Tencent
Not usable at all. No, you can't just install it with all of the requirement libraries and all of python versions on PC ...
Is there anyone who can install it on windows? I've installed most of the known neural networks... but this one is something like IMPOSSSUBBURUUU level for deploy You can't just do something..
I spent about seven hours trying to rearrange all versions of torch and all versions of CUDA, but all in vain. Nice job guys, you've done my day
btw : the installation of this neural network should be used as a test task for working on NASA
I also think, even if I Install :
python -m pip install ABSOLUTELY EVERYTHING, EVERY SINGLE PACKAGE
this one is not getting started anyway, hunyuan just generate any other error
Why do you use ..\venv\Scripts\python.exe ...? For me it does not look the correct way to use virtual environment.
Because I already used the recommended Install guide and it doesn't working, I also used custom wheels, pypa/build and two another methods to install this thing... It's just not working at all on windows, Shift+Delete helps, completely useless tool if you on windows, unfortunately :( Only burns the time
I don't use Window but others got it to run afaik.
With official guide to install you talking about https://github.com/Tencent/Hunyuan3D-2/blob/main/README.md?
I can't see a part telling you to use a virtual environment and how. In your case I would first clearify the use of your virtual environment. Maybe it works with direct call the python.exe in the path of your \venv, but the common way is to activate the venv in the terminal.
Then install pytorch like on the offical site and then go with the guide in the README.md (I guess your cuda is installed correct).
Before using venv, I already used default method, I told you 🙄 It doesn't work
Its a beast to install on windows - downgrade CUDA upgrade VS2022 - now conflicts the other way around. Ended up adding extra_compile_args={ 'cxx': ['-O2'], 'nvcc': [ '-O2', '--expt-relaxed-constexpr', '-allow-unsupported-compiler', '-D_ALLOW_COMPILER_AND_STL_VERSION_MISMATCH' ]to setup.py (..\hy3dgen\texgen\custom_rasterizer) and finally it worked. Gorgeous mesh generation - i also added the Low Vram flag or it just never finished generating texture)
Looks like python environment not correctly set. I personally prefer runing source venv/bin/activate first. In the file named "venv/bin/activate", you can see what environment variables are set.
