From f0cffc74449dd4bb027b6a0f3d445eb3d5d1ac11 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= Date: Fri, 18 Jul 2025 10:52:39 +0200 Subject: [PATCH] fix exaone4 gf usage --- src/llama-model.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/llama-model.cpp b/src/llama-model.cpp index cd3e456948cd2..46d5d32fc1b1b 100644 --- a/src/llama-model.cpp +++ b/src/llama-model.cpp @@ -13530,7 +13530,7 @@ struct llm_build_exaone : public llm_graph_context { template struct llm_build_exaone4 : public llm_graph_context { - llm_build_exaone4(const llama_model & model, const llm_graph_params & params, ggml_cgraph * gf) : llm_graph_context(params) { + llm_build_exaone4(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { const int64_t n_embd_head = hparams.n_embd_head_k; GGML_ASSERT(n_embd_head == hparams.n_embd_head_v); @@ -13603,7 +13603,7 @@ struct llm_build_exaone4 : public llm_graph_context { cb(Kcur, "Kcur", il); cb(Vcur, "Vcur", il); - cur = build_attn(inp_attn, gf, + cur = build_attn(inp_attn, model.layers[il].wo, NULL, Qcur, Kcur, Vcur, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il); cb(cur, "attn_out", il);