def test_nlp_erroneous_rnn_args_invalid(self):
num_words = 1024
(x_train,
y_train), (x_test,
y_test) = TestUtil.get_random_variable_length_dataset(
max_value=num_words)
explained_model = RandomForestClassifier(n_estimators=64,
max_depth=5,
random_state=1)
counter = CountVectoriser(num_words)
tfidf_transformer = TfidfTransformer()
explained_model = Pipeline([('counts', counter),
('tfidf', tfidf_transformer),
('model', explained_model)])
explained_model.fit(x_train, y_train)
with self.assertRaises(ValueError):
_ = RNNModelBuilder(
with_embedding=True,
verbose=0) # Must also specify the embedding_size argument.
model_builder = RNNModelBuilder(embedding_size=num_words,
with_embedding=True,
verbose=0)
input_layer = Input(shape=(10, 2))
with self.assertRaises(ValueError):
model_builder.build(input_layer)
input_layer = Input(shape=(10, 3))
with self.assertRaises(ValueError):
model_builder.build(input_layer)
def test_imdb_padded_valid(self):
num_samples = 32
num_words = 1024
(x_train, y_train), (x_test, y_test) = TestUtil.get_imdb(word_dictionary_size=num_words,
num_subsamples=num_samples)
explained_model = RandomForestClassifier(n_estimators=64, max_depth=5, random_state=1)
counter = CountVectoriser(num_words)
tfidf_transformer = TfidfTransformer()
explained_model = Pipeline([('counts', counter),
('tfidf', tfidf_transformer),
('model', explained_model)])
explained_model.fit(x_train, y_train)
model_builder = RNNModelBuilder(embedding_size=num_words, with_embedding=True,
num_layers=2, num_units=32, activation="relu", p_dropout=0.2, verbose=0,
batch_size=32, learning_rate=0.001, num_epochs=2, early_stopping_patience=128)
masking_operation = WordDropMasking()
loss = binary_crossentropy
explainer = CXPlain(explained_model, model_builder, masking_operation, loss)
x_train = pad_sequences(x_train, padding="post", truncating="post", dtype=int)
x_test = pad_sequences(x_test, padding="post", truncating="post", dtype=int, maxlen=x_train.shape[1])
explainer.fit(x_train, y_train)
eval_score = explainer.score(x_test, y_test)
train_score = explainer.get_last_fit_score()
median = explainer.predict(x_test)
self.assertTrue(median.shape == x_test.shape)
def test_causal_loss_padded_input(self):
models = TestUtil.get_classification_models()
batch_size = 32
num_samples = 1024
num_words = 1024
(x_train, y_train), (x_test, y_test) = \
TestUtil.get_random_variable_length_dataset(num_samples=num_samples, max_value=num_words)
x, y = np.concatenate([x_train, x_test],
axis=0), np.concatenate([y_train, y_test],
axis=0)
self.assertEqual(x.shape[0], num_samples)
for explained_model in models:
counter = CountVectoriser(num_words)
tfidf_transformer = TfidfTransformer()
explained_model = Pipeline([('counts', counter),
('tfidf', tfidf_transformer),
('model', explained_model)])
TestUtil.fit_proxy(explained_model, x, y)
masking = WordDropMasking()
x = pad_sequences(x, padding="post", truncating="post", dtype=int)
_, y_pred, all_y_pred_imputed = masking.get_predictions_after_masking(
explained_model,
x,
y,
batch_size=batch_size,
downsample_factors=(1, ),
flatten=False)
auxiliary_outputs = y_pred
all_but_one_auxiliary_outputs = all_y_pred_imputed
all_but_one_auxiliary_outputs = TestUtil.split_auxiliary_outputs_on_feature_dim(
all_but_one_auxiliary_outputs)
delta_errors = calculate_delta_errors(
y,
auxiliary_outputs,
all_but_one_auxiliary_outputs,
NumpyInterface.binary_crossentropy,
math_ops=NumpyInterface)
# Ensure correct delta error dimensionality.
self.assertEqual(delta_errors.shape, (num_samples, x.shape[1]))
def test_nlp_not_padded_invalid(self):
num_words = 1024
(x_train, y_train), (_, _) = TestUtil.get_random_variable_length_dataset(max_value=num_words)
explained_model = RandomForestClassifier(n_estimators=64, max_depth=5, random_state=1)
counter = CountVectoriser(num_words)
tfidf_transformer = TfidfTransformer()
explained_model = Pipeline([('counts', counter),
('tfidf', tfidf_transformer),
('model', explained_model)])
explained_model.fit(x_train, y_train)
model_builder = RNNModelBuilder(embedding_size=num_words, with_embedding=True,
num_layers=2, num_units=32, activation="relu", p_dropout=0.2, verbose=0,
batch_size=32, learning_rate=0.001, num_epochs=2, early_stopping_patience=128)
masking_operation = WordDropMasking()
loss = binary_crossentropy
explainer = CXPlain(explained_model, model_builder, masking_operation, loss)
with self.assertRaises(ValueError):
explainer.fit(x_train, y_train)