|
1654 |
+
},
|
1655 |
+
{
|
1656 |
+
"epoch": 0.37,
|
1657 |
+
"learning_rate": 0.00036238179005779176,
|
1658 |
+
"loss": 0.1154,
|
1659 |
+
"step": 55000
|
1660 |
+
},
|
1661 |
+
{
|
1662 |
+
"epoch": 0.37,
|
1663 |
+
"learning_rate": 0.00036232552101363446,
|
1664 |
+
"loss": 0.1115,
|
1665 |
+
"step": 55200
|
1666 |
+
},
|
1667 |
+
{
|
1668 |
+
"epoch": 0.37,
|
1669 |
+
"learning_rate": 0.0003622693363205972,
|
1670 |
+
"loss": 0.1203,
|
1671 |
+
"step": 55400
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"epoch": 0.37,
|
1675 |
+
"learning_rate": 0.00036221267137136654,
|
1676 |
+
"loss": 0.1241,
|
1677 |
+
"step": 55600
|
1678 |
+
},
|
1679 |
+
{
|
1680 |
+
"epoch": 0.38,
|
1681 |
+
"learning_rate": 0.00036215580806838815,
|
1682 |
+
"loss": 0.1113,
|
1683 |
+
"step": 55800
|
1684 |
+
},
|
1685 |
+
{
|
1686 |
+
"epoch": 0.38,
|
1687 |
+
"learning_rate": 0.0003620987464753372,
|
1688 |
+
"loss": 0.1275,
|
1689 |
+
"step": 56000
|
1690 |
+
},
|
1691 |
+
{
|
1692 |
+
"epoch": 0.38,
|
1693 |
+
"learning_rate": 0.0003620417734481885,
|
1694 |
+
"loss": 0.1238,
|
1695 |
+
"step": 56200
|
1696 |
+
},
|
1697 |
+
{
|
1698 |
+
"epoch": 0.38,
|
1699 |
+
"learning_rate": 0.0003619843164575562,
|
1700 |
+
"loss": 0.1158,
|
1701 |
+
"step": 56400
|
1702 |
+
},
|
1703 |
+
{
|
1704 |
+
"epoch": 0.38,
|
1705 |
+
"learning_rate": 0.00036192666136888646,
|
1706 |
+
"loss": 0.113,
|
1707 |
+
"step": 56600
|
1708 |
+
},
|
1709 |
+
{
|
1710 |
+
"epoch": 0.38,
|
1711 |
+
"learning_rate": 0.000361868808246741,
|
1712 |
+
"loss": 0.114,
|
1713 |
+
"step": 56800
|
1714 |
+
},
|
1715 |
+
{
|
1716 |
+
"epoch": 0.38,
|
1717 |
+
"learning_rate": 0.00036181075715590336,
|
1718 |
+
"loss": 0.1304,
|
1719 |
+
"step": 57000
|
1720 |
+
},
|
1721 |
+
{
|
1722 |
+
"epoch": 0.39,
|
1723 |
+
"learning_rate": 0.00036175250816137877,
|
1724 |
+
"loss": 0.1105,
|
1725 |
+
"step": 57200
|
1726 |
+
},
|
1727 |
+
{
|
1728 |
+
"epoch": 0.39,
|
1729 |
+
"learning_rate": 0.0003616940613283941,
|
1730 |
+
"loss": 0.1187,
|
1731 |
+
"step": 57400
|
1732 |
+
},
|
1733 |
+
{
|
1734 |
+
"epoch": 0.39,
|
1735 |
+
"learning_rate": 0.00036163541672239765,
|
1736 |
+
"loss": 0.1166,
|
1737 |
+
"step": 57600
|
1738 |
+
},
|
1739 |
+
{
|
1740 |
+
"epoch": 0.39,
|
1741 |
+
"learning_rate": 0.00036157657440905933,
|
1742 |
+
"loss": 0.1209,
|
1743 |
+
"step": 57800
|
1744 |
+
},
|
1745 |
+
{
|
1746 |
+
"epoch": 0.39,
|
1747 |
+
"learning_rate": 0.00036151753445427023,
|
1748 |
+
"loss": 0.1363,
|
1749 |
+
"step": 58000
|
1750 |
+
},
|
1751 |
+
{
|
1752 |
+
"epoch": 0.39,
|
1753 |
+
"learning_rate": 0.00036145829692414304,
|
1754 |
+
"loss": 0.1096,
|
1755 |
+
"step": 58200
|
1756 |
+
},
|
1757 |
+
{
|
1758 |
+
"epoch": 0.39,
|
1759 |
+
"learning_rate": 0.0003613988618850114,
|
1760 |
+
"loss": 0.1188,
|
1761 |
+
"step": 58400
|
1762 |
+
},
|
1763 |
+
{
|
1764 |
+
"epoch": 0.39,
|
1765 |
+
"learning_rate": 0.00036133922940343035,
|
1766 |
+
"loss": 0.1083,
|
1767 |
+
"step": 58600
|
1768 |
+
},
|
1769 |
+
{
|
1770 |
+
"epoch": 0.4,
|
1771 |
+
"learning_rate": 0.0003612793995461759,
|
1772 |
+
"loss": 0.1197,
|
1773 |
+
"step": 58800
|
1774 |
+
},
|
1775 |
+
{
|
1776 |
+
"epoch": 0.4,
|
1777 |
+
"learning_rate": 0.00036121937238024515,
|
1778 |
+
"loss": 0.1279,
|
1779 |
+
"step": 59000
|
1780 |
+
},
|
1781 |
+
{
|
1782 |
+
"epoch": 0.4,
|
1783 |
+
"learning_rate": 0.00036115914797285614,
|
1784 |
+
"loss": 0.1189,
|
1785 |
+
"step": 59200
|
1786 |
+
},
|
1787 |
+
{
|
1788 |
+
"epoch": 0.4,
|
1789 |
+
"learning_rate": 0.00036109872639144766,
|
1790 |
+
"loss": 0.1356,
|
1791 |
+
"step": 59400
|
1792 |
+
},
|
1793 |
+
{
|
1794 |
+
"epoch": 0.4,
|
1795 |
+
"learning_rate": 0.00036103810770367947,
|
1796 |
+
"loss": 0.1222,
|
1797 |
+
"step": 59600
|
1798 |
+
},
|
1799 |
+
{
|
1800 |
+
"epoch": 0.4,
|
1801 |
+
"learning_rate": 0.00036097729197743193,
|
1802 |
+
"loss": 0.1201,
|
1803 |
+
"step": 59800
|
1804 |
+
},
|
1805 |
+
{
|
1806 |
+
"epoch": 0.4,
|
1807 |
+
"learning_rate": 0.000360916279280806,
|
1808 |
+
"loss": 0.1156,
|
1809 |
+
"step": 60000
|
1810 |
+
},
|
1811 |
+
{
|
1812 |
+
"epoch": 0.41,
|
1813 |
+
"learning_rate": 0.0003608550696821234,
|
1814 |
+
"loss": 0.1168,
|
1815 |
+
"step": 60200
|
1816 |
+
},
|
1817 |
+
{
|
1818 |
+
"epoch": 0.41,
|
1819 |
+
"learning_rate": 0.00036079366324992627,
|
1820 |
+
"loss": 0.1174,
|
1821 |
+
"step": 60400
|
1822 |
+
},
|
1823 |
+
{
|
1824 |
+
"epoch": 0.41,
|
1825 |
+
"learning_rate": 0.00036073236855829993,
|
1826 |
+
"loss": 0.1202,
|
1827 |
+
"step": 60600
|
1828 |
+
},
|
1829 |
+
{
|
1830 |
+
"epoch": 0.41,
|
1831 |
+
"learning_rate": 0.00036067056964888823,
|
1832 |
+
"loss": 0.1073,
|
1833 |
+
"step": 60800
|
1834 |
+
},
|
1835 |
+
{
|
1836 |
+
"epoch": 0.41,
|
1837 |
+
"learning_rate": 0.00036060857411256375,
|
1838 |
+
"loss": 0.1288,
|
1839 |
+
"step": 61000
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 0.41,
|
1843 |
+
"learning_rate": 0.0003605463820187487,
|
1844 |
+
"loss": 0.1174,
|
1845 |
+
"step": 61200
|
1846 |
+
},
|
1847 |
+
{
|
1848 |
+
"epoch": 0.41,
|
1849 |
+
"learning_rate": 0.0003604839934370854,
|
1850 |
+
"loss": 0.1136,
|
1851 |
+
"step": 61400
|
1852 |
+
},
|
1853 |
+
{
|
1854 |
+
"epoch": 0.41,
|
1855 |
+
"learning_rate": 0.0003604214084374361,
|
1856 |
+
"loss": 0.112,
|
1857 |
+
"step": 61600
|
1858 |
+
},
|
1859 |
+
{
|
1860 |
+
"epoch": 0.42,
|
1861 |
+
"learning_rate": 0.00036035894148492016,
|
1862 |
+
"loss": 0.1091,
|
1863 |
+
"step": 61800
|
1864 |
+
},
|
1865 |
+
{
|
1866 |
+
"epoch": 0.42,
|
1867 |
+
"learning_rate": 0.0003602959648409783,
|
1868 |
+
"loss": 0.1329,
|
1869 |
+
"step": 62000
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 0.42,
|
1873 |
+
"learning_rate": 0.00036023279198960366,
|
1874 |
+
"loss": 0.1262,
|
1875 |
+
"step": 62200
|
1876 |
+
},
|
1877 |
+
{
|
1878 |
+
"epoch": 0.42,
|
1879 |
+
"learning_rate": 0.0003601694230015367,
|
1880 |
+
"loss": 0.1231,
|
1881 |
+
"step": 62400
|
1882 |
+
},
|
1883 |
+
{
|
1884 |
+
"epoch": 0.42,
|
1885 |
+
"learning_rate": 0.00036010585794773764,
|
1886 |
+
"loss": 0.1107,
|
1887 |
+
"step": 62600
|
1888 |
+
},
|
1889 |
+
{
|
1890 |
+
"epoch": 0.42,
|
1891 |
+
"learning_rate": 0.0003600420968993861,
|
1892 |
+
"loss": 0.1299,
|
1893 |
+
"step": 62800
|
1894 |
+
},
|
1895 |
+
{
|
1896 |
+
"epoch": 0.42,
|
1897 |
+
"learning_rate": 0.0003599781399278814,
|
1898 |
+
"loss": 0.1161,
|
1899 |
+
"step": 63000
|
1900 |
+
},
|
1901 |
+
{
|
1902 |
+
"epoch": 0.43,
|
1903 |
+
"learning_rate": 0.00035991398710484205,
|
1904 |
+
"loss": 0.115,
|
1905 |
+
"step": 63200
|
1906 |
+
},
|
1907 |
+
{
|
1908 |
+
"epoch": 0.43,
|
1909 |
+
"learning_rate": 0.00035984963850210587,
|
1910 |
+
"loss": 0.1223,
|
1911 |
+
"step": 63400
|
1912 |
+
},
|
1913 |
+
{
|
1914 |
+
"epoch": 0.43,
|
1915 |
+
"learning_rate": 0.0003597850941917301,
|
1916 |
+
"loss": 0.1154,
|
1917 |
+
"step": 63600
|
1918 |
+
},
|
1919 |
+
{
|
1920 |
+
"epoch": 0.43,
|
1921 |
+
"learning_rate": 0.00035972035424599094,
|
1922 |
+
"loss": 0.1167,
|
1923 |
+
"step": 63800
|
1924 |
+
},
|
1925 |
+
{
|
1926 |
+
"epoch": 0.43,
|
1927 |
+
"learning_rate": 0.0003596554187373838,
|
1928 |
+
"loss": 0.1129,
|
1929 |
+
"step": 64000
|
1930 |
+
},
|
1931 |
+
{
|
1932 |
+
"epoch": 0.43,
|
1933 |
+
"learning_rate": 0.000359590287738623,
|
1934 |
+
"loss": 0.1255,
|
1935 |
+
"step": 64200
|
1936 |
+
},
|
1937 |
+
{
|
1938 |
+
"epoch": 0.43,
|
1939 |
+
"learning_rate": 0.0003595249613226417,
|
1940 |
+
"loss": 0.1212,
|
1941 |
+
"step": 64400
|
1942 |
+
},
|
1943 |
+
{
|
1944 |
+
"epoch": 0.44,
|
1945 |
+
"learning_rate": 0.0003594594395625922,
|
1946 |
+
"loss": 0.1202,
|
1947 |
+
"step": 64600
|
1948 |
+
},
|
1949 |
+
{
|
1950 |
+
"epoch": 0.44,
|
1951 |
+
"learning_rate": 0.0003593937225318451,
|
1952 |
+
"loss": 0.1145,
|
1953 |
+
"step": 64800
|
1954 |
+
},
|
1955 |
+
{
|
1956 |
+
"epoch": 0.44,
|
1957 |
+
"learning_rate": 0.00035932781030399,
|
1958 |
+
"loss": 0.1076,
|
1959 |
+
"step": 65000
|
1960 |
+
},
|
1961 |
+
{
|
1962 |
+
"epoch": 0.44,
|
1963 |
+
"learning_rate": 0.0003592617029528349,
|
1964 |
+
"loss": 0.1167,
|
1965 |
+
"step": 65200
|
1966 |
+
},
|
1967 |
+
{
|
1968 |
+
"epoch": 0.44,
|
1969 |
+
"learning_rate": 0.00035919540055240643,
|
1970 |
+
"loss": 0.1182,
|
1971 |
+
"step": 65400
|
1972 |
+
},
|
1973 |
+
{
|
1974 |
+
"epoch": 0.44,
|
1975 |
+
"learning_rate": 0.00035912890317694947,
|
1976 |
+
"loss": 0.1249,
|
1977 |
+
"step": 65600
|
1978 |
+
},
|
1979 |
+
{
|
1980 |
+
"epoch": 0.44,
|
1981 |
+
"learning_rate": 0.00035906221090092747,
|
1982 |
+
"loss": 0.1267,
|
1983 |
+
"step": 65800
|
1984 |
+
},
|
1985 |
+
{
|
1986 |
+
"epoch": 0.44,
|
1987 |
+
"learning_rate": 0.0003589953237990218,
|
1988 |
+
"loss": 0.1111,
|
1989 |
+
"step": 66000
|
1990 |
+
},
|
1991 |
+
{
|
1992 |
+
"epoch": 0.45,
|
1993 |
+
"learning_rate": 0.0003589282419461323,
|
1994 |
+
"loss": 0.1155,
|
1995 |
+
"step": 66200
|
1996 |
+
},
|
1997 |
+
{
|
1998 |
+
"epoch": 0.45,
|
1999 |
+
"learning_rate": 0.0003588613022841523,
|
2000 |
+
"loss": 0.1156,
|
2001 |
+
"step": 66400
|
2002 |
+
},
|
2003 |
+
{
|
2004 |
+
"epoch": 0.45,
|
2005 |
+
"learning_rate": 0.0003587938321276814,
|
2006 |
+
"loss": 0.1381,
|
2007 |
+
"step": 66600
|
2008 |
+
},
|
2009 |
+
{
|
2010 |
+
"epoch": 0.45,
|
2011 |
+
"learning_rate": 0.0003587261674458557,
