I don't know what this is about But nerdy is having fun
@frapastique
2 жыл бұрын
Thanks for your work, really real great stuff
@jimimased1894
2 жыл бұрын
what a legend!
@NerdyRodent
2 жыл бұрын
Fun, isn’t it? 🙂
@llucaskaique
2 жыл бұрын
loved the video man, awesome!
@viber6268
2 жыл бұрын
Thank you legend.
@ll2pakll
Жыл бұрын
If anyone has an error like this `Could not find MSVC/GCC/CLANG`. You need to insert the compiler path in this file: `dnnlib\tflib\custom_ops.py`. Here is what the inserted line looks like, which worked for me 'c:\Program Files\Microsoft Visual Studio\\2022\Community\VC\Tools\MSVC\\14.33.31629\\bin\Hostx64\\x64' Also make sure you install Visual Studio and there in the "Development of classic applications in C++" install the module "MSVC version 143".
@catherinele8417
2 жыл бұрын
I like your .txt easier to read and understand than a detailed read.me . Thank you for sharing !
@nr.bln.
2 жыл бұрын
This is great
@ronpeleg6535
2 жыл бұрын
there some fun stuff over here hhh. subscribed!
@NerdyRodent
2 жыл бұрын
Lots of AI fun to be had indeed! 🙃
@junhyeonpark5239
2 жыл бұрын
Nice video! Can I adjust the feature like glasses, skin color..?
@moistweener
Жыл бұрын
Night sounds like a stupid question, but can these be implemented into a React app? Given the different language it’s written in/ capabilities? Idk but something that came to mind, haven’t seen many videos or content about it. I know it’s open source so there isn’t an api.
@Nothing-cx4jt
2 жыл бұрын
Thank you so much for this amazing content. Could you please make a tutorial on StyleGAN3 train on non-square image, please?
@secretsather
2 жыл бұрын
Do you know what the gamma setting affects? I cant seem to find much information online as it relates to training other than the suggested values in the docs MD file
@NerdyRodent
2 жыл бұрын
I’ve just been going by the docs file too. The technical stuff about gamma is described in the paper, but the general gist seems to be “play within these value ranges and see what happens” 🙃
@sananazari4797
2 жыл бұрын
it is the regularization value
@AlizerLeHaxor
2 жыл бұрын
Does resuming training from a .pkl file also resumes your kimg?
@TheChannelJ1
Жыл бұрын
Does StyleGAN3 work on Profile Angles of Face Portraits? I have only seen it work on Frontal Perspectives or 3/4th Perspectives? I have NEVER seen it do well with Side Profile Shots of people's faces.
@NerdyRodent
Жыл бұрын
Part of it is to do without your dataset. Make sure you have some profile angles if you want those.
@TheChannelJ1
Жыл бұрын
@@NerdyRodent Thank you.
@marverickbin
2 жыл бұрын
Can we change a given parameter and let all the other fixed? I mean, the same person, but we change only the smile, or the pose, or the gender.
@NerdyRodent
2 жыл бұрын
Previous techniques have a lot of entanglement, but Nvidia LACE seems pretty good for that!
@senadheeragesilva8140
Жыл бұрын
Hi, I am Sri Lankan student who is doing my final year research. I started to train styleGAN3 using colab pro since I cannot find any GPU machine in my area. I have some financial issues and still I need my model to be trained some more. If you don't mind can you help me to train my dataset. And I am really glad if you can help me with this. Thank you.
@NerdyRodent
Жыл бұрын
Colab pro is indeed your best bet!
@senadheeragesilva8140
Жыл бұрын
@@NerdyRodent Thanks Nerdy Appreciate your reply
@SpeedOfSpace
2 жыл бұрын
Hi there, fantastic vid I'm really keen to get this working, although, I'm getting an error: ResolvePackageNotFound: - cudatoolkit=11.1 when I try to create an environment from the .yml file. I downloaded cudatoolkit 11.1 and can see it's installed using nvcc -V, any tips? Download packages manually?
@NerdyRodent
2 жыл бұрын
I’ve upgraded to a later cudatoolkit. You could try “conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch”
@magemerlyn8887
2 жыл бұрын
I was having this issue as well and had to change the .yml file slightly. I added "conda-forge::" in front of cudatoolkit=11.1 so that one line looks like. "- conda-forge::cudatoolkit=11.1"
@sergioruiz2249
2 жыл бұрын
Will my computer explode if I try to run this with my RTX 2060i? It asks for 12 GB of VRAM, but I only have 6 hahah
@NerdyRodent
2 жыл бұрын
You may be able to train on smaller images, such as 128x128, with small gpu batches
@tiporight
2 жыл бұрын
Hi, this is great! I'm getting a "broken pipe" " crf not working" error while is generating the video with the clip script. any ideas?
@NerdyRodent
2 жыл бұрын
Sounds like a problem with your ffmpeg install.
@tiporight
2 жыл бұрын
@@NerdyRodent which version of ffmpeg is compatible with the clip script in Linux?
@NerdyRodent
2 жыл бұрын
@@tiporight I'm using the default version that comes with Ubuntu 20.04, which is ffmpeg version 4.2.4-1ubuntu0.1
@tiporight
2 жыл бұрын
@@NerdyRodent I was able to use sg3-clip, but the results are very weak. The animation is very similar to the original using any text prompt. I also checked the -help section. The results are not like your original clip model. Am I missing something?
@tiporight
2 жыл бұрын
I was able to make it work. How long is the training time for each tick in your configuration?
@marcoc2
2 жыл бұрын
How to train new models for StyleGAN3?
@NerdyRodent
2 жыл бұрын
00:00 to 13:00 covers training
@marcoc2
2 жыл бұрын
@@NerdyRodent thanks! I skiped that part because I watched it live
@andranikarakelov8139
2 жыл бұрын
are you coding now only in pytorch ?
@NerdyRodent
2 жыл бұрын
Sometimes I use tensorflow, but most stuff seems to be pytorch these days!
@huseyngorbani6544
2 жыл бұрын
Shall we know what VM you are using?
@NerdyRodent
2 жыл бұрын
Not sure what you’re asking there, bud? 😉
@huseyngorbani6544
2 жыл бұрын
@@NerdyRodent I was wondering if you are using any Virtual Machine with dedicated GPUs or not. I have a project in my mind but I am aware my 2018 MacBook pro won't survive training GanStyle 3 😅. Any suggestions for VM?
@NerdyRodent
2 жыл бұрын
@@huseyngorbani6544 No VMs - just running normally 😉
@huseyngorbani6544
2 жыл бұрын
@@NerdyRodent good for you 💪🏻
@huseyngorbani6544
2 жыл бұрын
@@NerdyRodent Finally, is it possible to train GanStyle3 so that it combines any 2 given images and outputs an abstract image out of those 2?
@Justobolsa
2 жыл бұрын
Need a collab tutorial pls
@NerdyRodent
2 жыл бұрын
Colab is more for interactive use than leaving a program running for several days.
Пікірлер: 54