[skip ci][Doc] Refine the Diffusion Features User Guide

[wtomin]

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

This PR majorly refines the user_guide of diffusion models:

• The main doc is docs/user_guide/diffusion_features.md, which maintains the supported features table, the supported models table, and feature compatibility table
• The other feature docs are:

  • Cache Acceleration:

    • TeaCache Configuration Guide - Parameter tuning, performance tips, troubleshooting
    • Cache-DiT Advanced Guide - DBCache, TaylorSeer, SCM techniques and optimization

  • Parallelism Methods:

    • Tensor Parallelism Guide - Shard DiT weights across GPUs to reduce per-GPU memory
    • Sequence Parallelism Guide - Ulysses-SP and Ring-Attention for long sequence handling
    • CFG Parallelism Guide - Distribute CFG positive/negative branches across GPUs
    • HSDP Guide - Hybrid Sharded Data Parallel, which is similar to PyTorch FSDP2 sharding
    • EP Guide - Expert Parallelism shards the MoE expert MLP blocks

  • Memory Optimization:

    • CPU Offload Guide - Offload model components to CPU, reduce GPU memory usage
    • VAE Patch Parallelism Guide - Distribute VAE decode tiling across GPUs for high-resolution images
    • Quantization Overview - Overview of quantization methods for diffusion models

  • Extensions:

    • LoRA Inference Guide - Low-Rank Adaptation for style customization and fine-tuning

  • Advanced Topics:

    • Feature Compatibility - How to combine cache + parallelism methods for maximum performance

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
• The test results. Please paste the results comparison before and after, or the e2e results.
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
• (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7e4b1e0f3c

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[princepride]

#1287 I noticed we have another pr about cache accelerate

[wtomin]

#1287 I noticed we have another pr about cache accelerate

Since it's not merged. I may not include this feature in this PR.

It occurs to me that this feature will need two documents:

• docs/user_guide/diffusion/cache/magcache.md: introducing how to use the this method;
• docs/design/feature/magcache.md: introducing how to adapt a model/pipeline to support this method.

Thanks for reminding.

[Gaohan123]

@wtomin Please resolve conflicts

[princepride]

@yuanheng-zhao PTAL

[hsliuustc0106]

lgtm