From 096d87d335e433a6994124987e76ca37ea0545b4 Mon Sep 17 00:00:00 2001 From: kevin Date: Fri, 31 Oct 2025 14:59:18 +0800 Subject: [PATCH] fix bug (#4679) --- fastdeploy/config.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/fastdeploy/config.py b/fastdeploy/config.py index 21af68219..6d550dbac 100644 --- a/fastdeploy/config.py +++ b/fastdeploy/config.py @@ -1556,6 +1556,10 @@ class FDConfig: self.graph_opt_config.use_cudagraph = False logger.info(f"CUDAGraph only support on GPU, current device type is {self.device_config.device_type}!") + if self.model_config.enable_mm and self.graph_opt_config.use_cudagraph: + self.cache_config.enable_prefix_caching = False + logger.info("Multi-modal models do not support prefix caching when using CUDAGraph!") + if self.scheduler_config.splitwise_role == "mixed": self.model_config.moe_phase = MoEPhase(phase="prefill") elif self.scheduler_config.splitwise_role == "prefill":