def __init__(self, M, data_store, num_latent=ScoringParams.num_latent_factors):
"""Construct a new instance.
:M: This parameter controls the number of recommendations that will
be served by the PMF model.
"""
AbstractRecommender.__init__(self)
self._M = M
self.num_latent = num_latent
self.user_matrix = None
self.latent_item_rep_mat = None
self.weight_matrix = None
self.s3_client = data_store
self._load_model_output_matrices(model_path=PMF_MODEL_PATH)
self._load_package_id_to_name_map()
self._package_tag_map = self.s3_client.read_json_file(PACKAGE_TAG_MAP)
self.item_ratings = load_rating(TRAINING_DATA_ITEMS, data_store)
self.user_stacks = load_rating(PRECOMPUTED_STACKS, data_store)
_logger.info("Created an instance of pmf-recommendation, loaded data from S3")
def test_load_rating(self):
"""Test the load_rating method."""
path = 'test_load_rating.txt'
test_datastore = LocalDataStore('tests/test_data')
r = load_rating(path, test_datastore)
self.assertListEqual(r, [[5409, 2309, 54909, 2054], list()])
# train the auto-encoder
cvae_model.fit(data, epochs=params_training.num_epochs,
batch_size=params_training.batch_size,
callbacks=[saver, tensorboard_config])
return self.encoder_z_mean
if __name__ == '__main__':
p = TrainNetwork()
logger.info("Preprocessing of data started.")
p.get_preprocess_data.preprocess_data()
x_train = np.load(os.path.join(TEMPORARY_DATA_PATH, 'content_matrix.npz'))
x_train = x_train['matrix']
input_dim = x_train.shape[1]
logger.info("size of training file is: ".format(len(x_train), len(x_train[0])))
user_to_item_matrix = load_rating(TEMPORARY_USER_ITEM_FILEPATH, TEMPORARY_DATASTORE)
item_to_user_matrix = load_rating(TEMPORARY_ITEM_USER_FILEPATH, TEMPORARY_DATASTORE)
logger.info("Shape of User and Item matrices:".format(np.shape(user_to_item_matrix),
np.shape(item_to_user_matrix)))
pretrain.fit(x_train)
encoder_weights = p.train(x_train)
logger.info("Shape of encoder weights are: ".format(tf.shape(encoder_weights),
len(encoder_weights),
len(encoder_weights[0])))
pmf_obj = PMFTraining(len(user_to_item_matrix), len(item_to_user_matrix), encoder_weights)
logger.debug("PMF model has been initialised")
pmf_obj(user_to_item_matrix=user_to_item_matrix,
item_to_user_matrix=item_to_user_matrix)
logger.debug("PMF model has been trained.")
pmf_obj.save_model()
p.get_preprocess_data.obj_.save_on_s3(TEMPORARY_DATA_PATH)