So, does this mean DAE is end of life as well?

Support
#1

I just got this from the Tensorflow developers…

@peardox We are checking to see if you still need help on this issue, as you are using an older version of tensorflow(1.x) which is officially considered as end of life. We recommend that you upgrade to 2.6 which is latest stable version of TF and let us know if the issue still persists in newer versions.Thanks!

As Tensorflow(1.x) is end of life DAE either to update to Tensorflow (2.x)

End of life in this context means its a dead version, much like Windows 7 and 8 (with Windows 10 being also slated to the great digital graveyard 25th Oct 2025)

#2

oh the Man cheebus I wish we could still use Win7. My favorite of them all. It was perfect, I thought. Win10 always seemed like a “Modern” skin for 7.

#3

Of course not. This has nothing todo with our desktop software.

How To: Install Deep Art Effects GPU
#4

It has everything to do with desktop

Cuda 9 supports 1xxx cards and lower
Cuda 10 supports 2xxx cards and lower
Cuda 11 supports 3xxx cards and lower

Tensorflow 1.x supports up to Cuda 10.0 (I’ve got a working version of that)
Tensorflow 2.x supports Cuda 10.1+

DAE uses Tensorflow to perform its AI stuff therefore the two are intertwined

DAE GPU as supplied (Cuda 9) will not work on any NVIDIA card released since the start of 2019

#5

In the future there will be more than one DAE version available. For older and newer OS systems.

1 Like
#6

I really hope so, that would be great. Peardox has been really helpful and knowledgeable on this subject and replied to me a few weeks ago with this detailed explanation of TF and CUDA version only supporting older cards… I bought a Radeon VII January 2020 to use this great software to then realise CUDA was needed. Finally got a 3080 2 months ago and was a bit gutted newer cards are not supported by DAE. Amazing software would be great to have a separate version for new cards. I purchased DAE and all the hardware upgrades to do 4k video within DAE using the GPU… newer cards support would be fantastic! :slight_smile:

#7

Have you tried the CPU version, just for the devil’s due?

#8

Yes that’s all I can use - the CPU version which is great don’t get me wrong! but only a tiny fraction of what could be done efficiently using the GPU version. If I want to experiment with any media it has to be a single image which is only a few minutes in basic mode. I did intend to use this great software for 4k drone footage and other high quality video footage… I am optimistic and hopeful the developers will support newer cards from the last 3 years, fingers crossed! If not my DAE and 3080 purchase will be in vain haha. However I’m hopeful they will shine thought with newer card support and newer features such as the upgrade to styleGAN2 that nvidia showed off 6 months back… with Non Texture Sticking! That would be Nice… Its good to Dream :smile: You never know.

#9

You could always give a “Russian PC” a go… At least it works with the GPU.

There are a few caveats…

  1. It cost money (15 EU for 25 hours)
  2. The GPU is 1/4 of the actual physical GPU - this is about the same as a 1060 with 6Gb
  3. You need my version of Tensorflow (I built it on a Russian PC) - no problem making that available…

On the bright side there’s a 1 hour free trial so you could set up DAE and CUDA and still have enough time to run some tests. There’s also a referral program that doubles your first credit so you’d get 50 hours for the 15 Euro-bucks (and I get 25 hours )

If someone wants to try this out I’ll make a YouTube Video of the entire process (new laptop - I can finally make proper videos again).

Another “Deal” they have is that you can have “Unlimited” usage from Midnight to Noon for 749 Roubles (GBP 7.49 by chance of the exchange rate when I tried it) - this is what I used to actually test everything and build the CUDA 10.0 version of Tensorflow so I’ve still got just short of 23 hours left on my credit.

One thing I did find, even using the night-time option, was that I needed some daytime credit as well as I would find myself needing something I’d left running while I had a kip.

They appear to have changed how their Cloud PCs look to non-Russians - I just briefly went in and it was all English (plus they have a language switching icon on the desktop)

While there are other cloud systems available they all cost more with no real benefits. Most of the cloud PC services are aimed at gamers and have no ability to load anything other than games from Steam (and sometimes Epic). The only other offering worth mentioning is Shadow which is 30 USD PCM and has a waiting list (and some bad feedback over still waiting - mind you, the Russians don’t have any so no comparison)

It won’t be a 3080 but it’ll work and the cost is reasonable.