This dataset is part of a larger dataset that was generated by rolling out a pretrained RL agent and saving trajectories as chunked .array_record shards (compatible with the DreamerV3/DreamerV4-style dataset format used
in dreamer4-jax-private/coinrun_data/generate_coinrun_dataset.py).
This dataset contains only the successful episodes from the original dataset.
For the original dataset, see edwhu/coinrun_hard_agent.… See the full description on the dataset page:
https://huggingface.co/datasets/NathanGavenski/coinrun_hard_agent_only_successful.
