metadata
dataset_info:
features:
- name: images
list:
- name: bytes
dtype: binary
- name: path
dtype: 'null'
- name: data_source
dtype: string
- name: reward_model
struct:
- name: ground_truth
dtype: string
- name: extra_info
struct:
- name: answer
dtype: string
- name: doc_id
dtype: string
- name: final_size
sequence: int64
- name: image_path
dtype: string
- name: original_problem
dtype: string
- name: original_size
sequence: int64
- name: question
dtype: string
- name: was_resized
dtype: bool
- name: prompt
list:
- name: content
dtype: string
- name: role
dtype: string
- name: ability
dtype: string
- name: env_name
dtype: string
splits:
- name: train
num_bytes: 8590651196
num_examples: 5729
download_size: 8582205021
dataset_size: 8590651196
configs:
- config_name: default
data_files:
- split: train
path: data/train-*