Maybe I’m wrong, but the extensive installation process based on the environment.yml file seems excessive. It can take a very long time and is probably not necessary for those who keep their environments current. If you concur, could you please consider including a comment to that effect in your installation instructions? If you do not concur, is there a practical way to reduce installation time? I am installing on a supercomputer in which I must request a very specific amount of time on the machine. Twice, now, I have underestimated the time needed. The second was 50 minutes and that was still not enough. I am now trying with 2 hours. These requests cost me Service Units which I can’t get back. Please, at least provide an estimate or estimates for various environments.
Depending on your needs, you can set up a Miniforge environment and only install those packages that you require. In particular, this mean just numpy and scipy. Then you can pip install mne, which does not have the optional dependencies contained in the conda package.
another way to speed up installation is to use mamba (a drop-in replacement for conda, for the download-and-install steps). But I also second what @cbrnr said, starting out with Miniforge and whittling down the install list to just the packages you really need can save a lot of install time.
Feel free to chime in on RFC: installation process instructions · Issue #9623 · mne-tools/mne-python · GitHub too, we’re actively discussing ways to improve our install process and installation instructions.