pytest-pytorch
pytest plugin for a better developer experience when working with the PyTorch test suite
Why do I need it?
Some testcases in the PyTorch test suite are only used as templates and will be instantiated at runtime. Unfortunately, [PyTorch]'s naming scheme for parametrizations differs from [pytest]'s. As a consequence, these tests cannot be selected by their names as written and one has to remember [PyTorch]'s naming scheme. This can be especially disrupting to your workflow if your IDE (PyCharm, VSCode) relies on [pytest]'s default selection syntax.
If this has ever been a source of frustration for you, worry no longer. pytest-pytorch was made for you.
How do I install it?
You can install pytest-pytorch with pip
$ pip install pytest-pytorch
or with conda:
$ conda install -c conda-forge pytest-pytorch
How do I use it?
With pytest-pytorch installed you can select test cases and tests by their names as written:
| Use case | Command |
|---|---|
| Run a test case against all devices | pytest test_foo.py::TestBar |
| Run a test against all devices | pytest test_foo.py::TestBar::test_baz |
Similar to a parametrization by @pytest.mark.parametrize you can use the -k flag to select a specific set of parameters:
| Use case | Command |
|---|---|
| Run a test case against one device | pytest test_foo.py::TestBar -k "$DEVICE" |
| Run a test against one device | pytest test_foo.py::TestBar::test_baz -k "$DEVICE" |
Can I have a little more background?
Sure, we have written a blog post about pytest-pytorch that goes into details.
