def main():
exp = nli.experiment(train_reader=nli.SNLITrainReader(SNLI_HOME,
samp_percentage=1.0),
assess_reader=nli.SNLIDevReader(SNLI_HOME,
samp_percentage=1.0),
phi=sentence_encoding_rnn_phi,
train_func=fit_bilstm_attention,
random_state=None,
vectorize=False)
print(exp)
"multinli_1.0_matched_annotations.txt",
"multinli_1.0_mismatched_annotations.txt"
])
def test_read_annotated_subset(src_filename):
src_filename = os.path.join(annotations_home, src_filename)
data = nli.read_annotated_subset(src_filename, multinli_home)
assert len(data) == 495
def test_build_dataset():
nli.build_dataset(reader=nli.SNLITrainReader(snli_home,
samp_percentage=0.01),
phi=lambda x, y: {"$UNK": 1},
vectorizer=None,
vectorize=True)
@pytest.mark.parametrize("assess_reader", [None, nli.SNLIDevReader(snli_home)])
def test_experiment(assess_reader):
def fit_maxent(X, y):
mod = LogisticRegression(solver='liblinear', multi_class='auto')
mod.fit(X, y)
return mod
nli.experiment(train_reader=nli.SNLITrainReader(snli_home,
samp_percentage=0.01),
phi=lambda x, y: {"$UNK": 1},
train_func=fit_maxent,
assess_reader=assess_reader,
random_state=42)
def test_read_annotated_subset(src_filename):
src_filename = os.path.join(
annotations_home, src_filename)
data = nli.read_annotated_subset(src_filename, multinli_home)
assert len(data) == 495
def test_build_dataset():
nli.build_dataset(
reader=nli.SNLITrainReader(snli_home, samp_percentage=0.01),
phi=lambda x, y: {"$UNK": 1},
vectorizer=None,
vectorize=True)
@pytest.mark.parametrize("assess_reader", [
None,
nli.SNLIDevReader(snli_home)
])
def test_experiment(assess_reader):
def fit_maxent(X, y):
mod = LogisticRegression(solver='liblinear', multi_class='auto')
mod.fit(X, y)
return mod
nli.experiment(
train_reader=nli.SNLITrainReader(snli_home, samp_percentage=0.01),
phi=lambda x, y: {"$UNK": 1},
train_func=fit_maxent,
assess_reader=assess_reader,
random_state=42)