iiinova — where your digital self keeps evolving

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

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

MLGym makes it super easy to set up complex tasks to be solved by LLMs. Honestly one of the most intuivite APIs I have ever seen in that space!

000

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

After that, your LLM reads these instructions, and outputs prompts with some thoughts. The commands are executed in the docker's bash, and the result is returned to the agent.

100

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

SnakeViz is a browser based graphical viewer for the output of Python's cProfile module.

Good old cProfile with snakeviz is pretty cool too jiffyclub.github.io/snakeviz/ Again, not for cuda ops, and not as fine-grained as line-profiler but quite useful for macro-tracking of compute time

jiffyclub.github.io

SnakeViz

020

torch.utils.benchmark.Timer is amazing to assess the runtime of a whole isolated piece of code, but be mindful that the way it plays with global variables isn't always obvious and may differ from time.time() on occasions

I use line_profiler to check the code line-by-line (careful: cuda ops re async, do not trust it for these!) - very useful to check cpu-overhead pypi.org/project/line...

pypi.org

line-profiler

Line-by-line profiler

PyTorch Profiler — PyTorch Tutorials 2.6.0+cu124 documentation

The profilers I use: PyTorch profiler to view the time spend doing the various ops of my code. It can reliably show you what's going on for a single iteration of your function. pytorch.org/tutorials/re...

pytorch.org

In general, in-place operations are not preferable to regular ones (you won't gain much mem improvement or speed-ups). Don't load your code with ReLU(inplace=True), mul_, add_ if not absolutely necessary.

110

Using hydra or similar fancy config objects: Avoid calling cfg.attribute often in the code. Instead, cache the args values in your script as global workspace variables.

If you have a tiny model (robotics, RL) cpu-overhead bound, avoid frequent calls to eval() or train() in eager mode, or model.parameters() or anything that goes through your model. Prefer cached versions of these calls.

Avoid calling tensor.item() in between cuda operations. This triggers a cuda synchronization and blocks your code. Do the logging after all code (forward / backward / optim) has completed. See how to find sync points here)

130

A guide on good usage of non_blocking and pin_memory() in PyTorch — PyTorch Tutorials 2.6.0+cu124 documentation

Avoid pinning memory in your code unless you thoroughly tested that it accelerates runtime (see this tutorial for more info). As an aside, pin_memory is also less safe! pytorch.org/tutorials/in...

pytorch.org

Don't send tensors to device using to(device) if you can instantiate them directly there. For instance, prefer randn((), device=device) to randn(()).to(device)

130

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

1122

I guess my point was that a proper name + definition is necessary to write good code. When I see “policy”, “critic”, “replay buffer”, “env” I know exactly what does and doesn’t belong to them. With agent is systematically a “hm yeah why not” - then you end up with ill-defined monster classes

000

If your agent is a policy call it policy, if it's a trainer call it trainer! If it's just a big undefined collection of methods, consider refactoring it...

Every time I meet with people and someone talks about agent, there's at least one person who asks "what do you mean by agent?" or "you should not call that an agent".

110

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

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

hard to tell, let's try :D

010

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 · 12/12/2024

Easily transform video files into PyTorch tensors with: 🎯User-friendly APIs 🎯Exceptional CPU and CUDA performance 🎯Advanced sampling capabilities tailored for ML training pipelines

000

Links: NeurIPS page: neurips.cc/virtual/2024... GitHub: github.com/facebookrese... Paper: arxiv.org/abs/2312.01472

010

Where: West Ballroom A-D poster 6510, Wednesday Dec. 11th from 11 a.m. PST to 2 p.m. PST We’d love to see you there — please come and say hi!

100

Built on TorchRL and PyTorch, BenchMARL ensures high performance and state-of-the-art implementations, while its flexible configuration and standardized reporting make it a breeze to use.

100

BenchMARL is a cutting-edge training library designed to bring standardized benchmarking to the world of Multi-Agent Reinforcement Learning (MARL). It allows for easy comparison across different algorithms, models, and environments, making it a game-changer for researchers and developers alike.

100