def get_user_global_predict(self, test_dataset=None):
if test_dataset is None:
preprocess = DataPreprocess()
test_dataset = preprocess.generate_test_data(batch_size=512)
# test_dataset = test_dataset.take(1)
container = np.random.choice(self.containers)
SERVER_URL = container + self.model_url
pred_y = []
true_y = []
for test_x, test_y in test_dataset:
input_x = get_input_data(test_x)
data = {"inputs": input_x}
input_data_json = json.dumps(data)
response = requests.post(SERVER_URL, data=input_data_json)
response = json.loads(response.text)
array = np.array(response['outputs'])
_array = np.where(array > 0.5, 1, 0)
pred_y.extend(_array)
true_y.extend(test_y.numpy()[:, np.newaxis])
get_metrics(true_y, pred_y, is_print=True)
def predict(self, test_dataset=None, is_print=True):
if test_dataset is None:
test_dataset = self.data_process.generate_test_data()
true_y = []
pred_y = []
for test_x, test_y in test_dataset:
data = get_input_tensor_data(test_x)
array = self.deepfm_model(data, training=False)
_array = np.where(array > 0.5, 1, 0)
pred_y.extend(_array)
true_y.extend(test_y.numpy()[:, np.newaxis])
get_metrics(true_y, pred_y, is_print=is_print)