sglang/docs/diffusion/performance/index.md

Performance Optimization

SGLang-Diffusion provides multiple performance optimization strategies to accelerate inference. This section covers all available performance tuning options.

Overview

Optimization	Type	Description
Cache-DiT	Caching	Block-level caching with DBCache, TaylorSeer, and SCM
TeaCache	Caching	Timestep-level caching using L1 similarity
Attention Backends	Kernel	Optimized attention implementations (FlashAttention, SageAttention, etc.)
Profiling	Diagnostics	PyTorch Profiler and Nsight Systems guidance

Caching Strategies

SGLang supports two complementary caching approaches:

Cache-DiT

Cache-DiT provides block-level caching with advanced strategies. It can achieve up to 1.69x speedup.

Quick Start:

SGLANG_CACHE_DIT_ENABLED=true \
sglang generate --model-path Qwen/Qwen-Image \
    --prompt "A beautiful sunset over the mountains"

Key Features:

• DBCache: Dynamic block-level caching based on residual differences
• TaylorSeer: Taylor expansion-based calibration for optimized caching
• SCM: Step-level computation masking for additional speedup

See Cache-DiT Documentation for detailed configuration.

TeaCache

TeaCache (Temporal similarity-based caching) accelerates diffusion inference by detecting when consecutive denoising steps are similar enough to skip computation entirely.

Quick Overview:

• Tracks L1 distance between modulated inputs across timesteps
• When accumulated distance is below threshold, reuses cached residual
• Supports CFG with separate positive/negative caches

Supported Models: Wan (wan2.1, wan2.2), Hunyuan (HunyuanVideo), Z-Image

See TeaCache Documentation for detailed configuration.

Attention Backends

Different attention backends offer varying performance characteristics depending on your hardware and model:

• FlashAttention: Fastest on NVIDIA GPUs with fp16/bf16
• SageAttention: Alternative optimized implementation
• xformers: Memory-efficient attention
• SDPA: PyTorch native scaled dot-product attention

See Attention Backends for platform support and configuration options.

Profiling

To diagnose performance bottlenecks, SGLang-Diffusion supports profiling tools:

• PyTorch Profiler: Built-in Python profiling
• Nsight Systems: GPU kernel-level analysis

See Profiling Guide for detailed instructions.

References

• Cache-DiT Repository
• TeaCache Paper