td_env

Defines, builds and installs a conda environment including the LSST Science Pipelines and Time Domain requested packages installable via conda-forge or pip.

Currently Supported Installations

• Native NERSC Installation on Cori also available in Jupyter
• Docker images available on DockerHub and in Shifter at NERSC
• Development versions are also available to test pre-releases

Running at NERSC

• To use the td_env installed at NERSC
  • source /global/cfs/cdirs/lsst/groups/TD/setup_td.sh
• OR access the desc-td-env Jupyter kernel by making sure you have DESC jupyter kernels installed
• OR use the available Shifter image
  • lsstdesc/td-env:stable

    shifter --image=lsstdesc/td-env:stable /bin/bash 
    source /global/cfs/cdirs/lsst/groups/TD/setup_td.sh

Current Stable Release

https://github.com/LSSTDESC/td_env/releases/latest

Running cosmosis + firecrown

Cosmosis is available in the td_env integration and dev builds. Users are encouraged to try this out.

The default setup_td_dev.sh will prepare the environment by settting $CSL_DIR, $FIRECROWN_DIR, $FIRECROWN_EXAMPLES_DIR, $FIRECROWN_SITE_PACKAGES variables which points to the current build directory. Currently the default setup does NOT entirely complete the cosmosis set up and does not run source cosmosis-configure. However, this can be enabled by doing: source setup_td_int.sh -c.

To run a test of cosmosis + firecrown after doing source setup_td_int.sh please try: cosmosis $FIRECROWN_DIR/examples/des_y1_3x2pt/des_y1_3x2pt.ini

To Request Additional Packages, Report Problems, or Submit Questions

Please open an issue on this repository.

To Add New Packages to the td_env environment

1. Clone this repo and checkout the "integration" branch
2. Add new package names to the td_env/conda/packlist.txt (for conda-forge) or td_env/conda/piplist.txt (for PyPI)
   • If this package is not installable from conda-forge or PyPI, please open an issue on this repository.
3. Commit and push your changes to the integration branch
4. Once changes are pushed to the integration branch automated builds will be triggered:
   • Docker builds are handled by GitHub actions and will be triggered immediately. The builds can be followed by viewing the Actions page.
     • A successful build will produce a new image available on Dockerhub: lsstdesc/td_env:integration
   • Every 4 hours, a build is triggered at NERSC.
     • Integration builds are installed in /global/common/software/lsst/cori-haswell-gcc/stack/td_env-int
     • The status of the build can be checked here on GitHub. Clicking "Details" will bring you to the GitLab build logs.
       5. If there are data files or set up commands needed for the new package, then please open an issue on this repository.

Known Issues

Due to changes in how cfitsion >=v4.0.0 handles version checking, some packages that depend on cfitsio are now issuing warnings like

WARNING: version mismatch between CFITSIO header (v4.000999999999999) and linked library (v4.01).

Those packages that depend on cfitsio, such as healpy, need to patch their code to deal with cfitsio's version handling which apparenly involves floats. This warning is safe to ignore, and when LSST Science Pipelines updates packages such as healpy, this issue should disappear.

For Developers

To export a full list of versions installed in the td_env conda environment without build info: conda env export --no-builds > desc-python-env-nersc-install-nobuildinfo.yml