|
2012 |
+
"loss": 0.1141,
|
2013 |
+
"step": 66800
|
2014 |
+
},
|
2015 |
+
{
|
2016 |
+
"epoch": 0.45,
|
2017 |
+
"learning_rate": 0.00035865830831444557,
|
2018 |
+
"loss": 0.1039,
|
2019 |
+
"step": 67000
|
2020 |
+
},
|
2021 |
+
{
|
2022 |
+
"epoch": 0.45,
|
2023 |
+
"learning_rate": 0.0003585902548094393,
|
2024 |
+
"loss": 0.1329,
|
2025 |
+
"step": 67200
|
2026 |
+
},
|
2027 |
+
{
|
2028 |
+
"epoch": 0.45,
|
2029 |
+
"learning_rate": 0.0003585220070070426,
|
2030 |
+
"loss": 0.1255,
|
2031 |
+
"step": 67400
|
2032 |
+
},
|
2033 |
+
{
|
2034 |
+
"epoch": 0.46,
|
2035 |
+
"learning_rate": 0.00035845356498367904,
|
2036 |
+
"loss": 0.12,
|
2037 |
+
"step": 67600
|
2038 |
+
},
|
2039 |
+
{
|
2040 |
+
"epoch": 0.46,
|
2041 |
+
"learning_rate": 0.00035838492881598946,
|
2042 |
+
"loss": 0.1142,
|
2043 |
+
"step": 67800
|
2044 |
+
},
|
2045 |
+
{
|
2046 |
+
"epoch": 0.46,
|
2047 |
+
"learning_rate": 0.00035831609858083215,
|
2048 |
+
"loss": 0.1199,
|
2049 |
+
"step": 68000
|
2050 |
+
},
|
2051 |
+
{
|
2052 |
+
"epoch": 0.46,
|
2053 |
+
"learning_rate": 0.0003582470743552828,
|
2054 |
+
"loss": 0.1206,
|
2055 |
+
"step": 68200
|
2056 |
+
},
|
2057 |
+
{
|
2058 |
+
"epoch": 0.46,
|
2059 |
+
"learning_rate": 0.00035817820278955816,
|
2060 |
+
"loss": 0.1298,
|
2061 |
+
"step": 68400
|
2062 |
+
},
|
2063 |
+
{
|
2064 |
+
"epoch": 0.46,
|
2065 |
+
"learning_rate": 0.0003581087917843051,
|
2066 |
+
"loss": 0.1344,
|
2067 |
+
"step": 68600
|
2068 |
+
},
|
2069 |
+
{
|
2070 |
+
"epoch": 0.46,
|
2071 |
+
"learning_rate": 0.00035803918702080075,
|
2072 |
+
"loss": 0.127,
|
2073 |
+
"step": 68800
|
2074 |
+
},
|
2075 |
+
{
|
2076 |
+
"epoch": 0.46,
|
2077 |
+
"learning_rate": 0.00035796938857698805,
|
2078 |
+
"loss": 0.1171,
|
2079 |
+
"step": 69000
|
2080 |
+
},
|
2081 |
+
{
|
2082 |
+
"epoch": 0.47,
|
2083 |
+
"learning_rate": 0.00035789939653102667,
|
2084 |
+
"loss": 0.1142,
|
2085 |
+
"step": 69200
|
2086 |
+
},
|
2087 |
+
{
|
2088 |
+
"epoch": 0.47,
|
2089 |
+
"learning_rate": 0.0003578292109612934,
|
2090 |
+
"loss": 0.119,
|
2091 |
+
"step": 69400
|
2092 |
+
},
|
2093 |
+
{
|
2094 |
+
"epoch": 0.47,
|
2095 |
+
"learning_rate": 0.0003577588319463815,
|
2096 |
+
"loss": 0.1232,
|
2097 |
+
"step": 69600
|
2098 |
+
},
|
2099 |
+
{
|
2100 |
+
"epoch": 0.47,
|
2101 |
+
"learning_rate": 0.0003576882595651009,
|
2102 |
+
"loss": 0.1206,
|
2103 |
+
"step": 69800
|
2104 |
+
},
|
2105 |
+
{
|
2106 |
+
"epoch": 0.47,
|
2107 |
+
"learning_rate": 0.0003576174938964781,
|
2108 |
+
"loss": 0.1093,
|
2109 |
+
"step": 70000
|
2110 |
+
},
|
2111 |
+
{
|
2112 |
+
"epoch": 0.47,
|
2113 |
+
"learning_rate": 0.00035754653501975596,
|
2114 |
+
"loss": 0.1352,
|
2115 |
+
"step": 70200
|
2116 |
+
},
|
2117 |
+
{
|
2118 |
+
"epoch": 0.47,
|
2119 |
+
"learning_rate": 0.0003574753830143938,
|
2120 |
+
"loss": 0.1207,
|
2121 |
+
"step": 70400
|
2122 |
+
},
|
2123 |
+
{
|
2124 |
+
"epoch": 0.48,
|
2125 |
+
"learning_rate": 0.00035740403796006713,
|
2126 |
+
"loss": 0.1077,
|
2127 |
+
"step": 70600
|
2128 |
+
},
|
2129 |
+
{
|
2130 |
+
"epoch": 0.48,
|
2131 |
+
"learning_rate": 0.00035733249993666767,
|
2132 |
+
"loss": 0.1159,
|
2133 |
+
"step": 70800
|
2134 |
+
},
|
2135 |
+
{
|
2136 |
+
"epoch": 0.48,
|
2137 |
+
"learning_rate": 0.00035726112815854356,
|
2138 |
+
"loss": 0.1196,
|
2139 |
+
"step": 71000
|
2140 |
+
},
|
2141 |
+
{
|
2142 |
+
"epoch": 0.48,
|
2143 |
+
"learning_rate": 0.0003571892054013809,
|
2144 |
+
"loss": 0.1162,
|
2145 |
+
"step": 71200
|
2146 |
+
},
|
2147 |
+
{
|
2148 |
+
"epoch": 0.48,
|
2149 |
+
"learning_rate": 0.0003571170899157135,
|
2150 |
+
"loss": 0.1439,
|
2151 |
+
"step": 71400
|
2152 |
+
},
|
2153 |
+
{
|
2154 |
+
"epoch": 0.48,
|
2155 |
+
"learning_rate": 0.00035704478178229574,
|
2156 |
+
"loss": 0.1113,
|
2157 |
+
"step": 71600
|
2158 |
+
},
|
2159 |
+
{
|
2160 |
+
"epoch": 0.48,
|
2161 |
+
"learning_rate": 0.00035697228108209786,
|
2162 |
+
"loss": 0.1195,
|
2163 |
+
"step": 71800
|
2164 |
+
},
|
2165 |
+
{
|
2166 |
+
"epoch": 0.49,
|
2167 |
+
"learning_rate": 0.00035689958789630546,
|
2168 |
+
"loss": 0.1121,
|
2169 |
+
"step": 72000
|
2170 |
+
},
|
2171 |
+
{
|
2172 |
+
"epoch": 0.49,
|
2173 |
+
"learning_rate": 0.0003568267023063201,
|
2174 |
+
"loss": 0.1024,
|
2175 |
+
"step": 72200
|
2176 |
+
},
|
2177 |
+
{
|
2178 |
+
"epoch": 0.49,
|
2179 |
+
"learning_rate": 0.0003567536243937585,
|
2180 |
+
"loss": 0.1103,
|
2181 |
+
"step": 72400
|
2182 |
+
},
|
2183 |
+
{
|
2184 |
+
"epoch": 0.49,
|
2185 |
+
"learning_rate": 0.00035668035424045273,
|
2186 |
+
"loss": 0.1306,
|
2187 |
+
"step": 72600
|
2188 |
+
},
|
2189 |
+
{
|
2190 |
+
"epoch": 0.49,
|
2191 |
+
"learning_rate": 0.0003566068919284503,
|
2192 |
+
"loss": 0.1145,
|
2193 |
+
"step": 72800
|
2194 |
+
},
|
2195 |
+
{
|
2196 |
+
"epoch": 0.49,
|
2197 |
+
"learning_rate": 0.0003565332375400137,
|
2198 |
+
"loss": 0.1157,
|
2199 |
+
"step": 73000
|
2200 |
+
},
|
2201 |
+
{
|
2202 |
+
"epoch": 0.49,
|
2203 |
+
"learning_rate": 0.0003564593911576208,
|
2204 |
+
"loss": 0.119,
|
2205 |
+
"step": 73200
|
2206 |
+
},
|
2207 |
+
{
|
2208 |
+
"epoch": 0.49,
|
2209 |
+
"learning_rate": 0.00035638535286396404,
|
2210 |
+
"loss": 0.1117,
|
2211 |
+
"step": 73400
|
2212 |
+
},
|
2213 |
+
{
|
2214 |
+
"epoch": 0.5,
|
2215 |
+
"learning_rate": 0.0003563114943695969,
|
2216 |
+
"loss": 0.117,
|
2217 |
+
"step": 73600
|
2218 |
+
},
|
2219 |
+
{
|
2220 |
+
"epoch": 0.5,
|
2221 |
+
"learning_rate": 0.00035623707346086925,
|
2222 |
+
"loss": 0.108,
|
2223 |
+
"step": 73800
|
2224 |
+
},
|
2225 |
+
{
|
2226 |
+
"epoch": 0.5,
|
2227 |
+
"learning_rate": 0.00035616246088982775,
|
2228 |
+
"loss": 0.106,
|
2229 |
+
"step": 74000
|
2230 |
+
},
|
2231 |
+
{
|
2232 |
+
"epoch": 0.5,
|
2233 |
+
"learning_rate": 0.0003560876567400229,
|
2234 |
+
"loss": 0.1316,
|
2235 |
+
"step": 74200
|
2236 |
+
},
|
2237 |
+
{
|
2238 |
+
"epoch": 0.5,
|
2239 |
+
"learning_rate": 0.0003560126610952201,
|
2240 |
+
"loss": 0.1121,
|
2241 |
+
"step": 74400
|
2242 |
+
},
|
2243 |
+
{
|
2244 |
+
"epoch": 0.5,
|
2245 |
+
"learning_rate": 0.0003559374740393988,
|
2246 |
+
"loss": 0.124,
|
2247 |
+
"step": 74600
|
2248 |
+
},
|
2249 |
+
{
|
2250 |
+
"epoch": 0.5,
|
2251 |
+
"learning_rate": 0.000355862095656753,
|
2252 |
+
"loss": 0.1163,
|
2253 |
+
"step": 74800
|
2254 |
+
},
|
2255 |
+
{
|
2256 |
+
"epoch": 0.51,
|
2257 |
+
"learning_rate": 0.0003557865260316909,
|
2258 |
+
"loss": 0.1191,
|
2259 |
+
"step": 75000
|
2260 |
+
},
|
2261 |
+
{
|
2262 |
+
"epoch": 0.51,
|
2263 |
+
"learning_rate": 0.0003557107652488348,
|
2264 |
+
"loss": 0.1189,
|
2265 |
+
"step": 75200
|
2266 |
+
},
|
2267 |
+
{
|
2268 |
+
"epoch": 0.51,
|
2269 |
+
"learning_rate": 0.00035563481339302114,
|
2270 |
+
"loss": 0.1118,
|
2271 |
+
"step": 75400
|
2272 |
+
},
|
2273 |
+
{
|
2274 |
+
"epoch": 0.51,
|
2275 |
+
"learning_rate": 0.00035555867054930034,
|
2276 |
+
"loss": 0.1242,
|
2277 |
+
"step": 75600
|
2278 |
+
},
|
2279 |
+
{
|
2280 |
+
"epoch": 0.51,
|
2281 |
+
"learning_rate": 0.0003554823368029366,
|
2282 |
+
"loss": 0.1064,
|
2283 |
+
"step": 75800
|
2284 |
+
},
|
2285 |
+
{
|
2286 |
+
"epoch": 0.51,
|
2287 |
+
"learning_rate": 0.00035540581223940794,
|
2288 |
+
"loss": 0.1078,
|
2289 |
+
"step": 76000
|
2290 |
+
},
|
2291 |
+
{
|
2292 |
+
"epoch": 0.51,
|
2293 |
+
"learning_rate": 0.000355329096944406,
|
2294 |
+
"loss": 0.1145,
|
2295 |
+
"step": 76200
|
2296 |
+
},
|
2297 |
+
{
|
2298 |
+
"epoch": 0.51,
|
2299 |
+
"learning_rate": 0.00035525219100383614,
|
2300 |
+
"loss": 0.122,
|
2301 |
+
"step": 76400
|
2302 |
+
},
|
2303 |
+
{
|
2304 |
+
"epoch": 0.52,
|
2305 |
+
"learning_rate": 0.00035517548046019095,
|
2306 |
+
"loss": 0.1132,
|
2307 |
+
"step": 76600
|
2308 |
+
},
|
2309 |
+
{
|
2310 |
+
"epoch": 0.52,
|
2311 |
+
"learning_rate": 0.0003550981944392052,
|
2312 |
+
"loss": 0.1183,
|
2313 |
+
"step": 76800
|
2314 |
+
},
|
2315 |
+
{
|
2316 |
+
"epoch": 0.52,
|
2317 |
+
"learning_rate": 0.00035502071803121456,
|
2318 |
+
"loss": 0.1086,
|
2319 |
+
"step": 77000
|
2320 |
+
},
|
2321 |
+
{
|
2322 |
+
"epoch": 0.52,
|
2323 |
+
"learning_rate": 0.0003549430513229765,
|
2324 |
+
"loss": 0.11,
|
2325 |
+
"step": 77200
|
2326 |
+
},
|
2327 |
+
{
|
2328 |
+
"epoch": 0.52,
|
2329 |
+
"learning_rate": 0.0003548651944014617,
|
2330 |
+
"loss": 0.1144,
|
2331 |
+
"step": 77400
|
2332 |
+
},
|
2333 |
+
{
|
2334 |
+
"epoch": 0.52,
|
2335 |
+
"learning_rate": 0.0003547871473538539,
|
2336 |
+
"loss": 0.1114,
|
2337 |
+
"step": 77600
|
2338 |
+
},
|
2339 |
+
{
|
2340 |
+
"epoch": 0.52,
|
2341 |
+
"learning_rate": 0.0003547089102675495,
|
2342 |
+
"loss": 0.1131,
|
2343 |
+
"step": 77800
|
2344 |
+
},
|
2345 |
+
{
|
2346 |
+
"epoch": 0.53,
|
2347 |
+
"learning_rate": 0.000354630483230158,
|
2348 |
+
"loss": 0.1178,
|
2349 |
+
"step": 78000
|
2350 |
+
},
|
2351 |
+
{
|
2352 |
+
"epoch": 0.53,
|
2353 |
+
"learning_rate": 0.0003545518663295014,
|
2354 |
+
"loss": 0.1184,
|
2355 |
+
"step": 78200
|
2356 |
+
},
|
2357 |
+
{
|
2358 |
+
"epoch": 0.53,
|
2359 |
+
"learning_rate": 0.0003544730596536143,
|
2360 |
+
"loss": 0.1131,
|
2361 |
+
"step": 78400
|
2362 |
+
},
|
2363 |
+
{
|
2364 |
+
"epoch": 0.53,
|
2365 |
+
"learning_rate": 0.000354394063290744,
|
2366 |
+
"loss": 0.1103,
|
2367 |
+
"step": 78600
|
2368 |
+
},
|
2369 |
+
{
|
2370 |
+
"epoch": 0.53,
|
2371 |
+
