STOK-Core implements State-Time Option Kernels end-to-end: a reinforcement learning toolkit modeling goal/constraint termination distributions for compositional planning. It pairs Burn-backed GPU acceleration with Task MDP solvers, Chapman-Kolmogorov composition, and factorized prediction kernels so hierarchical planners remain modular.
rust reinforcement-learning planning gpu-acceleration markov-decision-process hierarchical-rl compositional-planning
-
Updated
Jun 6, 2026 - Rust