[Bugfix] multimodal_gen(hunyuan3d): honor config precisions for delight/paint

python/sglang/multimodal_gen/runtime/pipelines_core/stages/hunyuan3d_paint.py

@@ -33,6 +33,7 @@
 )
 from sglang.multimodal_gen.runtime.server_args import ServerArgs
 from sglang.multimodal_gen.runtime.utils.logging_utils import init_logger
+from sglang.multimodal_gen.utils import PRECISION_TO_TYPE
 
 logger = init_logger(__name__)
 
@@ -300,16 +301,26 @@ def _load_delight_model(self, server_args: ServerArgs):
                 local_path = None
 
         if local_path and os.path.exists(local_path):
+            # Resolve precision from config with a simple fallback for CPU/MPS
+            dit_dtype = PRECISION_TO_TYPE.get(
+                getattr(self.config, "dit_precision", "fp16"), torch.float16
+            )
+            if self.device.type in ("cpu", "mps") and dit_dtype in (
+                torch.float16,
+                torch.bfloat16,
+            ):
+                # Avoid half/bfloat on CPU/MPS to be safe
+                dit_dtype = torch.float32
             pipeline = StableDiffusionInstructPix2PixPipeline.from_pretrained(
                 local_path,
-                torch_dtype=torch.float16,
+                torch_dtype=dit_dtype,
                 safety_checker=None,
             )
             pipeline.scheduler = EulerAncestralDiscreteScheduler.from_config(
                 pipeline.scheduler.config
             )
             pipeline.set_progress_bar_config(disable=True)
-            self._delight_pipeline = pipeline.to(self.device, torch.float16)
+            self._delight_pipeline = pipeline.to(self.device, dit_dtype)
             logger.info("Delight model loaded successfully")
         else:
             logger.warning(
@@ -347,6 +358,7 @@ def _run_delight(self, image):
 
         image = self._delight_pipeline(
             prompt=self.config.delight_prompt,
+            negative_prompt=getattr(self.config, "delight_negative_prompt", ""),
             image=image,
             generator=torch.manual_seed(42),
             height=512,
@@ -559,10 +571,28 @@ def _do_load_paint(self, server_args: ServerArgs) -> None:
         else:
             raise FileNotFoundError(f"No VAE weights in {vae_dir}")
         self.vae.load_state_dict(state_dict)
-        self.vae = self.vae.to(device=self.device, dtype=torch.float16).eval()
+        # Resolve VAE/DiT dtypes from config with simple CPU/MPS fallback
+        vae_dtype = PRECISION_TO_TYPE.get(
+            getattr(self.config, "vae_precision", "fp32"), torch.float32
+        )
+        if self.device.type in ("cpu", "mps") and vae_dtype in (
+            torch.float16,
+            torch.bfloat16,
+        ):
+            vae_dtype = torch.float32
+        dit_dtype = PRECISION_TO_TYPE.get(
+            getattr(self.config, "dit_precision", "fp16"), torch.float16
+        )
+        if self.device.type in ("cpu", "mps") and dit_dtype in (
+            torch.float16,
+            torch.bfloat16,
+        ):
+            dit_dtype = torch.float32
+
+        self.vae = self.vae.to(device=self.device, dtype=vae_dtype).eval()
         self.transformer = UNet2p5DConditionModel.from_pretrained(
             os.path.join(local_path, "unet"),
-            torch_dtype=torch.float16,
+            torch_dtype=dit_dtype,
         ).to(self.device)
         self.is_turbo = bool(getattr(self.config, "paint_turbo_mode", False))
         sched_path = os.path.join(local_path, "scheduler", "scheduler_config.json")