char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Daniel Beveniussampling : add support for backend sampling (#17004)d3dce4e0a2026-01-04

#include "common.cuh"

void ggml_cuda_op_top_k(ggml_backend_cuda_context & ctx, ggml_tensor * dst);