Hi thank you for the great work and congrats for the oral at NeurIPs.

I want to benchmark the inference speed as reported in the paper. However, the readme and the code structure are not very clear about how to reproduce the results. Can you provide some pointers?

Thank you.