"learning_rate": 0.00035431487732935006,
|
2372 |
+
"loss": 0.1099,
|
2373 |
+
"step": 78800
|
2374 |
+
},
|
2375 |
+
{
|
2376 |
+
"epoch": 0.53,
|
2377 |
+
"learning_rate": 0.00035423589920671934,
|
2378 |
+
"loss": 0.124,
|
2379 |
+
"step": 79000
|
2380 |
+
},
|
2381 |
+
{
|
2382 |
+
"epoch": 0.53,
|
2383 |
+
"learning_rate": 0.0003541563352613894,
|
2384 |
+
"loss": 0.1216,
|
2385 |
+
"step": 79200
|
2386 |
+
},
|
2387 |
+
{
|
2388 |
+
"epoch": 0.53,
|
2389 |
+
"learning_rate": 0.00035407658198374195,
|
2390 |
+
"loss": 0.1153,
|
2391 |
+
"step": 79400
|
2392 |
+
},
|
2393 |
+
{
|
2394 |
+
"epoch": 0.54,
|
2395 |
+
"learning_rate": 0.00035399663946308424,
|
2396 |
+
"loss": 0.107,
|
2397 |
+
"step": 79600
|
2398 |
+
},
|
2399 |
+
{
|
2400 |
+
"epoch": 0.54,
|
2401 |
+
"learning_rate": 0.00035391650778893543,
|
2402 |
+
"loss": 0.1105,
|
2403 |
+
"step": 79800
|
2404 |
+
},
|
2405 |
+
{
|
2406 |
+
"epoch": 0.54,
|
2407 |
+
"learning_rate": 0.00035383618705102626,
|
2408 |
+
"loss": 0.1361,
|
2409 |
+
"step": 80000
|
2410 |
+
},
|
2411 |
+
{
|
2412 |
+
"epoch": 0.54,
|
2413 |
+
"learning_rate": 0.00035375567733929953,
|
2414 |
+
"loss": 0.1221,
|
2415 |
+
"step": 80200
|
2416 |
+
},
|
2417 |
+
{
|
2418 |
+
"epoch": 0.54,
|
2419 |
+
"learning_rate": 0.0003536749787439093,
|
2420 |
+
"loss": 0.1085,
|
2421 |
+
"step": 80400
|
2422 |
+
},
|
2423 |
+
{
|
2424 |
+
"epoch": 0.54,
|
2425 |
+
"learning_rate": 0.0003535940913552215,
|
2426 |
+
"loss": 0.1115,
|
2427 |
+
"step": 80600
|
2428 |
+
},
|
2429 |
+
{
|
2430 |
+
"epoch": 0.54,
|
2431 |
+
"learning_rate": 0.00035351301526381307,
|
2432 |
+
"loss": 0.1107,
|
2433 |
+
"step": 80800
|
2434 |
+
},
|
2435 |
+
{
|
2436 |
+
"epoch": 0.55,
|
2437 |
+
"learning_rate": 0.00035343175056047263,
|
2438 |
+
"loss": 0.1211,
|
2439 |
+
"step": 81000
|
2440 |
+
},
|
2441 |
+
{
|
2442 |
+
"epoch": 0.55,
|
2443 |
+
"learning_rate": 0.0003533502973361997,
|
2444 |
+
"loss": 0.117,
|
2445 |
+
"step": 81200
|
2446 |
+
},
|
2447 |
+
{
|
2448 |
+
"epoch": 0.55,
|
2449 |
+
"learning_rate": 0.00035326865568220515,
|
2450 |
+
"loss": 0.1154,
|
2451 |
+
"step": 81400
|
2452 |
+
},
|
2453 |
+
{
|
2454 |
+
"epoch": 0.55,
|
2455 |
+
"learning_rate": 0.0003531868256899108,
|
2456 |
+
"loss": 0.1079,
|
2457 |
+
"step": 81600
|
2458 |
+
},
|
2459 |
+
{
|
2460 |
+
"epoch": 0.55,
|
2461 |
+
"learning_rate": 0.00035310480745094907,
|
2462 |
+
"loss": 0.1115,
|
2463 |
+
"step": 81800
|
2464 |
+
},
|
2465 |
+
{
|
2466 |
+
"epoch": 0.55,
|
2467 |
+
"learning_rate": 0.00035302260105716366,
|
2468 |
+
"loss": 0.1247,
|
2469 |
+
"step": 82000
|
2470 |
+
},
|
2471 |
+
{
|
2472 |
+
"epoch": 0.55,
|
2473 |
+
"learning_rate": 0.00035294061904054505,
|
2474 |
+
"loss": 0.119,
|
2475 |
+
"step": 82200
|
2476 |
+
},
|
2477 |
+
{
|
2478 |
+
"epoch": 0.56,
|
2479 |
+
"learning_rate": 0.00035285803755310787,
|
2480 |
+
"loss": 0.1219,
|
2481 |
+
"step": 82400
|
2482 |
+
},
|
2483 |
+
{
|
2484 |
+
"epoch": 0.56,
|
2485 |
+
"learning_rate": 0.00035277526818717824,
|
2486 |
+
"loss": 0.1113,
|
2487 |
+
"step": 82600
|
2488 |
+
},
|
2489 |
+
{
|
2490 |
+
"epoch": 0.56,
|
2491 |
+
"learning_rate": 0.00035269231103544056,
|
2492 |
+
"loss": 0.1112,
|
2493 |
+
"step": 82800
|
2494 |
+
},
|
2495 |
+
{
|
2496 |
+
"epoch": 0.56,
|
2497 |
+
"learning_rate": 0.00035260958238174535,
|
2498 |
+
"loss": 0.1099,
|
2499 |
+
"step": 83000
|
2500 |
+
},
|
2501 |
+
{
|
2502 |
+
"epoch": 0.56,
|
2503 |
+
"learning_rate": 0.00035252625087505373,
|
2504 |
+
"loss": 0.1519,
|
2505 |
+
"step": 83200
|
2506 |
+
},
|
2507 |
+
{
|
2508 |
+
"epoch": 0.56,
|
2509 |
+
"learning_rate": 0.00035244273186140206,
|
2510 |
+
"loss": 0.1178,
|
2511 |
+
"step": 83400
|
2512 |
+
},
|
2513 |
+
{
|
2514 |
+
"epoch": 0.56,
|
2515 |
+
"learning_rate": 0.0003523590254343144,
|
2516 |
+
"loss": 0.1095,
|
2517 |
+
"step": 83600
|
2518 |
+
},
|
2519 |
+
{
|
2520 |
+
"epoch": 0.56,
|
2521 |
+
"learning_rate": 0.00035227513168752453,
|
2522 |
+
"loss": 0.1161,
|
2523 |
+
"step": 83800
|
2524 |
+
},
|
2525 |
+
{
|
2526 |
+
"epoch": 0.57,
|
2527 |
+
"learning_rate": 0.0003521910507149763,
|
2528 |
+
"loss": 0.1091,
|
2529 |
+
"step": 84000
|
2530 |
+
},
|
2531 |
+
{
|
2532 |
+
"epoch": 0.57,
|
2533 |
+
"learning_rate": 0.00035210678261082275,
|
2534 |
+
"loss": 0.1176,
|
2535 |
+
"step": 84200
|
2536 |
+
},
|
2537 |
+
{
|
2538 |
+
"epoch": 0.57,
|
2539 |
+
"learning_rate": 0.00035202232746942697,
|
2540 |
+
"loss": 0.1197,
|
2541 |
+
"step": 84400
|
2542 |
+
},
|
2543 |
+
{
|
2544 |
+
"epoch": 0.57,
|
2545 |
+
"learning_rate": 0.0003519376853853612,
|
2546 |
+
"loss": 0.1163,
|
2547 |
+
"step": 84600
|
2548 |
+
},
|
2549 |
+
{
|
2550 |
+
"epoch": 0.57,
|
2551 |
+
"learning_rate": 0.0003518528564534069,
|
2552 |
+
"loss": 0.1152,
|
2553 |
+
"step": 84800
|
2554 |
+
},
|
2555 |
+
{
|
2556 |
+
"epoch": 0.57,
|
2557 |
+
"learning_rate": 0.0003517678407685551,
|
2558 |
+
"loss": 0.1086,
|
2559 |
+
"step": 85000
|
2560 |
+
},
|
2561 |
+
{
|
2562 |
+
"epoch": 0.57,
|
2563 |
+
"learning_rate": 0.0003516826384260057,
|
2564 |
+
"loss": 0.1023,
|
2565 |
+
"step": 85200
|
2566 |
+
},
|
2567 |
+
{
|
2568 |
+
"epoch": 0.58,
|
2569 |
+
"learning_rate": 0.00035159724952116777,
|
2570 |
+
"loss": 0.1111,
|
2571 |
+
"step": 85400
|
2572 |
+
},
|
2573 |
+
{
|
2574 |
+
"epoch": 0.58,
|
2575 |
+
"learning_rate": 0.0003515116741496592,
|
2576 |
+
"loss": 0.1195,
|
2577 |
+
"step": 85600
|
2578 |
+
},
|
2579 |
+
{
|
2580 |
+
"epoch": 0.58,
|
2581 |
+
"learning_rate": 0.0003514259124073067,
|
2582 |
+
"loss": 0.1031,
|
2583 |
+
"step": 85800
|
2584 |
+
},
|
2585 |
+
{
|
2586 |
+
"epoch": 0.58,
|
2587 |
+
"learning_rate": 0.00035133996439014564,
|
2588 |
+
"loss": 0.1232,
|
2589 |
+
"step": 86000
|
2590 |
+
},
|
2591 |
+
{
|
2592 |
+
"epoch": 0.58,
|
2593 |
+
"learning_rate": 0.00035125383019442015,
|
2594 |
+
"loss": 0.1214,
|
2595 |
+
"step": 86200
|
2596 |
+
},
|
2597 |
+
{
|
2598 |
+
"epoch": 0.58,
|
2599 |
+
"learning_rate": 0.0003511675099165826,
|
2600 |
+
"loss": 0.1129,
|
2601 |
+
"step": 86400
|
2602 |
+
},
|
2603 |
+
{
|
2604 |
+
"epoch": 0.58,
|
2605 |
+
"learning_rate": 0.0003510810036532938,
|
2606 |
+
"loss": 0.1312,
|
2607 |
+
"step": 86600
|
2608 |
+
},
|
2609 |
+
{
|
2610 |
+
"epoch": 0.58,
|
2611 |
+
"learning_rate": 0.00035099431150142305,
|
2612 |
+
"loss": 0.1087,
|
2613 |
+
"step": 86800
|
2614 |
+
},
|
2615 |
+
{
|
2616 |
+
"epoch": 0.59,
|
2617 |
+
"learning_rate": 0.00035090743355804756,
|
2618 |
+
"loss": 0.1117,
|
2619 |
+
"step": 87000
|
2620 |
+
},
|
2621 |
+
{
|
2622 |
+
"epoch": 0.59,
|
2623 |
+
"learning_rate": 0.0003508203699204526,
|
2624 |
+
"loss": 0.1073,
|
2625 |
+
"step": 87200
|
2626 |
+
},
|
2627 |
+
{
|
2628 |
+
"epoch": 0.59,
|
2629 |
+
"learning_rate": 0.0003507331206861315,
|
2630 |
+
"loss": 0.1101,
|
2631 |
+
"step": 87400
|
2632 |
+
},
|
2633 |
+
{
|
2634 |
+
"epoch": 0.59,
|
2635 |
+
"learning_rate": 0.000350646123587719,
|
2636 |
+
"loss": 0.1253,
|
2637 |
+
"step": 87600
|
2638 |
+
},
|
2639 |
+
{
|
2640 |
+
"epoch": 0.59,
|
2641 |
+
"learning_rate": 0.0003505585043800183,
|
2642 |
+
"loss": 0.1035,
|
2643 |
+
"step": 87800
|
2644 |
+
},
|
2645 |
+
{
|
2646 |
+
"epoch": 0.59,
|
2647 |
+
"learning_rate": 0.00035047069986882655,
|
2648 |
+
"loss": 0.1128,
|
2649 |
+
"step": 88000
|
2650 |
+
},
|
2651 |
+
{
|
2652 |
+
"epoch": 0.59,
|
2653 |
+
"learning_rate": 0.00035038271015246694,
|
2654 |
+
"loss": 0.1128,
|
2655 |
+
"step": 88200
|
2656 |
+
},
|
2657 |
+
{
|
2658 |
+
"epoch": 0.6,
|
2659 |
+
"learning_rate": 0.0003502945353294695,
|
2660 |
+
"loss": 0.1097,
|
2661 |
+
"step": 88400
|
2662 |
+
},
|
2663 |
+
{
|
2664 |
+
"epoch": 0.6,
|
2665 |
+
"learning_rate": 0.00035020617549857193,
|
2666 |
+
"loss": 0.1153,
|
2667 |
+
"step": 88600
|
2668 |
+
},
|
2669 |
+
{
|
2670 |
+
"epoch": 0.6,
|
2671 |
+
"learning_rate": 0.00035011763075871896,
|
2672 |
+
"loss": 0.1261,
|
2673 |
+
"step": 88800
|
2674 |
+
},
|
2675 |
+
{
|
2676 |
+
"epoch": 0.6,
|
2677 |
+
"learning_rate": 0.00035002890120906235,
|
2678 |
+
"loss": 0.1111,
|
2679 |
+
"step": 89000
|
2680 |
+
},
|
2681 |
+
{
|
2682 |
+
"epoch": 0.6,
|
2683 |
+
"learning_rate": 0.00034993998694896083,
|
2684 |
+
"loss": 0.121,
|
2685 |
+
"step": 89200
|
2686 |
+
},
|
2687 |
+
{
|
2688 |
+
"epoch": 0.6,
|
2689 |
+
"learning_rate": 0.00034985088807798,
|
2690 |
+
"loss": 0.1327,
|
2691 |
+
"step": 89400
|
2692 |
+
},
|
2693 |
+
{
|
2694 |
+
"epoch": 0.6,
|
2695 |
+
"learning_rate": 0.0003497616046958922,
|
2696 |
+
"loss": 0.1166,
|
2697 |
+
"step": 89600
|
2698 |
+
},
|
2699 |
+
{
|
2700 |
+
"epoch": 0.6,
|
2701 |
+
"learning_rate": 0.00034967213690267624,
|
2702 |
+
"loss": 0.1173,
|
2703 |
+
"step": 89800
|
2704 |
+
},
|
2705 |
+
{
|
2706 |
+
"epoch": 0.61,
|
2707 |
+
"learning_rate": 0.0003495824847985177,
|
2708 |
+
"loss": 0.1138,
|
2709 |
+
"step": 90000
|
2710 |
+
},
|
2711 |
+
{
|
2712 |
+
"epoch": 0.61,
|
2713 |
+
"learning_rate": 0.0003494926484838084,
|
2714 |
+
"loss": 0.1134,
|
2715 |
+
"step": 90200
|
2716 |
+
},
|
2717 |
+
{
|
2718 |
+
"epoch": 0.61,
|
2719 |
+
"learning_rate": 0.00034940262805914626,
|
2720 |
+
"loss": 0.1032,
|
2721 |
+
"step": 90400
|
2722 |
+
},
|
2723 |
+
{
|
2724 |
+
"epoch": 0.61,
|
2725 |
+
"learning_rate": 0.0003493124236253357,
|
2726 |
+
"loss": 0.1209,
|
2727 |
+
"step": 90600
|
2728 |
+
},
|
2729 |
+
{
|
2730 |
+
