skills.rest
dyyoungg
dyyounggCOMMUNITY
Explore all skills by this author →
GitHub

sglang-diffusion-modelopt-quant

Quantize diffusion DiTs with NVIDIA ModelOpt for FP8/NVFP4 SGLang Diffusion.

1|Updated May 8, 2026
Search Tags:#checkpoint#benchmark#quantization#diffusion#fp8#nvfp4#modelopt
One-click install
npx skills add https://github.com/dyyoungg/sglang-dev --skill sglang-diffusion-modelopt-quant-dyyoungg
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sglang-diffusion-modelopt-quant
Source: https://github.com/dyyoungg/sglang-dev/tree/main/python/sglang/multimodal_gen/.claude/skills/sglang-diffusion-modelopt-quant
Command: npx skills add https://github.com/dyyoungg/sglang-dev --skill sglang-diffusion-modelopt-quant-dyyoungg

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Quantize a diffusion DiT using NVIDIA ModelOpt and adapt the exported checkpoint for loading in SGLang Diffusion, enabling accurate FP8/NVFP4 performance and validated deployment.

Core Features & Use Cases

  • Quantize diffusion DiTs with ModelOpt for FP8/NVFP4 exports.
  • Adapt ModelOpt-exported checkpoints to the SGLang Diffusion runtime format.
  • Validate quality and benchmark performance across FP8/NVFP4 configurations.

Quick Start

Run the ModelOpt-based workflow to quantize a diffusion DiT and convert the exported weights for loading in SGLang Diffusion.

Frequently Asked Questions about sglang-diffusion-modelopt-quant

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

FAQPage Schema
How do I quantize a diffusion DiT to FP8 or NVFP4 for SGLang Diffusion deployment?▼

To quantize a diffusion DiT to FP8 or NVFP4 for SGLang Diffusion, apply the NVIDIA ModelOpt workflow to export weights and adapt the checkpoint to the SGLang Diffusion runtime format.

What is the process for adapting a ModelOpt checkpoint for SGLang Diffusion?▼

Adapting a ModelOpt checkpoint for SGLang Diffusion involves converting the exported quantized weights into the runtime format required by SGLang, ensuring proper loading and validated deployment.

Can I benchmark performance across FP8 and NVFP4 configurations using ModelOpt?▼

Yes, you can benchmark performance across FP8 and NVFP4 configurations using the runtime helpers provided to validate model quality and compare results after applying the ModelOpt quantization workflow.

Does the SGLang Diffusion quantization workflow require specific scripts from ModelOpt?▼

Yes, the workflow enforces official ModelOpt scripts as the source of truth for quantizing the diffusion DiT, providing runtime helpers to load, validate, and compare the adapted checkpoints.

How do I validate model quality after FP8 or NVFP4 quantization?▼

You validate model quality after FP8 or NVFP4 quantization using the provided runtime helpers to load and compare the adapted checkpoints against expected performance configurations.

Related Skills

Explore More Software Engineering Skills→
annealing-inversionannealing-inversion

sglang-diffusion-modelopt-quant

Quantize diffusion DiT models with NVIDIA ModelOpt for SGLang Diffusion.

Community
Advanced
sqjiansqjian

sglang-diffusion-modelopt-quant

Quantize diffusion transformer models with NVIDIA ModelOpt FP8 and NVFP4 workflows.

Community
Advanced
OpenMOSSOpenMOSS
410

sglang-diffusion-modelopt-quant

Quantize diffusion DiT models with NVIDIA ModelOpt to FP8 or NVFP4 checkpoints for SGLang Diffusion.

Official
Advanced
FutureMLS-LabFutureMLS-Lab
556

debug-cuda-crash

Diagnose SGLang CUDA crashes by recording kernel API calls and tensor metadata.

Official
Advanced