metadata

license: cc-by-4.0
language:
  - en
tags:
  - text-to-video

Data-Juicer Sandbox: A Comprehensive Suite for Multimodal Data-Model Co-development

Project description

The emergence of large-scale multi-modal generative models has drastically advanced artificial intelligence, introducing unprecedented levels of performance and functionality. However, optimizing these models remains challenging due to historically isolated paths of model-centric and data-centric developments, leading to suboptimal outcomes and inefficient resource utilization. In response, we present a novel sandbox suite tailored for integrated data-model co-development. This sandbox provides a comprehensive experimental platform, enabling rapid iteration and insight-driven refinement of both data and models. Our proposed "Probe-Analyze-Refine" workflow, validated through applications on T2V-Turbo and achieve a new state-of-the-art on VBench leaderboard with 1.09% improvement from T2V-Turbo. Our experiment code and model are released at Data-Juicer Sandbox.

Dataset Information

The whole dataset is available here (About 227.5GB).
Number of samples: 147,176 (Include videos and keep ~12.09% from the original dataset)
The original dataset totals 1,217k instances from InternVid (606k), Panda-70M (605k), and MSR-VTT (6k).

Refining Recipe

# global parameters
# global parameters
project_name: 'Data-Juicer-recipes-T2V-optimal'
dataset_path: '/path/to/your/dataset'  # path to your dataset directory or file
export_path: '/path/to/your/dataset.jsonl'

np: 4  # number of subprocess to process your dataset

# process schedule
# a list of several process operators with their arguments
process:
  - video_nsfw_filter:
      hf_nsfw_model: Falconsai/nsfw_image_detection
      score_threshold: 0.000195383
      frame_sampling_method: uniform
      frame_num: 3
      reduce_mode: avg
      any_or_all: any
      mem_required: '1GB'
  - video_frames_text_similarity_filter:
      hf_clip: openai/clip-vit-base-patch32
      min_score: 0.306337
      max_score: 1.0
      frame_sampling_method: uniform
      frame_num: 3
      horizontal_flip: false
      vertical_flip: false
      reduce_mode: avg
      any_or_all: any
      mem_required: '10GB'