Adam J. Stewart

Conda-forge still needs to be updated with NumPy 2 support: https://github.com/conda-forge/lightly-feedstock/issues/20

If there are no new features/API changes and only bug fixes, shouldn't it be 1.2.1?

From the [numpy 2.4.0rc1 release notes](https://github.com/numpy/numpy/releases/tag/v2.4.0rc1): > NumPy's C extension modules have begun to use multi-phase initialisation, as defined by PEP 489. As part of this, a new explicit check...

Can confirm this issue with numpy 2.4.0rc1: ``` > coverage run --source=torchgeo.datasets.landcoverai -m pytest tests/datasets/test_landcoverai.py ImportError while loading conftest '/Users/Adam/torchgeo/tests/conftest.py'. tests/conftest.py:7: in import matplotlib ../spack/var/spack/environments/default/.spack-env/view/lib/python3.13/site-packages/matplotlib/__init__.py:161: in from . import _api,...

Ten thumbs up. Compilers should be a "dependency" just like any other package. There are more than 4 languages in the world.

Whenever a Spack package like `python` depends on another package like `xz`, Spack queries the `xz` package to find out where it's libraries are installed. In the case of `xz`,...

In that case I think we can tell `spack external find` to resolve symlinks and use the full path to the executable. Let me see if I can figure out...

In the long run, we want to be able to store more information in `packages.yaml` like you suggest. For example, there are systems where executables are in `/usr/bin` but libraries...

You can create a fork of the repository and push your branch there.