Update tasks/audio.py

tasks/audio.py

@@ -228,13 +228,8 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     int8_model.qconfig = torch.quantization.get_default_qat_qconfig('fbgemm')  # Optional if not defined in saved model
     
     # Load the state dictionary
-    int8_model.load_state_dict(torch.load(quantized_model_path, weights_only=True))
+    int8_model.load_state_dict(torch.load(quantized_model_path, map_location=torch.device('cpu'), weights_only=True))
     int8_model.eval()  # Set to evaluation mode
-    
-    # Move model to device
-    device = 'cuda' if torch.cuda.is_available() else 'cpu'
-    int8_model.to(device)
-    
     #model.load_state_dict(torch.load("./best_blazeface_model_second.pth", map_location=torch.device('cpu'), weights_only=True))
 
     #model = torch.quantization.quantize_dynamic(model, {torch.nn.Linear}, dtype=torch.qint8)
@@ -250,7 +245,6 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     predictions = []
     with torch.inference_mode():
         for data, target in test_loader:
-            waveforms, labels = waveforms.to(device), labels.to(device)
             outputs = int8_model(waveforms)
             _, predicted = torch.max(outputs, 1)
             predictions.extend(predicted.tolist())