`error: option --cudagp not recognized` when trying to install on Linux

[lingdoc]

According to the docs, I should be able to compile a cuda-enabled python version on my Ubuntu machine, but apparently the --cudagp flag isn't recognized.

[lingdoc]

ok, well the flag still doesn't work, but I was able to build a version using the pip cmake command.

After cloning from github and building/installing with cmake & gcc on linux (see here), I navigated to the python-package directory and ran the following command (after upgrading pip):

pip install . --config-settings=cmake.define.USE_CUDA=ON

this seems to have worked.. at least when I run models with the GPU_use=True, argument I get the following info:

[GPBoost] [Info] CUDA device detected (1 device(s)). GPU acceleration will be used where supported.