fishlikeX/sm70-attn

FlashAttention brought back to Tesla V100 — a deep llama.cpp fork: SM 7.0 D256 kernels, SplitKV3, q4_0 KV cache, DFlash2 speculative decoding and multimodal fixes.

语言：C++ · GitHub 星标：19 · Fork：3

话题：cuda · flash-attention · llama-cpp · llm-inference · qwen · speculative-decoding · v100 · volta

在 GitHub 上查看原仓库