char-fork/llama.cpp

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

R0CKSTARmusa: enable fp16 mma (all) and cublas on qy2 (#13842)716301d1b

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
333 B12 linesraw
1#pragma once
2
3#include "ggml-cuda/common.cuh"
4#include "ggml.h"
5
6// Asynchronously copies data from src tensor to dst tensor using the provided context.
7// Returns a musaError_t indicating success or failure.
8musaError_t mudnnMemcpyAsync(
9    ggml_backend_cuda_context &ctx,
10    const ggml_tensor *dst,
11    const ggml_tensor *src
12);