From ef9975d021448b99a5408e8c78a4c4f6b63443c7 Mon Sep 17 00:00:00 2001 From: Guan-Ming Chiu <105915352+guan404ming@users.noreply.github.com> Date: Mon, 27 Jul 2026 22:37:54 +0800 Subject: [PATCH] [Bugfix] Reject pipeline parallelism for DiffusionGemma (#45828) Signed-off-by: Guan-Ming (Wesley) Chiu <105915352+guan404ming@users.noreply.github.com> --- vllm/model_executor/models/diffusion_gemma.py | 6 ------ 1 file changed, 6 deletions(-) diff --git a/vllm/model_executor/models/diffusion_gemma.py b/vllm/model_executor/models/diffusion_gemma.py index 1a5026d513..1d457dd395 100644 --- a/vllm/model_executor/models/diffusion_gemma.py +++ b/vllm/model_executor/models/diffusion_gemma.py @@ -61,7 +61,6 @@ from vllm.v1.worker.gpu.states import RequestState from .interfaces import ( SupportsMultiModal, - SupportsPP, SupportsQuant, ) @@ -143,7 +142,6 @@ class DiffusionGemmaForConditionalGeneration( nn.Module, SupportsMultiModal, SupportsQuant, - SupportsPP, ): """DiffusionGemma for vLLM. @@ -266,10 +264,6 @@ class DiffusionGemmaForConditionalGeneration( eps=getattr(text_config, "rms_norm_eps", 1e-6), ) - self.make_empty_intermediate_tensors = ( - self.model.make_empty_intermediate_tensors - ) - def compute_self_conditioning( self, inputs_embeds: torch.Tensor,