iris / llama.cpp /ggml /src /ggml-cuda /fattn-tile-f32.cuh
Mat17892's picture
llamacpp
b664585 verified
raw
history blame contribute delete
115 Bytes
#include "common.cuh"
void ggml_cuda_flash_attn_ext_tile_f32(ggml_backend_cuda_context & ctx, ggml_tensor * dst);