🚀 Excited to release Stoix! A new #OpenSource library for End-to-End Distributed (Synchronously) Single-Agent Reinforcement Learning in JAX. 🏛️. 🔗

RT @pfau: New paper accepted to ICML! We present a novel policy optimization algorithm for continuous control with a simple closed form whi….

RT @MattVMacfarlane: Thrilled to see our NeurIPS 2024 paper, Sequential Monte Carlo Policy Optimisation (, featured….

RT @DulhanJay: Efficient LLM reasoning over large data doesn't require massive contexts! 🫡. We show that a simple in-context method, PRISM,….

RT @pcastr: It's amazing two of the 2024 #NobelPrize were for AI! But as they say: it took a village. "We didn't win a Nobel", a parody of….

RT @instadeepai: Excited to share our latest work on Sequential Monte Carlo Policy Optimisation (SPO)🔥— a scalable, search-based RL algorit….

RT @ClementBonnet16: Introducing Latent Program Network (LPN), a new architecture for inductive program synthesis that builds in test-time….

Credit to the Mava team for helping as well as InstaDeeps amazing implementation for help as reference.

Inline with Stoix’s code philosophy, it's super hackable and easy to modify, perfect for flexible and high-performance RL research! . Currently only PPO is supported but many more to come. Let me know your thoughts.

🚀 Introducing Sebulba Systems in Stoix! Harness JAX's speed with non-JAX environments in your single-agent RL research. Sebulba splits the actors and learners, running environments on CPUs but performing batched inference/learning using GPUs/TPUs. 🔗

RT @pcastr: Can we fix the review process before we try to automate science?.

RT @callumtilbury: What happens when trying to learn multi-agent coordination from a static dataset? Catastrophe, if you’re not careful!. T….

RT @ChalumeauFelix: Excited to introduce our latest neural solver, MEMENTO! Enhancing problem-specific adaptation with an explicit memory.….

RT @AlexLaterre: Got lost in the #ICLR2024 poster maze? Don't worry, we've got your covered! 🛟. Here is @DonalByrne2, Senior Research Engin….

RT @callumtilbury: Curious about this diagram? Join us later today as we discuss growing the MARL ecosystem in JAX! 🤖🍿. @instadeepai @ruanj….

RT @ClementBonnet16: Excited to announce Jumanji v1.0, now featuring 22 fast, flexible, and scalable environments!. Fully written in JAX, J….

RT @ClementBonnet16: If you haven't yet, please check out the amazing works from the JAX community. E.g. environments: Brax (@OlivierBachem….

Important to mention is that Stoix can be seen as a single-agent counter part to Mava . Stoix actually started as a Mava clone when I wanted an End-to-End JAX PPO with all the extra features Mava provides. If you’re into in MARL, can’t recommend it enough.

🧪 Robust evaluation and logging ready for statistical testing with and 🚀 Optimised for speed with JAX's pmap & jit allowing for quick and easy scaling. Any contribution and feedback is welcome. 🤝 #RL #JAX.

🍬 Environment wrappers for most JAX-native environment suites (Gymnax @RobertTLange , Jumanji @ClementBonnet16 , Brax @GoogleOSS, XMinigrid @how_uhh, Craftax @mitrma and even JAXMarl @alexrutherford0 (using Centralised Controllers).

🌟 Features:. 🥑 Ready-to-use RL algorithms like PPO, DQN, SAC & more to serve as baselines and research starting points.