char-fork/llama.cpp

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

Frosty40sycl: add fused top-k MoE (#25217)efb3036c1

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
479 B12 linesraw
1#ifndef GGML_SYCL_TOPK_MOE_HPP
2#define GGML_SYCL_TOPK_MOE_HPP
3
4#include "common.hpp"
5
6// Detect a fusable op subgraph starting at cgraph node `i` and, if found, dispatch the fused
7// kernel. Returns the number of *following* nodes consumed (0 = no fusion applies at i).
8int ggml_sycl_fuse(ggml_backend_sycl_context & ctx, ggml_cgraph * cgraph, int i);
9
10int ggml_sycl_fuse_topk_moe(ggml_backend_sycl_context & ctx, ggml_cgraph * cgraph, int i);
11
12#endif  // GGML_SYCL_TOPK_MOE_HPP