Update README: GPU/CPU auto-detection support

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

README.md

@@ -14,7 +14,7 @@ license: mit
 
 Detect anime faces and 28 facial landmarks using YOLOv3 + HRNetV2.
 
-This Space uses **Zero GPU** for fast inference without constant GPU allocation costs.
+Supports both **GPU** (Zero GPU) and **CPU** inference. Auto-detects available hardware.
 
 ## Features