From 008de19e16d94f5553fa3ca823e53f4c49e1c765 Mon Sep 17 00:00:00 2001
From: SCDESPERTATE <74419971+SCDESPERTATE@users.noreply.github.com>
Date: Fri, 12 Dec 2025 11:42:33 +0800
Subject: [PATCH] [fix](kt-kernel): drop the weights held in Python for loading
 weights operation in C++ (#1695)

---
 kt-kernel/python/utils/llamafile.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kt-kernel/python/utils/llamafile.py b/kt-kernel/python/utils/llamafile.py
index f39704b..68dce64 100644
--- a/kt-kernel/python/utils/llamafile.py
+++ b/kt-kernel/python/utils/llamafile.py
@@ -217,3 +217,6 @@ class LlamafileMoEWrapper(BaseMoEWrapper):
         # Load weights
         self.cpu_infer.submit(self.moe.load_weights_task(physical_to_logical_map_cpu.data_ptr()))
         self.cpu_infer.sync()
+
+        # Drop original weights after loading
+        self.weights_to_keep = None
\ No newline at end of file