From cf711918e273e231d1155a80e0f33aff669ee187 Mon Sep 17 00:00:00 2001
From: Samuel Oliveira Alves
 <107287165+SamuelOliveirads@users.noreply.github.com>
Date: Thu, 13 Aug 2026 13:02:08 -0300
Subject: [PATCH] Synch DFlash Tokens ID (#2303)

* llama: synchronize DFlash argmax readback

* llama: synchronize DFlash token access
---
 src/llama.cpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/llama.cpp b/src/llama.cpp
index 4e477a55..7945d9db 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -11729,6 +11729,8 @@ float * llama_get_logits_ith(struct llama_context * ctx, int32_t i) {
 }
 
 llama_token llama_get_dflash_draft_token_ith(struct llama_context * ctx, int32_t i) {
+    llama_synchronize(ctx);
+
     if ((size_t) i >= ctx->dflash.draft_tokens.size()) {
         return LLAMA_TOKEN_NULL;
     }