xlite-dev/Awesome-DiT-Inference

Organizes diffusion inference optimization papers across five technical domains—sampling algorithms (DPM-Solver variants, parallel denoising), KV-cache strategies (DeepCache, layer caching), tensor parallelism and distributed training, quantization methods, and attention optimizations. Covers both UNet and Transformer-based (DiT) architectures with implementation links and research citations. Targets the broader ML inference acceleration ecosystem, complementing a parallel LLM inference resource for practitioners optimizing generative model deployments.

526 stars. Actively maintained with 1 commit in the last 30 days.