DirectML

Discussions about research, Faceswapping and things that don't fit in the other categories here.


Locked
User avatar
MaxHunter
Posts: 193
Joined: Thu May 26, 2022 6:02 am
Has thanked: 177 times
Been thanked: 13 times

DirectML

Post by MaxHunter »

I noticed in the recent update DirectML was mentioned/implemented. Is this new? Will it speed up training if I switch to that platform?

User avatar
bryanlyon
Site Admin
Posts: 793
Joined: Fri Jul 12, 2019 12:49 am
Answers: 44
Location: San Francisco
Has thanked: 4 times
Been thanked: 218 times
Contact:

Re: DirectML

Post by bryanlyon »

DirectML is replacing PlaidML for AMD support on Windows. If you're using an Nvidia card we HIGHLY recommend you stick with Cuda and the Nvidia installation options (though DirectML should work for Nvidia cards, it will be slower and have more overhead).

DirectML will be completely replacing PlaidML support on Windows soon. The one potential issue with this for most users is that PlaidML trained models will not work with the DIrectML version of Faceswap. However, going forward, the DirectML and Nvidia versions of models will work interchangeably. Before PlaidML is completely retired from FaceSwap, Torzdf will make a "final" version containing PlaidML which will be archived for users to use old models but not supported going forward.

For Linux, the best AMD solution is ROCm which beats DirectML and PlaidML. This solution can be hard to implement however due to the compilation steps that ROCm requires.

TLDR:
If you're on Nvidia, keep using Faceswap the way that you have been.
If you're on AMD on Windows, we recommend upgrading to DirectML for all future model training.
If you're on AMD on Linux, we recommend migrating to ROCm for all future model training if you haven't already.

User avatar
MaxHunter
Posts: 193
Joined: Thu May 26, 2022 6:02 am
Has thanked: 177 times
Been thanked: 13 times

Re: DirectML

Post by MaxHunter »

Thanks, man. I will switch back to the master. 😉

Locked