From 0308f5e3d7bf9879f818b1a4ae589ff36b242af5 Mon Sep 17 00:00:00 2001
From: compilade <113953597+compilade@users.noreply.github.com>
Date: Thu, 28 Mar 2024 08:05:54 -0400
Subject: [PATCH] llama : fix command-r inference when omitting outputs (#6367)

---
 llama.cpp | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/llama.cpp b/llama.cpp
index 892d46fbcfcec..77ec9b7a1935d 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -9152,8 +9152,9 @@ struct llm_build_context {
             if (il == n_layer - 1) {
                 // skip computing output for unused tokens
                 struct ggml_tensor * inp_out_ids = build_inp_out_ids();
-                cur  = ggml_get_rows(ctx0,  cur, inp_out_ids);
-                inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
+                cur     = ggml_get_rows(ctx0,     cur, inp_out_ids);
+                inpL    = ggml_get_rows(ctx0,    inpL, inp_out_ids);
+                ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
             }
 
             struct ggml_tensor * attn_out = cur;