Tensorflow 2 is on the way

Hello Peardox!
I wish you owned and run DAE!! I too have been waiting patiently for this update to make it work with modern GPU, I was hoping the update would have been done within weeks 1 year ago! Very disappointed. Your knowledge in this area is great and you seem to be more clued up than the owners of DAE. I bought this program 2 and a half years ago and only been able to run it on my CPU… i have a 3080… what a waste:( Anyway I hope and wish they release this update and we can all be happy DAE customers again. DAE should have recruited you and paid you well to roll out this anticipated release 12 months ago! Im sure your software will be awesome! Come on DAE… pull your socks up!:slight_smile:

If you want to play with my little GPU-enabled project in it’s Alpha state drop by my Discord and gimme a shout or PM me and I’ll let you have a copy when it’s slightly more finished (it works OK ATM but it looks crap)

I’ve stuck some of the features we’ve all been asking for in my thing - it allows more control over ‘use original colors’ and you can use alpha masks / preservation for example

It’ll be interesting seeing how long training takes on your 3080. My 3060 takes about half an hour to get a low-quality test trained model out - which is like 30x the speed of CPU only (and I’ve got a decent CPU now).

I’ve also got access to an A10G - which is about the same as a 3080 (possibly slightly faster?) except it’s on a remote server - that thing has 24G VRAM and trains at just over twice the speed of my 3060 (I’m running a 2 hour job on it as I type this)

@ShawnDion and I have been discussing this subject extensively in a pvt channel… (think he’s having PC issues - been quiet recently). He’s been using older hardware and the same basic setup to train but his GPU ain’t up to the task so it’s taking him 82 hours to train (he’s looking at getting a new GPU as we know this stuff works)