thu-ml/SpargeAttn · AI Repo StatsTHthu-ml/SpargeAttn
[ICML2025] SpargeAttention: A training-free sparse attention that accelerates any model inference.
Suy luận & chạy LLM cục bộCuda Apache-2.0 108tạo 2025-02-25 #ai-infra#attention#inference-acceleration#llm#mlsys#quantization#sageattention#sparse-attention#video-generation#vision-transformer#vit