Open SourceApr 27, 2026 · 7 min read
Jaxpot: Train self-play RL agents FAST by parallelizing environments on GPU
Fast self-play RL with GPU environments - how we built Jaxpot for PPO, AlphaZero-style training, and imperfect-information games like Dark Hex.
Karol Kłusek
ML Engineer @ bards.ai



