---
dataset_info:
  features:
  - name: images
    sequence: image
  - name: texts
    list:
    - name: user
      dtype: string
    - name: assistant
      dtype: string
    - name: source
      dtype: string
  splits:
  - name: train
    num_bytes: 244951255658.16818
    num_examples: 565009
  download_size: 145422811605
  dataset_size: 244951255658.16818
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

[Docmatix](https://huggingface.co/datasets/HuggingFaceM4/Docmatix), but with multi-image samples filtered out.