Contribution tips

Installing from source

Create a code directory where you will install the virtual env and other TORAX dependencies.

`shell mkdir /path/to/torax_dir && cd "$_" ` Where /path/to/torax_dir should be replaced by a path of your choice.

Create a TORAX virtual env.

`shell python3 -m venv toraxvenv source toraxvenv/bin/activate `

Download and install the TORAX codebase via http:

`shell git clone https://github.com/google-deepmind/torax.git ` or ssh (ensure that you have the appropriate SSH key uploaded to github).

`shell git clone git@github.com:google-deepmind/torax.git ` Enter the TORAX directory and pip install the dependencies.

`shell cd torax; pip install -e . `

If you want to install with the dev dependencies (useful for running pytest and installing pyink for lint checking), then run with the [dev]:

`shell cd torax; pip install -e .[dev] `

Optional: Install additional GPU support for JAX if your machine has a GPU: https://jax.readthedocs.io/en/latest/installation.html#supported-platforms

Code reviews

All submissions, including submissions by project members, require review. We use GitHub pull requests for this purpose. Consult [GitHub Help] for more information on using pull requests.

Documentation

If appropriate for the change you are making, please update the documentation as part of your contribution.

TORAX’s documentation is written in [reStructuredText].

All documentation is found in the torax/docs directory, in .rst files. Please either modify the existing files or create new files and links as appropriate.

Once you have written your documentation, you can stage it to see how it looks before you commit it to the repository. To stage your documentation, run the following command from the docs folder in the TORAX repository:

make html

This will generate the documentation in the docs/_build/html directory. You can then view the documentation by opening the index.html file in your browser. Run pwd to find the path to the docs/_build/html directory and then point your browser to file:///path/to/torax/docs/_build/html/index.html.

The staged documentation is cleaned up from your local repository by running the command:

make clean

Following the submission of your pull request, the documentation will be automatically regenerated and published to TORAX’s readthedocs server.

Testing

TORAX’s tests are written in the [pytest] framework

While all tests are run automatically on GitHub when you push a pull request, it is a good idea to run them locally while you are developing your feature. Ensure that you have installed the required developer dependencies (including pytest) by having run the following command from the TORAX root directory:

pip install -e .[dev]

To run all tests, run the following command from the TORAX root directory:

pytest -n <num_workers>

with <num_workers> being the number of processes that you want to use in parallel.

It is recommended to run tests with the environment variable TORAX_ERRORS_ENABLED=True to enable full test coverage. However, it is then recommended to revert back to TORAX_ERRORS_ENABLED=False when running TORAX in production mode, to enable the persistent JAX cache.

To run a specific test, run the following command from the TORAX root directory, in this case running all the geometry tests.

pytest torax/_src/geometry/tests/geometry_test.py

Further filtering is possible, for example running only the test_face_to_cell test (in geometry.py):

pytest -k face_to_cell

Which runs any test containing the string expression face_to_cell.

Where appropiate, please add tests for your changes.

An important class of test is the sim test. These are integration tests running the configs in the torax/tests/test_data/ directory, and comparing to the ground-truth .nc TORAX outputs found in the same directory. Sim tests can be triggered separately by a command (from the TORAX root directory) such as:

pytest -n <num_workers> torax/tests/sim_test.py

If any sim tests fail, they write their output to the /tmp/torax_failed_sim_test_outputs/<test_name>.nc. This is useful for debugging, and also to stage new output files for replacing the ground-truth files, if you expect that your change to the code produces different outputs.

To compare the absolute and relative differences between the failed sim tests to the ground-truth files, run the following command from the TORAX root directory:

python3 torax/tests/scripts/compare_sim_tests.py

These command has the optional flag --failed_test_output_dir <dir> which takes a directory containing the failed test outputs, instead of the default directory /tmp/torax_failed_sim_test_outputs.

It is sometimes useful to plot the difference between the ground-truth and a failed TORAX sim test, either for debugging or to verify that the magnitude of difference is as expected. To do this, run the following command from the root of the TORAX repository. Using test_qlknnheat as an example:

plot_torax --outfile torax/tests/test_data/test_qlknnheat.nc /tmp/torax_failed_sim_test_outputs/test_qlknnheat.nc

If it is deemed that the new outputs should replace the ground-truth files, they can be copied over using the following command, again with this example working when run from the TORAX repository root:

python3 torax/tests/scripts/copy_sim_tests.py

Where we also have the optional flag --failed_test_output_dir <dir> which takes a directory containing the failed test outputs, instead of the default directory /tmp/torax_failed_sim_test_outputs.

Finally, there are use-cases where it is desirable to rerun all the sim tests, even if the tests are passing. An example is when the output API changes and we wish to keep all the test .nc files up-to-date. In this case, run the following command from the TORAX root directory:

python3 torax/tests/scripts/run_and_save_all_benchmarks.py

This script has the following optional flags:

• --output_dir (default /tmp/torax_sim_outputs): directory where to save the outputs

• --num_proc (default 16): number of processes to use

The compare_sim_tests.py can be used for sanity checking the outputs, and the copy_sim_tests.py can be used to replace the ground-truth files. Note that the --failed_test_output_dir flag in the compare and copy scripts needs to be set to the same output directory as the run_and_save_all_benchmarks.py script.

Important

When making changes to the output structure, e.g. adding fields, a subset of the sim tests will fail. To pass these specific tests, it is required to update implicit.nc, test_changing_config_before.nc, and test_changing_config_after.nc. However, the recommended workflow when changing output API is to run the run_and_save_all_benchmarks.py script, which also updates the aforementioned files. When doing so, it is further strongly recommended to afterwards run the compare_sim_tests.py script to verify that the changes to the ground-truth files are as expected. For pure output API changes, these should be zero. Results of compare_sim_tests.py should be shared in the pull request discussion.