happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

handle broadcast

This commit is contained in:
Aman Gupta 2026-03-16 09:30:32 +01:00
parent a57641d189
commit 2b685f2f05
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/ggml-cpu/ggml-cpu.c
View File

@ -3078,7 +3078,8 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
struct ggml_tensor * glu = cgraph->nodes[node_n + 2];
// Fused path for `--n-cpu-moe` when n_tokens = 1
if (node->src[1] == node1->src[1] && node->src[2] == node1->src[2] &&
node->src[1]->ne[2] == 1 && ggml_get_glu_op(glu) == GGML_GLU_OP_SWIGLU) {
ggml_nrows(node->src[1]) == 1 &&
ggml_get_glu_op(glu) == GGML_GLU_OP_SWIGLU) {
ggml_compute_forward_fused_moe_silu(&params, node, node1, glu);
fused_nodes = 2;
}