def test_model_not_training(self):
data_tensor = torch.tensor(
[
[
[8, 4, 0, 0, 1, 2],
[9, 7, 2, 7, 3, 8],
[8, 4, 0, 0, 1, 2]
],
[
[8, 4, 0, 0, 1, 2],
[9, 7, 2, 7, 3, 8],
[8, 4, 0, 0, 1, 2]
]
],
dtype=torch.float32
)
dropout_p = 0.9
dim = 0
is_model_training = False
result = pwF.sub_tensor_dropout(data_tensor, dropout_p, dim, is_model_training)
self.assertListEqual(result.tolist(), data_tensor.tolist())
def test_large_p_model_training(self):
data_tensor = torch.tensor(
[
[
[8, 4, 0, 0, 1, 2],
[9, 7, 2, 7, 3, 8],
[8, 4, 0, 0, 1, 2]
],
[
[8, 4, 0, 0, 1, 2],
[9, 7, 2, 7, 3, 8],
[8, 4, 0, 0, 1, 2]
]
],
dtype=torch.float32
)
dropout_p = 0.9999999
dim = 0
is_model_training = True
result = pwF.sub_tensor_dropout(data_tensor, dropout_p, dim, is_model_training)
self.assertListEqual(result[0].tolist(), result[1].tolist())