"epoch": 0.61,
|
2731 |
+
"learning_rate": 0.0003492224876824008,
|
2732 |
+
"loss": 0.1307,
|
2733 |
+
"step": 90800
|
2734 |
+
},
|
2735 |
+
{
|
2736 |
+
"epoch": 0.61,
|
2737 |
+
"learning_rate": 0.0003491319164523126,
|
2738 |
+
"loss": 0.1144,
|
2739 |
+
"step": 91000
|
2740 |
+
},
|
2741 |
+
{
|
2742 |
+
"epoch": 0.61,
|
2743 |
+
"learning_rate": 0.00034904116151621703,
|
2744 |
+
"loss": 0.1174,
|
2745 |
+
"step": 91200
|
2746 |
+
},
|
2747 |
+
{
|
2748 |
+
"epoch": 0.62,
|
2749 |
+
"learning_rate": 0.00034895067812499057,
|
2750 |
+
"loss": 0.1099,
|
2751 |
+
"step": 91400
|
2752 |
+
},
|
2753 |
+
{
|
2754 |
+
"epoch": 0.62,
|
2755 |
+
"learning_rate": 0.00034885955699922493,
|
2756 |
+
"loss": 0.1104,
|
2757 |
+
"step": 91600
|
2758 |
+
},
|
2759 |
+
{
|
2760 |
+
"epoch": 0.62,
|
2761 |
+
"learning_rate": 0.00034876825247243814,
|
2762 |
+
"loss": 0.1115,
|
2763 |
+
"step": 91800
|
2764 |
+
},
|
2765 |
+
{
|
2766 |
+
"epoch": 0.62,
|
2767 |
+
"learning_rate": 0.0003486767646468723,
|
2768 |
+
"loss": 0.1149,
|
2769 |
+
"step": 92000
|
2770 |
+
},
|
2771 |
+
{
|
2772 |
+
"epoch": 0.62,
|
2773 |
+
"learning_rate": 0.0003485850936249749,
|
2774 |
+
"loss": 0.1115,
|
2775 |
+
"step": 92200
|
2776 |
+
},
|
2777 |
+
{
|
2778 |
+
"epoch": 0.62,
|
2779 |
+
"learning_rate": 0.00034849323950939855,
|
2780 |
+
"loss": 0.124,
|
2781 |
+
"step": 92400
|
2782 |
+
},
|
2783 |
+
{
|
2784 |
+
"epoch": 0.62,
|
2785 |
+
"learning_rate": 0.00034840120240300075,
|
2786 |
+
"loss": 0.1309,
|
2787 |
+
"step": 92600
|
2788 |
+
},
|
2789 |
+
{
|
2790 |
+
"epoch": 0.63,
|
2791 |
+
"learning_rate": 0.0003483089824088442,
|
2792 |
+
"loss": 0.1079,
|
2793 |
+
"step": 92800
|
2794 |
+
},
|
2795 |
+
{
|
2796 |
+
"epoch": 0.63,
|
2797 |
+
"learning_rate": 0.000348216579630196,
|
2798 |
+
"loss": 0.1093,
|
2799 |
+
"step": 93000
|
2800 |
+
},
|
2801 |
+
{
|
2802 |
+
"epoch": 0.63,
|
2803 |
+
"learning_rate": 0.00034812399417052833,
|
2804 |
+
"loss": 0.1159,
|
2805 |
+
"step": 93200
|
2806 |
+
},
|
2807 |
+
{
|
2808 |
+
"epoch": 0.63,
|
2809 |
+
"learning_rate": 0.00034803122613351773,
|
2810 |
+
"loss": 0.1172,
|
2811 |
+
"step": 93400
|
2812 |
+
},
|
2813 |
+
{
|
2814 |
+
"epoch": 0.63,
|
2815 |
+
"learning_rate": 0.00034793827562304517,
|
2816 |
+
"loss": 0.1127,
|
2817 |
+
"step": 93600
|
2818 |
+
},
|
2819 |
+
{
|
2820 |
+
"epoch": 0.63,
|
2821 |
+
"learning_rate": 0.000347845142743196,
|
2822 |
+
"loss": 0.1152,
|
2823 |
+
"step": 93800
|
2824 |
+
},
|
2825 |
+
{
|
2826 |
+
"epoch": 0.63,
|
2827 |
+
"learning_rate": 0.0003477518275982598,
|
2828 |
+
"loss": 0.1123,
|
2829 |
+
"step": 94000
|
2830 |
+
},
|
2831 |
+
{
|
2832 |
+
"epoch": 0.63,
|
2833 |
+
"learning_rate": 0.00034765833029273027,
|
2834 |
+
"loss": 0.1321,
|
2835 |
+
"step": 94200
|
2836 |
+
},
|
2837 |
+
{
|
2838 |
+
"epoch": 0.64,
|
2839 |
+
"learning_rate": 0.00034756465093130503,
|
2840 |
+
"loss": 0.1116,
|
2841 |
+
"step": 94400
|
2842 |
+
},
|
2843 |
+
{
|
2844 |
+
"epoch": 0.64,
|
2845 |
+
"learning_rate": 0.0003474707896188856,
|
2846 |
+
"loss": 0.1063,
|
2847 |
+
"step": 94600
|
2848 |
+
},
|
2849 |
+
{
|
2850 |
+
"epoch": 0.64,
|
2851 |
+
"learning_rate": 0.00034737721712853624,
|
2852 |
+
"loss": 0.1339,
|
2853 |
+
"step": 94800
|
2854 |
+
},
|
2855 |
+
{
|
2856 |
+
"epoch": 0.64,
|
2857 |
+
"learning_rate": 0.0003472829931380883,
|
2858 |
+
"loss": 0.1165,
|
2859 |
+
"step": 95000
|
2860 |
+
},
|
2861 |
+
{
|
2862 |
+
"epoch": 0.64,
|
2863 |
+
"learning_rate": 0.0003471885875120447,
|
2864 |
+
"loss": 0.1187,
|
2865 |
+
"step": 95200
|
2866 |
+
},
|
2867 |
+
{
|
2868 |
+
"epoch": 0.64,
|
2869 |
+
"learning_rate": 0.0003470940003561202,
|
2870 |
+
"loss": 0.1201,
|
2871 |
+
"step": 95400
|
2872 |
+
},
|
2873 |
+
{
|
2874 |
+
"epoch": 0.64,
|
2875 |
+
"learning_rate": 0.0003469992317762327,
|
2876 |
+
"loss": 0.1115,
|
2877 |
+
"step": 95600
|
2878 |
+
},
|
2879 |
+
{
|
2880 |
+
"epoch": 0.65,
|
2881 |
+
"learning_rate": 0.0003469042818785036,
|
2882 |
+
"loss": 0.1031,
|
2883 |
+
"step": 95800
|
2884 |
+
},
|
2885 |
+
{
|
2886 |
+
"epoch": 0.65,
|
2887 |
+
"learning_rate": 0.000346809150769257,
|
2888 |
+
"loss": 0.1212,
|
2889 |
+
"step": 96000
|
2890 |
+
},
|
2891 |
+
{
|
2892 |
+
"epoch": 0.65,
|
2893 |
+
"learning_rate": 0.00034671383855502014,
|
2894 |
+
"loss": 0.1148,
|
2895 |
+
"step": 96200
|
2896 |
+
},
|
2897 |
+
{
|
2898 |
+
"epoch": 0.65,
|
2899 |
+
"learning_rate": 0.000346618345342523,
|
2900 |
+
"loss": 0.1273,
|
2901 |
+
"step": 96400
|
2902 |
+
},
|
2903 |
+
{
|
2904 |
+
"epoch": 0.65,
|
2905 |
+
"learning_rate": 0.0003465226712386982,
|
2906 |
+
"loss": 0.1204,
|
2907 |
+
"step": 96600
|
2908 |
+
},
|
2909 |
+
{
|
2910 |
+
"epoch": 0.65,
|
2911 |
+
"learning_rate": 0.00034642681635068095,
|
2912 |
+
"loss": 0.1101,
|
2913 |
+
"step": 96800
|
2914 |
+
},
|
2915 |
+
{
|
2916 |
+
"epoch": 0.65,
|
2917 |
+
"learning_rate": 0.0003463307807858089,
|
2918 |
+
"loss": 0.1171,
|
2919 |
+
"step": 97000
|
2920 |
+
},
|
2921 |
+
{
|
2922 |
+
"epoch": 0.65,
|
2923 |
+
"learning_rate": 0.00034623456465162207,
|
2924 |
+
"loss": 0.1197,
|
2925 |
+
"step": 97200
|
2926 |
+
},
|
2927 |
+
{
|
2928 |
+
"epoch": 0.66,
|
2929 |
+
"learning_rate": 0.00034613816805586253,
|
2930 |
+
"loss": 0.1153,
|
2931 |
+
"step": 97400
|
2932 |
+
},
|
2933 |
+
{
|
2934 |
+
"epoch": 0.66,
|
2935 |
+
"learning_rate": 0.0003460415911064746,
|
2936 |
+
"loss": 0.1155,
|
2937 |
+
"step": 97600
|
2938 |
+
},
|
2939 |
+
{
|
2940 |
+
"epoch": 0.66,
|
2941 |
+
"learning_rate": 0.00034594483391160453,
|
2942 |
+
"loss": 0.1134,
|
2943 |
+
"step": 97800
|
2944 |
+
},
|
2945 |
+
{
|
2946 |
+
"epoch": 0.66,
|
2947 |
+
"learning_rate": 0.0003458478965796003,
|
2948 |
+
"loss": 0.1058,
|
2949 |
+
"step": 98000
|
2950 |
+
},
|
2951 |
+
{
|
2952 |
+
"epoch": 0.66,
|
2953 |
+
"learning_rate": 0.00034575077921901166,
|
2954 |
+
"loss": 0.1172,
|
2955 |
+
"step": 98200
|
2956 |
+
},
|
2957 |
+
{
|
2958 |
+
"epoch": 0.66,
|
2959 |
+
"learning_rate": 0.0003456534819385901,
|
2960 |
+
"loss": 0.1182,
|
2961 |
+
"step": 98400
|
2962 |
+
},
|
2963 |
+
{
|
2964 |
+
"epoch": 0.66,
|
2965 |
+
"learning_rate": 0.00034555600484728827,
|
2966 |
+
"loss": 0.1244,
|
2967 |
+
"step": 98600
|
2968 |
+
},
|
2969 |
+
{
|
2970 |
+
"epoch": 0.67,
|
2971 |
+
"learning_rate": 0.0003454583480542605,
|
2972 |
+
"loss": 0.1117,
|
2973 |
+
"step": 98800
|
2974 |
+
},
|
2975 |
+
{
|
2976 |
+
"epoch": 0.67,
|
2977 |
+
"learning_rate": 0.000345360511668862,
|
2978 |
+
"loss": 0.1088,
|
2979 |
+
"step": 99000
|
2980 |
+
},
|
2981 |
+
{
|
2982 |
+
"epoch": 0.67,
|
2983 |
+
"learning_rate": 0.0003452624958006496,
|
2984 |
+
"loss": 0.1136,
|
2985 |
+
"step": 99200
|
2986 |
+
},
|
2987 |
+
{
|
2988 |
+
"epoch": 0.67,
|
2989 |
+
"learning_rate": 0.0003451643005593805,
|
2990 |
+
"loss": 0.1148,
|
2991 |
+
"step": 99400
|
2992 |
+
},
|
2993 |
+
{
|
2994 |
+
"epoch": 0.67,
|
2995 |
+
"learning_rate": 0.0003450659260550133,
|
2996 |
+
"loss": 0.105,
|
2997 |
+
"step": 99600
|
2998 |
+
},
|
2999 |
+
{
|
3000 |
+
"epoch": 0.67,
|
3001 |
+
"learning_rate": 0.00034496786561145394,
|
3002 |
+
"loss": 0.1195,
|
3003 |
+
"step": 99800
|
3004 |
+
},
|
3005 |
+
{
|
3006 |
+
"epoch": 0.67,
|
3007 |
+
"learning_rate": 0.0003448691338065063,
|
3008 |
+
"loss": 0.1182,
|
3009 |
+
"step": 100000
|
3010 |
+
},
|
3011 |
+
{
|
3012 |
+
"epoch": 0.68,
|
3013 |
+
"learning_rate": 0.0003447702230689861,
|
3014 |
+
"loss": 0.1163,
|
3015 |
+
"step": 100200
|
3016 |
+
},
|
3017 |
+
{
|
3018 |
+
"epoch": 0.68,
|
3019 |
+
"learning_rate": 0.000344671133509653,
|
3020 |
+
"loss": 0.1245,
|
3021 |
+
"step": 100400
|
3022 |
+
},
|
3023 |
+
{
|
3024 |
+
"epoch": 0.68,
|
3025 |
+
"learning_rate": 0.0003445718652394668,
|
3026 |
+
"loss": 0.1046,
|
3027 |
+
"step": 100600
|
3028 |
+
},
|
3029 |
+
{
|
3030 |
+
"epoch": 0.68,
|
3031 |
+
"learning_rate": 0.0003444724183695874,
|
3032 |
+
"loss": 0.1182,
|
3033 |
+
"step": 100800
|
3034 |
+
},
|
3035 |
+
{
|
3036 |
+
"epoch": 0.68,
|
3037 |
+
"learning_rate": 0.0003443727930113748,
|
3038 |
+
"loss": 0.1207,
|
3039 |
+
"step": 101000
|
3040 |
+
},
|
3041 |
+
{
|
3042 |
+
"epoch": 0.68,
|
3043 |
+
"learning_rate": 0.00034427298927638874,
|
3044 |
+
"loss": 0.1079,
|
3045 |
+
"step": 101200
|
3046 |
+
},
|
3047 |
+
{
|
3048 |
+
"epoch": 0.68,
|
3049 |
+
"learning_rate": 0.0003441730072763888,
|
3050 |
+
"loss": 0.104,
|
3051 |
+
"step": 101400
|
3052 |
+
},
|
3053 |
+
{
|
3054 |
+
"epoch": 0.68,
|
3055 |
+
"learning_rate": 0.0003440728471233341,
|
3056 |
+
"loss": 0.1102,
|
3057 |
+
"step": 101600
|
3058 |
+
},
|
3059 |
+
{
|
3060 |
+
"epoch": 0.69,
|
3061 |
+
"learning_rate": 0.000343973011063044,
|
3062 |
+
"loss": 0.1252,
|
3063 |
+
"step": 101800
|
3064 |
+
},
|
3065 |
+
{
|
3066 |
+
"epoch": 0.69,
|
3067 |
+
"learning_rate": 0.0003438724958299181,
|
3068 |
+
"loss": 0.1109,
|
3069 |
+
"step": 102000
|
3070 |
+
},
|
3071 |
+
{
|
3072 |
+
"epoch": 0.69,
|
3073 |
+
"learning_rate": 0.00034377230668762864,
|
3074 |
+
"loss": 0.1208,
|
3075 |
+
"step": 102200
|
3076 |
+
},
|
3077 |
+
{
|
3078 |
+
"epoch": 0.69,
|
3079 |
+
"learning_rate": 0.00034367143682240804,
|
3080 |
+
"loss": 0.1033,
|
3081 |
+
"step": 102400
|
3082 |
+
},
|
3083 |
+
{
|
3084 |
+
"epoch": 0.69,
|
3085 |
+
"learning_rate": 0.0003435703893657879,
|
3086 |
+
"loss": 0.1112,
|
3087 |
+
"step": 102600
