From: makomk <redacted>
Date: Sat, 13 Jan 2024 14:16:11 +0000 (+0000)
Subject: server : fix crash with multimodal models without BOS token (#4904)
X-Git-Tag: upstream/0.0.4488~2637
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=ee8243adaa9a9f51ff449213383874e49efe368f;p=pkg%2Fggml%2Fsources%2Fllama.cpp

server : fix crash with multimodal models without BOS token (#4904)
---

diff --git a/examples/server/server.cpp b/examples/server/server.cpp
index c1ab8f9d..7b33aea1 100644
--- a/examples/server/server.cpp
+++ b/examples/server/server.cpp
@@ -1835,7 +1835,7 @@ struct llama_server_context
 
                     slot.cache_tokens = prompt_tokens;
 
-                    if (slot.n_past == slot.num_prompt_tokens)
+                    if (slot.n_past == slot.num_prompt_tokens && slot.n_past > 0)
                     {
                         // we have to evaluate at least 1 token to generate logits.
                         LOG_TEE("slot %d : we have to evaluate at least 1 token to generate logits\n", slot.id);