def get_default_preprocessor(
cls,
truncated_mode: str = 'pre',
truncated_length_left: typing.Optional[int] = None,
truncated_length_right: typing.Optional[int] = None,
filter_mode: str = 'df',
filter_low_freq: float = 1,
filter_high_freq: float = float('inf'),
remove_stop_words: bool = False,
ngram_size: typing.Optional[int] = 3,
) -> BasePreprocessor:
"""
Model default preprocessor.
The preprocessor's transform should produce a correctly shaped data
pack that can be used for training.
:return: Default preprocessor.
"""
return preprocessors.BasicPreprocessor(
truncated_mode=truncated_mode,
truncated_length_left=truncated_length_left,
truncated_length_right=truncated_length_right,
filter_mode=filter_mode,
filter_low_freq=filter_low_freq,
filter_high_freq=filter_high_freq,
remove_stop_words=remove_stop_words,
ngram_size=ngram_size
)
def test_drmm_padding(train_raw):
preprocessor = preprocessors.BasicPreprocessor()
data_preprocessed = preprocessor.fit_transform(train_raw, verbose=0)
embedding_matrix = load_from_file(embeddings.EMBED_10_GLOVE, mode='glove')
term_index = preprocessor.context['vocab_unit'].state['term_index']
embedding_matrix = embedding_matrix.build_matrix(term_index)
histgram_callback = callbacks.Histogram(embedding_matrix=embedding_matrix,
bin_size=30,
hist_mode='LCH')
dataset = Dataset(data_preprocessed,
mode='point',
callbacks=[histgram_callback])
pre_fixed_padding = callbacks.DRMMPadding(fixed_length_left=5,
fixed_length_right=5,
pad_mode='pre')
dataloader = DataLoader(dataset, batch_size=5, callback=pre_fixed_padding)
for batch in dataloader:
assert batch[0]['text_left'].shape == (5, 5)
assert batch[0]['text_right'].shape == (5, 5)
assert batch[0]['match_histogram'].shape == (5, 5, 30)
post_padding = callbacks.DRMMPadding(pad_mode='post')
dataloader = DataLoader(dataset, batch_size=5, callback=post_padding)
for batch in dataloader:
max_left_len = max(batch[0]['length_left'].detach().cpu().numpy())
max_right_len = max(batch[0]['length_right'].detach().cpu().numpy())
assert batch[0]['text_left'].shape == (5, max_left_len)
assert batch[0]['text_right'].shape == (5, max_right_len)
assert batch[0]['match_histogram'].shape == (5, max_left_len, 30)
def get_default_preprocessor(cls) -> BasePreprocessor:
"""
Model default preprocessor.
The preprocessor's transform should produce a correctly shaped data
pack that can be used for training.
:return: Default preprocessor.
"""
return preprocessors.BasicPreprocessor()
def get_default_preprocessor(cls,
truncated_mode: str = 'pre',
truncated_length_left: int = 10,
truncated_length_right: int = 40,
filter_mode: str = 'df',
filter_low_freq: float = 1,
filter_high_freq: float = float('inf'),
remove_stop_words: bool = False,
ngram_size: int = 3):
""":return: Default preprocessor."""
return preprocessors.BasicPreprocessor(
truncated_mode=truncated_mode,
truncated_length_left=truncated_length_left,
truncated_length_right=truncated_length_right,
filter_mode=filter_mode,
filter_low_freq=filter_low_freq,
filter_high_freq=filter_high_freq,
remove_stop_words=remove_stop_words,
ngram_size=ngram_size)
def test_basic_padding(train_raw):
preprocessor = preprocessors.BasicPreprocessor()
data_preprocessed = preprocessor.fit_transform(train_raw, verbose=0)
dataset = Dataset(data_preprocessed, mode='point')
pre_fixed_padding = callbacks.BasicPadding(fixed_length_left=5,
fixed_length_right=5,
pad_mode='pre')
dataloader = DataLoader(dataset, batch_size=5, callback=pre_fixed_padding)
for batch in dataloader:
assert batch[0]['text_left'].shape == (5, 5)
assert batch[0]['text_right'].shape == (5, 5)
post_padding = callbacks.BasicPadding(pad_mode='post')
dataloader = DataLoader(dataset, batch_size=5, callback=post_padding)
for batch in dataloader:
max_left_len = max(batch[0]['length_left'].numpy())
max_right_len = max(batch[0]['length_right'].numpy())
assert batch[0]['text_left'].shape == (5, max_left_len)
assert batch[0]['text_right'].shape == (5, max_right_len)
