skills.rest
hao-ai-lab
hao-ai-labOFFICIAL
Explore all skills by this author →
Website𝕏@haoailabGitHub

seed-ssim-references

Generate, verify, and upload SSIM reference artefacts to Hugging Face datasets.

3.9k|398|Updated Oct 24, 2024
Search Tags:#testing#reference#video#dataset#ssim#hf
One-click install
npx skills add https://github.com/hao-ai-lab/FastVideo --skill seed-ssim-references
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: seed-ssim-references
Source: https://github.com/hao-ai-lab/FastVideo/tree/main/.agents/skills/seed-ssim-references
Command: npx skills add https://github.com/hao-ai-lab/FastVideo --skill seed-ssim-references

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires modal, huggingface_hub, torch, imageio, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill streamlines the process of generating, verifying, and uploading SSIM reference artefacts for new video similarity tests, ensuring consistent and accurate benchmarking.

Core Features & Use Cases

  • Artefact Generation: Runs SSIM tests on Modal's GPU to produce pixel (.mp4) or latent (.pt) reference files for new video similarity tests.
  • Verification & Quality Check: Pauses for user verification of visual quality or numerical sanity, preventing erroneous references from being uploaded.
  • Uploading & Seeding: Safely uploads verified references to Hugging Face HF datasets, with safeguards against overwriting existing data, enabling reproducible and robust test baseline establishment.
  • Use Case: When adding a new SSIM test in fastvideo/tests/ssim/, you run this Skill to generate correct reference artefacts and seed them into the shared dataset for consistent test validation.

Quick Start

Prompt the user for the test file path, execute the modal run to generate artefacts, inspect the output files for quality, and then upload them to Hugging Face dataset.

Frequently Asked Questions about seed-ssim-references

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

FAQPage Schema
How do I generate SSIM reference artefacts for video similarity tests?▼

You generate SSIM reference artefacts by running your test file on Modal GPU resources to produce pixel-level `.mp4` or latent `.pt` reference files for validating video similarity tests.

Can I upload SSIM references directly to a Hugging Face dataset without verification?▼

No, you cannot upload SSIM references directly without verification; the process pauses for visual or numerical sanity checks and uses upload guards to prevent overwriting existing data on Hugging Face.

Do I need Modal and Torch to run SSIM video reference tests?▼

Yes, you need Modal and Torch to execute SSIM video reference tests, along with Hugging Face Hub and Imageio dependencies, to support GPU computation and handle both `.mp4` and `.pt` artefacts.

What is the best way to maintain consistent reference datasets for reproducible video testing?▼

The best way to maintain consistent reference datasets for reproducible video testing is to automate the generation, verification, and uploading of SSIM reference artefacts to a shared Hugging Face dataset.

Why should I pause to inspect generated `.mp4` or `.pt` files before seeding a test baseline?▼

You should pause to inspect generated `.mp4` or `.pt` files before seeding a test baseline to prevent erroneous references from being uploaded, which ensures accurate and reproducible benchmarking.

Related Skills

Explore More Software Engineering Skills→
tony-ng-vntony-ng-vn

huggingface-jobs

Run Python workloads on Hugging Face Jobs with GPU/CPU resources.

Community
Advanced
aayushsoamaayushsoam
1

huggingface-hub

Manage models and datasets on the Hugging Face Hub.

Community
Intermediate
buildoakbuildoak
15

video-gen

Orchestrate fal.ai video generation and assembly across multi-scene pipelines.

Community
Advanced
tayyabexetayyabexe
3

hf-mcp

Search Hugging Face Hub models, datasets, and Spaces, and run cloud compute jobs.

Community
Advanced