call_script() python-based API seems incomplete

Great project! The GPU memory management design is super cool!

I’m interested in using the pure PyTorch ops (no CUDA) path via call_script(). However, when I run __binning_script(), it crashes.

Are you currently working on the call_script() functions (or planning to soon)? Or is the Python-based API deprecated/out-of-date?

Thanks a lot.