iiinova — where your digital self keeps evolving

vmoens @vmoens.bsky.social · 30/11/2025

At NeurIPS this week? DM if you want to meet!

vmoens @vmoens.bsky.social · 02/10/2025

Happy to announce that I've joined Periodic Labs as member of technical staff. We're a mission driven startup aimed at accelerating scientific discovery using AI, with a strong focus on material science (discovery of new materials such as superconductors and such). We're hiring: periodic.com

vmoens @vmoens.bsky.social · 06/03/2025

The fact that all LLM libraries don't have the same data format is as surprising as the fact that there is more than one sign language dialect

120

vmoens @vmoens.bsky.social · 06/03/2025

Ray is an excellent way of testing if all your `__repr__` are coded properly (but it shouldn't be)

020

vmoens @vmoens.bsky.social · 03/03/2025

Just stumbled upon RouteRL: a multiagent RL framework to facilitate the testing and development of efficient route choice strategies coexistence-project.github.io/RouteRL/ Looks pretty cool!

coexistence-project.github.io

RouteRL 1.0.0 documentationContentsMenuExpandLight modeDark modeAuto light/dark, in light modeAuto light/dark, in dark mode

040

Reposted by vmoens

Xuan Son Nguyen @ngxson.hf.co · 27/02/2025

What is GGUF, Safetensors, PyTorch, ONNX? In this blog post, let's discover common formats for storing an AI model. huggingface.co/blog/ngxson/...

huggingface.co

Common AI Model Formats

A Blog post by Xuan-Son Nguyen on Hugging Face

053

vmoens @vmoens.bsky.social · 21/02/2025

Today we're opensourcing MLGym, an API for AI research agents. MLGym relies on a gym environment that wraps a docker image. Each env has a task specified as a YAML file, telling in plain english what you want your LLM to achieve 👇

130

vmoens @vmoens.bsky.social · 19/02/2025

A few tips I share when I talk about perf with PyTorch in eager mode (with focus on small models): 🪢

1122

vmoens @vmoens.bsky.social · 11/02/2025

I stand by my point that the word "agent" should be avoided at all costs. At least in RL, anytime I see an "Agent" class it's meant to be a "whatever doesn't fit in any other bucket in my codebase".

250

vmoens @vmoens.bsky.social · 06/02/2025

Everyone's like "hey I just coded and trained a SOTA LLM in my garage last week, also wrote a blogpost about it and opensourced the repo" and the only thing I did in the meantime was fix a CI and configure a remote interpreter on a server... 😢

050

vmoens @vmoens.bsky.social · 28/01/2025

Clap hands if you were doing RL before it was cool

030

vmoens @vmoens.bsky.social · 26/01/2025

You’ve never really understood PyTorch until you’ve figured out what torch.scatter exactly does. I’ve never really understood PyTorch.

161

Reposted by vmoens

MuJoCo.org @mujoco.bsky.social · 16/01/2025

Introducing playground.mujoco.org Combining MuJoCo’s rich and thriving ecosystem, massively parallel GPU-accelerated simulation, and real-world results across a diverse range of robot platforms: quadrupeds, humanoids, dexterous hands, and arms. Get started today: pip install playground

playground.mujoco.org

MuJoCo Playground

An open-source framework for GPU-accelerated robot learning and sim-to-real transfer

17520

vmoens @vmoens.bsky.social · 09/01/2025

One of the bests Mindscape episodes I've heard! I've always been interested in the epistemology of controversial scientific concepts such as emergence or consciousness. Vaguely defined concepts (or diverging definitions) are often at the root of many pointless debates. Kudos on clarifying things!

000

vmoens @vmoens.bsky.social · 21/12/2024

New Year’s resolutions: - eat healthier - exercise more - no more “is all you need” papers

2101

Reposted by vmoens

Stone Tao @stonet2000.bsky.social · 20/12/2024

Yesterday the hyped Genesis simulator released. But it's up to 10x slower than existing GPU sims, not 10-80x faster or 430,000x faster than realtime since they benchmark mostly static environments blog post with corrected open source benchmarks & details: stoneztao.substack.com/p/the-new-hy...

