Datasets:

eagle0504
/

llava-video-text-dataset

configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
dataset_info:
  features:
    - name: video_url
      dtype: string
    - name: conversation
      list:
        - name: content
          list:
            - name: text
              dtype: string
            - name: type
              dtype: string
        - name: role
          dtype: string
    - name: num_frames
      dtype: int64
  splits:
    - name: train
      num_bytes: 578
      num_examples: 2
    - name: validation
      num_bytes: 578
      num_examples: 2
  download_size: 7696
  dataset_size: 1156

eagle0504/llava-video-text-dataset

This is a tiny LLaVA dataset with exactly four video samples for training.

Field video_url: Video URLs (MP4/GIF format)
Field conversation: LLaVA conversation format with user/assistant roles
Field num_frames: Number of frames per video (5)

Dataset Structure

Each sample contains a conversation in LLaVA format:

{
  "video_url": "https://example.com/video.mp4",
  "conversation": [
    {
      "role": "user",
      "content": [
        {"type": "text", "text": "What is in this video?"},
        {"type": "image"},
        {"type": "image"},
        {"type": "image"},
        {"type": "image"},
        {"type": "image"}
      ]
    },
    {
      "role": "assistant", 
      "content": [{"type": "text", "text": "There is a cat in the video."}]
    }
  ],
  "num_frames": 5
}

Usage

from datasets import load_dataset

dataset = load_dataset("eagle0504/llava-video-text-dataset")

Model Compatibility

This dataset is designed for LLaVA models that support video input through multiple image frames.