|
3088 |
+
},
|
3089 |
+
{
|
3090 |
+
"epoch": 0.69,
|
3091 |
+
"learning_rate": 0.0003434691644309205,
|
3092 |
+
"loss": 0.1153,
|
3093 |
+
"step": 102800
|
3094 |
+
},
|
3095 |
+
{
|
3096 |
+
"epoch": 0.69,
|
3097 |
+
"learning_rate": 0.00034336776213115676,
|
3098 |
+
"loss": 0.1053,
|
3099 |
+
"step": 103000
|
3100 |
+
},
|
3101 |
+
{
|
3102 |
+
"epoch": 0.7,
|
3103 |
+
"learning_rate": 0.0003432661825800463,
|
3104 |
+
"loss": 0.1188,
|
3105 |
+
"step": 103200
|
3106 |
+
},
|
3107 |
+
{
|
3108 |
+
"epoch": 0.7,
|
3109 |
+
"learning_rate": 0.00034316442589133725,
|
3110 |
+
"loss": 0.1069,
|
3111 |
+
"step": 103400
|
3112 |
+
},
|
3113 |
+
{
|
3114 |
+
"epoch": 0.7,
|
3115 |
+
"learning_rate": 0.000343062492178976,
|
3116 |
+
"loss": 0.1195,
|
3117 |
+
"step": 103600
|
3118 |
+
},
|
3119 |
+
{
|
3120 |
+
"epoch": 0.7,
|
3121 |
+
"learning_rate": 0.00034296038155710726,
|
3122 |
+
"loss": 0.1099,
|
3123 |
+
"step": 103800
|
3124 |
+
},
|
3125 |
+
{
|
3126 |
+
"epoch": 0.7,
|
3127 |
+
"learning_rate": 0.0003428580941400738,
|
3128 |
+
"loss": 0.1185,
|
3129 |
+
"step": 104000
|
3130 |
+
},
|
3131 |
+
{
|
3132 |
+
"epoch": 0.7,
|
3133 |
+
"learning_rate": 0.0003427556300424164,
|
3134 |
+
"loss": 0.1247,
|
3135 |
+
"step": 104200
|
3136 |
+
},
|
3137 |
+
{
|
3138 |
+
"epoch": 0.7,
|
3139 |
+
"learning_rate": 0.0003426529893788736,
|
3140 |
+
"loss": 0.111,
|
3141 |
+
"step": 104400
|
3142 |
+
},
|
3143 |
+
{
|
3144 |
+
"epoch": 0.7,
|
3145 |
+
"learning_rate": 0.00034255017226438175,
|
3146 |
+
"loss": 0.1043,
|
3147 |
+
"step": 104600
|
3148 |
+
},
|
3149 |
+
{
|
3150 |
+
"epoch": 0.71,
|
3151 |
+
"learning_rate": 0.00034244717881407483,
|
3152 |
+
"loss": 0.1183,
|
3153 |
+
"step": 104800
|
3154 |
+
},
|
3155 |
+
{
|
3156 |
+
"epoch": 0.71,
|
3157 |
+
"learning_rate": 0.00034234452542979544,
|
3158 |
+
"loss": 0.1165,
|
3159 |
+
"step": 105000
|
3160 |
+
},
|
3161 |
+
{
|
3162 |
+
"epoch": 0.71,
|
3163 |
+
"learning_rate": 0.00034224118053428676,
|
3164 |
+
"loss": 0.1178,
|
3165 |
+
"step": 105200
|
3166 |
+
},
|
3167 |
+
{
|
3168 |
+
"epoch": 0.71,
|
3169 |
+
"learning_rate": 0.0003421376596489696,
|
3170 |
+
"loss": 0.1283,
|
3171 |
+
"step": 105400
|
3172 |
+
},
|
3173 |
+
{
|
3174 |
+
"epoch": 0.71,
|
3175 |
+
"learning_rate": 0.00034203396288976624,
|
3176 |
+
"loss": 0.1091,
|
3177 |
+
"step": 105600
|
3178 |
+
},
|
3179 |
+
{
|
3180 |
+
"epoch": 0.71,
|
3181 |
+
"learning_rate": 0.00034193009037279546,
|
3182 |
+
"loss": 0.1101,
|
3183 |
+
"step": 105800
|
3184 |
+
},
|
3185 |
+
{
|
3186 |
+
"epoch": 0.71,
|
3187 |
+
"learning_rate": 0.00034182604221437305,
|
3188 |
+
"loss": 0.1147,
|
3189 |
+
"step": 106000
|
3190 |
+
},
|
3191 |
+
{
|
3192 |
+
"epoch": 0.72,
|
3193 |
+
"learning_rate": 0.0003417218185310113,
|
3194 |
+
"loss": 0.1052,
|
3195 |
+
"step": 106200
|
3196 |
+
},
|
3197 |
+
{
|
3198 |
+
"epoch": 0.72,
|
3199 |
+
"learning_rate": 0.00034161741943941935,
|
3200 |
+
"loss": 0.1089,
|
3201 |
+
"step": 106400
|
3202 |
+
},
|
3203 |
+
{
|
3204 |
+
"epoch": 0.72,
|
3205 |
+
"learning_rate": 0.0003415128450565025,
|
3206 |
+
"loss": 0.1223,
|
3207 |
+
"step": 106600
|
3208 |
+
},
|
3209 |
+
{
|
3210 |
+
"epoch": 0.72,
|
3211 |
+
"learning_rate": 0.00034140809549936245,
|
3212 |
+
"loss": 0.1067,
|
3213 |
+
"step": 106800
|
3214 |
+
},
|
3215 |
+
{
|
3216 |
+
"epoch": 0.72,
|
3217 |
+
"learning_rate": 0.00034130317088529706,
|
3218 |
+
"loss": 0.1182,
|
3219 |
+
"step": 107000
|
3220 |
+
},
|
3221 |
+
{
|
3222 |
+
"epoch": 0.72,
|
3223 |
+
"learning_rate": 0.0003411980713318002,
|
3224 |
+
"loss": 0.1026,
|
3225 |
+
"step": 107200
|
3226 |
+
},
|
3227 |
+
{
|
3228 |
+
"epoch": 0.72,
|
3229 |
+
"learning_rate": 0.00034109279695656154,
|
3230 |
+
"loss": 0.114,
|
3231 |
+
"step": 107400
|
3232 |
+
},
|
3233 |
+
{
|
3234 |
+
"epoch": 0.72,
|
3235 |
+
"learning_rate": 0.00034098734787746675,
|
3236 |
+
"loss": 0.1046,
|
3237 |
+
"step": 107600
|
3238 |
+
},
|
3239 |
+
{
|
3240 |
+
"epoch": 0.73,
|
3241 |
+
"learning_rate": 0.00034088172421259687,
|
3242 |
+
"loss": 0.114,
|
3243 |
+
"step": 107800
|
3244 |
+
},
|
3245 |
+
{
|
3246 |
+
"epoch": 0.73,
|
3247 |
+
"learning_rate": 0.00034077592608022865,
|
3248 |
+
"loss": 0.1079,
|
3249 |
+
"step": 108000
|
3250 |
+
},
|
3251 |
+
{
|
3252 |
+
"epoch": 0.73,
|
3253 |
+
"learning_rate": 0.00034066995359883405,
|
3254 |
+
"loss": 0.1134,
|
3255 |
+
"step": 108200
|
3256 |
+
},
|
3257 |
+
{
|
3258 |
+
"epoch": 0.73,
|
3259 |
+
"learning_rate": 0.00034056380688708037,
|
3260 |
+
"loss": 0.1188,
|
3261 |
+
"step": 108400
|
3262 |
+
},
|
3263 |
+
{
|
3264 |
+
"epoch": 0.73,
|
3265 |
+
"learning_rate": 0.00034045748606383004,
|
3266 |
+
"loss": 0.1123,
|
3267 |
+
"step": 108600
|
3268 |
+
},
|
3269 |
+
{
|
3270 |
+
"epoch": 0.73,
|
3271 |
+
"learning_rate": 0.00034035099124814027,
|
3272 |
+
"loss": 0.109,
|
3273 |
+
"step": 108800
|
3274 |
+
},
|
3275 |
+
{
|
3276 |
+
"epoch": 0.73,
|
3277 |
+
"learning_rate": 0.0003402443225592633,
|
3278 |
+
"loss": 0.1105,
|
3279 |
+
"step": 109000
|
3280 |
+
},
|
3281 |
+
{
|
3282 |
+
"epoch": 0.74,
|
3283 |
+
"learning_rate": 0.00034013748011664594,
|
3284 |
+
"loss": 0.117,
|
3285 |
+
"step": 109200
|
3286 |
+
},
|
3287 |
+
{
|
3288 |
+
"epoch": 0.74,
|
3289 |
+
"learning_rate": 0.00034003099955202997,
|
3290 |
+
"loss": 0.1196,
|
3291 |
+
"step": 109400
|
3292 |
+
},
|
3293 |
+
{
|
3294 |
+
"epoch": 0.74,
|
3295 |
+
"learning_rate": 0.0003399238108283236,
|
3296 |
+
"loss": 0.112,
|
3297 |
+
"step": 109600
|
3298 |
+
},
|
3299 |
+
{
|
3300 |
+
"epoch": 0.74,
|
3301 |
+
"learning_rate": 0.00033981644870978374,
|
3302 |
+
"loss": 0.1267,
|
3303 |
+
"step": 109800
|
3304 |
+
},
|
3305 |
+
{
|
3306 |
+
"epoch": 0.74,
|
3307 |
+
"learning_rate": 0.000339709451424421,
|
3308 |
+
"loss": 0.1266,
|
3309 |
+
"step": 110000
|
3310 |
+
},
|
3311 |
+
{
|
3312 |
+
"epoch": 0.74,
|
3313 |
+
"learning_rate": 0.0003396017437425494,
|
3314 |
+
"loss": 0.1283,
|
3315 |
+
"step": 110200
|
3316 |
+
},
|
3317 |
+
{
|
3318 |
+
"epoch": 0.74,
|
3319 |
+
"learning_rate": 0.0003394938630264929,
|
3320 |
+
"loss": 0.1104,
|
3321 |
+
"step": 110400
|
3322 |
+
},
|
3323 |
+
{
|
3324 |
+
"epoch": 0.75,
|
3325 |
+
"learning_rate": 0.0003393858093970554,
|
3326 |
+
"loss": 0.1183,
|
3327 |
+
"step": 110600
|
3328 |
+
},
|
3329 |
+
{
|
3330 |
+
"epoch": 0.75,
|
3331 |
+
"learning_rate": 0.0003392775829752349,
|
3332 |
+
"loss": 0.1186,
|
3333 |
+
"step": 110800
|
3334 |
+
},
|
3335 |
+
{
|
3336 |
+
"epoch": 0.75,
|
3337 |
+
"learning_rate": 0.00033916918388222233,
|
3338 |
+
"loss": 0.1141,
|
3339 |
+
"step": 111000
|
3340 |
+
},
|
3341 |
+
{
|
3342 |
+
"epoch": 0.75,
|
3343 |
+
"learning_rate": 0.0003390606122394025,
|
3344 |
+
"loss": 0.114,
|
3345 |
+
"step": 111200
|
3346 |
+
},
|
3347 |
+
{
|
3348 |
+
"epoch": 0.75,
|
3349 |
+
"learning_rate": 0.000338951868168353,
|
3350 |
+
"loss": 0.1064,
|
3351 |
+
"step": 111400
|
3352 |
+
},
|
3353 |
+
{
|
3354 |
+
"epoch": 0.75,
|
3355 |
+
"learning_rate": 0.00033884295179084477,
|
3356 |
+
"loss": 0.1095,
|
3357 |
+
"step": 111600
|
3358 |
+
},
|
3359 |
+
{
|
3360 |
+
"epoch": 0.75,
|
3361 |
+
"learning_rate": 0.00033873386322884173,
|
3362 |
+
"loss": 0.1064,
|
3363 |
+
"step": 111800
|
3364 |
+
},
|
3365 |
+
{
|
3366 |
+
"epoch": 0.75,
|
3367 |
+
"learning_rate": 0.0003386246026045003,
|
3368 |
+
"loss": 0.1207,
|
3369 |
+
"step": 112000
|
3370 |
+
},
|
3371 |
+
{
|
3372 |
+
"epoch": 0.76,
|
3373 |
+
"learning_rate": 0.00033851517004016994,
|
3374 |
+
"loss": 0.1147,
|
3375 |
+
"step": 112200
|
3376 |
+
},
|
3377 |
+
{
|
3378 |
+
"epoch": 0.76,
|
3379 |
+
"learning_rate": 0.0003384055656583925,
|
3380 |
+
"loss": 0.1238,
|
3381 |
+
"step": 112400
|
3382 |
+
},
|
3383 |
+
{
|
3384 |
+
"epoch": 0.76,
|
3385 |
+
"learning_rate": 0.0003382957895819021,
|
3386 |
+
"loss": 0.1208,
|
3387 |
+
"step": 112600
|
3388 |
+
},
|
3389 |
+
{
|
3390 |
+
"epoch": 0.76,
|
3391 |
+
"learning_rate": 0.00033818584193362537,
|
3392 |
+
"loss": 0.1013,
|
3393 |
+
"step": 112800
|
3394 |
+
},
|
3395 |
+
{
|
3396 |
+
"epoch": 0.76,
|
3397 |
+
"learning_rate": 0.00033807572283668094,
|
3398 |
+
"loss": 0.1144,
|
3399 |
+
"step": 113000
|
3400 |
+
},
|
3401 |
+
{
|
3402 |
+
"epoch": 0.76,
|
3403 |
+
"learning_rate": 0.0003379654324143793,
|
3404 |
+
"loss": 0.1062,
|
3405 |
+
"step": 113200
|
3406 |
+
},
|
3407 |
+
{
|
3408 |
+
"epoch": 0.76,
|
3409 |
+
"learning_rate": 0.00033785497079022296,
|
3410 |
+
"loss": 0.1134,
|
3411 |
+
"step": 113400
|
3412 |
+
},
|
3413 |
+
{
|
3414 |
+
"epoch": 0.77,
|
3415 |
+
"learning_rate": 0.00033774433808790615,
|
3416 |
+
"loss": 0.1087,
|
3417 |
+
"step": 113600
|
3418 |
+
},
|
3419 |
+
{
|
3420 |
+
"epoch": 0.77,
|
3421 |
+
"learning_rate": 0.0003376335344313145,
|
3422 |
+
"loss": 0.1135,
|
3423 |
+
"step": 113800
|
3424 |
+
},
|
3425 |
+
{
|
3426 |
+
"epoch": 0.77,
|
3427 |
+
"learning_rate": 0.0003375225599445254,
|
3428 |
+
"loss": 0.1042,
|
3429 |
+
"step": 114000
|
3430 |
+
},
|
3431 |
+
{
|
3432 |
+
"epoch": 0.77,
|
3433 |
+
"learning_rate": 0.0003374114147518071,
|
3434 |
+
"loss": 0.1081,
|
3435 |
+
"step": 114200
|
3436 |
+
},
|
3437 |
+
{
|
3438 |
+
"epoch": 0.77,
|
3439 |
+
"learning_rate": 0.0003373006559806056,
|
3440 |
+
"loss": 0.1206,
|
3441 |
+
"step": 114400
|
3442 |
+
},
|
3443 |
+
{
|
3444 |
+
"epoch": 0.77,
|
3445 |