48922

vmoens @vmoens.bsky.social · 19/12/2024

Wrong answers only: What does this `Human-computer` sticker seen at neurips hide?

250

vmoens @vmoens.bsky.social · 14/12/2024

Check out Motivo, a behavioral foundation model for humanoid control by FAIR. It's a one-of-its-kind unsupervised RL project, and it comes with a demo that is SO fun to play with! metamotivo.metademolab.com (for the record, they use compile and cudagraphs -> github.com/facebookrese...)

1305

vmoens @vmoens.bsky.social · 13/12/2024

I’m 100% sure this button never does anything

020

vmoens @vmoens.bsky.social · 11/12/2024

Tomorrow with Matteo Bettini we'll be presenting BenchMARL at #NeurIPS (@neuripsconf.bsky.social) in #Vancouver

180

vmoens @vmoens.bsky.social · 09/12/2024

I'm excited to be in #Vancouver for #NeurIPS 2024 where I brought a bunch of @LEGO_Group bricks in my bag! *no, I'm not Santa!

2101

vmoens @vmoens.bsky.social · 07/12/2024

We’re looking for an intern (research scientist/PhD) to join the PyTorch team in NYC this summer and work on GPU kernel generation, more info here: www.metacareers.com/jobs/6044463... You’ll be working with @soumithchintala.bsky.social, @marksaroufim.bsky.social and myself

metacareers.com

Research Scientist Intern, PyTorch Core (PhD)

Meta's mission is to build the future of human connection and the technology that makes it possible.

1143

vmoens @vmoens.bsky.social · 06/12/2024

There are two kinds of developers: - Those who know what they would have done differently had they known what a pain it was going to be - Those who change things accordingly and ruin everyone's life forever

020

vmoens @vmoens.bsky.social · 04/12/2024

My turn to go on the MLOps podcast! We talked about pytorch, compile, RL and lots of other stuff. Check it out wherever you listen to podcast, as they say 🍏 podcasts.apple.com/gb/podcast/m... ᯤ open.spotify.com/show/7wZygk3...

open.spotify.com

MLOps.community

Podcast · Demetrios Brinkmann · Weekly talks and fireside chats about everything that has to do with the new space emerging around DevOps for Machine Learning aka MLOps aka Machine Learning Operations...

030

vmoens @vmoens.bsky.social · 27/11/2024

bsky pro-tip: To check what lists you're part of: clearsky.app/%3Cid%3E.bsk... You're welcome!

clearsky.app

ClearSky

120

vmoens @vmoens.bsky.social · 27/11/2024

👉 PyTorch starter pack! go.bsky.app/P3bo7mp

0132

vmoens @vmoens.bsky.social · 27/11/2024

You know this annoying thing where you're thinking "I'd like to see the code of this pytorch class!" and you cick on source and it leads you to some weird looking html page? Fear no more! Now (on the `main` doc) you'll end up right on github!

0110

Reposted by vmoens

Sung Kim @sungkim.bsky.social · 26/11/2024

PyTorch internals by @ezyang.bsky.social This post is a long form essay version of a talk about PyTorch internals, that he gave at the PyTorch NYC meetup on May 14, 2019. blog.ezyang.com/2019/05/pyto...

