PyTorch Inference on CPU with torch.no_grad()

Always wrap inference code in torch.no_grad() context manager. It disables gradient computation and reduces memory consumption, which is important when deploying models on edge devices like Raspberry Pi.

with torch.no_grad():
    output = model(input_tensor)