+
"learning_rate": 0.00033718917060157275,
|
3446 |
+
"loss": 0.1112,
|
3447 |
+
"step": 114600
|
3448 |
+
},
|
3449 |
+
{
|
3450 |
+
"epoch": 0.77,
|
3451 |
+
"learning_rate": 0.0003370775148899379,
|
3452 |
+
"loss": 0.1075,
|
3453 |
+
"step": 114800
|
3454 |
+
},
|
3455 |
+
{
|
3456 |
+
"epoch": 0.77,
|
3457 |
+
"learning_rate": 0.0003369656889707324,
|
3458 |
+
"loss": 0.1091,
|
3459 |
+
"step": 115000
|
3460 |
+
},
|
3461 |
+
{
|
3462 |
+
"epoch": 0.78,
|
3463 |
+
"learning_rate": 0.0003368536929691781,
|
3464 |
+
"loss": 0.1104,
|
3465 |
+
"step": 115200
|
3466 |
+
},
|
3467 |
+
{
|
3468 |
+
"epoch": 0.78,
|
3469 |
+
"learning_rate": 0.0003367415270106874,
|
3470 |
+
"loss": 0.1055,
|
3471 |
+
"step": 115400
|
3472 |
+
},
|
3473 |
+
{
|
3474 |
+
"epoch": 0.78,
|
3475 |
+
"learning_rate": 0.00033662919122086295,
|
3476 |
+
"loss": 0.11,
|
3477 |
+
"step": 115600
|
3478 |
+
},
|
3479 |
+
{
|
3480 |
+
"epoch": 0.78,
|
3481 |
+
"learning_rate": 0.00033651668572549757,
|
3482 |
+
"loss": 0.1178,
|
3483 |
+
"step": 115800
|
3484 |
+
},
|
3485 |
+
{
|
3486 |
+
"epoch": 0.78,
|
3487 |
+
"learning_rate": 0.0003364040106505742,
|
3488 |
+
"loss": 0.1063,
|
3489 |
+
"step": 116000
|
3490 |
+
},
|
3491 |
+
{
|
3492 |
+
"epoch": 0.78,
|
3493 |
+
"learning_rate": 0.00033629116612226564,
|
3494 |
+
"loss": 0.1232,
|
3495 |
+
"step": 116200
|
3496 |
+
},
|
3497 |
+
{
|
3498 |
+
"epoch": 0.78,
|
3499 |
+
"learning_rate": 0.0003361781522669344,
|
3500 |
+
"loss": 0.1143,
|
3501 |
+
"step": 116400
|
3502 |
+
},
|
3503 |
+
{
|
3504 |
+
"epoch": 0.79,
|
3505 |
+
"learning_rate": 0.0003360655355470882,
|
3506 |
+
"loss": 0.1233,
|
3507 |
+
"step": 116600
|
3508 |
+
},
|
3509 |
+
{
|
3510 |
+
"epoch": 0.79,
|
3511 |
+
"learning_rate": 0.0003359521842626105,
|
3512 |
+
"loss": 0.1191,
|
3513 |
+
"step": 116800
|
3514 |
+
},
|
3515 |
+
{
|
3516 |
+
"epoch": 0.79,
|
3517 |
+
"learning_rate": 0.00033583866403069976,
|
3518 |
+
"loss": 0.1114,
|
3519 |
+
"step": 117000
|
3520 |
+
},
|
3521 |
+
{
|
3522 |
+
"epoch": 0.79,
|
3523 |
+
"learning_rate": 0.0003357249749784751,
|
3524 |
+
"loss": 0.1075,
|
3525 |
+
"step": 117200
|
3526 |
+
},
|
3527 |
+
{
|
3528 |
+
"epoch": 0.79,
|
3529 |
+
"learning_rate": 0.0003356111172332447,
|
3530 |
+
"loss": 0.1084,
|
3531 |
+
"step": 117400
|
3532 |
+
},
|
3533 |
+
{
|
3534 |
+
"epoch": 0.79,
|
3535 |
+
"learning_rate": 0.0003354970909225057,
|
3536 |
+
"loss": 0.1037,
|
3537 |
+
"step": 117600
|
3538 |
+
},
|
3539 |
+
{
|
3540 |
+
"epoch": 0.79,
|
3541 |
+
"learning_rate": 0.0003353828961739441,
|
3542 |
+
"loss": 0.1187,
|
3543 |
+
"step": 117800
|
3544 |
+
},
|
3545 |
+
{
|
3546 |
+
"epoch": 0.79,
|
3547 |
+
"learning_rate": 0.00033526853311543423,
|
3548 |
+
"loss": 0.1078,
|
3549 |
+
"step": 118000
|
3550 |
+
},
|
3551 |
+
{
|
3552 |
+
"epoch": 0.8,
|
3553 |
+
"learning_rate": 0.00033515400187503924,
|
3554 |
+
"loss": 0.1083,
|
3555 |
+
"step": 118200
|
3556 |
+
},
|
3557 |
+
{
|
3558 |
+
"epoch": 0.8,
|
3559 |
+
"learning_rate": 0.00033503930258101024,
|
3560 |
+
"loss": 0.1163,
|
3561 |
+
"step": 118400
|
3562 |
+
},
|
3563 |
+
{
|
3564 |
+
"epoch": 0.8,
|
3565 |
+
"learning_rate": 0.0003349244353617869,
|
3566 |
+
"loss": 0.101,
|
3567 |
+
"step": 118600
|
3568 |
+
},
|
3569 |
+
{
|
3570 |
+
"epoch": 0.8,
|
3571 |
+
"learning_rate": 0.00033480997593825657,
|
3572 |
+
"loss": 0.1162,
|
3573 |
+
"step": 118800
|
3574 |
+
},
|
3575 |
+
{
|
3576 |
+
"epoch": 0.8,
|
3577 |
+
"learning_rate": 0.00033469477409273286,
|
3578 |
+
"loss": 0.1088,
|
3579 |
+
"step": 119000
|
3580 |
+
},
|
3581 |
+
{
|
3582 |
+
"epoch": 0.8,
|
3583 |
+
"learning_rate": 0.00033457940470781543,
|
3584 |
+
"loss": 0.1168,
|
3585 |
+
"step": 119200
|
3586 |
+
},
|
3587 |
+
{
|
3588 |
+
"epoch": 0.8,
|
3589 |
+
"learning_rate": 0.0003344638679126941,
|
3590 |
+
"loss": 0.1154,
|
3591 |
+
"step": 119400
|
3592 |
+
},
|
3593 |
+
{
|
3594 |
+
"epoch": 0.81,
|
3595 |
+
"learning_rate": 0.0003343481638367461,
|
3596 |
+
"loss": 0.1118,
|
3597 |
+
"step": 119600
|
3598 |
+
},
|
3599 |
+
{
|
3600 |
+
"epoch": 0.81,
|
3601 |
+
"learning_rate": 0.0003342322926095362,
|
3602 |
+
"loss": 0.1446,
|
3603 |
+
"step": 119800
|
3604 |
+
},
|
3605 |
+
{
|
3606 |
+
"epoch": 0.81,
|
3607 |
+
"learning_rate": 0.0003341162543608161,
|
3608 |
+
"loss": 0.1079,
|
3609 |
+
"step": 120000
|
3610 |
+
},
|
3611 |
+
{
|
3612 |
+
"epoch": 0.81,
|
3613 |
+
"learning_rate": 0.0003340000492205248,
|
3614 |
+
"loss": 0.1049,
|
3615 |
+
"step": 120200
|
3616 |
+
},
|
3617 |
+
{
|
3618 |
+
"epoch": 0.81,
|
3619 |
+
"learning_rate": 0.0003338836773187878,
|
3620 |
+
"loss": 0.1123,
|
3621 |
+
"step": 120400
|
3622 |
+
},
|
3623 |
+
{
|
3624 |
+
"epoch": 0.81,
|
3625 |
+
"learning_rate": 0.00033376713878591777,
|
3626 |
+
"loss": 0.1151,
|
3627 |
+
"step": 120600
|
3628 |
+
},
|
3629 |
+
{
|
3630 |
+
"epoch": 0.81,
|
3631 |
+
"learning_rate": 0.00033365043375241366,
|
3632 |
+
"loss": 0.0998,
|
3633 |
+
"step": 120800
|
3634 |
+
},
|
3635 |
+
{
|
3636 |
+
"epoch": 0.82,
|
3637 |
+
"learning_rate": 0.000333533562348961,
|
3638 |
+
"loss": 0.1189,
|
3639 |
+
"step": 121000
|
3640 |
+
},
|
3641 |
+
{
|
3642 |
+
"epoch": 0.82,
|
3643 |
+
"learning_rate": 0.0003334165247064315,
|
3644 |
+
"loss": 0.118,
|
3645 |
+
"step": 121200
|
3646 |
+
},
|
3647 |
+
{
|
3648 |
+
"epoch": 0.82,
|
3649 |
+
"learning_rate": 0.0003332993209558833,
|
3650 |
+
"loss": 0.1189,
|
3651 |
+
"step": 121400
|
3652 |
+
},
|
3653 |
+
{
|
3654 |
+
"epoch": 0.82,
|
3655 |
+
"learning_rate": 0.0003331819512285602,
|
3656 |
+
"loss": 0.113,
|
3657 |
+
"step": 121600
|
3658 |
+
},
|
3659 |
+
{
|
3660 |
+
"epoch": 0.82,
|
3661 |
+
"learning_rate": 0.0003330650037460781,
|
3662 |
+
"loss": 0.1162,
|
3663 |
+
"step": 121800
|
3664 |
+
},
|
3665 |
+
{
|
3666 |
+
"epoch": 0.82,
|
3667 |
+
"learning_rate": 0.00033294730328792147,
|
3668 |
+
"loss": 0.1105,
|
3669 |
+
"step": 122000
|
3670 |
+
},
|
3671 |
+
{
|
3672 |
+
"epoch": 0.82,
|
3673 |
+
"learning_rate": 0.0003328294372471769,
|
3674 |
+
"loss": 0.1041,
|
3675 |
+
"step": 122200
|
3676 |
+
},
|
3677 |
+
{
|
3678 |
+
"epoch": 0.82,
|
3679 |
+
"learning_rate": 0.00033271140575583016,
|
3680 |
+
"loss": 0.1064,
|
3681 |
+
"step": 122400
|
3682 |
+
},
|
3683 |
+
{
|
3684 |
+
"epoch": 0.83,
|
3685 |
+
"learning_rate": 0.000332593208946052,
|
3686 |
+
"loss": 0.1035,
|
3687 |
+
"step": 122600
|
3688 |
+
},
|
3689 |
+
{
|
3690 |
+
"epoch": 0.83,
|
3691 |
+
"learning_rate": 0.0003324748469501984,
|
3692 |
+
"loss": 0.1061,
|
3693 |
+
"step": 122800
|
3694 |
+
},
|
3695 |
+
{
|
3696 |
+
"epoch": 0.83,
|
3697 |
+
"learning_rate": 0.00033235631990081034,
|
3698 |
+
"loss": 0.1199,
|
3699 |
+
"step": 123000
|
3700 |
+
},
|
3701 |
+
{
|
3702 |
+
"epoch": 0.83,
|
3703 |
+
"learning_rate": 0.00033223762793061363,
|
3704 |
+
"loss": 0.1074,
|
3705 |
+
"step": 123200
|
3706 |
+
},
|
3707 |
+
{
|
3708 |
+
"epoch": 0.83,
|
3709 |
+
"learning_rate": 0.00033211877117251865,
|
3710 |
+
"loss": 0.1195,
|
3711 |
+
"step": 123400
|
3712 |
+
},
|
3713 |
+
{
|
3714 |
+
"epoch": 0.83,
|
3715 |
+
"learning_rate": 0.00033199974975962047,
|
3716 |
+
"loss": 0.1132,
|
3717 |
+
"step": 123600
|
3718 |
+
},
|
3719 |
+
{
|
3720 |
+
"epoch": 0.83,
|
3721 |
+
"learning_rate": 0.0003318811601638971,
|
3722 |
+
"loss": 0.1206,
|
3723 |
+
"step": 123800
|
3724 |
+
},
|
3725 |
+
{
|
3726 |
+
"epoch": 0.84,
|
3727 |
+
"learning_rate": 0.0003317618106630228,
|
3728 |
+
"loss": 0.1195,
|
3729 |
+
"step": 124000
|
3730 |
+
},
|
3731 |
+
{
|
3732 |
+
"epoch": 0.84,
|
3733 |
+
"learning_rate": 0.0003316428948842102,
|
3734 |
+
"loss": 0.1057,
|
3735 |
+
"step": 124200
|
3736 |
+
},
|
3737 |
+
{
|
3738 |
+
"epoch": 0.84,
|
3739 |
+
"learning_rate": 0.000331523217827277,
|
3740 |
+
"loss": 0.1067,
|
3741 |
+
"step": 124400
|
3742 |
+
},
|
3743 |
+
{
|
3744 |
+
"epoch": 0.84,
|
3745 |
+
"learning_rate": 0.00033140337678243725,
|
3746 |
+
"loss": 0.1185,
|
3747 |
+
"step": 124600
|
3748 |
+
},
|
3749 |
+
{
|
3750 |
+
"epoch": 0.84,
|
3751 |
+
"learning_rate": 0.0003312833718838879,
|
3752 |
+
"loss": 0.1089,
|
3753 |
+
"step": 124800
|
3754 |
+
},
|
3755 |
+
{
|
3756 |
+
"epoch": 0.84,
|
3757 |
+
"learning_rate": 0.00033116320326600986,
|
3758 |
+
"loss": 0.104,
|
3759 |
+
"step": 125000
|
3760 |
+
},
|
3761 |
+
{
|
3762 |
+
"epoch": 0.84,
|
3763 |
+
"learning_rate": 0.000331042871063367,
|
3764 |
+
"loss": 0.1254,
|
3765 |
+
"step": 125200
|
3766 |
+
},
|
3767 |
+
{
|
3768 |
+
"epoch": 0.84,
|
3769 |
+
"learning_rate": 0.00033092237541070647,
|
3770 |
+
"loss": 0.1085,
|
3771 |
+
"step": 125400
|
3772 |
+
},
|
3773 |
+
{
|
3774 |
+
"epoch": 0.85,
|
3775 |
+
"learning_rate": 0.0003308017164429586,
|
3776 |
+
"loss": 0.1132,
|
3777 |
+
"step": 125600
|
3778 |
+
},
|
3779 |
+
{
|
3780 |
+
"epoch": 0.85,
|
3781 |
+
"learning_rate": 0.00033068089429523636,
|
3782 |
+
"loss": 0.0999,
|
3783 |
+
"step": 125800
|
3784 |
+
},
|
3785 |
+
{
|
3786 |
+
"epoch": 0.85,
|
3787 |
+
"learning_rate": 0.00033055990910283567,
|
3788 |
+
"loss": 0.106,
|
3789 |
+
"step": 126000
|
3790 |
+
},
|
3791 |
+
{
|
3792 |
+
"epoch": 0.85,
|
3793 |
+
"learning_rate": 0.00033043876100123486,
|
3794 |
+
"loss": 0.1066,
|
3795 |
+
"step": 126200
|
3796 |
+
},
|
3797 |
+
{
|
3798 |
+
"epoch": 0.85,
|
3799 |
+
"learning_rate": 0.0003303174501260947,
|
3800 |
+
"loss": 0.1109,
|