0394

vmoens @vmoens.bsky.social · 25/11/2024

Your worst fear in three words? OpenReview Data Leak

031

Reposted by vmoens

Jakob Foerster @jfoerst.bsky.social · 23/11/2024

Hello BlueSky! Joao Henriques (joao.science) and I are hiring a fully funded PhD student (UK/international) for the FAIR-Oxford program. The student will spend 50% of their time @UniofOxford and 50% @MetaAI (FAIR) in London, while completing a DPhil (Oxford PhD). Deadline: 2nd of Dec AOE!!

joao.science

João F. Henriques

Research of Joao F. Henriques

15118

vmoens @vmoens.bsky.social · 23/11/2024

For those who wonder about the best way to start contributing to pytorch or open-source projects, here are the top three pointers I'd share: 1. The Ultimate Guide to PyTorch Contributions github.com/pytorch/pyto... For pytorch core that should be the n1 item on your list.

github.com

The Ultimate Guide to PyTorch Contributions

Tensors and Dynamic neural networks in Python with strong GPU acceleration - pytorch/pytorch

2183

Reposted by vmoens

Glen Berseth @glenberseth.bsky.social · 22/11/2024

RLC will be held at the Univ. of Alberta, Edmonton, in 2025. I'm happy to say that we now have the conference's website out: rl-conference.cc/index.html Looking forward to seeing you all there! @rl-conference.bsky.social #reinforcementlearning

26019

vmoens @vmoens.bsky.social · 22/11/2024

I have a deep, dark secret ... I write most of my docstrings using LLMs: I copy paste the function and ask for the docstrings. Minor edits and you're done (usually I write the example myself, that's why they're so bad). Most of the time it's *much* better than whatever I would have done!

2100

vmoens @vmoens.bsky.social · 22/11/2024

One of my fav projects: LeanRL, a simple RL library that provides recipes for fast RL training using torch.compile and cudagraphs. Using these, we got >6x speed-ups compared to the original CleanRL implementations. github.com/pytorch-labs...

2325

vmoens @vmoens.bsky.social · 20/11/2024

Not sure this makes it "10x better" than anything else, but I wrote a tuto on how to export trained policies in torchrl to use them on any edge device or hardware (from RaspberryPi to your smart toaster) The tuto is short bc it's easy Feedback is welcome!! pytorch.org/rl/main/tuto...

pytorch.org

Exporting TorchRL modules — torchrl main documentation

010

vmoens @vmoens.bsky.social · 20/11/2024

Just got started with the PyTorch Starter Pack Who did I forget?

032

Reposted by vmoens

Antonin Raffin @araffin.bsky.social · 18/11/2024

Stable-Baselines3 v2.4.0 has been released! (along with SB3-Contrib and RL Zoo) We added a new algorithm in SB3-Contrib, CrossQ, a sample-efficient alternative to SAC, and added support for Gymnasium v1.0. Full changelog: github.com/DLR-RM/stabl...

github.com

Release Stable-Baselines3 v2.4.0: New algorithm (CrossQ in SB3-Contrib) and Gymnasium v1.0 support · DLR-RM/stable-baselines3

WarningStable-Baselines3 (SB3) v2.4.0 will be the last one supporting Python 3.8 (end of life in October 2024) and PyTorch < 2.3. We highly recommended you to upgrade to Python >= 3.9 and PyTorch >...

2219

vmoens @vmoens.bsky.social · 19/11/2024

That’s so true! There’s some stigma associated with RL and I see a bunch of people doing everything they can to avoid putting that label on what they do, even if they have MDPs, rewards, value fn and all the rest. There’s so much domain knowledge being lost, and so much friction for collaborations.

020

Reposted by vmoens

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 18/11/2024

The RL (and some non-RL folks) starter pack is almost full. Pretty clear that the academic move here has succeeded go.bsky.app/3WPHcHg

1210632

vmoens @vmoens.bsky.social · 19/11/2024

Hello bluesky! Happy to be here and not there I'm a research engineer at #Meta / #PyTorch (order tbd). I work on torchrl github.com/pytorch/rl (sort of a swiss knife for RL in PyTorch) but what I'm secretly most proud of is tensordict github.com/pytorch/tens... - I can be pretty annoying about it.

github.com

GitHub - pytorch/rl: A modular, primitive-first, python-first PyTorch library for Reinforcement Learning.

A modular, primitive-first, python-first PyTorch library for Reinforcement Learning. - pytorch/rl

290