Improve MPS handling and add VAE slicing/tiling

concept-maker_gui.py

@@ -1907,7 +1907,8 @@ class App(TkinterDnD.Tk):  # type: ignore
             raise RuntimeError(f"Model file not found: {model_path}")
         has_mps = bool(getattr(torch.backends, "mps", None) and torch.backends.mps.is_available())
         device = "cuda" if torch.cuda.is_available() else "mps" if has_mps else "cpu"
-        dtype = torch.float16 if device in {"cuda", "mps"} else torch.float32
+        # MPS often produces black outputs in float16; keep float32 there for stability
+        dtype = torch.float16 if device == "cuda" else torch.float32
         pipe = StableDiffusionXLPipeline.from_single_file(
             str(model_path),
             torch_dtype=dtype,
@@ -1924,6 +1925,11 @@ class App(TkinterDnD.Tk):  # type: ignore
             pipe.enable_attention_slicing()
         except Exception:
             pass
+        try:
+            pipe.enable_vae_slicing()
+            pipe.enable_vae_tiling()
+        except Exception:
+            pass
         try:
             pipe.set_progress_bar_config(disable=True)
         except Exception: