Search found 790 matches

by bryanlyon
Mon Mar 11, 2024 9:27 pm
Forum: General Discussion
Topic: Will installing new Python version mess with app?
Replies: 1
Views: 102

Re: Will installing new Python version mess with app?

If you installed Faceswap properly (using the installer or proper guide) then installing other Python instances shouldn't do anything to FaceSwap. However, since you've already got Conda installed with Faceswap I suggest that you install that other application in a proper conda environment. See the ...
by bryanlyon
Mon Feb 12, 2024 11:52 pm
Forum: Training Discussion
Topic: A Short Hello with some questions
Replies: 3
Views: 1727

Re: A Short Hello with some questions

Hello and Welcome. * if one has a video of oneself and wants to put their face onto an "original" where the person is wearing glasses, does the software recognize it? Or does it pixelate the object (the glasses), or even cause image distortions? As long as a decent chunk of the training da...
by bryanlyon
Fri Jan 05, 2024 8:25 pm
Forum: Installation Support
Topic: GUI Glitching
Topic: GUI Glitching
Replies: 5
Views: 2735

Re: GUI Glitching

This actually is not a problem with FaceSwap but instead originates on your system. I'm assuming it's something due to drivers or the specific setup of Kali, WSL, Windows and your local system. Since there are so many variables there I wouldn't even know where to begin to find the cause. This is why...
by bryanlyon
Thu Nov 30, 2023 10:12 pm
Forum: Installation Support
Topic: FaceSwap Linux running, but suddenly stops "Proces Exited" with no Errors
Replies: 4
Views: 4511

Re: FaceSwap Linux running, but suddenly stops "Proces Exited" with no Errors

"Violation of the segment ('core' generated)" This implies that something is failing to get a memory allocation it's seeking. It's also clearly outside of python and in the realm of compiled code. Probably high up in a library. Unfortunately without more information the best we can do is g...
by bryanlyon
Mon Nov 20, 2023 4:49 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 83
Views: 130726

Re: Hardware best practices

The only 30xx card with the nvlink fingers is the 3090. All others cannot do any sort of "linking". That said, you can still use multiple GPUs in Faceswap. It is diminishing returns where each additional GPU slows down the collective more, and like torzdf says, only the batch size can be i...
by bryanlyon
Mon Nov 13, 2023 11:26 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 83
Views: 130726

Re: Hardware best practices

The trouble with hardware recommendations since the 20xx series has been that you always have to add "if you can get it for a good price" to the end of ANY recommendation. Will a 4060 TI with 16gb of vram work? Definitely. Will it beat a $100 1080? Yes. Is it worthwhile to buy it for FaceS...
by bryanlyon
Tue Oct 31, 2023 11:52 pm
Forum: Training Discussion
Topic: Location of “forward pass” and/or Where to implement a Custom Training Loop
Replies: 6
Views: 2627

Re: Location of “forward pass” and/or Where to implement a Custom Training Loop

You don't have to hook into the passes at all to add your own architecture. You can add new models by following the plugin architecture here: https://github.com/deepfakes/faceswap/tree/master/plugins/train/model or add a new training paradigm following the plugins here: https://github.com/deepfakes/...
by bryanlyon
Tue Oct 17, 2023 11:29 pm
Forum: Training Support
Topic: Tensorflow error happens randomly while training
Replies: 1
Views: 2261

Re: Tensorflow error happens randomly while training

As the previous message said, this is a driver issue. Nothing to do with Faceswap and nothing we can do to prevent or work around it. As far as I've been able to find on it though, it's almost 100% an overclocking issue. Chances are the card you have is factory overclocked and it's hitting instabili...
by bryanlyon
Tue Sep 26, 2023 9:18 pm
Forum: Installation Support
Topic: GPU not working (RTX3070 mobile)
Replies: 3
Views: 2955

Re: GPU not working (RTX3070 mobile)

CPU usage being high, especially on extraction, is very normal. How did you measure the GPU usage? If you're using the Task Manager it's useless (see that FAQ item). The best way to know if you're using GPU is the speed. If it's extracting at > 10 seconds per frame you're probably on CPU otherwise I...
by bryanlyon
Mon Sep 25, 2023 10:30 pm
Forum: Installation Support
Topic: GPU not working (RTX3070 mobile)
Replies: 3
Views: 2955

Re: GPU not working (RTX3070 mobile)

What makes you think that it's not using the GPU? Most often it is working on the GPU just fine but you're checking the wrong place. See: https://forum.faceswap.dev/app.php/faqpage#f0r3

by bryanlyon
Mon Aug 28, 2023 3:55 pm
Forum: Hardware
Topic: AMD 7000 Series GPU's
Replies: 2
Views: 3624

Re: AMD 7000 Series GPU's

The AI accelerators in the 7000 series DO support FP16. However, we know almost nothing about their use. ROCm is supposed to support them, but we wont know how much it helps until we see them released. DirectML does not yet support them at all, and it's an open question if/when DirectML will add sup...
by bryanlyon
Fri Aug 11, 2023 4:25 pm
Forum: General Discussion
Topic: A question for someone who is from The fruity cargo cult
Replies: 3
Views: 4468

Re: A question for someone who is from The fruity cargo cult

Apple Silicon is impressive and while I'm not a fan of their ecosystem, I must admit to being intrigued by their M2 laptops. That said, by the benchmarks it's about 6-10x slower than a decent Nvidia GPU. I wouldn't buy it if you want a system to train on all the time, it just isn't the economical op...
by bryanlyon
Wed Jun 21, 2023 4:50 pm
Forum: Extract Support
Topic: OSError: Unable to open file (truncated file: eof = 71163904, sblock->base_addr = 0, stored_eof = 90000392)
Replies: 3
Views: 3556

Re: OSError: Unable to open file (truncated file: eof = 71163904, sblock->base_addr = 0, stored_eof = 90000392)

The path is there, you might have to enable hidden files to see it as files that start with a "." are typically hidden on *nix based systems. The command line is able to go to the folder as well.

by bryanlyon
Thu Jun 15, 2023 4:03 pm
Forum: Training Discussion
Topic: No Warp question
Replies: 3
Views: 3342

Re: Training question

I suggest turning it off a bit before fit training. It's important to note that warping is partially to avoid overtraining the model, so shouldn't be turned off TOO early. Turning it off only during fit training is probably not the best option either since fit training deliberately starves it for da...
by bryanlyon
Thu Jun 15, 2023 3:56 pm
Forum: Convert Discussion
Topic: Just an idea for getting better resolution.
Replies: 2
Views: 3264

Re: Just an idea for getting better resolution.

Sadly, no. This technique works fine for non-generative tasks such as color correction and things like that. But slight differences in the swap output will not line up properly when applied in a dithered fashion. This leads to inconsistent swap results.

by bryanlyon
Wed Jun 14, 2023 3:39 pm
Forum: General Discussion
Topic: Can I generate a video from another driving video?
Replies: 1
Views: 2443

Re: Can I generate a video from another driving video?

Not with FaceSwap. FaceSwap only swaps faces. You'd have to use another project like First Order Model for something like what you're asking. (Note that by default they don't have a model trained on doing close up whole body so you'd have to train your own.)

by bryanlyon
Fri May 26, 2023 10:26 pm
Forum: Training Support
Topic: Made it about 15000 its in, then crashing every few hundred or so. File in use error?
Replies: 9
Views: 4153

Re: Made it about 15000 its in, then crashing every few hundred or so. File in use error?

Probably caused by something reading the image at the same time we try to write it. Maybe thumbnailer?

by bryanlyon
Fri May 19, 2023 6:25 pm
Forum: Training Discussion
Topic: Global encoders?
Replies: 3
Views: 1474

Re: Global encoders?

There are datasets with over a million faces that have been used to train the newest model architectures to encode faces. They've been very good at starting training, but have never been able to beat a 1:1 trained encoder. Is it theoretically possible for a generic encoder? Almost definitely. Has it...
by bryanlyon
Fri May 19, 2023 5:59 pm
Forum: Training Discussion
Topic: Global encoders?
Replies: 3
Views: 1474

Re: Global encoders?

In theory this would be possible, in practice, only sort-of. The problem is that the encoder will eventually need to focus on the specific faces that you're training to get the best results. That said, we've enabled the ability to copy encoders and even have pre-trained encoders available with Phaze...
by bryanlyon
Tue May 16, 2023 6:14 pm
Forum: Extract Support
Topic: WARNING The AMD backend has been deprecated
Replies: 7
Views: 3573

Re: WARNING The AMD backend has been deprecated

It's important to note that if you have any models that you have started training or want to use, those will NOT work in the DirectML version. You'll have to start again if you do that. Existing extracts and data will work fine though. This is why AMD was kept available but deprecated. AMD trained m...