From acec1b1d84c428ced19d53935979107eeab922a2 Mon Sep 17 00:00:00 2001 From: Guancheng Fu <110874468+gc-fu@users.noreply.github.com> Date: Sat, 10 Oct 2026 12:27:35 +0800 Subject: [PATCH] [Bugfix][XPU] Remove stale W4A8 weight after repacking (#57511) Signed-off-by: Guancheng Fu Co-authored-by: Kunshang Ji --- vllm/model_executor/kernels/linear/mixed_precision/xpu.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/kernels/linear/mixed_precision/xpu.py b/vllm/model_executor/kernels/linear/mixed_precision/xpu.py index c6545f312a..c82e862e58 100644 --- a/vllm/model_executor/kernels/linear/mixed_precision/xpu.py +++ b/vllm/model_executor/kernels/linear/mixed_precision/xpu.py @@ -186,7 +186,7 @@ class XPUW4A8IntLinearKernel(MPLinearKernel): # Free the original unpacked int8 weight (still registered as "weight") # to avoid double-storing both int8 [N, K] and int32 [N, K/8] in memory. - layer.register_parameter("weight", None) + delattr(layer, "weight") def apply_weights( self,