The last time we discussed this, we thought about dropping it when either of the two happens:
We see < 10% Python 2.7 downloads
The pain threshold for maintaining 2.7 support is so great that PyTorch as a product loses significant development time
Neither of them have happened, or are expected to happen by June of next year as far as I can tell.
My answer, as you see, is not giving you a commitment, but giving you the thought process of the team.
We haven’t committed to an LTS release at this point (we aren’t even 1.0 stable until December this year)
Thank you the fair team for your work, Pytorch is a pleasure to use.
I have a few questions:
If I correctly understand, we should expect a stable 1.0 version for december?
If I start a project using the release candidate version, will the migration to the forthcoming stable version represents a lot of work?