[Bugfix] Add Fake VllmConfig on NPU and add slicing/tiling args in Qwen-Image (#145)

Signed-off-by: gcanlin <canlinguosdu@gmail.com>

examples/offline_inference/qwen_image/gradio_demo.py

@@ -5,7 +5,7 @@ import gradio as gr
 import torch
 
 from vllm_omni.entrypoints.omni import Omni
-from vllm_omni.utils.platform_utils import detect_device_type
+from vllm_omni.utils.platform_utils import detect_device_type, is_npu
 
 ASPECT_RATIOS: dict[str, tuple[int, int]] = {
     "1:1": (1328, 1328),
@@ -59,7 +59,14 @@ def parse_args() -> argparse.Namespace:
 
 @lru_cache(maxsize=1)
 def get_omni(model_name: str) -> Omni:
-    return Omni(model=model_name)
+    # Enable VAE memory optimizations on NPU
+    vae_use_slicing = is_npu()
+    vae_use_tiling = is_npu()
+    return Omni(
+        model=model_name,
+        vae_use_slicing=vae_use_slicing,
+        vae_use_tiling=vae_use_tiling,
+    )
 
 
 def build_demo(args: argparse.Namespace) -> gr.Blocks:

examples/offline_inference/qwen_image/text_to_image.py

@@ -7,7 +7,7 @@ from pathlib import Path
 import torch
 
 from vllm_omni.entrypoints.omni import Omni
-from vllm_omni.utils.platform_utils import detect_device_type
+from vllm_omni.utils.platform_utils import detect_device_type, is_npu
 
 
 def parse_args() -> argparse.Namespace:
@@ -49,7 +49,15 @@ def main():
     device = detect_device_type()
     generator = torch.Generator(device=device).manual_seed(args.seed)
 
-    omni = Omni(model=args.model)
+    # Enable VAE memory optimizations on NPU
+    vae_use_slicing = is_npu()
+    vae_use_tiling = is_npu()
+
+    omni = Omni(
+        model=args.model,
+        vae_use_slicing=vae_use_slicing,
+        vae_use_tiling=vae_use_tiling,
+    )
     images = omni.generate(
         args.prompt,
         height=args.height,