Optimize model device placement by moving to load_model function

[Copilot]

Summary

Fixed inefficient model device placement in the Gradio inference application. The model is now moved to the device once during loading instead of repeatedly on every prediction call.

Problem

Previously, the model was being moved to the device on every prediction:

def predict_image(image: np.ndarray, model_type: str) -> tuple[dict, np.ndarray]:
    model = load_model(model_type)
    model.to(device)  # ❌ Inefficient: device transfer on every prediction
    # ... rest of prediction logic

This meant that every time a user requested a prediction, the model would be transferred to the device (CPU or GPU), even though it was already on the correct device from the previous prediction. This added unnecessary overhead to each inference call.

Solution

Moved the model.to(device) call from predict_image() to load_model():

def load_model(model_type: str = "efficientvit") -> nn.Module:
    # ... model creation and weight loading
    model.to(device)  # ✅ Efficient: device transfer happens once
    model.eval()
    return model

def predict_image(image: np.ndarray, model_type: str) -> tuple[dict, np.ndarray]:
    model = load_model(model_type)
    # model.to(device) removed - no longer needed
    # ... rest of prediction logic

Impact

• Performance: Eliminates redundant device transfers on every prediction call
• Code Quality: Device placement now happens in the logical place (during model loading)
• Minimal Changes: Only 2 lines modified, ensuring surgical precision

Fixes issue raised in PR #1

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Moving the model to device on every prediction call is inefficient. The model should be moved to device once during loading in the load_model function, not repeatedly in predict_image.</issue_title>
<issue_description>Moving the model to device on every prediction call is inefficient. The model should be moved to device once during loading in the load_model function, not repeatedly in predict_image.

Originally posted by @copilot in #1 (comment)
</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #2

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.