-
Notifications
You must be signed in to change notification settings - Fork 59
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Windows support? #27
Comments
There are some precompiled Triton wheels for Windows here. I got Sana working on Windows when it came out, it should still be possible now. |
The current inference and training code is not using Triton. You should run on your windows smoothly. |
@lawrence-cj if not using, let's disable it first? |
If you are looking for complete guide for Windows installation |
It would be better to check if the triton is available to import rather than disable it directly. @Lyken17 |
Does not work when you follow the instructions step by step. |
What is not working. Post the error log from command prompt. |
where will the |
In ./diffusion/data/wids/wids_dl.py file line 18 |
Yes that was one of the problems that prevented me from using Sana the fcntl once that is solved I will try using it again. |
So far any attempts of using Sana under Windows 11 with a Nvidia Geforce RTX 4060 TI 16GB VRAM and 128GB System Ram have failed because of installation problems not even the so called claimed made for Windows versions of this work and I followed every installation instructions to the letter (some of the instructions are incomplete). |
Seems to disable wids related module during the inference should work. Hi @endman100 @nitinmukesh , in which step you are blocked? Installation or inferecne, if the later, can you share the inference scripts? |
i had to make a lot of pipeline fixes for to make it work on windows - so many only linux available libraries used |
@FurkanGozukara can you share your solution and (if possible) draft PR? I agree that we may put too many linux pkgs during development and hence want to trim the dependency |
I'm not blocked, I made it work within 1 hour on Windows (didn't bragged about it) and using it. Here is the repo with all these changes To fix the Linux specific issues
In and Please see if portalocker library is supported on Linux , replace fcntl. In This code create issues Make sure this should be taken as input from user OR defaults to log folder within project directory OR can be disabled using a switch |
He will not share or contribute for free he paywalls most of his stuff money is what drives him even tough he uses so much from the open source community who do things for free he needs to make a profit out of it. |
Seriously though. I don't even get how he isn't banned from Github and why I'm seing him even though I blocked him. He's so disrespectfull it's even more irritating that he is promoting his scamy content on Github in addition to HuggingFace and Reddit. |
Are you still struggling to make it work on Windows? I will look at it again. |
Yes I still have not managed to make it work and I follow the instructions exactly. (sana) K:\Sana_win>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth |
pull the latest code and maybe you can try to uninstall triton if the latest code still not work for you. |
I tried does not work: git clone https://github.com/NVlabs/Sana.git |
Trying to run the software anyways will give more dependencies that are not installed. (sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth |
@lawrence-cj It should not be this difficult for you to make a working version for Windows? |
@nitinmukesh your installation of Sana for Windows also does not work similar problems. |
It does not look like anymore this model is even worth spending all that time on getting to work after days it still does not. Their devs clearly do not even know how to get it running on Windows 11. pip install gradio (sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth (sana) K:\Sana>pip install spaces (sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth (sana) K:\Sana>pip install torchvision ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. |
for who cant solve how to install i have 1 click installer that works all you need is having installed python 3.10.11, cuda 12.x, c++ tools - public post : https://www.patreon.com/posts/nvidia-labs-sana-116531341 requirements tutorial that shows everything step by step (python, cuda, c++ tools, cudnn) : https://youtu.be/DrhUHnYfwC0 I spent 12 hours to make this installer + the improved gradio |
The error stops the process of installing other dependencies, like I like remove the triton related part, like: Line 44 in 5765093
And I'll try to remove the dependency of triton later when I'm free. The bandwidth is full. @CRCODE22 |
pip install gradio |
Ok it would really if you can remove the dependency of triton do I still need to install xformers? |
If it where only that simple once you have installed that as I mentioned and shown in my earlier post it will keep asking for other dependencies that should go automatic. I will first try to remove that line 44 as @lawrence-cj suggested and see if dependencies are automatically installed. |
@lawrence-cj I did a reinstall and removed the line 44 related to triton on Sana/pyproject.toml]( Line 44 in 5765093
(sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth (sana) K:\Sana>pip install -U xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121 (sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth |
I solved most of the problems but the triton requirement is still hard coded it seems and prevents Sana from working on Windows 11 (sana) K:\Sana>python app\app_sana.py --config configs\sana_config\1024ms\Sana_1600M_img1024.yaml --model_path Efficient-Large-Model\Sana_1600M_1024px\checkpoints\Sana_1600M_1024px.pth |
This is a very good project. Thank you for the team's contribution.
I would like to know if there will be Windows support without Triton?
Thanks!
The text was updated successfully, but these errors were encountered: