fattn.cuh source code [llama.cpp/ggml/src/ggml-cuda/fattn.cuh]

1	#include "common.cuh"
2
3	void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
4
5	bool ggml_cuda_flash_attn_ext_supported(int device, const ggml_tensor * dst);
6

Browse the source code of llama.cpp/ggml/src/ggml-cuda/fattn.cuh