[diffusion] chore: remove default post-denoising dit offload in local mode (#15573)

python/sglang/multimodal_gen/runtime/pipelines_core/stages/denoising.py

@@ -754,20 +754,6 @@ class DenoisingStage(PipelineStage):
                 torch.mps.current_allocated_memory(),
             )
 
-        # In offline local mode (`sglang generate`), offload transformer weights to CPU
-        # after denoising to reduce peak VRAM during VAE decoding.
-        if current_platform.is_cuda_alike() and server_args.is_local_mode:
-            for model in (self.transformer, self.transformer_2):
-                if model is not None:
-                    model.to("cpu")
-            logger.info(
-                "Offloaded denoiser transformer weights to CPU after denoising to reduce peak VRAM during VAE decoding."
-            )
-            try:
-                torch.cuda.empty_cache()
-            except Exception:
-                pass
-
     def _preprocess_sp_latents(self, batch: Req, server_args: ServerArgs):
         """Shard latents for Sequence Parallelism if applicable."""
         if get_sp_world_size() <= 1: