Running on windows
Does this not install in windows?
Cant seem to get the pip install xcodec2 working.
Also having issues with this -
" AssertionError: Unable to pre-compile ops without torch installed. Please install torch before attempting to pre-compile ops."
But torch is installed...
You're going to need to have Python 3.9 which is why he uses Anaconda environment. Anaconda's official Windows build is a buggy mess. Alternatively, create a Python 3.9 venv. Second answer is the best for Windows: https://stackoverflow.com/questions/70422866/how-to-create-a-venv-with-a-different-python-version
The error you're getting is from the deepspeed package. It should handle dependencies properly in the package deps, but the packager didn't know what they're doing. Nothing HKUST Audio can do about that.
Does this not install in windows?
Cant seem to get the pip install xcodec2 working.
Nvidia NCCL is not available for windows, so it looks like xcodec2 is not currently able to be installed on the windows OS. There is a version of NCCL rewritten by MyCaffe to be compatible for windows, but it's unclear if this project could make use of it.
I got xcodec2 working on Windows by ignoring all the dependencies (--no-deps) and compiling DeepSpeed from the source. There's a convenient build_win.bat script for it in the DS repo but it might require MS Build Tools.
I packaged the xcodec inference code as a library, give it a try.
too much deps like a hell...give it up...
Thank you all for your comments. As an update, there were a few packages that prevented this to install in windows out of the box. As @MaGoogus pointed out, nvidia nccl is not available for windows. Furthermore, I had issues with other packages not being 'found' by pip.
In the end, I decided to just install linux. That is a far manageable hassle than trying to do this in windows.
The actual requirements are transformers and torch actually, and both can run on CPU. The other requirements are installed when you install torch on a CUDA computer.
I removed those to make it easier. https://pypi.org/project/xcodec2-infer-lib/0.1.1/
If you don't trust it (and you probably shouldn't), here is the github for the library: https://github.com/uetuluk/xcodec2-infer-lib
You only need to install these: https://github.com/uetuluk/xcodec2-infer-lib/blob/master/requirements.txt