From 59298ee29e2f8368d02bd904081e66a27611dc7d Mon Sep 17 00:00:00 2001 From: ddchenhao66 Date: Thu, 30 Oct 2025 08:48:15 +0000 Subject: [PATCH] [XPU] xpu currently disable prefix cache for VL model --- fastdeploy/engine/args_utils.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/fastdeploy/engine/args_utils.py b/fastdeploy/engine/args_utils.py index 9112efa7d30..6e5f52225bb 100644 --- a/fastdeploy/engine/args_utils.py +++ b/fastdeploy/engine/args_utils.py @@ -1078,6 +1078,10 @@ def create_engine_config(self, port_availability_check=True) -> FDConfig: all_dict = asdict(self) model_cfg = ModelConfig(all_dict) + # XPU currently disable prefix cache for VL model + if current_platform.is_xpu() and (self.enable_mm or model_cfg.enable_mm): + self.enable_prefix_caching = False + if not model_cfg.is_unified_ckpt and hasattr(model_cfg, "tensor_parallel_size"): self.tensor_parallel_size = model_cfg.tensor_parallel_size