Fix: use s_scale=0 when IP-Adapter loaded but no face requested

When IP-Adapter FaceID is initialized, it modifies the pipeline's UNet
cross-attention layers. Calling raw pipeline() without face embeddings
leaves these layers in a broken state, causing corrupted output.

Solution: When IP-Adapter is loaded but no face_image provided, call
ip_model.generate() with s_scale=0.0 and zero embeddings to properly
disable face conditioning while satisfying the modified layers.

requirements.txt

@@ -28,6 +28,7 @@ insightface==0.7.3
 
 # IP-Adapter from GitHub
 git+https://github.com/tencent-ailab/IP-Adapter.git
+einops  # Required by IP-Adapter
 
 # Utilities
 pydantic==2.6.0