is numpy 2 supported?

[heather999]

I've been trying to mamba install extinction into a python 3.12.11 environment with numpy 2.2.6 already installed, but it consistently tries to downgrade numpy - is this expected?

(lsst-scipipe-10.0.0-exact) heatherk@perlmutter:login02:/global/common/software/lsst/gitlab/td_env-dev/manualbuilds/td_env> python --version
Python 3.12.11
(lsst-scipipe-10.0.0-exact) heatherk@perlmutter:login02:/global/common/software/lsst/gitlab/td_env-dev/manualbuilds/td_env> mamba install -c conda-forge extinction
conda-forge/linux-64                                45.2MB @  53.6MB/s  0.8s

Pinned packages:

  - python=3.12


Transaction

  Prefix: /global/common/software/lsst/gitlab/td_env-dev/2025-07-22-51-51/conda/envs/lsst-scipipe-10.0.0-exact

  Updating specs:

   - extinction


  Package                Version  Build             Channel           Size
────────────────────────────────────────────────────────────────────────────
  Install:
────────────────────────────────────────────────────────────────────────────

  + extinction             0.4.6  py312hc7c0aa3_4   conda-forge      112kB

  Remove:
────────────────────────────────────────────────────────────────────────────

  - rubin-env             10.0.0  py312h5fef794_15  conda-forge     Cached
  - rubin-env-nosysroot   10.0.0  py312hed51c57_15  conda-forge     Cached

  Change:
────────────────────────────────────────────────────────────────────────────

  - assist                 1.1.9  py312h41a817b_1   conda-forge     Cached
  + assist                 1.1.9  py312hd2a09a5_2   conda-forge       62kB

  Upgrade:
────────────────────────────────────────────────────────────────────────────

  - spiceypy               6.0.1  pyhd8ed1ab_0      conda-forge     Cached
  + spiceypy               6.0.3  pyhd8ed1ab_0      conda-forge      162kB

  Downgrade:
────────────────────────────────────────────────────────────────────────────

  - numpy                  2.2.6  py312h72c5963_0   conda-forge     Cached
  + numpy                 1.26.4  py312heda63a1_0   conda-forge        7MB
  - qp                     1.0.1  pyha770c72_0      conda-forge     Cached
  + qp                     0.9.3  ha770c72_0        conda-forge        7kB
  - qp-prob                1.0.1  pyhd8ed1ab_0      conda-forge     Cached
  + qp-prob                0.9.3  pyhd8ed1ab_0      conda-forge      103kB

  Summary:

  Install: 1 packages
  Remove: 2 packages
  Change: 1 packages
  Upgrade: 1 packages
  Downgrade: 3 packages

[lukeshingles]

Hi, I don't know anything about mamba, but this doesn't happen with pip or uv on my setup. Seems like it's a mamba bug, environment config issue, or one of those other packages caused the downgrade. Try running the command in verbose mode (-v usually) to see if you can figure out exactly what triggered the numpy downgrade.