ik-llama-cpp-sys 0.1.4

Low level bindings to ik_llama.cpp (ikawrakow's SOTA-quant fork)
Documentation
1
2
3
4
5
6
7
#pragma once

#include "common.cuh"

void ggml_cuda_flash_attn_ext_vec_f16(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

bool ggml_cuda_fattn_vec_f16_is_supported(ggml_backend_cuda_context & ctx, const ggml_tensor * dst);