metadata
dataset_info:
features:
- name: observation.images.image
dtype: video_frame
- name: language_instruction
dtype: string
- name: observation.state
sequence: float32
length: 1
- name: action
sequence: float32
length: 4
- name: timestamp
dtype: float32
- name: episode_index
dtype: int64
- name: frame_index
dtype: int64
- name: next.reward
dtype: float32
- name: next.done
dtype: bool
- name: index
dtype: int64
splits:
- name: train
num_bytes: 17712500
num_examples: 100000
download_size: 1140164
dataset_size: 17712500
configs:
- config_name: default
data_files:
- split: train
path: data/train-*