Small changes to allow classifier to run on embeddings with different dtypes.
Tests in chirp/inference/tests/classify_test.py test for different combinations of embedding and model dtypes and all combinations pass. However, it seems that the problem necessitating this change arises only when running a model of one dtype on a tfrecord dataset of a different type, as in https://github.com/QutEcoacoustics/perch/blob/8b24db69e8e5b980f6431c9b38b5e8b7cb414194/chirp/inference/tf_examples.py#L125, and the current classify_test.py does not include that. To be thorough, a test that does this should be added.
This PR also includes a correction to the baw_utils api to allow the domain to be specified
Small changes to allow classifier to run on embeddings with different dtypes.
Tests in
chirp/inference/tests/classify_test.pytest for different combinations of embedding and model dtypes and all combinations pass. However, it seems that the problem necessitating this change arises only when running a model of one dtype on a tfrecord dataset of a different type, as in https://github.com/QutEcoacoustics/perch/blob/8b24db69e8e5b980f6431c9b38b5e8b7cb414194/chirp/inference/tf_examples.py#L125, and the current classify_test.py does not include that. To be thorough, a test that does this should be added.This PR also includes a correction to the baw_utils api to allow the domain to be specified