To run all the workshop notebooks on your own computer, please be sure your machine is configured with the packages in the installation check file. These packages are the ones we use to verify that the notebooks are working as expected.
These instructions describe setup using git
and Miniconda
. It is not strictly necessary to use either of these.
If you have any problems with any of these steps, please check if your problem has already been reported as an issue. If not, please create a new issue to ask your question.
If you are using Windows, we now recommend using the Windows Subsystem for Linux (WSL) instead of using native Windows tools. WSL is now fully supported by Microsoft and tends to result in fewer install headaches, and lets you use tools that were developed for Linux seamlessly in windows. While you still may be able to use the Windows-native installation of Miniconda, these instructions focus on the WSL approach for the above reasons.
To install WSL, you should follow the instructions provided by Microsoft:
https://docs.microsoft.com/en-us/windows/wsl/install
While you may choose an alternative Linux distribution from the default Ubuntu, the instructions below have been tested on Ubuntu, so unless you have a specific reason, we suggest you stick with the default. Once you reach the point in the instructions with a working Linux terminal prompt, you can proceed to step 1 of these instructions.
Optional While you can run a WSL terminal with the command prompt built into Windows, it's rather bare-bones and you may not have the best experience. For WSL on Windows you'll probably want to install Windows Terminal to have a terminal experience closer to what you'd see on Mac or Linux.
Miniconda is a free minimal installer for
conda
. It is a small, bootstrap version of Anaconda that includes onlyconda
, Python, the packages they depend on, and a small number of other useful packages likepip
,zlib
etc. If you have already installed Miniconda or Anaconda, you can skip to the next step.
In a terminal window, check if Miniconda is already installed:
conda info
If Miniconda is not already installed, follow these instructions for your operating system:
https://conda.io/projects/conda/en/latest/user-guide/install/index.html
Please be sure to install a 64-bit version of Miniconda to ensure that all packages work correctly.
(On native Windows, you might also need additional compilers, although this should not be necessary in WSL).
Miniconda includes an environment manager called
conda
. An environment manager allows you to have multiple installations of Python, including packages and versions, installed on your computer. You can create, export, list, remove, and update environments that have different versions of Python and / or packages installed in them. For this workshop, we will configure an environment using theconda
command line utility.
Open a terminal window and verify that conda is working:
conda info
If you are having trouble, check your shell in a terminal window:
echo $SHELL
then run the initialization, if needed, in that same terminal window:
conda init $SHELL
On Windows, open the Anaconda Prompt
terminal app.
(An alternative to using conda is mamba which is a reimplementation of the conda package manager in C++.)
Note: you will need
conda
version4.6
or later. You can update yourconda
installation withconda update conda
At the prompt, check whether Git is already installed:
git --version
If the output shows a Git version, skip to the next step. Otherwise, install Git:
conda install git
If using git
, clone the workshop repository using
git:
git clone https://github.com/spacetelescope/roman-data-workshop.git
If you elect not to use git
, you can download the ZIP file by opening the green Code
button at
https://github.com/spacetelescope/roman-data-workshop and selecting Download ZIP
.
Miniconda includes an environment manager called conda. Environments allow you to have multiple sets of Python packages installed at the same time, making reproducibility and upgrades easier. You can create, export, list, remove, and update environments that have different versions of Python and/or packages installed in them.
Create a conda environment for this workshop using a yml file
. The python version and all needed packages are listed in
environment.yml
.
Open a terminal window using the appropriate one for your operating system.
Now navigate to this directory in the terminal:
cd roman-data-workshop
And finally, on any platform, to install and activate the roman-data-workshop-env
environment, type:
conda env create --file 00_install/environment.yml
conda activate roman-data-workshop-env
The name of the new conda environment created above should now be displayed next to the terminal
prompt: (roman-data-workshop-env)
In the terminal you used in the preceding step, run the verify_install.py
script to check the Python environment and
some required dependencies:
python 00_install/verify_install.py
If the script reports that some versions do not match for a specific package (for example numpy
), check first whether
the package was installed using conda
or pip
by using conda list <package>
:
conda list numpy
If the package was installed with pip
, the Build
and Channel
columns will include pypi
:
# packages in environment at /opt/miniconda3/envs/test:
#
# Name Version Build Channel
numpy 1.22.4 pypi_0 pypi
and then you can upgrade with pip install --upgrade <package>
:
pip install --upgrade numpy
If you need a pre-release version from PyPI, add
--pre
to thepip install
command.
Otherwise, if the package is installed with conda
, it will show something similar to the following:
# packages in environment at /opt/miniconda3/envs/test:
#
# Name Version Build Channel
numpy 1.22.3 py310hdcd3fac_0
numpy-base 1.22.3 py310hfd2de13_0
and you can update with conda update <package>
:
conda update numpy
Making sure to activate the conda
environment (conda activate roman-data-workshop-env
), run the following to
download the data files used by these notebooks:
python data/download.py
Making sure your terminal is in the roman-data-workshop
directory (you can use pwd
to check), you can then start the
Jupyter server on your local computer, with which you can view the Jupyter notebooks:
jupyter notebook
If successful, your browser will open a new page / tab pointing to localhost
, showing a listing of the current
directory (including subdirectories).
Click into one of the notebook directories, double-click on a notebook, and wait for it to launch. In the top right
corner, if you see a blue Kernel Ready
message appear and disappear, then all is well.
If you see a red Kernel Error
in the top right corner, click on it and scroll down to see the error message. If it
says FileNotFoundError
, shut down the notebook server on your terminal and run this command:
python -m ipykernel install --user
Now, try running jupyter notebook
again as above, and the Kernel Error
should be fixed. You can try running the first cell (usually an import
) to check.