MrDeepFakes Forums
  • New and improved dark forum theme!
  • Guests can now comment on videos on the tube.
   
iperovFaceswap and multi gpu training - fake
#11
(08-04-2019, 02:30 PM)iperov Wrote: You are not allowed to view links. Register or Login to view.@dpfks why this man still not banned for breaking forum rules? i cannot understand why so many unreasonable insults from him

Based on what I've seen in my short time here. Ghost can be blunt, and at times a dick. But he openly says what a lot of people are too polite, or afraid to say.

It's much easier to tear down someone else, than to build yourself up. Would it not be better to let your work speak for itself? Your work on DFL proves you're a talented coder. Why not spend the time working with Keras to implement Multi-GPU in DFL, instead of taking shots at others?
#12
I just warning faceswap users about multigpu: second card is heating in vain.
topic can be closed
#13
(08-04-2019, 04:05 PM)iperov Wrote: You are not allowed to view links. Register or Login to view.I just warning faceswap users about multigpu: second card is heating in vain.
topic can be closed

So I asked the devs specifically, and it looks like there's some proof of keras working with multi-GPUs; although the growth is not linear.

You are not allowed to view links. Register or Login to view.

Any thoughts on this?
~ Fake it till you make it ~
You are not allowed to view links. Register or Login to view.
You are not allowed to view links. Register or Login to view.
#14
(08-04-2019, 02:30 PM)iperov Wrote: You are not allowed to view links. Register or Login to view.@dpfks why this man still not banned for breaking forum rules? i cannot understand why so many unreasonable insults from him

You talking about yourself Sire ?

Learn to respect other opinions. You learnt coding well but you terriblly lack in giving respect to others project developrs.
#15
Hey guys lets be bigger than @iperov if her wants to show up out of nowhere and whine like a baby thats his prerogative, we dont have to stoop to his level. lets just ignore him like he ignores his fanbase.
#16
(08-04-2019, 04:32 PM)dpfks Wrote: You are not allowed to view links. Register or Login to view.
(08-04-2019, 04:05 PM)iperov Wrote: You are not allowed to view links. Register or Login to view.I just warning faceswap users about multigpu: second card is heating in vain.
topic can be closed

So I asked the devs specifically, and it looks like there's some proof of keras working with multi-GPUs; although the growth is not linear.

You are not allowed to view links. Register or Login to view.

Any thoughts on this?
Multi-gpu trainining does not result in a linear increase in training speed, it's about an 80% increase with a second matching card.  Multi-gpu allows training of larger batch sizes in each iteration, thus reducing overall training time.  Larger, more complex models can be trained.  Multi-GPU benefits are limited by the slower card or the card with less VRAM if you have two different GTX cards (e.g. 1080ti and 1070) in your system.  The fact that multi-GPU training does not result in linear gains does not make faceswap's implementation "fake," as iperov falsely asserts.
#17
Like I said, @iperov acts like a god and just popped up to shit on another project which implemented something properly that he could not. Lets just ignore him. He likely used faceswap and saw, 1. that it worked, and 2. that it had an easy to use GUI and wanted to give a reason for his oh so precious followers to not switch to another project because thats where he derives his self worth from. Lets not be like him, let him wallow in his shitty life alone. better to leave the baby to starve and finally shut up.
#18
(08-04-2019, 05:39 PM)JustAnnonymouse Wrote: You are not allowed to view links. Register or Login to view.
(08-04-2019, 04:32 PM)dpfks Wrote: You are not allowed to view links. Register or Login to view.
(08-04-2019, 04:05 PM)iperov Wrote: You are not allowed to view links. Register or Login to view.I just warning faceswap users about multigpu: second card is heating in vain.
topic can be closed

So I asked the devs specifically, and it looks like there's some proof of keras working with multi-GPUs; although the growth is not linear.

You are not allowed to view links. Register or Login to view.

Any thoughts on this?
Multi-gpu trainining does not result in a linear increase in training speed, it's about an 80% increase with a second matching card.  Multi-gpu allows training of larger batch sizes in each iteration, thus reducing overall training time.  Larger, more complex models can be trained.  Multi-GPU benefits are limited by the slower card or the card with less VRAM if you have two different GTX cards (e.g. 1080ti and 1070) in your system.  The fact that multi-GPU training does not result in linear gains does not make faceswap's implementation "fake," as iperov falsely asserts.

Yes I totally understand this. I didn't get a chance to test this in DFL back when it was still available, but now I have 2 GPUs. Iperov's argument has always been that it does not work at all which is why it was removed. I think this is one of the key factors currently of why a user would use Faceswap vs. DeepFaceLab. There are obviously advantages and disadvantages to each app though.

Even if multi-GPU gives a modest improvement I think it would be worth while to add back to DeepFaceLab so we can train with higher settings. Hope @iperov can look into this again.
~ Fake it till you make it ~
You are not allowed to view links. Register or Login to view.
You are not allowed to view links. Register or Login to view.
#19
@dpfks this man should be banned for breaking the forum rules

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.what a dick.

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.yeah hes just a dick.

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.fuck off man.

this is why i dont like this forum.
A lot of trolls, people insulting developer, and even an admin who does not want to fulfill his duties.

(08-04-2019, 05:58 PM)dpfks Wrote: You are not allowed to view links. Register or Login to view.Even if multi-GPU gives a modest improvement I think it would be worth while to add back to DeepFaceLab so we can train with higher settings. Hope @iperov can look into this again.

as I said, multi-gpu cannot be enabled by adding just a line of code keras.multi_gpu_model as in faceswap. Users think it's work, but it's not.
Enabling multi-gpu requires a lot of complicated tensorflow-only code
#20
(08-05-2019, 10:33 AM)iperov Wrote: You are not allowed to view links. Register or Login to view.@dpfks this man should be banned for breaking the forum rules

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.what a dick.

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.yeah hes just a dick.

(08-04-2019, 08:31 AM)GhostTears Wrote: You are not allowed to view links. Register or Login to view.fuck off man.

this is why i dont like this forum.
A lot of trolls, people insulting developer, and even an admin who does not want to fulfill his duties.

(08-04-2019, 05:58 PM)dpfks Wrote: You are not allowed to view links. Register or Login to view.Even if multi-GPU gives a modest improvement I think it would be worth while to add back to DeepFaceLab so we can train with higher settings. Hope @iperov can look into this again.

as I said, multi-gpu cannot be enabled by adding just a line of code keras.multi_gpu_model as in faceswap. Users think it's work, but it's not.
Enabling multi-gpu requires a lot of complicated tensorflow-only code

you should be banned for straight up lying about another project to make your sorry ass look good. You win anyway I quit. I refuse to use software made by such an asshole. People can insult you, its not YOUR forum, I know free speech isnt a think in your chinkshit country but it is elsewhere. and dpfks has no duties to you. Nobody does. eat a bag of horse shit you fucking twat. I know this post wil get removed but who fucking cares I quit.

Forum Jump:

Users browsing this thread: 1 Guest(s)