3801 |
+
"step": 126400
|
3802 |
+
},
|
3803 |
+
{
|
3804 |
+
"epoch": 0.85,
|
3805 |
+
"learning_rate": 0.00033019597661325845,
|
3806 |
+
"loss": 0.1097,
|
3807 |
+
"step": 126600
|
3808 |
+
},
|
3809 |
+
{
|
3810 |
+
"epoch": 0.85,
|
3811 |
+
"learning_rate": 0.00033007434059875116,
|
3812 |
+
"loss": 0.1142,
|
3813 |
+
"step": 126800
|
3814 |
+
},
|
3815 |
+
{
|
3816 |
+
"epoch": 0.86,
|
3817 |
+
"learning_rate": 0.00032995254221878013,
|
3818 |
+
"loss": 0.1156,
|
3819 |
+
"step": 127000
|
3820 |
+
},
|
3821 |
+
{
|
3822 |
+
"epoch": 0.86,
|
3823 |
+
"learning_rate": 0.00032983058160973426,
|
3824 |
+
"loss": 0.1143,
|
3825 |
+
"step": 127200
|
3826 |
+
},
|
3827 |
+
{
|
3828 |
+
"epoch": 0.86,
|
3829 |
+
"learning_rate": 0.00032970845890818425,
|
3830 |
+
"loss": 0.1193,
|
3831 |
+
"step": 127400
|
3832 |
+
},
|
3833 |
+
{
|
3834 |
+
"epoch": 0.86,
|
3835 |
+
"learning_rate": 0.00032958617425088226,
|
3836 |
+
"loss": 0.1052,
|
3837 |
+
"step": 127600
|
3838 |
+
},
|
3839 |
+
{
|
3840 |
+
"epoch": 0.86,
|
3841 |
+
"learning_rate": 0.00032946372777476184,
|
3842 |
+
"loss": 0.1078,
|
3843 |
+
"step": 127800
|
3844 |
+
},
|
3845 |
+
{
|
3846 |
+
"epoch": 0.86,
|
3847 |
+
"learning_rate": 0.0003293411196169377,
|
3848 |
+
"loss": 0.1017,
|
3849 |
+
"step": 128000
|
3850 |
+
},
|
3851 |
+
{
|
3852 |
+
"epoch": 0.86,
|
3853 |
+
"learning_rate": 0.00032921834991470565,
|
3854 |
+
"loss": 0.1029,
|
3855 |
+
"step": 128200
|
3856 |
+
},
|
3857 |
+
{
|
3858 |
+
"epoch": 0.86,
|
3859 |
+
"learning_rate": 0.0003290960338623602,
|
3860 |
+
"loss": 0.0962,
|
3861 |
+
"step": 128400
|
3862 |
+
},
|
3863 |
+
{
|
3864 |
+
"epoch": 0.87,
|
3865 |
+
"learning_rate": 0.00032897294228992666,
|
3866 |
+
"loss": 0.1078,
|
3867 |
+
"step": 128600
|
3868 |
+
},
|
3869 |
+
{
|
3870 |
+
"epoch": 0.87,
|
3871 |
+
"learning_rate": 0.00032884968958536763,
|
3872 |
+
"loss": 0.1129,
|
3873 |
+
"step": 128800
|
3874 |
+
},
|
3875 |
+
{
|
3876 |
+
"epoch": 0.87,
|
3877 |
+
"learning_rate": 0.00032872627588670074,
|
3878 |
+
"loss": 0.1107,
|
3879 |
+
"step": 129000
|
3880 |
+
},
|
3881 |
+
{
|
3882 |
+
"epoch": 0.87,
|
3883 |
+
"learning_rate": 0.00032860270133212367,
|
3884 |
+
"loss": 0.1232,
|
3885 |
+
"step": 129200
|
3886 |
+
},
|
3887 |
+
{
|
3888 |
+
"epoch": 0.87,
|
3889 |
+
"learning_rate": 0.0003284789660600145,
|
3890 |
+
"loss": 0.1185,
|
3891 |
+
"step": 129400
|
3892 |
+
},
|
3893 |
+
{
|
3894 |
+
"epoch": 0.87,
|
3895 |
+
"learning_rate": 0.00032835507020893096,
|
3896 |
+
"loss": 0.111,
|
3897 |
+
"step": 129600
|
3898 |
+
},
|
3899 |
+
{
|
3900 |
+
"epoch": 0.87,
|
3901 |
+
"learning_rate": 0.000328231013917611,
|
3902 |
+
"loss": 0.1082,
|
3903 |
+
"step": 129800
|
3904 |
+
},
|
3905 |
+
{
|
3906 |
+
"epoch": 0.88,
|
3907 |
+
"learning_rate": 0.00032810741880645453,
|
3908 |
+
"loss": 0.1128,
|
3909 |
+
"step": 130000
|
3910 |
+
},
|
3911 |
+
{
|
3912 |
+
"epoch": 0.88,
|
3913 |
+
"learning_rate": 0.000327983042852058,
|
3914 |
+
"loss": 0.1224,
|
3915 |
+
"step": 130200
|
3916 |
+
},
|
3917 |
+
{
|
3918 |
+
"epoch": 0.88,
|
3919 |
+
"learning_rate": 0.0003278585068740186,
|
3920 |
+
"loss": 0.1127,
|
3921 |
+
"step": 130400
|
3922 |
+
},
|
3923 |
+
{
|
3924 |
+
"epoch": 0.88,
|
3925 |
+
"learning_rate": 0.00032773381101179096,
|
3926 |
+
"loss": 0.1122,
|
3927 |
+
"step": 130600
|
3928 |
+
},
|
3929 |
+
{
|
3930 |
+
"epoch": 0.88,
|
3931 |
+
"learning_rate": 0.00032760895540500854,
|
3932 |
+
"loss": 0.103,
|
3933 |
+
"step": 130800
|
3934 |
+
},
|
3935 |
+
{
|
3936 |
+
"epoch": 0.88,
|
3937 |
+
"learning_rate": 0.000327483940193484,
|
3938 |
+
"loss": 0.1129,
|
3939 |
+
"step": 131000
|
3940 |
+
},
|
3941 |
+
{
|
3942 |
+
"epoch": 0.88,
|
3943 |
+
"learning_rate": 0.00032735876551720825,
|
3944 |
+
"loss": 0.1118,
|
3945 |
+
"step": 131200
|
3946 |
+
},
|
3947 |
+
{
|
3948 |
+
"epoch": 0.89,
|
3949 |
+
"learning_rate": 0.00032723343151635126,
|
3950 |
+
"loss": 0.1141,
|
3951 |
+
"step": 131400
|
3952 |
+
},
|
3953 |
+
{
|
3954 |
+
"epoch": 0.89,
|
3955 |
+
"learning_rate": 0.00032710793833126114,
|
3956 |
+
"loss": 0.1043,
|
3957 |
+
"step": 131600
|
3958 |
+
},
|
3959 |
+
{
|
3960 |
+
"epoch": 0.89,
|
3961 |
+
"learning_rate": 0.0003269822861024642,
|
3962 |
+
"loss": 0.1223,
|
3963 |
+
"step": 131800
|
3964 |
+
},
|
3965 |
+
{
|
3966 |
+
"epoch": 0.89,
|
3967 |
+
"learning_rate": 0.0003268564749706651,
|
3968 |
+
"loss": 0.1129,
|
3969 |
+
"step": 132000
|
3970 |
+
},
|
3971 |
+
{
|
3972 |
+
"epoch": 0.89,
|
3973 |
+
"learning_rate": 0.0003267305050767462,
|
3974 |
+
"loss": 0.1159,
|
3975 |
+
"step": 132200
|
3976 |
+
},
|
3977 |
+
{
|
3978 |
+
"epoch": 0.89,
|
3979 |
+
"learning_rate": 0.00032660437656176775,
|
3980 |
+
"loss": 0.1026,
|
3981 |
+
"step": 132400
|
3982 |
+
},
|
3983 |
+
{
|
3984 |
+
"epoch": 0.89,
|
3985 |
+
"learning_rate": 0.00032647808956696767,
|
3986 |
+
"loss": 0.1107,
|
3987 |
+
"step": 132600
|
3988 |
+
},
|
3989 |
+
{
|
3990 |
+
"epoch": 0.89,
|
3991 |
+
"learning_rate": 0.00032635164423376107,
|
3992 |
+
"loss": 0.1001,
|
3993 |
+
"step": 132800
|
3994 |
+
},
|
3995 |
+
{
|
3996 |
+
"epoch": 0.9,
|
3997 |
+
"learning_rate": 0.0003262250407037408,
|
3998 |
+
"loss": 0.1046,
|
3999 |
+
"step": 133000
|
4000 |
+
},
|
4001 |
+
{
|
4002 |
+
"epoch": 0.9,
|
4003 |
+
"learning_rate": 0.0003260982791186765,
|
4004 |
+
"loss": 0.1198,
|
4005 |
+
"step": 133200
|
4006 |
+
},
|
4007 |
+
{
|
4008 |
+
"epoch": 0.9,
|
4009 |
+
"learning_rate": 0.000325971359620515,
|
4010 |
+
"loss": 0.1114,
|
4011 |
+
"step": 133400
|
4012 |
+
},
|
4013 |
+
{
|
4014 |
+
"epoch": 0.9,
|
4015 |
+
"learning_rate": 0.0003258442823513799,
|
4016 |
+
"loss": 0.1076,
|
4017 |
+
"step": 133600
|
4018 |
+
},
|
4019 |
+
{
|
4020 |
+
"epoch": 0.9,
|
4021 |
+
"learning_rate": 0.00032571704745357134,
|
4022 |
+
"loss": 0.1073,
|
4023 |
+
"step": 133800
|
4024 |
+
},
|
4025 |
+
{
|
4026 |
+
"epoch": 0.9,
|
4027 |
+
"learning_rate": 0.00032558965506956623,
|
4028 |
+
"loss": 0.1153,
|
4029 |
+
"step": 134000
|
4030 |
+
},
|
4031 |
+
{
|
4032 |
+
"epoch": 0.9,
|
4033 |
+
"learning_rate": 0.0003254621053420176,
|
4034 |
+
"loss": 0.111,
|
4035 |
+
"step": 134200
|
4036 |
+
},
|
4037 |
+
{
|
4038 |
+
"epoch": 0.91,
|
4039 |
+
"learning_rate": 0.00032533503733919646,
|
4040 |
+
"loss": 0.1122,
|
4041 |
+
"step": 134400
|
4042 |
+
},
|
4043 |
+
{
|
4044 |
+
"epoch": 0.91,
|
4045 |
+
"learning_rate": 0.0003252071741381574,
|
4046 |
+
"loss": 0.1042,
|
4047 |
+
"step": 134600
|
4048 |
+
},
|
4049 |
+
{
|
4050 |
+
"epoch": 0.91,
|
4051 |
+
"learning_rate": 0.0003250791540218744,
|
4052 |
+
"loss": 0.1103,
|
4053 |
+
"step": 134800
|
4054 |
+
},
|
4055 |
+
{
|
4056 |
+
"epoch": 0.91,
|
4057 |
+
"learning_rate": 0.0003249509771337035,
|
4058 |
+
"loss": 0.11,
|
4059 |
+
"step": 135000
|
4060 |
+
},
|
4061 |
+
{
|
4062 |
+
"epoch": 0.91,
|
4063 |
+
"learning_rate": 0.0003248226436171763,
|
4064 |
+
"loss": 0.1291,
|
4065 |
+
"step": 135200
|
4066 |
+
},
|
4067 |
+
{
|
4068 |
+
"epoch": 0.91,
|
4069 |
+
"learning_rate": 0.0003246941536159999,
|
4070 |
+
"loss": 0.1095,
|
4071 |
+
"step": 135400
|
4072 |
+
},
|
4073 |
+
{
|
4074 |
+
"epoch": 0.91,
|
4075 |
+
"learning_rate": 0.0003245655072740564,
|
4076 |
+
"loss": 0.1084,
|
4077 |
+
"step": 135600
|
4078 |
+
},
|
4079 |
+
{
|
4080 |
+
"epoch": 0.91,
|
4081 |
+
"learning_rate": 0.0003244367047354031,
|
4082 |
+
"loss": 0.0981,
|
4083 |
+
"step": 135800
|
4084 |
+
},
|
4085 |
+
{
|
4086 |
+
"epoch": 0.92,
|
4087 |
+
"learning_rate": 0.00032430839132516974,
|
4088 |
+
"loss": 0.1288,
|
4089 |
+
"step": 136000
|
4090 |
+
},
|
4091 |
+
{
|
4092 |
+
"epoch": 0.92,
|
4093 |
+
"learning_rate": 0.00032417927760514917,
|
4094 |
+
"loss": 0.1036,
|
4095 |
+
"step": 136200
|
4096 |
+
},
|
4097 |
+
{
|
4098 |
+
"epoch": 0.92,
|
4099 |
+
"learning_rate": 0.0003240500081209162,
|
4100 |
+
"loss": 0.1092,
|
4101 |
+
"step": 136400
|
4102 |
+
},
|
4103 |
+
{
|
4104 |
+
"epoch": 0.92,
|
4105 |
+
"learning_rate": 0.00032392058301722597,
|
4106 |
+
"loss": 0.1103,
|
4107 |
+
"step": 136600
|
4108 |
+
},
|
4109 |
+
{
|
4110 |
+
"epoch": 0.92,
|
4111 |
+
"learning_rate": 0.00032379100243900774,
|
4112 |
+
"loss": 0.1238,
|
4113 |
+
"step": 136800
|
4114 |
+
},
|
4115 |
+
{
|
4116 |
+
"epoch": 0.92,
|
4117 |
+
"learning_rate": 0.000323661266531365,
|
4118 |
+
"loss": 0.1125,
|
4119 |
+
"step": 137000
|
4120 |
+
},
|
4121 |
+
{
|
4122 |
+
"epoch": 0.92,
|
4123 |
+
"learning_rate": 0.0003235313754395751,
|
4124 |
+
"loss": 0.1181,
|
4125 |
+
"step": 137200
|
4126 |
+
},
|
4127 |
+
{
|
4128 |
+
"epoch": 0.93,
|
4129 |
+
"learning_rate": 0.0003234013293090893,
|
4130 |
+
"loss": 0.1107,
|
4131 |
+
"step": 137400
|
4132 |
+
},
|
4133 |
+
{
|
4134 |
+
"epoch": 0.93,
|
4135 |
+
"learning_rate": 0.00032327112828553224,
|
4136 |
+
"loss": 0.1189,
|
4137 |
+
"step": 137600
|
4138 |
+
},
|
4139 |
+
{
|
4140 |
+
"epoch": 0.93,
|
4141 |
+
"learning_rate": 0.00032314077251470224,
|
4142 |
+
"loss": 0.1177,
|
4143 |
+
"step": 137800
|
4144 |
+
},
|
4145 |
+
{
|
4146 |
+
"epoch": 0.93,
|
4147 |
+
"learning_rate": 0.00032301026214257077,
|
4148 |
+
"loss": 0.1147,
|
4149 |
+
"step": 138000
|
4150 |
+
},
|
4151 |
+
{
|
4152 |
+
"epoch": 0.93,
|
4153 |
+
"learning_rate": 0.00032287959731528245,
|
4154 |
+
"loss": 0.1098,
|
4155 |
+
"step": 138200
|
4156 |
+
},
|
4157 |
+
{
|
4158 |
+
