Add support for Python 3.14.3 and PyTorch 2.10.0

[chameleon-ai]

As far as I know, this is the minimal set of changes needed to bring everything up to date with the latest python and pytorch.

This also addresses:
#2108
#2107
#1897

And probably a host of other issues.

Exact instructions to install for my platform:

python -m venv venv
source venv/bin/activate
pip install --upgrade pip wheel
pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm7.1
pip install -r requirements.txt

Also tested on python 3.13.12, 3.12.12, 3.11.14, and 3.10.19
Possibly deprecating:
#1830

I am only able to test this on Linux + AMD (ROCm 7.1)
Note that requirements.txt has been completely redone from scratch, so platform specific dependency issues are not reflected.
Not tested on versions of python older than 3.10

[iMartyn]

Data point: tested and working on Linux + nvidia

[chiajunjie-digipen]

Second vouch, build, downloading MDX model and actual separation works

[suurt8ll]

I'm running up-to-date Arch Linux, both the build process and the app itself work without problems!

[xnr32]

works on ubuntu 24.04, thanks!

[RattusSolarus]

Works on Debian Trixie. Spent a while failing with the original master before finding this.
Actually I'm amazed how well it works. I'm mixing a project from a live performance which has a lot of drum spill in the vocals which hasn't been able to be solved before. Running the vocal track with MDX Kim_Vocal_2 removed virtually all the drums with almost no residual artefacts. Brilliant!
Thanks for your wonderful work. 👍

[nikolas]

Congrats on getting this working - You may want to pin all these at the current versions they are working at, but that is your call.

I am not sure where the maintainer is, but it would be good to get these changes merged.