---
dataset_info:
  features:
  - name: images
    list: image
  - name: problem
    dtype: string
  - name: answer
    dtype: string
  - name: mask_image
    dtype: bool
  - name: rollout_batch
    struct:
    - name: attention_mask
      list:
        list: int64
    - name: input_ids
      list:
        list: int64
    - name: position_ids
      list:
        list:
          list: int64
    - name: prompts
      list:
        list: int64
    - name: response_mask
      list:
        list: int64
    - name: responses
      list:
        list: int64
  splits:
  - name: train
    num_bytes: 29455947344.0
    num_examples: 16960
  - name: val
    num_bytes: 7361777686.0
    num_examples: 4240
  download_size: 1017601095
  dataset_size: 36817725030.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
---