def get_inputs(cls, num_prediction_steps, num_results=None):
#Arrange
model = load_test_model()
args = get_args(model_name, Path())
input_params = InputParameters(args)
image_generation_params = ImageGenerationParameters(args)
input_data = get_input_data()
#Update input data parameters
num_classes = len(set(input_data[image_generation_params.label_col]))
image_generation_params_update = dict(num_classes=num_classes,
image_cols=['Image'])
update_params(image_generation_params,
**image_generation_params_update)
#Num results
num_results = num_results or image_generation_params.batch_size * num_prediction_steps
#Mocks
prediction_results = np.zeros((num_results, num_classes))
for row_id in range(num_results):
prediction_results[row_id, row_id % num_classes] = 1
model.predict_generator = MagicMock()
model.predict_generator.return_value = prediction_results
return model, input_data, input_params, image_generation_params, prediction_results
def test_configure_base_model(self):
#Arrange
op = Operation(num_unfrozen_layers, configure_base=True, base_level=0)
model = load_test_model()
#Act
model = op.configure(model)
#Assert
self.verify_unfrozen_layers(model, num_unfrozen_layers)
def test_set_model(self):
#Arrange
checkpoint, _, _, _ = get_checkpoint()
model = load_test_model()
#Act
checkpoint.set_model(model)
#Assert
self.assertIsNotNone(checkpoint._model)
def test_save(self):
#Arrange
model = load_test_model()
model_input = ModelInput(model_name)
#Mocks
model.save = MagicMock()
#Act
model_input.save(model, batch_id, epoch_id)
#Assert
model.save.assert_called_with(str(model_input.file_name(batch_id, epoch_id)))
def test_on_batch_end_save_called(self):
#Arrange
batch_id = 1
checkpoint, batch_input_files, _, _ = get_checkpoint()
input_data = get_input_data()
model = load_test_model()
checkpoint.set_model(model)
checkpoint.set_input_data(input_data)
checkpoint.on_epoch_begin(epoch_id)
checkpoint.on_batch_begin(batch_id)
#Act & Assert
self.on_batch_end(checkpoint, batch_id, batch_input_files)
def get_train_args():
input_data = get_input_data()
model = load_test_model()
input_params, training_params, image_generation_params, transformation_params = get_params()
image_generation_params.num_classes = 64
trainer = ImageTraining(
input_params,
training_params,
image_generation_params,
transformation_params,
MagicMock(),
summary = False)
return model, input_data, trainer
def test_on_epoch_end(self):
#Arrange
checkpoint, _, _, epoch_end_input_files = get_checkpoint()
model = load_test_model()
input_data = get_input_data()
result_file = epoch_end_input_files[0]
result_file.save = MagicMock()
checkpoint.set_model(model)
checkpoint.set_input_data(input_data)
checkpoint.on_epoch_begin(epoch_id)
checkpoint.on_batch_begin(batch_id)
#Act
checkpoint.on_epoch_end(epoch_id)
#Assert
self.on_result_file_save(checkpoint, result_file)