Skip to content

Commit

Permalink
update contents in parameter size figure
Browse files Browse the repository at this point in the history
  • Loading branch information
kaisugi committed Dec 10, 2023
1 parent 1796f53 commit 64e6546
Show file tree
Hide file tree
Showing 3 changed files with 21 additions and 19 deletions.
Binary file modified figures/parameter_size_overview.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
35 changes: 19 additions & 16 deletions figures/scripts/parameter_size_overview.csv
Original file line number Diff line number Diff line change
@@ -1,23 +1,26 @@
Model,Lab,Parameters(B),Announced,Type,Source(JP-unavailable)
LLM-jp-13B,LLM-jp,13,2023/10/20,JP-available,
PLaMo-13B,Preferred Networks,13,2023/09/28,JP-available,
Stockmark-13b,Stockmark,13,2023/10/27,JP-available,
Weblab-10B,Matsuo Lab,10,2023/08/22,JP-available,
JSLM Alpha,Stability AI,7,2023/08/10,JP-available,
CALM2,CyberAgent,7,2023/11/02,JP-available,
OpenCALM,CyberAgent,6.8,2023/05/17,JP-available,
,rinna,3.8,2023/07/31,JP-available,
,rinna,3.6,2023/05/17,JP-available,
,rinna,1.3,2022/01/26,JP-available,
,rinna,0.3,2021/08/25,JP-available,
,rinna,0.1,2021/08/25,JP-available,
japanese-large-lm,LINE,3.6,2023/08/14,JP-available,
,レトリバ,3,2023/05/12,JP-available,
,ABEJA,2.7,2022/07/27,JP-available,
Model,Lab,Parameters(B),Announced,Type,Source(JP)
LLM-jp-13B,LLM-jp,13,2023/10/20,JP-available,https://www.nii.ac.jp/news/release/2023/1020.html
PLaMo-13B,Preferred Networks,13,2023/09/28,JP-available,https://www.preferred.jp/ja/news/pr20230928/
Stockmark-13b,Stockmark,13,2023/10/27,JP-available,https://stockmark.co.jp/news/20231027
Weblab-10B,Matsuo Lab,10,2023/08/22,JP-available,https://www.t.u-tokyo.ac.jp/press/pr2023-08-18-001
JSLM Alpha,Stability AI,7,2023/08/10,JP-available,https://ja.stability.ai/blog/japanese-stablelm-alpha
CALM2,CyberAgent,7,2023/11/02,JP-available,https://www.cyberagent.co.jp/news/detail/id=29479
OpenCALM,CyberAgent,6.8,2023/05/17,JP-available,https://www.cyberagent.co.jp/news/detail/id=28817
,rinna,3.8,2023/07/31,JP-available,https://rinna.co.jp/news/2023/07/20230731.html
,rinna,3.6,2023/05/17,JP-available,https://rinna.co.jp/news/2023/05/20230507.html
,rinna,1.3,2022/01/26,JP-available,https://rinna.co.jp/news/2022/01/2022012601.html
,rinna,0.3,2021/04/07,JP-available,https://rinna.co.jp/news/2021/04/20210407.html
,rinna,0.1,2021/08/25,JP-available,https://rinna.co.jp/news/2021/08/20210825.html
japanese-large-lm,LINE,3.6,2023/08/14,JP-available,https://engineering.linecorp.com/ja/blog/3.6-billion-parameter-japanese-language-model
,レトリバ,3,2023/05/12,JP-available,https://note.com/retrieva/n/n7b4186dc5ada
,ABEJA,2.7,2022/07/27,JP-available,https://tech-blog.abeja.asia/entry/abeja-gpt-project-202207
,ABEJA,6.7,2022/07/27,JP-unavailable,https://tech-blog.abeja.asia/entry/abeja-gpt-project-202207
,ABEJA,13,2022/07/27,JP-unavailable,https://tech-blog.abeja.asia/entry/abeja-gpt-project-202207
tsuzumi,NTT,7,2023/11/01,JP-unavailable,https://group.ntt/jp/newsrelease/2023/11/01/231101a.html
,NEC,13,2023/07/06,JP-unavailable,https://jpn.nec.com/press/202307/20230706_02.html
,NICT,40,2023/07/04,JP-unavailable,https://www.nict.go.jp/press/2023/07/04-1.html
PolySphere-1,AI inside,14,2023/06/08,JP-unavailable,https://inside.ai/news/2023/06/08/aiinside-xresearch/
,RICOH,6,2023/03/15,JP-unavailable,https://www.anlp.jp/proceedings/annual_meeting/2023/pdf_dir/H9-4.pdf
LHTM-2,オルツ,160,2023/02/14,JP-unavailable,"https://alt.ai/news/news-1892/, https://xtech.nikkei.com/atcl/nxt/column/18/02423/053100030/"
HyperCLOVA,NAVER & LIINE,82,2022/11/30,JP-unavailable,https://www.youtube.com/watch?v=I4o7X3-aqJk
HyperCLOVA,NAVER & LIINE,39,2021/11/10,JP-unavailable,https://www.youtube.com/watch?v=V4pZulIWHpY
Expand Down
5 changes: 2 additions & 3 deletions figures/scripts/parameter_size_overview_generate.py
Original file line number Diff line number Diff line change
Expand Up @@ -2,9 +2,8 @@
パラメータサイズの推移を描画するスクリプト
CSVデータ作成に関するメモ
1. 日本語公開モデルに関しては、この記事の情報から作成
2. 日本語非公開モデルに関しては、プレスリリースから作成
3. 英語モデルに関しては、LifeArchitect.ai からデータを抽出。具体的には、
1. 日本語公開モデルに関しては、プレスリリースから作成
2. 英語モデルに関しては、LifeArchitect.ai からデータを抽出。具体的には、
a. Public? が緑色のものと、黄色・赤色のものでそれぞれフィルターをかける
b. GPT-3 以前のモデルは落とす
c. まだ非公開のモデルは落とす
Expand Down

0 comments on commit 64e6546

Please sign in to comment.