Search found 9 matches

by blackomen
Thu Mar 12, 2020 8:30 pm
Forum: General Chat
Topic: Way to notify training failure?
Replies: 5
Views: 1773

Re: Way to notify training failure?

Maybe the devs can offer an option of adding IFTTT integration...
by blackomen
Tue Feb 25, 2020 4:46 pm
Forum: Training
Topic: Speed/Quality and Batch size
Replies: 4
Views: 1650

Re: Speed/Quality and Batch size

Training speed is not just iterations. It's actually best to use the number of images in a time period. So Batch Size * Iterations is a better calculation So 50 * 240 = 12,000 images/hour 30 * 500 = 15,000 images/hour 6 * 900 = 5,400 images/hour (I however, believe you've likely mistaken the speeds...
by blackomen
Tue Feb 25, 2020 2:17 am
Forum: Training
Topic: Crash: Inappropriate ioctl for device
Replies: 5
Views: 782

Re: Crash: Inappropriate ioctl for device

Yeah, it works when ran normally from the terminal alright. However, I don't like the idea that losing my SSH connection will terminate my training. I'll look into maybe running this on Jupyter instead (esp since I got it to work on colab but it'll be nice to be able to train multiple models simulta...
by blackomen
Mon Feb 24, 2020 9:37 pm
Forum: Training
Topic: Crash: Inappropriate ioctl for device
Replies: 5
Views: 782

Re: Crash: Inappropriate ioctl for device

No, I was trying to run it from the command line from my cloud service (after failing to get the GUI to work via VNC.) I used the Generate button on my local computer's copy to get the commands and uploaded my faces and alignments to my cloud service and adjusted the paths in the generated command. ...
by blackomen
Sun Feb 23, 2020 6:46 pm
Forum: Training
Topic: Crash: Inappropriate ioctl for device
Replies: 5
Views: 782

Crash: Inappropriate ioctl for device

Here's what I'm getting when I try to run this on a cloud service: 02/23/2020 13:43:13 INFO Loading data, this may take a while... 02/23/2020 13:43:13 INFO Loading Model from Dfaker plugin... 02/23/2020 13:43:13 INFO No existing state file found. Generating. 02/23/2020 13:43:14 ERROR Got Exception o...
by blackomen
Fri Feb 21, 2020 3:04 pm
Forum: Training
Topic: Sunglasses
Replies: 1
Views: 721

Sunglasses

OK, I just rediscovered faceswap after playing with it a few years ago when it first came out and I remember having a lot of trouble swapping faces of people wearing sunglasses. For example, if my A video contained a normal face and my B video contained a face that's wearing sunglasses the whole tim...
by blackomen
Fri Feb 21, 2020 2:31 pm
Forum: Artificial Intelligence
Topic: Cyclical learning rates and learning rate finder?
Replies: 1
Views: 1166

Cyclical learning rates and learning rate finder?

Does anyone know if the gradient descent algorithm used by faceswap is using cyclical learning rates?

Also, can we implement a learning rate finder (similar to the one in the Fastai library)?
by blackomen
Fri Feb 21, 2020 2:27 pm
Forum: Hardware
Topic: Highest quality model to use on a 4GB card?
Replies: 3
Views: 1624

Re: Highest quality model to use on a 4GB card?

I have a GTX 960 MX with 4GB and I was able to get it to train with a BS of 8,size 256,and the first 2 optimizers checked using Dlight. I trained for about 8 hours before my A and B losses fell consistently below 0.02 each on a particular pair of videos Previously, I was using an older version of fa...