Hiroyuki V Yamazaki

Engineer at Preferred Networks. Computer Science at KTH, Royal Institute of Technology in Stockholm and Keio University in Tokyo.

[OptunaAutoML @OptunaAutoML]

Optuna v4.9.0 has been released! 🚀

This release enhances GPSampler with the Kriging Believer strategy, significantly improving parallel efficiency for multi-objective and constrained optimization.

👇 Check out the full release notes:
https://t.co/wBoa2Bw7hB

[CuPy @CuPy_Team]

https://t.co/OLqT5QTZye

[Preferred Networks @PreferredNetJP]

【中途採用】現在PFNで開発中の生成AI推論向け半導体 MN-Core™ LシリーズのLLMサービングエンジン開発チームでエンジニアの募集を開始しました！
MN-Coreシリーズのコンパイラエンジニア、ランタイムソフトウェアエンジニアも引き続き募集中です（リプライ参照）👇

Lシリーズは生成AIの推論に必要な高帯域幅を実現するため、3D積層DRAM技術を採用したアクセラレータです。本LLMサービングエンジンは、コンパイラが生成したプログラム、ランタイム環境、DRAM上でのデータ処理を緊密に協調させることで、超低遅延なLLM推論を実現することを目的としています。

ハードウェア開発者と密接に連携しながら最先端のLLM推論アクセラレータ向けソフトウェアの開発に携わってみたいという方は、募集の詳細をご覧ください！
https://t.co/5sdyCQkCAX

[Preferred Networks @PreferredNet]

Hiring now: We started looking for engineers who can develop the LLM serving engine with us for our MN-Core™ L series of proprietary inference AI chip.
See reply for other MN-Core-related positions👇 https://t.co/heSbL4cm7n

[OptunaAutoML @OptunaAutoML]

Optuna v4.7.0 has been released! This is a maintenance release with various minor bug fixes and improvements to the documentation and more.

👇See what’s new
https://t.co/5syvxZEPfv

[OptunaAutoML @OptunaAutoML]

Optuna v4.5 has been released!

⛏️GPSampler for constrained multi-objective optimization
🚀Significant speedup of TPESampler and plot_hypervolume_history
🦾CmaEsSampler now supports 1D search space
🐍The optunahub library is published on conda-forge

https://t.co/QS314M0DaI

[NumFOCUS @NumFOCUS]

Optuna v4.4 will be released this month, and the roadmap for their next exciting major release- Optuna v5- has just been published! Read more on their blog here: https://t.co/7RYfb1OZPT

[OptunaAutoML @OptunaAutoML]

Optuna v4.4 has been released with various new features, bug fixes, and enhancements.
🚀Optuna MCP server, which is our first LLM intensive toolchain
✅Gaussian process-based algorithm now supports multi-objective optimization
🌀A lot of new features in OptunaHub

[OptunaAutoML @OptunaAutoML]

We've published the development roadmap for Optuna v5, the next major release! Scheduled for release next summer, v5 will focus on making Optuna even more powerful and user-friendly — especially at the intersection of generative AI and black-box optimization.

[CNBC Int'l PR @cnbcipr]

. @ChristineCNBC takes us into the world of innovative Japanese companies that are truly making an impact. In the first episode of this two-part series, she speaks to the co-founders of @PreferredNetJP, one of Japan’s largest unicorns, using deep learning and AI to solve real-world problems across industries.
Premiers on @CNBCi from tonight, 2300 CET.

[CuPy @CuPy_Team]

📣 After 6 months, CuPy v13.4 is finally here! 🎉

Packed with exciting updates, including:
✅ NVIDIA CUDA 12.8 & Blackwell architecture support
🦾 AMD ROCm 6 compatibility
🐍 Python 3.13 binary packages
🔄 DLPack v1

Check out the full release notes ⬇️

[OptunaAutoML @OptunaAutoML]

🚀 In Optuna v4.1, major optimizations were made to speed up RDBStorage up to 63%. Check out our latest blog post for details!

🔗 https://t.co/bSWS7wqRH8

[OptunaAutoML @OptunaAutoML]

Optuna 4.2.0 and OptunaHub 0.2.0 are out! Try the newest Optuna and OptunaHub!

👇See what’s new
https://t.co/i8gcC3ZHWJ

[Shuhei Watanabe @nabenabe_0928]

Optuna now supports a more beginner-friendly sampler:)

[OptunaAutoML @OptunaAutoML]

AutoSampler is out! It automatically chooses an appropriate algorithm based on the evaluation budget and search space. Users can get better average performance without knowing details of each algorithm. Check out the article by @y0zaki!

https://t.co/K4YTXq31e6

[Toshihiko YANASE @toshihikoyanase]

Just change three lines in your Optuna code:

from optunahub import load_module
s = load_module("samplers/auto_sampler").AutoSampler()
study = optuna.create_study(sampler=s)

[OptunaAutoML @OptunaAutoML]

Optuna 4.0, the new major release, is here!

Huge thanks to all contributors and users! We are glad to see Optuna has become a leading hyperparameter optimization OSS!

See what's new: https://t.co/OEVCaXXgv9

[CuPy @CuPy_Team]

🎉 CuPy v13.3 is just out! This release includes performance improvements (“one-time only warm-up” eliminated), support for CUDA 12.5 & 12.6, and better interoperability with NumPy 2. Check out the release notes for details!
https://t.co/I6lnt2wOqL

[OptunaAutoML @OptunaAutoML]

OptunaHub now supports the SOTA evolutionary method, CatCMA, by Masahiro Nomura @cyberagent_ai. It is powerful in hyperparameter tuning with a mixed search space of continuous and categorical variables. For more details, check out the blog by @mamurai1208. https://t.co/fPkGMHFdBw

[NumFOCUS @NumFOCUS]

.@OptunaAutoML has reached 10,000 stars on GitHub! ⭐ We’re thrilled to see so much love and support being shown, and we’re incredibly proud of all the hard working people working to make Optuna so special 💙