From 940b1e8f636e24007f659e0f54c9c03cb6f61898 Mon Sep 17 00:00:00 2001
From: Douglas Hanley <thesecretaryofwar@gmail.com>
Date: Mon, 24 Jun 2024 00:55:16 -0500
Subject: [PATCH] fix output counting comment

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

diff --git a/llama.cpp b/llama.cpp
index a0c9d1d745c986..0566adca51a79a 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -12618,8 +12618,7 @@ static int llama_decode_internal(
     std::vector<llama_seq_id *>            seq_id_arr;
     std::vector<std::vector<llama_seq_id>> seq_id;
 
-    // this indicates we are doing pooling on an embedding model. non-embedding models always
-    // use "output_ids" so we need to preserve all outputs in that case (somewhat inefficiently)
+    // this indicates we are doing pooled embedding, so we ignore batch.logits and output all tokens
     bool embed_pooled = cparams.embeddings && cparams.pooling_type != LLAMA_POOLING_TYPE_NONE;
 
     // count outputs