skills.rest
sovr610
sovr610COMMUNITY
Explore all skills by this author →
WebsiteGitHub

V-JEPA 2 Data Pipeline

Build and validate a scalable V-JEPA 2 video data pipeline with YAML-config-driven DataLoader construction.

Updated Feb 28, 2026
Search Tags:#augmentation#data-pipeline#multisource#yaml-config#vjepa#video-data#distributed-sampling
One-click install
npx skills add https://github.com/sovr610/refffiy --skill v-jepa-2-data-pipeline
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: V-JEPA 2 Data Pipeline
Source: https://github.com/sovr610/refffiy/tree/main/brain-ai-dev/skills/vjepa2-data-pipeline
Command: npx skills add https://github.com/sovr610/refffiy --skill v-jepa-2-data-pipeline

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, torch, pyyaml, Pillow, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill helps ML engineers build and validate a scalable V-JEPA 2 video data pipeline, coordinating video decoding, clip sampling, augmentation, multi-source data mixing, and deterministic data loading.

Core Features & Use Cases

  • Multi-source VideoDataset support with fps, duration, and frame_step clip modes
  • YAML-config driven augmentation and data loading via DataConfig and AugConfig
  • Deterministic, reproducible DataLoader pipelines with per-source weighting and distributed sampling
  • Ready-made templates for data management, transforms, and evaluation, plus synthetic data fallbacks for testing

Quick Start

Create a DataConfig with your data_paths and an AugConfig for augmentation, then instantiate DataManager to build the train loader.

Frequently Asked Questions about V-JEPA 2 Data Pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a scalable video data pipeline for PyTorch distributed training?▼

A V-JEPA 2 video data pipeline coordinates multi-source loading, clip sampling, and augmentation via YAML configs. It builds a deterministic PyTorch DataLoader with DistributedWeightedSampler for reproducible distributed training.

How do I configure multi-source video dataset loading and sampling in PyTorch?▼

Configure multi-source video loading via DataConfig YAML files specifying data_paths, clip modes, and per-source weights. The pipeline applies DistributedWeightedSampler to manage data mixing across distributed training workers.

Can I use YAML to configure video augmentation transforms for machine learning models?▼

Yes, YAML AugConfig files drive the VideoTransformPipeline composition for video augmentation. This configures deterministic transformations to ensure reproducible data processing across distributed training runs.

Does this video data pipeline support distributed sampling and deterministic worker seeding?▼

Yes, the pipeline implements DistributedWeightedSampler for multi-source weighting and deterministic worker seeding. This ensures reproducible data loading and augmentation across multiple distributed training workers.

How do I test a video data pipeline without large datasets?▼

Test without large datasets by using the pipeline's synthetic data fallbacks. These generate mock video data to validate DataConfig, AugConfig, and DataLoader construction locally before scaling to full distributed training.

Related Skills

Explore More Data & Analytics Skills→
jongwonyjongwony
4

veo

Generate high-fidelity videos via Veo 3.1 API with text, image, or video input.

Community
Advanced
sovr610sovr610

V-JEPA 2 Distributed Training

Distribute V-JEPA 2 training across GPUs and nodes with NCCL and SLURM.

Community
Advanced
buda-aibuda-ai
25

keyframe-extractor

Detect scene changes and extract keyframes with metadata from video files.

Official
Intermediate
HermeticOrmusHermeticOrmus

PyTorch Patterns

Implement custom PyTorch datasets, modules, and autograd functions.

Community
Advanced