def test_fit_predict(self):
"""
Ensure model training does not error out
Ensure model returns predictions of the right type
"""
model = Comparison(**self.default_config())
n_samples = 10
model.fit([[
"Transformers was a terrible movie but a great model",
"Transformers are a great model but a terrible movie"
]] * n_samples, ['yes'] * n_samples)
test_data = [[
"Transformers was a terrible movie but a great model",
"Transformers are a great model but a terrible movie"
]]
predictions = model.predict(test_data)
for prediction in predictions:
self.assertIsInstance(prediction, (str, bytes))
probabilities = model.predict_proba(test_data)
for proba in probabilities:
self.assertIsInstance(proba, dict)
def test_fit_predict(self):
"""
Ensure model training does not error out
Ensure model returns predictions of the right type
"""
model = Comparison(**self.default_config())
n_samples = 10
model.fit(["Indico is the best"] * n_samples, ["Indico is the bestestestest"] * n_samples, ['yes'] * n_samples)
predictions = model.predict(["Is indico the best?"], ["Indico is the bestestestest"])
for prediction in predictions:
self.assertIsInstance(prediction, (str, bytes))
def test_comparison_auxiliary(self):
"""
Ensure model training does not error out
Ensure model returns reasonable predictions
"""
model = Comparison(**self.default_config(
chunk_long_sequences=False, max_length=50, batch_size=4))
trainX = [['i like apples', 'i like apples']] * 4
trainY = ['A', 'B', 'C', 'D']
train_context = [[self.train_context[i], self.train_context[j]]
for i in [0, 1] for j in [0, 1]]
model.fit(trainX, trainY, context=train_context)
preds = model.predict(trainX, context=train_context)
def __init__(self, filename=None, **kwargs):
super().__init__(filename=(filename or DATA_PATH), **kwargs)
@property
def md5(self):
return CHECKSUM
def download(self):
"""
Download quora duplicate questions dataset.
"""
path = Path(self.filename)
path.parent.mkdir(parents=True, exist_ok=True)
comparison_download(
url="http://qim.ec.quoracdn.net/quora_duplicate_questions.tsv",
text_column1="question1",
text_column2="question2",
target_column="is_duplicate",
filename=QUORA_SIMILARITY
)
if __name__ == "__main__":
# Train and evaluate on SST
dataset = QuoraDuplicate(nrows=5000).dataframe
model = Comparison(verbose=True, n_epochs=3)
trainX1, testX1, trainX2, testX2, trainY, testY = train_test_split(dataset.Text1, dataset.Text2, dataset.Target, test_size=0.3, random_state=42)
model.fit(trainX1, trainX2, trainY)
accuracy = np.mean(model.predict(testX1, testX2) == testY)
class_balance = np.mean(testY)
print('Test Accuracy: {:0.2f} for a {:0.2f} class balance'.format(accuracy, class_balance))
return CHECKSUM
def download(self):
"""
Download quora duplicate questions dataset.
"""
path = Path(self.filename)
path.parent.mkdir(parents=True, exist_ok=True)
comparison_download(url="https://s3.amazonaws.com/enso-data/Quora.csv",
text_column1="Text1",
text_column2="Text2",
target_column="Target",
filename=QUORA_SIMILARITY)
if __name__ == "__main__":
# Train and evaluate on SST
dataset = QuoraDuplicate(nrows=5000).dataframe
model = Comparison(n_epochs=1)
trainX1, testX1, trainX2, testX2, trainY, testY = train_test_split(
dataset.Text1.values,
dataset.Text2.values,
dataset.Target.values,
test_size=0.3,
random_state=42)
model.fit(list(zip(trainX1, trainX2)), trainY)
accuracy = np.mean(model.predict(list(zip(testX1, testX2))) == testY)
class_balance = np.mean(testY)
print('Test Accuracy: {:0.2f} for a {:0.2f} class balance'.format(
accuracy, class_balance))
def __init__(self, filename=None, **kwargs):
super().__init__(filename=(filename or DATA_PATH), **kwargs)
@property
def md5(self):
return CHECKSUM
def download(self):
"""
Download quora duplicate questions dataset.
"""
path = Path(self.filename)
path.parent.mkdir(parents=True, exist_ok=True)
comparison_download(
url="http://qim.ec.quoracdn.net/quora_duplicate_questions.tsv",
text_column1="question1",
text_column2="question2",
target_column="is_duplicate",
filename=QUORA_SIMILARITY
)
if __name__ == "__main__":
# Train and evaluate on SST
dataset = QuoraDuplicate(nrows=5000).dataframe
model = Comparison(verbose=True, n_epochs=1)
trainX1, testX1, trainX2, testX2, trainY, testY = train_test_split(dataset.Text1, dataset.Text2, dataset.Target, test_size=0.3, random_state=42)
model.fit(list(zip(trainX1, trainX2)), trainY)
accuracy = np.mean(model.predict(list(zip(testX1, testX2))) == testY)
class_balance = np.mean(testY)
print('Test Accuracy: {:0.2f} for a {:0.2f} class balance'.format(accuracy, class_balance))
