Search found 21 matches

by Linhaohoward
Sat Feb 08, 2020 2:40 pm
Forum: Training
Topic: Any advise for training based on highest quality possible?
Replies: 1
Views: 794

Any advise for training based on highest quality possible?

Any idea what’s the best model and configuration that I can train on based on my computer specs? Even batch size 2 will be fine, objective here is for highest quality possible.

Below is my specs.
Ryzen 9 3900x 12core 24 threads
32gb DRAM
2 x RTX 2070 Super 8gb each
by Linhaohoward
Wed Feb 05, 2020 4:52 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

Assuming you have used the installer (and it looks like you have) I would just let conda handle it. Off the top of my head (so I apologize if I miss a step) Start > Anaconda Prompt Inside your anaconda prompt: conda activate faceswap conda remove tensorflow* conda remove cudatoolkit conda remove cu...
by Linhaohoward
Wed Feb 05, 2020 3:39 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

should i also downgrade my CUDA? since faceswap auto downloads the needed configs and pip downloaded 1.15 is it because i'm running CUDA 10.2?
by Linhaohoward
Wed Feb 05, 2020 3:34 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

torzdf wrote:
Wed Feb 05, 2020 2:48 pm
Ok, the problem is with Keras and tf 1.15 with multi-gpu....

If you downgrade Tensorflow to 1.13.1 you should get it working fine.
Is there somewhere that can teach me how to downgrade tensorflow to the said version?
by Linhaohoward
Wed Feb 05, 2020 2:47 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

Hi, I can start dlight on 256 output with single gpu at bs 16 but i can't get mgpu to work. i have 2xrtx 2070s. Once i change to 2 gpu the model will not even start. I also tried training 384 output on dlight with single gpu since i succeeded with 256 single gpu but have failed to start as well. my ...
by Linhaohoward
Wed Feb 05, 2020 1:42 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

I don't believe that to be the case. Afaik there is nothing intrinsic to dlight to stop it behaving like any other model. What is the specific issue you are having? Hi torzdf, above is what i'm seeing. Below is the crash log 02/05/2020 21:37:54 MainProcess _training_0 nn_blocks set_default_initiali...
by Linhaohoward
Wed Feb 05, 2020 1:40 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

Re: problem starting dlight model

Loading... Setting Faceswap backend to NVIDIA 02/05/2020 21:36:55 INFO Log level set to: INFO Using TensorFlow backend. 02/05/2020 21:36:57 INFO Model A Directory: D:\Faceswap Images\A_src\A_aligned384x384 02/05/2020 21:36:57 INFO Model B Directory: D:\Faceswap Images\fan_jilamika_src\combined_face...
by Linhaohoward
Tue Feb 04, 2020 7:17 pm
Forum: Training
Topic: problem starting dlight model
Replies: 10
Views: 1485

problem starting dlight model

Hi, just want to double verify the problem as i think i might have read somewhere that dlight can't run with optimizing savings and multi gpu. Is this the case? cause i can't seem to get my dlight model to start.
by Linhaohoward
Tue Feb 04, 2020 2:38 pm
Forum: Extract
Topic: 384x384 extraction. Any advantages?
Replies: 1
Views: 775

384x384 extraction. Any advantages?

Hi,

Can i know the advantages and disadvantages of using a 384x384 face extraction vs a standard 256

Thank you in advance
by Linhaohoward
Tue Feb 04, 2020 2:04 pm
Forum: Hardware
Topic: Has deepfaking for long periods caused anyone long-term hardware issues?
Replies: 9
Views: 2717

Re: Has deepfaking for long periods caused anyone long-term hardware issues?

No, overclocking stresses the chip and can cause it to artifact or glitch. For gaming and graphics that's not a big deal since a frame with a glitched polygon probably wont even be noticeable, but a glitch during CUDA processing can corrupt that whole calculation, which we can't catch so at the wro...
by Linhaohoward
Mon Feb 03, 2020 6:22 pm
Forum: Training
Topic: what type of face photo sources should i extract if i want to train vgg obstructed?
Replies: 1
Views: 657

what type of face photo sources should i extract if i want to train vgg obstructed?

