Antonin Raffin

@araffin.bsky.social

3413 Followers 324 Following

Researcher in robotics and machine learning (Reinforcement Learning). Maintainer of Stable-Baselines (SB3). https://araffin.github.io/

An animation about Q-value overestimation and TD3 Q-value clipping to reduce the issue.
A screenshot of car dodging gym
a spurious correlation example
a robot arm support a robot humanoid on a treadmill
the DLR quadruped bert
a screenshot of the html version with prompt injection written white on black
a screenshot of SB3 downloads
Learning curves for PPO SB3 and PPO Jax where one is learning and the other not. Showing the one-line code diff
A small code snippet showing how to use the `n_steps` parameter with SB3 lib
A GUI showing how changing variables in the latent space affects the decoder output of an auto-encoder
A screenshot of output from the blenderproc tool, with associated options. Left: Mesh Middle: Point Cloud Right: Depth
Code to create a dm control env and convert it to a gym env compatible with SB3
An elastic quadruped robot jumping happily
An example of throwing more informative errors. Where name is a tuple, it also prints the variable instead of just complaining it is not a string: f"'name' must be a string, not of type {type(name)}: {name}" vs "'name' must be a string"
Types of reinforcement learning papers, using xkcd original artwork