Enable 16bit tensor inference

[dan64]

Hi Jantic,

I'd like to know if there is any plan to enable 16bit tensor inference in Deoldify.
This update should speed-up the coloring process significantly.

Thanks,
Dan

P.S.
I'd like to inform you that has been released the version 2.0 of filter vs-deoldify

[jantic]

I'd like to inform you that has been released the version 2.0 of filter vs-deoldify

Very cool!

re: 16bit inference: I don't plan on modifying the repo going forward other than maintenance things like broken dependencies. I've moved on to other things. That being said, it should be pretty easy to modify the code to do this in a fork with the code as is and it should probably just work based on what I've done elsewhere.