Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit5029545

Browse files
committed
chore
1 parent1954d1e commit5029545

File tree

65 files changed

+1705
-189
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

65 files changed

+1705
-189
lines changed

‎README.md

Lines changed: 35 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,3 +3,38 @@
33
poetry install
44
<!-- poetry run pip install --upgrade --force-reinstall --no-deps "apache-beam[gcp]" "multiprocess==0.70.14"-->
55
poetry run pip install --upgrade --force-reinstall --no-deps "apache-beam[gcp]" "multiprocess==0.70.14" "dill==0.3.1.1"
6+
7+
8+
9+
| base models| batch size| lr| JSICK (val)| JSICK (test)| JSTS (train)| JSTS (val)|
10+
| -------------------------------------------------------------------------------------------------------------------------| :--------:| :---:| :---------:| :----------:| :----------:| :--------:|
11+
|[cl-tohoku/bert-base-japanese-v2](https://huggingface.co/cl-tohoku/bert-base-japanese-v2)|||||||
12+
|[cl-tohoku/bert-base-japanese-char-v2](https://huggingface.co/cl-tohoku/bert-base-japanese-char-v2)|||||||
13+
|[cl-tohoku/bert-base-japanese](https://huggingface.co/cl-tohoku/bert-base-japanese)|||||||
14+
|[cl-tohoku/bert-base-japanese-whole-word-masking](https://huggingface.co/cl-tohoku/bert-base-japanese-whole-word-masking)|||||||
15+
|[cl-tohoku/bert-base-japanese-char](https://huggingface.co/cl-tohoku/bert-base-japanese-char)|||||||
16+
|[ku-nlp/roberta-base-japanese-char-wwm](https://huggingface.co/ku-nlp/roberta-base-japanese-char-wwm)|||||||
17+
|[studio-ousia/luke-japanese-base-lite](https://huggingface.co/studio-ousia/luke-japanese-base-lite)|||||||
18+
||||||||
19+
|[ku-nlp/deberta-v2-base-japanese](https://huggingface.co/ku-nlp/deberta-v2-base-japanese)|||||||
20+
|[nlp-waseda/roberta-base-japanese](https://huggingface.co/nlp-waseda/roberta-base-japanese)|||||||
21+
|[megagonlabs/roberta-long-japanese](https://huggingface.co/megagonlabs/roberta-long-japanese)|||||||
22+
||||||||
23+
|[bert-base-multilingual-cased](https://huggingface.co/bert-base-multilingual-cased)|||||||
24+
|[xlm-roberta-base](https://huggingface.co/xlm-roberta-base)|||||||
25+
|[microsoft/mdeberta-v3-base](https://huggingface.co/microsoft/mdeberta-v3-base)|||||||
26+
|[studio-ousia/mluke-base-lite](https://huggingface.co/studio-ousia/mluke-base-lite)|||||||
27+
28+
29+
30+
| large models| batch size| lr| JSICK (val)| JSICK (test)| JSTS (train)| JSTS (val)|
31+
| -------------------------------------------------------------------------------------------------------| :--------:| :---:| :---------:| :----------:| :----------:| :--------:|
32+
|[cl-tohoku/bert-large-japanese](https://huggingface.co/cl-tohoku/bert-large-japanese)|||||||
33+
|[ku-nlp/roberta-large-japanese-char-wwm](https://huggingface.co/ku-nlp/roberta-large-japanese-char-wwm)|||||||
34+
|[studio-ousia/luke-japanese-large-lite](https://huggingface.co/studio-ousia/luke-japanese-large-lite)|||||||
35+
||||||||
36+
|[nlp-waseda/roberta-large-japanese](https://huggingface.co/nlp-waseda/roberta-large-japanese)|||||||
37+
|[ku-nlp/deberta-v2-large-japanese](https://huggingface.co/ku-nlp/deberta-v2-large-japanese)|||||||
38+
||||||||
39+
|[xlm-roberta-large](https://huggingface.co/xlm-roberta-large)|||||||
40+
|[studio-ousia/mluke-large-lite](https://huggingface.co/studio-ousia/mluke-large-lite)|||||||

‎scripts/10/0.sh

Lines changed: 18 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,25 @@
11
device="cuda:0"
2-
model_name="cl-tohoku/bert-large-japanese"
32

