From 854e1295cca982e16478045a2b4b7386bc49491b Mon Sep 17 00:00:00 2001
From: Aleksandras Kostarevas <aleks076@protonmail.com>
Date: Tue, 28 Nov 2023 16:20:33 +0000
Subject: [PATCH] Fix problem with n_tokens==0

---
 native/jni/src/ggml/context.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/native/jni/src/ggml/context.cpp b/native/jni/src/ggml/context.cpp
index e2a94857c..44e8a706d 100644
--- a/native/jni/src/ggml/context.cpp
+++ b/native/jni/src/ggml/context.cpp
@@ -15,7 +15,7 @@ std::pair<token_sequence, token_sequence::size_type> transformer_context_fastfor
 
     // Handle the case when we have a shorter input than active context, requiring the last
     // token to be recomputed to get up-to-date logits
-    if((npast == next_context.size()) && (next_context.size() < ctx.active_context.size())) {
+    if((npast == next_context.size()) && (next_context.size() <= ctx.active_context.size())) {
         npast -= 1;
     }