char-fork/llama.cpp

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

Pranesh GonegandlaCUDA: dedup MoE gate/up activation quantization (#25441)5839ba352

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
278 B5 linesraw
1#pragma once
2
3void ggml_cuda_launch_mm_ids_helper(
4        const int32_t * ids, int32_t * ids_src1, int32_t * ids_dst, int32_t * expert_bounds,
5        int n_experts, int n_tokens, int n_expert_used, int nchannels_y, int si1, int sis1, bool write_inverse, cudaStream_t stream);