Yolande1994/CUDA_GPT2_Inference
纯 CUDA 实现 GPT2 推理,逐步优化算子与架构,最终在单请求下推理速度领先 vLLM 约 1.5 倍。含详细 NCU 分析与 README 优化报告。
GitHub repository with 12 stars and 0 forks.
Language: Cuda
Topics: cpp, cuda, cuda-graphs, cuda-kernels, flash-attention, flash-decoding, gemv, gpt-2, kv-cache, llm-inference