The easiest way to install the latest grizli release into a fresh virtualenv or conda environment is:

pip install grizli

If you are installing grizli for the first time, make sure to also set up directories and download reference files. If you will be working with HST data, the following will install all necessary libraries:

pip install "grizli[hst]"
conda install hstcal

If you will be working with JWST data, the following is the recommended installation process:

pip install "grizli[jwst]"

More detailed instructions are available below.

Development Environment

Grizli has been developed within a miniconda Python environment. Module dependencies, including general utilities like numpy, scipy, matplotlib, astronomy tools like astropy and specific software for dealing with space-telescope data (, drizzlepac, etc.) are all installed using the standard pip install method (see below). Most development is done in a python 3.9 environment on a MacBook Pro running Mojave 10.14.6. The basic build is tested in (Linux) python 3.7, 3.8 and 3.9 with the GitHub actions continuous integration (CI) tools, but the current test suite does not yet test all of the full functionality of the code.

Setting up a Local Environment

We recommend that grizli be installed in a new virtual environment to minimize conflicts with dependencies for other packages. Possible virtual environment managers are conda, venv, pyenv, etc. Here we give example with setting up a conda environment.

  • Generate a conda environment named grizli39 (or anything else you prefer). This will just provide the base Python distribution (along with pip):

conda create --name grizli39 python=3.9
  • Activate the environment. If you chose a name other than grizli39, substitute that below:

conda activate grizli39


The activation needs to be done each time you start a new terminal. Alternatively, if you want it automatically done for every new terminal, you need to put the above command in your ~/.bashrc file.

Preferred installation with pip

The latest release of grizli can be installed with pip:

pip install grizli

There are five available options for installing dependencies: hst, jwst, aws, test and docs. These can be installed as follows:

pip install "grizli[hst]"


pip install "grizli[jwst]"


pip install "grizli[jwst,test]"

To minimize conflict of dependencies, install only the ones that you need.

Additional dependencies

pip will install all needed dependencies. If you will be working with HST data, you will also need the hstcal library which is only available via conda:

conda install hstcal


If you are planning to run simultaneous fits to grism spectra plus photometry using the eazy-py connection, install eazy-py to ensure that you get its dependencies and templates.

git clone
cd eazy-py
pip install .
  • Download the templates (in a Python interpreter):

import eazy
  • Optional: Run basic tests with pytest. Note that the pysynphot failure is not critical:


Set up directories and fetch additional files

grizli requires a several environmental variables to be set that point to the directory location of configuration files. The export lines below can be put into the ~/.bashrc or ~/.bash_profile setup files so that the system variables are set automatically when you start a new terminal/shell session.

export GRIZLI="${HOME}/grizli" # or anywhere else
export iref="${GRIZLI}/iref/"  # for WFC3 calibration files
export jref="${GRIZLI}/jref/"  # for ACS calibration files
  • Create these directories, assuming that they do not already exist:

mkdir $GRIZLI
mkdir $GRIZLI/CONF      # needed for grism configuration files
mkdir $GRIZLI/templates # for redshift fits
mkdir $iref
mkdir $jref
  • Download the calibration and configuration files not provided with the code repository. Helper scripts are provided to download files that are currently hard-coded. HST calibrations will be downloaded to the $iref and $jref directories. Set get_acs=True to get files necessary for G800L processing:

import grizli.utils

Configuration files will be downliaded to the $GRIZLI/CONF directory. Set get_jwst=True to get config files for JWST processing:

grizli.utils.fetch_config_files(get_acs=False, get_jwst=False)
  • The grism redshift fits require galaxy SED templates that are provided with the repository but that need to be in a specific directory, $GRIZLI/templates. This is done so that users can modify/add templates in that directory without touching the files in the repository itself. For default processing they can by symlinked from the repository. Set force=True to symlink files even if they already exist in $GRIZLI/templates/:

import grizli.utils
  • Run basic tests with pytest:

pip install ".[test]"

Installing grizli from source

If you need to install ``grizli` form a specific branch or need an editable version of the library, you can do this directly from the repository.

  • Create a dedicated environment. See instructions above.

  • Change into a directory where the grizli repo will live.

  • Fetch the grizli repo and change into the newly cloned directory:

git clone
cd grizli
  • If you are installing from a branch, checkout the branch.

  • Compile and install the grizli module. This only needs to be done once (on initial clone), or after updating the repository (e.g., after a git pull).

pip install -e .

The -e flag stands for editable. Or to install the optional dependencies:

pip install -e ".[jwst,test]"

See above for the additional dependencies that need to be installed.

Using HST Files Staged on AWS

grizli can automatically pull FITS files from the public AWS S3 bucket mirror of the HST archive, which can be useful when running the full HST reduction pipeline. This requires that the AWS command line tools and the boto3 and awscli modules be installed:

# Put your AWS credentials, etc. in ~/.aws
pip install grizli ".[aws]"