Skip to content

Dataset · EpicPinkPenguin

PPO Agent Expert Trajectories in 16 Procgen Environments

EpicPinkPenguin/procgen

Massive set of images and actions from a trained RL agent in procedurally generated Procgen benchmark games.

What's inside

The dataset contains expert trajectories (game frames) recorded using a PPO RL agent across all 16 procedurally generated environments of the Procgen Benchmark in easy difficulty mode. The repository contains 100 million to 1 billion English examples, loaded standardly via the datasets library.

What you can do with it

  • Training behavioral cloning models (imitation learning) in games
  • Testing offline reinforcement learning algorithms on visual environments
Tasks
Reinforcement learning
Languages
English
Size
100M–1B rows
Downloads this month
25.3K
Likes
6

How to load it

from datasets import load_dataset

ds = load_dataset("EpicPinkPenguin/procgen")