def __init__(self,
train_biom,
test_biom,
valid_biom,
metadata=None,
batch_category=None,
batch_size=10,
num_workers=1):
super().__init__()
self.train_biom = train_biom
self.test_biom = test_biom
self.val_biom = valid_biom
self.batch_size = batch_size
self.num_workers = num_workers
if metadata is not None:
self.metadata = pd.read_table(metadata, dtype=str)
index_name = self.metadata.columns[0]
self.metadata = self.metadata.set_index(index_name)
else:
self.metadata = None
self.batch_category = batch_category
if self.batch_category is None:
self.collate_f = collate_single_f
else:
self.collate_f = collate_batch_f
# collect class mappings if they exist
if batch_category is not None:
train_dataset = BiomDataset(load_table(self.train_biom),
metadata=self.metadata,
batch_category=self.batch_category)
self.batch_categories = train_dataset.batch_cats
def test_dataloader(self):
test_dataset = BiomDataset(load_table(self.hparams.test_biom))
test_dataloader = DataLoader(test_dataset,
batch_size=self.hparams.batch_size,
collate_fn=collate_single_f,
shuffle=False,
num_workers=self.hparams.num_workers,
drop_last=True,
pin_memory=True)
return test_dataloader
def test_biom_getitem(self):
data = BiomDataset(self.table, self.metadata, batch_category='batch')
exp_sample = np.array([
65., 66., 12., 94., 37., 43., 97., 69., 6., 22., 87., 43., 87., 5.,
51., 53., 26., 54., 51., 76., 15., 92., 30., 43., 97., 98., 7.,
43., 25., 51., 75., 39., 13., 90., 89., 48., 60., 79., 9., 97.,
35., 47., 13., 44., 70., 94., 80., 62., 99., 73.
])
batch = data[0]
npt.assert_allclose(batch[0], exp_sample)
npt.assert_allclose(batch[1], np.array(0))
def test_dataloader(self):
test_dataset = BiomDataset(load_table(self.test_biom),
metadata=self.metadata,
batch_category=self.batch_category)
test_dataloader = DataLoader(test_dataset,
batch_size=self.batch_size,
collate_fn=self.collate_f,
shuffle=False,
num_workers=self.num_workers,
drop_last=True,
pin_memory=True)
return test_dataloader
def val_dataloader(self):
val_dataset = BiomDataset(load_table(self.val_biom),
metadata=self.metadata,
batch_category=self.batch_category)
batch_size = min(len(val_dataset) - 1, self.batch_size)
val_dataloader = DataLoader(val_dataset,
batch_size=batch_size,
collate_fn=self.collate_f,
shuffle=False,
num_workers=self.num_workers,
drop_last=True,
pin_memory=True)
return val_dataloader
def test_biom(self):
data = BiomDataset(self.table)