Skip to content

Latest commit

 

History

History
157 lines (114 loc) · 7.48 KB

INSTALL.md

File metadata and controls

157 lines (114 loc) · 7.48 KB
Raw

Installation instructions

Table of contents

  1. Basic installation
  2. Access to the proper environment on cori.nersc.gov
  3. An alternative installation at NERSC or at CC-IN2P3 for DESC members
  4. Making a local copy of CLMM

Basic procedure

Here we provide a quick guide for a basic instalation, this will install all the packages in your current environment. To create a specific conda environment for CLMM, we recommend you to check the begining of section Access to the proper environment on cori.nersc.gov.

Theory backend installation

First, choose and install a theory backend for CLMM. This can be CCL (versions between 2.7.1.dev10+gf81b59a4 and 3), NumCosmo (v0.19 or later), or cluster_toolkit and they are installable as follows.

To install CCL as the theory/cosmology backend, run

    conda install -c conda-forge pyccl

or

    pip install pyccl

To install NumCosmo, run

    conda install -c conda-forge numcosmo

Now, to install cluster-toolkit, cluster-toolkit has a gsl dependency, you'll also need gsl.

    conda install gsl
    git clone https://github.com/tmcclintock/cluster_toolkit.git
    cd cluster_toolkit
    python setup.py install
    cd ..

Note: While cluster-toolkit mentions the potential need to install CAMB/CLASS for all cluster-toolkit functionality, you do not need to install these to run CLMM.

Note, you may choose to install some or all of the ccl, numcosmo, and/or cluster_toolkit packages. You need at least one. If you install cluster_toolkit and others, then you need to install cluster_toolkit last. If you have already installed cluster_toolkit before the other packages, simply run, pip uninstall cluster_toolkit then re-install cluster_toolkit.

CLMM and dependency installation

Now, you can install CLMM and its dependencies as

    pip install numpy scipy astropy matplotlib
    pip install pytest sphinx sphinx_rtd_theme
    pip install jupyter  # need to have jupyter notebook tied to this environment, you can then see the environment in jupyter.nersc.gov
    git clone https://github.com/LSSTDESC/CLMM.git  # If you'd like to contribute but don't have edit permissions to the CLMM repo, see below how to fork the repo instead.
    cd CLMM
    python setup.py install     # build from source

Local environment for CLMM

Alternatively, you can create a new local environment by running

    conda env create -f environment.yml
    conda activate clmm

You can now install CLMM in a local and stable environment with the usual procedure.

Access to the proper environment on cori.nersc.gov

If you have access to NERSC, this will likely be the easiest to make sure you have the appropriate environment. After logging into cori.nersc.gov, you will need to execute the following. We recommend executing line-by-line to avoid errors:

    module load python  # Also loads anaconda
    conda create --name clmmenv  # Create an anaconda environment for clmm
    source activate clmmenv  # switch to your newly created environment
    conda install pip  # need pip to install everything else necessary for clmm
    conda install ipython # need to have the ipython tied to this environment
    conda install -c conda-forge firefox  # Need a browser to view jupyter notebooks

Note, for regular contributions and use, we recommend adding module load python to your ~/.bashrc so you have anaconda installed every time you log in. You will subseqeuntly also want to be in the correct environment whenever working with clmm, which means running source activate clmmenv at the start of each session.

Once in your CLMM conda env, you may follow the basic procedure to install CLMM and its dependencies.

The above allows you to develop at NERSC and run pytest. Your workflow as a developer would be to make your changes, do a python setup.py install then pytest to make sure your changes did not break any tests.

If you are a DESC member you may also add to your CLMM environment the GCR and GCRCatalog packages to access the DC2 datasets at NERSC. To run the DC2 example notebooks provided in CLMM, the following need to be installed in your CLMM environment at NERSC. Once in your CLMM environment (source activate clmmenv), run

    pip install pandas
    pip install pyarrow
    pip install healpy
    pip install h5py
    pip install GCR
    pip install https://github.com/LSSTDESC/gcr-catalogs/archive/master.zip
    pip install FoFCatalogMatching

To open up a notebook from NERSC in your browser, you will need to go to the nersc jupyter portal and sign in. You will need to make this conda environment available in the kernel list:

    python -m ipykernel install --user --name=conda-clmmenv

Clicking on the upper right corner of the notebook will provide options for your kernel. Choose the kernel conda-clmmenv that you just created. You will need to do a temporary install of both cluster_toolkit and clmm in the first cell of your jupyter notebook:

def install_clmm_pipeline(upgrade=False):
    import sys
    try:
        import clmm
        import cluster_toolkit
        installed = True
    except ImportError:
        installed = False
    if not upgrade:
        print('clmm is already installed and upgrade is False')
    else:
        !{sys.executable} -m pip install --user --upgrade git+https://github.com/tmcclintock/cluster_toolkit.git
        !{sys.executable} -m pip install --user --upgrade git+https://github.com/LSSTDESC/CLMM
install_clmm_pipeline(upgrade=True)  # Comment this if you do not need to adjust your environment, but this is useful in cori

An alternative installation at NERSC or at CC-IN2P3 for DESC members

The LSST-DESC collabration has setup a specific conda python environment at both NERSC and CC-IN2P3. See instructions there. This conda environment comes with ready access to DESC specific ressources and software such as DC2 catalogs, GCRCatalogs or CCL. Cloning that conda environment and proceeding from there makes the installation procedure lighter as some packages won't need to be installed.

Making a local copy of CLMM

As a newcomer, you likely will not have edit access to the main CLMM repository. Without edit privileges, you won't be able to create or push changes to branches in the base repository. You can get around this by creating a fork, a linked copy of the CLMM repository under your Github username. You can then push code changes to your fork which can later be merged with the base repository.

To create a fork, navigate to the CLMM home page and click 'Fork' in the upper right hand corner. The fork has been created under your username on Github's remote server and can now be cloned to your local repository with

    git clone [email protected]:YOUR-USERNAME/CLMM.git
    cd CLMM
    git remote add base [email protected]:LSSTDESC/CLMM.git

If you do have edit privileges to CLMM, it may be easier to simply clone the base CLMM repository.

    git clone [email protected]:LSSTDESC/CLMM.git