That's great! I'm going to mark this as solved and lock this thread.
Feel free to open another one with any other issues.
That's great! I'm going to mark this as solved and lock this thread.
Feel free to open another one with any other issues.
Yes. You've run out of system RAM. How many images are you processing?
If you check the sort section of the extraction guide you will see specific mention of this.
There appears to be an issue with the latest Tensorflow and S3FD on some setups,
It will still work if you run it with "single-process" enabled
Could you point me to where it says this? That is outdated information.
It is highly advised that you create your own image sets. For one, it will help you familiarize with the process. For another, you will need the alignments file for training/converting.
Glad you got it working
Ultimately if FS says you're out of VRAM, you're out of VRAM.
Without knowing the specifics of the model you're training, its settings, the batch size, it is impossible to give you any more than this.
Please provide the crash report from your faceswap folder. There isn't enough to go on here.
Do they give better results? Yes. Do they train in a shorter time? No. They take longer because they are heavier models.
I think you found it! But for anyone else it is the "check" mark with the other buttons next to the avatar.
Get the latest code. This is a bug which has been fixed
I do use Linux. Specifically Ubuntu 18.04
Exactly
3) my converted faces from A with my model (so I have a set of mix faces from A to B)
This is wrong. You should not have converted faces.
You convert THE VIDEO, not THE FACES.
Hard to say, and depends on a lot of things (model, CPU/GPU, quality of data etc). Leaving it going some more and see if it's improving
Ok, 2 things:
1) Remove your system wide installed Cuda. How to do this will depend on how you installed Cuda in the first place. Basically, it will probably conflict with the Conda installed version.
2) Update your Nvidia Drivers.