Handling Big Models For Inference