text-vision-2k-test / README.md
Nanobit's picture
Update README.md
690af00 verified
metadata
dataset_info:
  features:
    - name: messages
      list:
        - name: content
          list:
            - name: path
              dtype: string
            - name: text
              dtype: string
            - name: type
              dtype: string
        - name: role
          dtype: string
  splits:
    - name: train
      num_bytes: 393500
      num_examples: 2000
  download_size: 5523
  dataset_size: 393500
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

A 2k sample dataset for testing multimodal (text+vision) format. This is compatible with HF's processor apply_chat_template.

Load in Axolotl via:

datasets:
  - path: Nanobit/text-vision-2k-test
    type: chat_template

Make sure to download the image via:

wget https://huggingface.co/datasets/Nanobit/text-vision-2k-test/resolve/main/African_elephant.jpg

Image source: https://upload.wikimedia.org/wikipedia/commons/e/ec/African_elephant.jpg

Each sample has the following format and is repeated 2k times:

[
    {
        "role": "user",
        "content": [
            {"type": "image", "path": LOCAL_IMAGE_PATH},
            {
                "type": "text",
                "text": "How would you describe the above image in one phrase?",
            },
        ],
    },
    {
        "role": "assistant",
        "content": [
            {
                "type": "text",
                "text": "The image is about an African elephant.",
            }
        ],
    },
]