43
foriin 0 1 2;do
5-
forbatch_sizein32 64 128 256 512;do
4+
formodel_nameinstudio-ousia/luke-japanese-large-lite studio-ousia/luke-japanese-base-lite;do
65
forlrin 1e-5 3e-5 5e-5;do
7-
poetry run python src/train_unsup.py \
8-
--dataset_name wiki40b \
9-
--model_name$model_name \
10-
--batch_size$batch_size \
11-
--lr$lr \
12-
--device$device
6+
forbatch_sizein 512;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name jsnli+nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--gradient_checkpointing \
13+
--device$device
14+
done
15+
forbatch_sizein 256 128 64;do
16+
poetry run python src/train_sup.py \
17+
--dataset_name jsnli+nu-snli \
18+
--model_name$model_name \
19+
--batch_size$batch_size \
20+
--lr$lr \
21+
--device$device
22+
done
1323
done
1424
done
1525
done

‎scripts/10/1.sh

Lines changed: 18 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,25 @@
11
device="cuda:1"
2-
model_name="studio-ousia/luke-japanese-large-lite"
32

43
foriin 0 1 2;do
5-
forbatch_sizein32 64 128 256 512;do
4+
formodel_nameincl-tohoku/bert-large-japanese ku-nlp/roberta-large-japanese-char-wwm;do
65
forlrin 1e-5 3e-5 5e-5;do
7-
poetry run python src/train_unsup.py \
8-
--dataset_name wiki40b \
9-
--model_name$model_name \
10-
--batch_size$batch_size \
11-
--lr$lr \
12-
--device$device
6+
forbatch_sizein 512;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name jsnli+nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--gradient_checkpointing \
13+
--device$device
14+
done
15+
forbatch_sizein 256 128 64;do
16+
poetry run python src/train_sup.py \
17+
--dataset_name jsnli+nu-snli \
18+
--model_name$model_name \
19+
--batch_size$batch_size \
20+
--lr$lr \
21+
--device$device
22+
done
1323
done
1424
done
1525
done

‎scripts/10/2.sh

Lines changed: 20 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,27 @@
11
device="cuda:2"
2-
model_name="cl-tohoku/bert-base-japanese-v2"
32

43
foriin 0 1 2;do
5-
forbatch_sizein32 64 128 256 512;do
4+
formodel_nameinnlp-waseda/roberta-large-japanese ku-nlp/deberta-v2-large-japanese;do
65
forlrin 1e-5 3e-5 5e-5;do
7-
poetry run python src/train_unsup.py \
8-
--dataset_name wiki40b \
9-
--model_name$model_name \
10-
--batch_size$batch_size \
11-
--lr$lr \
12-
--device$device
6+
forbatch_sizein 512;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name jsnli+nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--use_jumanpp \
13+
--gradient_checkpointing \
14+
--device$device
15+
done
16+
forbatch_sizein 256 128 64;do
17+
poetry run python src/train_sup.py \
18+
--dataset_name jsnli+nu-snli \
19+
--model_name$model_name \
20+
--batch_size$batch_size \
21+
--lr$lr \
22+
--use_jumanpp \
23+
--device$device
24+
done
1325
done
1426
done
1527
done

‎scripts/10/3.sh

Lines changed: 18 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,25 @@
11
device="cuda:3"
2-
model_name="ku-nlp/deberta-v2-large-japanese"
32

43
foriin 0 1 2;do
5-
forbatch_sizein32 64 128 256 512;do
4+
formodel_nameinxlm-roberta-large studio-ousia/mluke-large-lite;do
65
forlrin 1e-5 3e-5 5e-5;do
7-
poetry run python src/train_unsup.py \
8-
--dataset_name wiki40b \
9-
--model_name$model_name \
10-
--batch_size$batch_size \
11-
--lr$lr \
12-
--device$device
6+
forbatch_sizein 512;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name jsnli+nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--gradient_checkpointing \
13+
--device$device
14+
done
15+
forbatch_sizein 256 128 64;do
16+
poetry run python src/train_sup.py \
17+
--dataset_name jsnli+nu-snli \
18+
--model_name$model_name \
19+
--batch_size$batch_size \
20+
--lr$lr \
21+
--device$device
22+
done
1323
done
1424
done
1525
done

