From 803b9d7881cd3a8482aaa1e6bf990193b55c6331 Mon Sep 17 00:00:00 2001 From: Wei Zhao <51183510+wzhao18@users.noreply.github.com> Date: Wed, 29 Apr 2026 00:08:16 -0400 Subject: [PATCH] [Bugfix] Fix Deepseek V4 import error due to AOT compile cache loading (#41090) Signed-off-by: wzhao18 Signed-off-by: Wei Zhao <51183510+wzhao18@users.noreply.github.com> --- vllm/model_executor/models/deepseek_v4.py | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/vllm/model_executor/models/deepseek_v4.py b/vllm/model_executor/models/deepseek_v4.py index d6edf0789f..d41a8b666d 100644 --- a/vllm/model_executor/models/deepseek_v4.py +++ b/vllm/model_executor/models/deepseek_v4.py @@ -1097,6 +1097,11 @@ class DeepseekV4DecoderLayer(nn.Module): aux_stream_list: list[torch.cuda.Stream] | None = None, ): super().__init__() + + # Lazy import to avoid top-level tilelang dependency. + # Registers both torch.ops.vllm.mhc_pre and mhc_post + import vllm.model_executor.layers.mhc # noqa: F401 + config = vllm_config.model_config.hf_config self.hidden_size = config.hidden_size @@ -1167,11 +1172,6 @@ class DeepseekV4DecoderLayer(nn.Module): hc_scale: torch.Tensor, hc_base: torch.Tensor, ): - # Lazy import to avoid top-level tilelang dependency. - # Registers both torch.ops.vllm.mhc_pre and mhc_post, - # so hc_post() doesn't need its own import. - import vllm.model_executor.layers.mhc # noqa: F401 - post_mix, res_mix, layer_input = torch.ops.vllm.mhc_pre( residual=x, fn=hc_fn,