def test_check_equal():
"""
Make sure our helper function for checking tensor equality actually works.
"""
a = torch.tensor([[1, 2], [3, 4]])
b = torch.tensor([[1, 2], [3, 4]])
utils.assert_equal(a, b)
def test_target(dataset):
"""Make sure the `target` attr contains what we expect."""
assert len(dataset.target) == 4
assert_equal(dataset.target[0], torch.tensor([0, 0]))
assert_equal(dataset.target[1], torch.tensor([0, 0, 0, 1]))
assert_equal(dataset.target[2], torch.tensor([0, 0, 1, 2]))
assert_equal(dataset.target[3], torch.tensor([0]))
def test_source(dataset):
"""Make sure the `source` attr contains what we expect."""
assert len(dataset.source) == 4
# Check the first item, which comes from the sentence:
# [("hi", "O"), ("there", "O")]
words, word_lens, idxs, word_idxs, context = dataset.source[0]
assert isinstance(words, torch.Tensor)
assert isinstance(word_lens, torch.Tensor)
assert isinstance(idxs, torch.Tensor)
assert isinstance(word_idxs, torch.Tensor)
assert context is None
assert list(words.size()) == [2, 5]
assert_equal(word_lens, torch.tensor([5, 2]))
assert_equal(idxs, torch.tensor([1, 0]))
assert list(word_idxs.size()) == [2]
def test_sent2tensor(vocab, sent):
"""Check that Vocab.sent2tensor has the correct output format."""
char_tensors, word_lengths, word_idxs, word_tensors, context = \
vocab.sent2tensor(sent)
check_lens = [len(s) for s in sent]
check_sorted_lens = sorted(check_lens, reverse=True)
check_idxs = sorted(range(len(sent)),
reverse=True,
key=lambda i: check_lens[i])
# Verify sizes.
assert isinstance(char_tensors, torch.Tensor)
assert list(char_tensors.size()) == [len(sent), max(check_lens)]
assert isinstance(word_tensors, torch.Tensor)
assert list(word_tensors.size()) == [len(sent)]
# Verify order of word lengths and idxs.
assert_equal(word_lengths, torch.tensor(check_sorted_lens))
assert_equal(word_idxs, torch.tensor(check_idxs))
for i, word_tensor in enumerate(char_tensors):
check_word = sent[check_idxs[i]]
check_word_tensor = torch.tensor(
[vocab.chars_stoi[c]
for c in check_word] + [0] * (max(check_lens) - len(check_word)))
assert_equal(word_tensor, check_word_tensor)
def test_labs2tensor(vocab, labs, is_test, check):
res = vocab.labs2tensor(labs, test=is_test)
assert_equal(res, check)
def test_unsort(inputs, check):
utils.assert_equal(utils.unsort(*inputs), check)
def test_sort_and_pad(inputs, check):
padded, lens, idx = utils.sort_and_pad(*inputs)
utils.assert_equal(padded, check[0])
utils.assert_equal(lens, check[1])
utils.assert_equal(idx, check[2])
def test_pad(inputs, check):
utils.assert_equal(utils.pad(*inputs), check)
def test_sequence_mask(inp, chk, max_len):
"""Test `sequence_mask()` method."""
res = utils.sequence_mask(inp, max_len=max_len)
utils.assert_equal(res, chk)