‎scripts/10/prev/2023-03-10/0.sh

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
device="cuda:0"
2+
3+
foriin 0 1 2;do
4+
formodel_namein cl-tohoku/bert-base-japanese-v2 cl-tohoku/bert-base-japanese-char-v2 cl-tohoku/bert-base-japanese cl-tohoku/bert-base-japanese-whole-word-masking;do
5+
forbatch_sizein 64 128 256 512;do
6+
forlrin 1e-5 3e-5 5e-5;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--device$device
13+
14+
poetry run python src/train_sup.py \
15+
--dataset_name nu-mnli \
16+
--model_name$model_name \
17+
--batch_size$batch_size \
18+
--lr$lr \
19+
--device$device
20+
21+
poetry run python src/train_sup.py \
22+
--dataset_name nu-snli+mnli \
23+
--model_name$model_name \
24+
--batch_size$batch_size \
25+
--lr$lr \
26+
--device$device
27+
done
28+
done
29+
done
30+
done

‎scripts/10/prev/2023-03-10/1.sh

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
device="cuda:1"
2+
3+
foriin 0 1 2;do
4+
formodel_namein cl-tohoku/bert-large-japanese ku-nlp/roberta-large-japanese-char-wwm;do
5+
forbatch_sizein 64 128 256 512;do
6+
forlrin 1e-5 3e-5 5e-5;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--device$device
13+
14+
poetry run python src/train_sup.py \
15+
--dataset_name nu-mnli \
16+
--model_name$model_name \
17+
--batch_size$batch_size \
18+
--lr$lr \
19+
--device$device
20+
21+
poetry run python src/train_sup.py \
22+
--dataset_name nu-snli+mnli \
23+
--model_name$model_name \
24+
--batch_size$batch_size \
25+
--lr$lr \
26+
--device$device
27+
done
28+
done
29+
done
30+
done

‎scripts/10/prev/2023-03-10/2.sh

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,33 @@
1+
device="cuda:2"
2+
3+
foriin 0 1 2;do
4+
formodel_namein nlp-waseda/roberta-large-japanese ku-nlp/deberta-v2-large-japanese;do
5+
forbatch_sizein 64 128 256 512;do
6+
forlrin 1e-5 3e-5 5e-5;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--use_jumanpp \
13+
--device$device
14+
15+
poetry run python src/train_sup.py \
16+
--dataset_name nu-mnli \
17+
--model_name$model_name \
18+
--batch_size$batch_size \
19+
--lr$lr \
20+
--use_jumanpp \
21+
--device$device
22+
23+
poetry run python src/train_sup.py \
24+
--dataset_name nu-snli+mnli \
25+
--model_name$model_name \
26+
--batch_size$batch_size \
27+
--lr$lr \
28+
--use_jumanpp \
29+
--device$device
30+
done
31+
done
32+
done
33+
done

‎scripts/10/prev/2023-03-10/3.sh

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
device="cuda:3"
2+
3+
foriin 0 1 2;do
4+
formodel_namein xlm-roberta-large studio-ousia/mluke-large-lite;do
5+
forbatch_sizein 64 128 256 512;do
6+
forlrin 1e-5 3e-5 5e-5;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--device$device
13+
14+
poetry run python src/train_sup.py \
15+
--dataset_name nu-mnli \
16+
--model_name$model_name \
17+
--batch_size$batch_size \
18+
--lr$lr \
19+
--device$device
20+
21+
poetry run python src/train_sup.py \
22+
--dataset_name nu-snli+mnli \
23+
--model_name$model_name \
24+
--batch_size$batch_size \
25+
--lr$lr \
26+
--device$device
27+
done
28+
done
29+
done
30+
done

‎scripts/10/prev/2023-03-11/0.sh

Lines changed: 25 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,25 @@
1+
device="cuda:0"
2+
3+
forbatch_sizein 1024;do
4+
foriin 0 1;do
5+
formodel_namein cl-tohoku/bert-large-japanese ku-nlp/roberta-large-japanese-char-wwm;do
6+
forlrin 1e-5 3e-5 5e-5;do
7+
poetry run python src/train_sup.py \
8+
--dataset_name nu-snli \
9+
--model_name$model_name \
10+
--batch_size$batch_size \
11+
--lr$lr \
12+
--gradient_checkpointing \
13+
--device$device
14+
15+
poetry run python src/train_sup.py \
16+
--dataset_name nu-snli+mnli \
17+
--model_name$model_name \
18+
--batch_size$batch_size \
19+
--lr$lr \
20+
--gradient_checkpointing \
21+
--device$device
22+
done
23+
done
24+
done
25+
done

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp