def convert_to_config_list(initial_list):
if initial_list is None or initial_list == []:
raise ValueError(
"manifest_filepaths and tarred_audio_filepaths must not be empty.")
if not isinstance(initial_list, ListConfig):
initial_list = ListConfig([initial_list])
for list_idx, list_val in enumerate(initial_list):
if type(list_val) != type(initial_list[0]):
raise ValueError(
"manifest_filepaths and tarred_audio_filepaths need to be a list of lists for bucketing or just a list of strings"
)
if type(initial_list[0]) is not ListConfig:
initial_list = ListConfig([initial_list])
return initial_list
def test_lottery_transform(self):
"""
test the lottery transform when params are indicated in the yaml
"""
pos = torch.randn(10000, 3)
x = torch.randn(10000, 6)
dummy = torch.randn(10000, 6)
data = Data(pos=pos, x=x, dummy=dummy)
conf = ListConfig([{"transform": "GridSampling3D", "params": {"size": 0.1}}, {"transform": "Center"},])
tr = LotteryTransform(transform_options=conf)
tr(data)
self.assertIsInstance(tr.random_transforms.transforms[0], GridSampling3D)
self.assertIsInstance(tr.random_transforms.transforms[1], T.Center)
def test_InstantiateTransforms(self):
conf = ListConfig([
{
"transform": "GridSampling",
"params": {
"size": 0.1
}
},
{
"transform": "Center"
},
])
t = instantiate_transforms(conf)
self.assertIsInstance(t.transforms[0], GridSampling)
self.assertIsInstance(t.transforms[1], T.Center)