From 471e026327cca9f6f58aeefe32129a6cb9390f4f Mon Sep 17 00:00:00 2001
From: Aaron Lee <lee.aaron.65@gmail.com>
Date: Tue, 19 Aug 2025 23:10:56 -0400
Subject: [PATCH] fixed vram leak

---
 src/llama-context.cpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/llama-context.cpp b/src/llama-context.cpp
index a09ac6d447..62d7898b5f 100644
--- a/src/llama-context.cpp
+++ b/src/llama-context.cpp
@@ -3050,5 +3050,7 @@ void llama_build_and_execute_mtp_graph(struct llama_context * ctx,
     if (logits_mtp) {
         ctx->set_logits_ith(logits_mtp, sched, last_tok_idx);
     }
+
+    ggml_backend_sched_free(sched);
 }