Should i purposely include some photos of B with certain obstructions? eg, drinking from a straw, wearing glasses and ect.
If yes, what percentage of the photos should be of such? 5%?
by Linhaohoward
Sat Feb 01, 2020 1:55 pm
Forum: Hardware
Topic: GPU Training Speed
Replies: 9
Views: 4625

Re: GPU Training Speed

My Sys info: CPU AMD 3900x, 32GB 3200 RAM, m.2 SSD storage, Windows 10 No Overclocking done Trainer Dfl-SAE 160Input Size, Df achitecture 512 autoencoder, 45 encoder dims, 25 decoder dims, Vgg-obstructed, Batchsize 20 Optimizer Savings Save interval 1000 GPU: MultiGPUs 2x Zotac RTX 2070 Super NVlink...
by Linhaohoward
Fri Jan 31, 2020 8:48 pm
Forum: Training
Topic: Can you keep using 1 single model?
Replies: 1
Views: 634

Can you keep using 1 single model?

Hi, can i keep using the same model to train different targets? The model is used to train actor A face onto actor B. Can i continue training the same model to put actor A face onto actor C?
by Linhaohoward
Fri Jan 31, 2020 4:43 pm
Forum: Hardware
Topic: Has deepfaking for long periods caused anyone long-term hardware issues?
Replies: 9
Views: 2717

Re: Has deepfaking for long periods caused anyone long-term hardware issues?

Most cards should be thermally safe at their maximum operation speed. They should be okay indefinitely. That said, if you can add some cooling to reduce the temperature it's not a bad idea. If your card hovers in the 90c range you might want to underclock it to keep it cooler. There shouldn't be an...
by Linhaohoward
Wed Dec 25, 2019 8:10 am
Forum: Hardware
Topic: Hardware best practices
Replies: 27
Views: 12878

Re: Hardware best practices

Just switched to Faceswap yesterday. Is there any difference for multi gpu with nvlink on Faceswap? As compared to just multi gpu without nvlink.
by Linhaohoward
Tue Dec 24, 2019 8:47 am
Forum: Hardware
Topic: Hardware best practices
Replies: 27
Views: 12878

Re: Hardware best practices

For Faceswap i recommend the multiple smaller gpus and a mid/high end cpu. Ryzen 5/7 and 1-4 2060 supers is a great setup. Thank you for all the tips. Just made the purchase of a Ryzen 5 with 2x2070 Super and nvlink bridge. Im happy now though my credit card is crying 😂 does Faceswap support nvlink?
by Linhaohoward
Mon Dec 23, 2019 5:42 am
Forum: Hardware
Topic: Hardware best practices
Replies: 27
Views: 12878

Re: Hardware best practices

This is the forum for faceswap, not DFL. DFL questions should be directed to the appropriate locations, as we cannot advise. Sorry my bad I’m actually intending on changing to Faceswap. Since dfl doesn’t support multi gpu in the first place. Just hoping to get my hardwares right once and for all so...
by Linhaohoward
Sun Dec 22, 2019 10:12 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 27
Views: 12878

Re: Hardware best practices

It's important to note that DFL does not support multiple gpus, only Faceswap can use more than one GPU. Cpu is still heavily used for things like augmentation, but it's far less important than the gpu. So do you recommend 2x rtx2070s over a single 2080ti in Faceswap? Can you advise if Ryzen 5 woul...
by Linhaohoward
Sun Dec 22, 2019 7:56 pm
Forum: Hardware
Topic: Hardware best practices
Replies: 27
Views: 12878

Re: Hardware best practices

20xx series cards don't use SLI, but instead use nvlink. This is actually better, but it's not critical. Communication between the cards can happen through the pci-e bus or nvlink but sli won't work for non-gaming tasks. However, consumer nvlink will only work between 2 cards anyway. So having 4 ca...
by Linhaohoward
Sat Dec 21, 2019 1:39 pm
Forum: Training
Topic: Single GPU vs Multiple GPU
Replies: 7
Views: 2479

Re: Single GPU vs Multiple GPU

I’m noob, just wondering why 3x2070s over 4x2060s?