From 31b6aec899f7045d9733bca54506057a62c80b02 Mon Sep 17 00:00:00 2001 From: Anirban Kar Date: Sat, 25 Jul 2026 03:42:13 +0530 Subject: [PATCH] feat: wire MoE expert cache into laguna graph --- src/models/laguna.cpp | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/models/laguna.cpp b/src/models/laguna.cpp index fb55ec12f934..4d209c0b1a8b 100644 --- a/src/models/laguna.cpp +++ b/src/models/laguna.cpp @@ -286,7 +286,10 @@ llama_model_laguna::graph::graph(const llama_model & model, const llm_graph_para hparams.expert_weights_norm, hparams.expert_weights_scale, (llama_expert_gating_func_type) hparams.expert_gating_func, - il); + il, + nullptr, nullptr, + nullptr, nullptr, nullptr, + nullptr, &model.layers[il]); cb(moe_out, "ffn_moe_out", il); // Always-on shared expert, summed in parallel.