A simple wrapper to analyse and visualise reinforcement learning agents’ behaviour in the environment
Visrl (pronounced “visceral”) is a simple wrapper to analyse and visualise reinforcement learning agents’ behaviour in the environment. Reinforcement learning requires a lot of overhead code to inspect an agent’s behaviour visually, typically through env.render(). Visrl allows users to easily intervene and switch between agent control and human control, and allows inserting a breakpoint in the game state to pause only at a relevant state of interest. Features Set action hotkeys Human intervention: Take actions 1 step at a time […]
Read more