Pierre

This repo is mirror-only - nothing will be accepted here for legitimate reasons. Please see: https://github.com/torvalds/linux/pull/17#issuecomment-5654674 > I don't do github pull requests. > > github throws away all the...

It might be worth looking into the pytorch variable allocation, which can have a big impact on required GPU memory[[1](https://discuss.pytorch.org/t/cuda-out-of-memory/449/2)]. But I'm not really seeing any major issues in this...

This does not occur on linux (no CUDA).

> One idea is to use the included SQLite database instead of a filesystem. That's actually what I had in mind. I've written a filesystem that's backed by SQLite (/comdb2)...

Woooh, awesome. Thanks for implementing this :)

@chrisgannon or @eeropic can either one of you please detail the workflow you've used to create and import these shapes? A short bullet point summary with code samples would be...

@chrisgannon Thx, was able to adapt that technique to vectors created with https://vectr.com.

@chrisgannon I chose vectr because its one of the few free online vector tools. I usually work in Inkscape locally. I tested your tool with a few shapes copied from...

@charlesr1971 Related: #6 > Right now there's no easy way to access a child of a parent copied with copyGraph.

Wrong person @charlesr1971