A new version of jax caused new mypy errors to show up, requiring e.g. explicit casting to float
A new version of jax appears to be more strict about datatypes, complaining about operations using complex64 and complex128. This is fixed by explicitly casting some types in a jax.numpy.linalg.solve to ensure that all operands share a common type.
The batch-compatible amplitudes_for_fields was overly strict about shapes, requiring fields to have a batch dimension. This caused existing tests to fail. We update the shape validation and add more tests.
A test in fields module failed due to tolerances.
Here, we fix all these errors, and also make some updates to the CI
split CI into multiple jobs run in parallel, so that it completes more quickly
Run CI on a schedule, to ensure that new errors are caught
CI has been failing for several reasons
complex64andcomplex128. This is fixed by explicitly casting some types in ajax.numpy.linalg.solveto ensure that all operands share a common type.amplitudes_for_fieldswas overly strict about shapes, requiring fields to have a batch dimension. This caused existing tests to fail. We update the shape validation and add more tests.fieldsmodule failed due to tolerances.Here, we fix all these errors, and also make some updates to the CI