def build(cls, **kwargs):
ParameterValidator.assert_keys_present(
list(kwargs.keys()),
['metadata_file', 'name', 'repertoire_ids', 'metadata_fields'],
RepertoireDataset.__name__, "repertoire dataset")
repertoires = []
metadata_df = pd.read_csv(kwargs['metadata_file'],
comment=Constants.COMMENT_SIGN)
for index, row in metadata_df.iterrows():
filename = Path(kwargs['metadata_file']).parent / row['filename']
if not filename.is_file() and 'repertoires' in str(filename):
filename = filename.parent.parent / Path(row['filename']).name
repertoire = Repertoire(data_filename=filename,
metadata_filename=filename.parent /
f'{filename.stem}_metadata.yaml',
identifier=row['identifier'])
repertoires.append(repertoire)
if "repertoire_ids" in kwargs.keys(
) and "repertoires" not in kwargs.keys(
) and kwargs['repertoire_ids'] is not None:
assert all(rep.identifier == kwargs['repertoire_ids'][i] for i, rep in enumerate(repertoires)), \
f"{RepertoireDataset.__name__}: repertoire ids from the iml_dataset file and metadata file don't match for the dataset " \
f"{kwargs['name']} with identifier {kwargs['identifier']}."
return RepertoireDataset(**{**kwargs, **{"repertoires": repertoires}})
def test_run(self):
dataset = RepertoireDataset(repertoires=[
Repertoire(Path("0.npy"), None, "0"),
Repertoire(Path("0.npy"), None, "8"),
Repertoire(Path("0.npy"), None, "1"),
Repertoire(Path("0.npy"), None, "9"),
Repertoire(Path("0.npy"), None, "2"),
Repertoire(Path("0.npy"), None, "10"),
Repertoire(Path("0.npy"), None, "3"),
Repertoire(Path("0.npy"), None, "11"),
Repertoire(Path("0.npy"), None, "4"),
Repertoire(Path("0.npy"), None, "12"),
Repertoire(Path("0.npy"), None, "5"),
Repertoire(Path("0.npy"), None, "13"),
Repertoire(Path("0.npy"), None, "6"),
Repertoire(Path("0.npy"), None, "14"),
Repertoire(Path("0.npy"), None, "7")
])
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i) for i in range(5)
]
for path in paths:
PathBuilder.build(path)
df = pd.DataFrame(
data={
"key1": [0, 0, 1, 1, 2, 2, 0, 0, 1, 1, 2, 2, 0, 0, 1],
"filename": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
})
df.to_csv(EnvironmentSettings.root_path /
"test/tmp/datasplitter/metadata.csv")
dataset.metadata_file = EnvironmentSettings.root_path / "test/tmp/datasplitter/metadata.csv"
training_percentage = 0.7
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
training_percentage=training_percentage,
split_strategy=SplitType.RANDOM,
split_count=5,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(10, len(trains[0].get_data()))
self.assertEqual(5, len(tests[0].get_data()))
self.assertEqual(5, len(trains))
self.assertEqual(5, len(tests))
self.assertEqual(10, len(trains[0].repertoires))
trains2, tests2 = DataSplitter.run(
DataSplitterParams(dataset=dataset,
training_percentage=training_percentage,
split_strategy=SplitType.RANDOM,
split_count=5,
paths=paths))
self.assertEqual(trains[0].get_repertoire_ids(),
trains2[0].get_repertoire_ids())
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i)
for i in range(dataset.get_example_count())
]
for path in paths:
PathBuilder.build(path)
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
split_strategy=SplitType.LOOCV,
split_count=-1,
training_percentage=-1,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(14, len(trains[0].get_data()))
self.assertEqual(1, len(tests[0].get_data()))
self.assertEqual(15, len(trains))
self.assertEqual(15, len(tests))
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i) for i in range(5)
]
for path in paths:
PathBuilder.build(path)
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
split_strategy=SplitType.K_FOLD,
split_count=5,
training_percentage=-1,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(len(trains[0].get_data()), 12)
self.assertEqual(len(tests[0].get_data()), 3)
self.assertEqual(5, len(trains))
self.assertEqual(5, len(tests))
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
split_strategy=SplitType.STRATIFIED_K_FOLD,
split_count=3,
training_percentage=-1,
paths=paths,
label_config=LabelConfiguration(
[Label("key1", [0, 1, 2])])))
self.assertEqual(len(trains[0].get_data()), 10)
self.assertEqual(len(tests[0].get_data()), 5)
self.assertEqual(3, len(trains))
self.assertEqual(3, len(tests))
for train in trains:
self.assertTrue(
all(cls in train.get_metadata(["key1"])["key1"]
for cls in [0, 1, 2]))
for test in tests:
self.assertTrue(
all(cls in test.get_metadata(["key1"])["key1"]
for cls in [0, 1, 2]))
shutil.rmtree(EnvironmentSettings.root_path / "test/tmp/datasplitter/")
def test_run(self):
dataset = RepertoireDataset(repertoires=[
Repertoire("0.npy", None, "0"),
Repertoire("0.npy", None, "1"),
Repertoire("0.npy", None, "2"),
Repertoire("0.npy", None, "3"),
Repertoire("0.npy", None, "4"),
Repertoire("0.npy", None, "5"),
Repertoire("0.npy", None, "6"),
Repertoire("0.npy", None, "7")
])
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i) for i in range(5)
]
for path in paths:
PathBuilder.build(path)
df = pd.DataFrame(data={
"key1": [0, 0, 1, 1, 1, 2, 2, 0],
"filename": [0, 1, 2, 3, 4, 5, 6, 7]
})
df.to_csv(EnvironmentSettings.root_path /
"test/tmp/datasplitter/metadata.csv")
dataset.metadata_file = EnvironmentSettings.root_path / "test/tmp/datasplitter/metadata.csv"
training_percentage = 0.7
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
training_percentage=training_percentage,
split_strategy=SplitType.RANDOM,
split_count=5,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(len(trains[0].get_data()), 5)
self.assertEqual(len(tests[0].get_data()), 3)
self.assertEqual(5, len(trains))
self.assertEqual(5, len(tests))
self.assertEqual(5, len(trains[0].repertoires))
trains2, tests2 = DataSplitter.run(
DataSplitterParams(dataset=dataset,
training_percentage=training_percentage,
split_strategy=SplitType.RANDOM,
split_count=5,
paths=paths))
self.assertEqual(trains[0].get_repertoire_ids(),
trains2[0].get_repertoire_ids())
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i)
for i in range(dataset.get_example_count())
]
for path in paths:
PathBuilder.build(path)
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
split_strategy=SplitType.LOOCV,
split_count=-1,
training_percentage=-1,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(len(trains[0].get_data()), 7)
self.assertEqual(len(tests[0].get_data()), 1)
self.assertEqual(8, len(trains))
self.assertEqual(8, len(tests))
paths = [
EnvironmentSettings.root_path /
"test/tmp/datasplitter/split_{}".format(i) for i in range(5)
]
for path in paths:
PathBuilder.build(path)
trains, tests = DataSplitter.run(
DataSplitterParams(dataset=dataset,
split_strategy=SplitType.K_FOLD,
split_count=5,
training_percentage=-1,
paths=paths))
self.assertTrue(isinstance(trains[0], RepertoireDataset))
self.assertTrue(isinstance(tests[0], RepertoireDataset))
self.assertEqual(len(trains[0].get_data()), 6)
self.assertEqual(len(tests[0].get_data()), 2)
self.assertEqual(5, len(trains))
self.assertEqual(5, len(tests))
shutil.rmtree(EnvironmentSettings.root_path / "test/tmp/datasplitter/")