Dave Rumph

Yes. I just updated to macOS Monterey 12.3 and cormeltools 5.2, and it crashed in essentially the same way as before when I enabled use of multiprocessing.

In my scenario there are two models involved, and two instances trained differently of one of the models. I haven't tried recently to simplify down to a minimal "reproduce the...

It looks like I already constructed a pretty minimal model. (Sorry, I last worked on this six months ago.). The example code is setting up a processing pipeline for video...

Also, I rebuilt the detector.mlmodel file from detector.pb using coremltools 5.2, and some of the symbol names seem to have changed. I needed to change "Multibox/Sigmoid" to "Multibox_Sigmoid" on detector_machinery.py...

I'll put together a more simplified version this morning.

Here is a much-simplified version: https://caltech.box.com/s/3wcgpzd7uqxauvg50wp3tz09cgcdvpmo It depends only on multiprocessing and numpy (and coremltools, of course). You'll still need to have "detector.mlmodel" in your filesystem. Run as follows: ```...