README.md

metadata

dataset_info:
  features:
    - name: observation.image
      dtype: video_frame
    - name: observation.state
      sequence: float32
      length: 4
    - name: action
      sequence: float32
      length: 3
    - name: episode_index
      dtype: int64
    - name: frame_index
      dtype: int64
    - name: timestamp
      dtype: float32
    - name: next.reward
      dtype: float32
    - name: next.done
      dtype: bool
    - name: index
      dtype: int64
  splits:
    - name: train
      num_bytes: 2222500
      num_examples: 20000
  download_size: 1057600
  dataset_size: 2222500
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*