"epoch": 0.93,
|
4159 |
+
"learning_rate": 0.00032274877817915483,
|
4160 |
+
"loss": 0.1114,
|
4161 |
+
"step": 138400
|
4162 |
+
},
|
4163 |
+
{
|
4164 |
+
"epoch": 0.93,
|
4165 |
+
"learning_rate": 0.0003226178048806784,
|
4166 |
+
"loss": 0.1136,
|
4167 |
+
"step": 138600
|
4168 |
+
},
|
4169 |
+
{
|
4170 |
+
"epoch": 0.94,
|
4171 |
+
"learning_rate": 0.000322486677566516,
|
4172 |
+
"loss": 0.1111,
|
4173 |
+
"step": 138800
|
4174 |
+
},
|
4175 |
+
{
|
4176 |
+
"epoch": 0.94,
|
4177 |
+
"learning_rate": 0.0003223553963835033,
|
4178 |
+
"loss": 0.1106,
|
4179 |
+
"step": 139000
|
4180 |
+
},
|
4181 |
+
{
|
4182 |
+
"epoch": 0.94,
|
4183 |
+
"learning_rate": 0.0003222239614786479,
|
4184 |
+
"loss": 0.1227,
|
4185 |
+
"step": 139200
|
4186 |
+
},
|
4187 |
+
{
|
4188 |
+
"epoch": 0.94,
|
4189 |
+
"learning_rate": 0.0003220923729991298,
|
4190 |
+
"loss": 0.1115,
|
4191 |
+
"step": 139400
|
4192 |
+
},
|
4193 |
+
{
|
4194 |
+
"epoch": 0.94,
|
4195 |
+
"learning_rate": 0.00032196063109230086,
|
4196 |
+
"loss": 0.119,
|
4197 |
+
"step": 139600
|
4198 |
+
},
|
4199 |
+
{
|
4200 |
+
"epoch": 0.94,
|
4201 |
+
"learning_rate": 0.0003218287359056848,
|
4202 |
+
"loss": 0.111,
|
4203 |
+
"step": 139800
|
4204 |
+
},
|
4205 |
+
{
|
4206 |
+
"epoch": 0.94,
|
4207 |
+
"learning_rate": 0.00032169668758697683,
|
4208 |
+
"loss": 0.1146,
|
4209 |
+
"step": 140000
|
4210 |
+
},
|
4211 |
+
{
|
4212 |
+
"epoch": 0.94,
|
4213 |
+
"learning_rate": 0.0003215651476708621,
|
4214 |
+
"loss": 0.0971,
|
4215 |
+
"step": 140200
|
4216 |
+
},
|
4217 |
+
{
|
4218 |
+
"epoch": 0.95,
|
4219 |
+
"learning_rate": 0.0003214327942955547,
|
4220 |
+
"loss": 0.1157,
|
4221 |
+
"step": 140400
|
4222 |
+
},
|
4223 |
+
{
|
4224 |
+
"epoch": 0.95,
|
4225 |
+
"learning_rate": 0.0003213002882315282,
|
4226 |
+
"loss": 0.1192,
|
4227 |
+
"step": 140600
|
4228 |
+
},
|
4229 |
+
{
|
4230 |
+
"epoch": 0.95,
|
4231 |
+
"learning_rate": 0.00032116762962716206,
|
4232 |
+
"loss": 0.1189,
|
4233 |
+
"step": 140800
|
4234 |
+
},
|
4235 |
+
{
|
4236 |
+
"epoch": 0.95,
|
4237 |
+
"learning_rate": 0.0003210348186310064,
|
4238 |
+
"loss": 0.113,
|
4239 |
+
"step": 141000
|
4240 |
+
},
|
4241 |
+
{
|
4242 |
+
"epoch": 0.95,
|
4243 |
+
"learning_rate": 0.0003209018553917821,
|
4244 |
+
"loss": 0.1198,
|
4245 |
+
"step": 141200
|
4246 |
+
},
|
4247 |
+
{
|
4248 |
+
"epoch": 0.95,
|
4249 |
+
"learning_rate": 0.00032076874005838044,
|
4250 |
+
"loss": 0.1082,
|
4251 |
+
"step": 141400
|
4252 |
+
},
|
4253 |
+
{
|
4254 |
+
"epoch": 0.95,
|
4255 |
+
"learning_rate": 0.0003206354727798631,
|
4256 |
+
"loss": 0.1188,
|
4257 |
+
"step": 141600
|
4258 |
+
},
|
4259 |
+
{
|
4260 |
+
"epoch": 0.96,
|
4261 |
+
"learning_rate": 0.00032050205370546186,
|
4262 |
+
"loss": 0.112,
|
4263 |
+
"step": 141800
|
4264 |
+
},
|
4265 |
+
{
|
4266 |
+
"epoch": 0.96,
|
4267 |
+
"learning_rate": 0.0003203684829845785,
|
4268 |
+
"loss": 0.1139,
|
4269 |
+
"step": 142000
|
4270 |
+
},
|
4271 |
+
{
|
4272 |
+
"epoch": 0.96,
|
4273 |
+
"learning_rate": 0.0003202347607667846,
|
4274 |
+
"loss": 0.1087,
|
4275 |
+
"step": 142200
|
4276 |
+
},
|
4277 |
+
{
|
4278 |
+
"epoch": 0.96,
|
4279 |
+
"learning_rate": 0.00032010222668614745,
|
4280 |
+
"loss": 0.1351,
|
4281 |
+
"step": 142400
|
4282 |
+
},
|
4283 |
+
{
|
4284 |
+
"epoch": 0.96,
|
4285 |
+
"learning_rate": 0.00031996820343515567,
|
4286 |
+
"loss": 0.1126,
|
4287 |
+
"step": 142600
|
4288 |
+
},
|
4289 |
+
{
|
4290 |
+
"epoch": 0.96,
|
4291 |
+
"learning_rate": 0.00031983402913548365,
|
4292 |
+
"loss": 0.1261,
|
4293 |
+
"step": 142800
|
4294 |
+
},
|
4295 |
+
{
|
4296 |
+
"epoch": 0.96,
|
4297 |
+
"learning_rate": 0.0003196997039373788,
|
4298 |
+
"loss": 0.1123,
|
4299 |
+
"step": 143000
|
4300 |
+
},
|
4301 |
+
{
|
4302 |
+
"epoch": 0.96,
|
4303 |
+
"learning_rate": 0.0003195652279912577,
|
4304 |
+
"loss": 0.1252,
|
4305 |
+
"step": 143200
|
4306 |
+
},
|
4307 |
+
{
|
4308 |
+
"epoch": 0.97,
|
4309 |
+
"learning_rate": 0.00031943060144770555,
|
4310 |
+
"loss": 0.1031,
|
4311 |
+
"step": 143400
|
4312 |
+
},
|
4313 |
+
{
|
4314 |
+
"epoch": 0.97,
|
4315 |
+
"learning_rate": 0.00031929582445747623,
|
4316 |
+
"loss": 0.1078,
|
4317 |
+
"step": 143600
|
4318 |
+
},
|
4319 |
+
{
|
4320 |
+
"epoch": 0.97,
|
4321 |
+
"learning_rate": 0.0003191608971714921,
|
4322 |
+
"loss": 0.1157,
|
4323 |
+
"step": 143800
|
4324 |
+
},
|
4325 |
+
{
|
4326 |
+
"epoch": 0.97,
|
4327 |
+
"learning_rate": 0.00031902581974084374,
|
4328 |
+
"loss": 0.107,
|
4329 |
+
"step": 144000
|
4330 |
+
},
|
4331 |
+
{
|
4332 |
+
"epoch": 0.97,
|
4333 |
+
"learning_rate": 0.00031889059231679,
|
4334 |
+
"loss": 0.1033,
|
4335 |
+
"step": 144200
|
4336 |
+
},
|
4337 |
+
{
|
4338 |
+
"epoch": 0.97,
|
4339 |
+
"learning_rate": 0.00031875521505075774,
|
4340 |
+
"loss": 0.1127,
|
4341 |
+
"step": 144400
|
4342 |
+
},
|
4343 |
+
{
|
4344 |
+
"epoch": 0.97,
|
4345 |
+
"learning_rate": 0.0003186196880943413,
|
4346 |
+
"loss": 0.119,
|
4347 |
+
"step": 144600
|
4348 |
+
},
|
4349 |
+
{
|
4350 |
+
"epoch": 0.98,
|
4351 |
+
"learning_rate": 0.00031848401159930296,
|
4352 |
+
"loss": 0.1187,
|
4353 |
+
"step": 144800
|
4354 |
+
},
|
4355 |
+
{
|
4356 |
+
"epoch": 0.98,
|
4357 |
+
"learning_rate": 0.00031834818571757227,
|
4358 |
+
"loss": 0.1131,
|
4359 |
+
"step": 145000
|
4360 |
+
},
|
4361 |
+
{
|
4362 |
+
"epoch": 0.98,
|
4363 |
+
"learning_rate": 0.0003182122106012462,
|
4364 |
+
"loss": 0.1141,
|
4365 |
+
"step": 145200
|
4366 |
+
},
|
4367 |
+
{
|
4368 |
+
"epoch": 0.98,
|
4369 |
+
"learning_rate": 0.00031807608640258873,
|
4370 |
+
"loss": 0.1158,
|
4371 |
+
"step": 145400
|
4372 |
+
},
|
4373 |
+
{
|
4374 |
+
"epoch": 0.98,
|
4375 |
+
"learning_rate": 0.00031793981327403085,
|
4376 |
+
"loss": 0.112,
|
4377 |
+
"step": 145600
|
4378 |
+
},
|
4379 |
+
{
|
4380 |
+
"epoch": 0.98,
|
4381 |
+
"learning_rate": 0.0003178033913681702,
|
4382 |
+
"loss": 0.1089,
|
4383 |
+
"step": 145800
|
4384 |
+
},
|
4385 |
+
{
|
4386 |
+
"epoch": 0.98,
|
4387 |
+
"learning_rate": 0.0003176668208377711,
|
4388 |
+
"loss": 0.1117,
|
4389 |
+
"step": 146000
|
4390 |
+
},
|
4391 |
+
{
|
4392 |
+
"epoch": 0.98,
|
4393 |
+
"learning_rate": 0.0003175301018357644,
|
4394 |
+
"loss": 0.1141,
|
4395 |
+
"step": 146200
|
4396 |
+
},
|
4397 |
+
{
|
4398 |
+
"epoch": 0.99,
|
4399 |
+
"learning_rate": 0.000317393234515247,
|
4400 |
+
"loss": 0.1145,
|
4401 |
+
"step": 146400
|
4402 |
+
},
|
4403 |
+
{
|
4404 |
+
"epoch": 0.99,
|
4405 |
+
"learning_rate": 0.0003172562190294821,
|
4406 |
+
"loss": 0.1055,
|
4407 |
+
"step": 146600
|
4408 |
+
},
|
4409 |
+
{
|
4410 |
+
"epoch": 0.99,
|
4411 |
+
"learning_rate": 0.0003171197417173119,
|
4412 |
+
"loss": 0.1074,
|
4413 |
+
"step": 146800
|
4414 |
+
},
|
4415 |
+
{
|
4416 |
+
"epoch": 0.99,
|
4417 |
+
"learning_rate": 0.0003169824311004134,
|
4418 |
+
"loss": 0.1166,
|
4419 |
+
"step": 147000
|
4420 |
+
},
|
4421 |
+
{
|
4422 |
+
"epoch": 0.99,
|
4423 |
+
"learning_rate": 0.0003168449727782822,
|
4424 |
+
"loss": 0.1075,
|
4425 |
+
"step": 147200
|
4426 |
+
},
|
4427 |
+
{
|
4428 |
+
"epoch": 0.99,
|
4429 |
+
"learning_rate": 0.00031670736690484315,
|
4430 |
+
"loss": 0.0993,
|
4431 |
+
"step": 147400
|
4432 |
+
},
|
4433 |
+
{
|
4434 |
+
"epoch": 0.99,
|
4435 |
+
"learning_rate": 0.0003165696136341865,
|
4436 |
+
"loss": 0.1093,
|
4437 |
+
"step": 147600
|
4438 |
+
},
|
4439 |
+
{
|
4440 |
+
"epoch": 1.0,
|
4441 |
+
"learning_rate": 0.0003164317131205673,
|
4442 |
+
"loss": 0.109,
|
4443 |
+
"step": 147800
|
4444 |
+
},
|
4445 |
+
{
|
4446 |
+
"epoch": 1.0,
|
4447 |
+
"learning_rate": 0.00031629366551840567,
|
4448 |
+
"loss": 0.1129,
|
4449 |
+
"step": 148000
|
4450 |
+
},
|
4451 |
+
{
|
4452 |
+
"epoch": 1.0,
|
4453 |
+
"learning_rate": 0.00031615547098228636,
|
4454 |
+
"loss": 0.1125,
|
4455 |
+
"step": 148200
|
4456 |
+
},
|
4457 |
+
{
|
4458 |
+
"epoch": 1.0,
|
4459 |
+
"learning_rate": 0.0003160171296669587,
|
4460 |
+
"loss": 0.1057,
|
4461 |
+
"step": 148400
|
4462 |
+
},
|
4463 |
+
{
|
4464 |
+
"epoch": 1.0,
|
4465 |
+
"eval_accuracy": 0.9420512498988455,
|
4466 |
+
"eval_auc": 0.8522129916967822,
|
4467 |
+
"eval_f1": 0.35228549223252686,
|
4468 |
+
"eval_loss": 0.4365999102592468,
|
4469 |
+
"eval_mcc": 0.39725936154292724,
|
4470 |
+
"eval_precision": 0.2294216317399737,
|
4471 |
+
"eval_recall": 0.7584797572261781,
|
4472 |
+
"eval_runtime": 9338.8173,
|
4473 |
+
"eval_samples_per_second": 23.624,
|
4474 |
+
"eval_steps_per_second": 3.937,
|
4475 |
+
"step": 148440
|
4476 |
+
}
|
4477 |
+
],
|
4478 |
+
"logging_steps": 200,
|
4479 |
+
"max_steps": 593760,
|
4480 |
+
"num_train_epochs": 4,
|
4481 |
+
"save_steps": 500,
|
4482 |
+
"total_flos": 1.784375613961479e+17,
|
4483 |
+
"trial_name": null,
|
4484 |
+
"trial_params": null
|
4485 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7ca181af64980131bfbd4dfa37d8632db34574aa82dbc3a73caa97c7b19fb2c6
|
3 |
+
size 4091
|
vocab.txt
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<cls>
|
2 |
+
<pad>
|
3 |
+
<eos>
|
4 |
+
<unk>
|
5 |
+
L
|
6 |
+
A
|
7 |
+
G
|
8 |
+
V
|
9 |
+
S
|
10 |
+
E
|
11 |
+
R
|
12 |
+
T
|
13 |
+
I
|
14 |
+
D
|
15 |
+
P
|
16 |
+
K
|
17 |
+
Q
|
18 |
+
N
|
19 |
+
F
|
20 |
+
Y
|
21 |
+
M
|
22 |
+
H
|
23 |
+
W
|
24 |
+
C
|
25 |
+
X
|
26 |
+
B
|
27 |
+
U
|
28 |
+
Z
|
29 |
+
O
|
30 |
+
.
|
31 |
+
-
|
32 |
+
<null_1>
|
33 |
+
<mask>
|