---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
dataset_info:
  features:
  - name: panels
    list: image
  - name: choices
    list: image
  - name: relation_structure_encoded
    dtype:
      array2_d:
        shape:
        - 4
        - 12
        dtype: uint8
  - name: relation_structure
    dtype:
      array2_d:
        shape:
        - 1
        - 3
        dtype: string
  - name: meta_target
    dtype:
      array2_d:
        shape:
        - 1
        - 12
        dtype: uint8
  - name: target
    dtype: uint8
  - name: id
    dtype: int32
  splits:
  - name: train
    num_bytes: 26850203831.0
    num_examples: 1200000
  - name: validation
    num_bytes: 602510542.0
    num_examples: 20000
  - name: test
    num_bytes: 4475789847.0
    num_examples: 200000
  download_size: 44244925294
  dataset_size: 31928504220.0
---
# Dataset Card for "PGM"

Dataset for the paper [Measuring abstract reasoning in neural networks
](https://arxiv.org/abs/1807.04225).
Only the `neutral` config is present at this point.