Update app.py

app.py

@@ -17,18 +17,16 @@ def load_model_once():
     if model is None:
         print("Loading Dia model... This may take a few minutes.")
         try:
-            # Load model with correct parameters for Dia
+            # Load model without trying to move it manually to GPU
+            # The Dia model handles GPU placement internally
             model = Dia.from_pretrained("nari-labs/Dia-1.6B", compute_dtype="float32")
             
-            # Move model to GPU if available
+            print("Model loaded successfully!")
             if torch.cuda.is_available():
-                model = model.cuda()
-                print(f"Model loaded on GPU: {torch.cuda.get_device_name()}")
+                print(f"CUDA is available: {torch.cuda.get_device_name()}")
             else:
-                print("Model loaded on CPU")
+                print("CUDA is not available, using CPU")
                 
-            print("Model loaded successfully!")
-            
         except Exception as e:
             print(f"Error loading model: {e}")
             raise e
@@ -61,7 +59,7 @@ def generate_audio(text, seed=42):
                 
         print(f"Generating speech for: {text[:100]}...")
         
-        # Generate audio - disable torch compile for stability
+        # Generate audio - disable torch compile for T4 stability
         with torch.no_grad():
             audio_output = current_model.generate(
                 text,
@@ -95,7 +93,7 @@ def generate_audio(text, seed=42):
         print(error_msg)
         return None, error_msg
 
-# Create the Gradio interface - simplified to avoid OAuth triggers
+# Create the Gradio interface
 demo = gr.Blocks(title="Dia TTS Demo")
 
 with demo: