Faceswap Forum

Thanks for this post! This is exactly the kind of information and research that I would like to see from users.

I look forward to seeing how your model gets on

MaxHunter wrote: ↑Thu Aug 11, 2022 1:40 am
Wow!! I went this entire time without knowing there were presets?!!! Is this in the tutorial?

Yes. Linked from the very first post
viewtopic.php?p=5367#p5367

Ok, this is a system issue, which I cannot directly solve. However, there appears to be an SSL issue on your machine. A bit of googling comes up with some potential solutions: https://www.google.com/search?q=git+error+openssl+ssl_read+connection+was+reset+errno+10054&client=ubuntu&hs=NQ7&...

I couldn't tell you batch size, just based on settings. See here for finding the best batch size yourself:

viewtopic.php?p=388

Your use is very edge-case, as I have not seen it raised before.

Short answer: no, and not something I'm likely to add, I'm afraid.

Corrupted model.

Lower learning rate + rollback 50k

That leaves me with a jumbled folder of source faces from which I remove unwanted faces for training and alignments files that exist but which are mostly from the original extract (extract n= usually 15-40 depending on fps, length and quality of video) and which would be difficult to tie back to sp...

It's been a long time since I used DLight, but this is not an issue I have seen before.

Generally that model needs a lower LR (₃.5e-5) as it tends to collapse otherwise.

That may solve your issue, it may not.

This is a connection issue with your setup/conda.

Do this and try again. Hopefully it will resolve any Conda related issues:
app.php/faqpage#f1r1

The easiest way to see it is visually. Look at the "red" area in the training previews. These are masked out. Features are badly learned in the masked out area (this is the point of masking without 'learn mask enabled', it tells the NN which parts of the image it is interested in, and whic...

Thank you for your advise , Would you please suggest some setting about 1080p video for using V2 with rtx 3080 n 3090 ? because for now , v2-s with stojo present , 1080p video , some close-up shot the face go blur , but v2-l with stojo present much better . suggestion like : extraction px phaze-A o...

Ok, good to know. Thanks. I will look at the potential for choosing to put that on cpu.

Looks like model corruption to me.

You'd need to try rolling back 50k or so, lowering learning rate a bit and continuing.

Just an update. I'm noticing that there is a much longer delay between iterations compared with, for example, the 256 StoJo model. The GPU is not used to its full potential. For example, this is a graph for I believe the slowdown is caused by the larger amount of data (larger image sizes - which ar...

Removing the bug tag from this as -d, --distributed is deprecated in favour of -D, --distribution-strategy

No. It is purely a numbers thing. I doubt I would implement anything like this, because if you're at the point of reviewing every single frame, then you might as well review every single face. At which point, you're manually going through everything anyway, so could build your training set that way.

Which mask were you using? If landmarks based mask, then the GPU isn't used at all. The mask is generated from landmarks.

If it was an NN based mask, it would seem unlikely. However, let me know which one, because if it is as fast on CPU, then that is something I would want to look at.

Yes, there are huge benefits for multiple reasons. Off the top of my head: 1) Landmarks based masks are only as good as the landmarks. It really requires very precise landmark alignments for the mask to be effective. For face alignment, only the core landmarks are used (all landmarks, excluding the ...

Gradient clipping is a mechanism to help prevent exploding/vanishing gradients (that is numbers that go to +/- infinity or to 0). Both of these will cause a model to NaN (Mixed Precision is more prone to this, as infinity in limited precision space is a smaller number than infinity in full precision...

Protip: Don't report the forum owner for spam.

It is in the convert guide. Literally covered in the first section. Also this:
viewtopic.php?t=2090

Faceswap Forum

Search found 1847 matches

Re: [Guide] Introducing - Phaze-A

Re: [Guide] Introducing - Phaze-A

Re: help me!“Some applications failed to install.Process Aborted.Check Details.”

Re: [Guide] Introducing - Phaze-A

Re: Possible to Delete Source Pictures / Clean Alignments File?

Re: empty face

Re: Dlight worked to about 6000 its, then flatlined - addt'l questions

Re: Dlight worked to about 6000 its, then flatlined

Re: Windows Installer error: “Some applications failed to install.Process Aborted.Check Details.”

Re: Clarificaton on Training options: Mask selection

Re: [Guide] Introducing - Phaze-A

Re: Masking with CPU

Re: wired highlight patches

Re: Potential VRAM Saving techniques

Re: Bug: ValueError: Please use `tf.keras.losses.Reduction.SUM` or `tf.keras.losses.Reduction.NONE`

Re: saving individual frames from manual

Re: Masking with CPU

Re: Clarificaton on Training options: Mask selection

Re: AutoClip: Any Feedback From Users?

Re: How do I change the face of only one of two people in one picture?