diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py index 390cb0a22..67a65b473 100644 --- a/python/sglang/srt/model_executor/model_runner.py +++ b/python/sglang/srt/model_executor/model_runner.py @@ -78,7 +78,6 @@ from sglang.srt.layers.cp.utils import ( get_cp_strategy, ) from sglang.srt.layers.logits_processor import LogitsProcessorOutput -from sglang.srt.layers.moe.dwdp import DwdpManager from sglang.srt.layers.sampler import create_sampler from sglang.srt.layers.torchao_utils import apply_torchao_config_to_model from sglang.srt.layers.utils.cp_utils import is_mla_prefill_cp_enabled @@ -1027,6 +1026,8 @@ class ModelRunner: return if self.server_args.dwdp_size <= 1: return + from sglang.srt.layers.moe.dwdp import DwdpManager + manager = DwdpManager(self.server_args) set_global_dwdp_manager(manager) manager.setup(self.model)