def test_get_class_counts(test_output_dirs: OutputFolderForTests) -> None:
"""
Test training and testing of sequence models that predicts at multiple time points,
when it is started via run_ml.
"""
dataset_contents = _get_multi_label_sequence_dataframe()
config = ToyMultiLabelSequenceModel(should_validate=False)
assert config.get_target_indices() == [1, 2, 3]
expected_prediction_targets = ["Seq_pos 01", "Seq_pos 02", "Seq_pos 03"]
assert len(config.get_target_indices()) == len(
expected_prediction_targets) # type: ignore
config.set_output_to(test_output_dirs.root_dir)
config.dataset_data_frame = dataset_contents
config.pre_process_dataset_dataframe()
splits = config.get_dataset_splits()
train_dataset = config.create_torch_datasets(splits)[
ModelExecutionMode.TRAIN]
class_counts = train_dataset.get_class_counts()
assert class_counts == {0: 2}
def test_run_ml_with_multi_label_sequence_in_crossval(test_output_dirs: TestOutputDirectories) -> None:
"""
Test training and testing of sequence models that predicts at multiple time points,
including aggregation of cross validation results.
"""
logging_to_stdout()
config = ToyMultiLabelSequenceModel(should_validate=False)
assert config.get_target_indices() == [1, 2, 3]
expected_prediction_targets = ["Seq_pos 01", "Seq_pos 02", "Seq_pos 03"]
target_indices = config.get_target_indices()
assert target_indices
assert len(target_indices) == len(expected_prediction_targets)
config.set_output_to(test_output_dirs.root_dir)
config.dataset_data_frame = _get_multi_label_sequence_dataframe()
config.pre_process_dataset_dataframe()
config.num_epochs = 1
config.number_of_cross_validation_splits = 2
azure_config = get_default_azure_config()
azure_config.train = True
MLRunner(config, azure_config).run()