From 489554bc1126bcacd93fbf0b395d6c3062fe0d64 Mon Sep 17 00:00:00 2001 From: Kawrakow Date: Mon, 10 Nov 2025 17:38:46 +0200 Subject: [PATCH] Minor: remove unnecesssary calls to build_inp_out_ids (#935) Co-authored-by: Iwan Kawrakow --- src/llama-build-context.cpp | 2 -- 1 file changed, 2 deletions(-) diff --git a/src/llama-build-context.cpp b/src/llama-build-context.cpp index f7838a66..96d39b24 100644 --- a/src/llama-build-context.cpp +++ b/src/llama-build-context.cpp @@ -6474,7 +6474,6 @@ ggml_cgraph * llm_build_context::build_glm4_moe() { // crop output on last layer if (il == n_transformer_layers - 1 && inp_out_ids) { // skip computing output for unused tokens - ggml_tensor * inp_out_ids = build_inp_out_ids(); cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); } @@ -7776,7 +7775,6 @@ ggml_cgraph * llm_build_context::build_ernie4_5() { } if (il == n_layer - 1 && inp_out_ids) { - ggml_tensor * inp_out_ids = build_inp_out_ids(); cur = ggml_get_rows(ctx0, cur, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); }