Merge pull request #16 from audiohacking/copilot/fix-decoding-audio-error

Fix MPS dtype mismatch in audio decoding and app relaunch on close

Author: lmangani

backend/app/services/music_service.py

@@ -822,14 +822,16 @@ def _pad_audio_token(token):
         if callback is not None:
             callback(95, "Decoding audio...")
 
+        # Get codec dtype once for consistency
+        codec_dtype = getattr(pipeline, 'codec_dtype', torch.float32)
+        
         # Lazy codec loading: Load HeartCodec only when needed (for 12GB GPU mode)
         lazy_codec = getattr(pipeline, '_lazy_codec', False)
         if lazy_codec and pipeline._codec is None:
             print("[Lazy Loading] Loading HeartCodec for decoding...", flush=True)
             codec_path = getattr(pipeline, '_codec_path', None)
             if codec_path:
                 # Use the same dtype as specified in the pipeline for consistency
-                codec_dtype = getattr(pipeline, 'codec_dtype', torch.float32)
                 pipeline._codec = HeartCodec.from_pretrained(
                     codec_path,
                     device_map=pipeline.codec_device,
@@ -842,7 +844,8 @@ def _pad_audio_token(token):
             else:
                 raise RuntimeError("Cannot load HeartCodec: codec_path not available")
 
-        frames_for_codec = frames.to(pipeline.codec_device)
+        # Convert frames to codec device and dtype (important for MPS float16)
+        frames_for_codec = frames.to(device=pipeline.codec_device, dtype=codec_dtype)
         wav = pipeline.codec.detokenize(frames_for_codec)
 
         # Cleanup codec if using lazy loading (free VRAM for next generation)

launcher.py

@@ -210,17 +210,25 @@ def run_server():
         print("Warning: pywebview not available, falling back to browser")
         import webbrowser
         webbrowser.open("http://127.0.0.1:8000")
-        # Keep the server running
-        while True:
-            time.sleep(1)
+        # Keep the server running - wait for user to terminate
+        print("Server is running. Press Ctrl+C to stop.")
+        try:
+            while True:
+                time.sleep(1)
+        except KeyboardInterrupt:
+            print("\nStopping server...")
     except Exception as e:
         print(f"Error launching window: {e}")
         print("Falling back to browser...")
         import webbrowser
         webbrowser.open("http://127.0.0.1:8000")
-        # Keep the server running
-        while True:
-            time.sleep(1)
+        # Keep the server running - wait for user to terminate
+        print("Server is running. Press Ctrl+C to stop.")
+        try:
+            while True:
+                time.sleep(1)
+        except KeyboardInterrupt:
+            print("\nStopping server...")
 
 def main():
     """Main entry point."""