def setUp(self):
self.model = LearnerModel(categories=10,
learning_rate=0.01,
logger=logging.getLogger("test"),
beta=0.005)
self.model.build_graph()
self.model.initialize()
def setUp(self):
self.model = LearnerModel(categories=10,
use_sigmoid=True,
learning_rate=1.5,
logger=logging.getLogger("test"))
self.model.build_graph()
self.model.initialize()
print("Testing sigmoid functionality")
def test_save(self):
print("Testing saving functionality")
print(self.model.predict(np.zeros([1, 2048, 32])))
self.model.train(np.zeros([800, 2048, 32]), np.zeros([800, 10]))
print(self.model.predict(np.zeros([1, 2048, 32])))
self.model.save("saved/test.model")
self.model.close()
tf.reset_default_graph()
self.model = LearnerModel(categories=10,
logger=logging.getLogger("test"))
self.model.build_graph()
self.model.load("saved/test.model-0")
print(self.model.predict(np.zeros([1, 2048, 32])))
class TestLearner(unittest.TestCase):
"""A collection of tests against the model to ensure it works"""
def setUp(self):
self.model = LearnerModel(categories=10,
logger=logging.getLogger("test"))
def test_build(self):
"""Test whether the graph can be built correctly"""
self.model.build_graph()
def tearDown(self):
self.model.close()
tf.reset_default_graph()
class TestSigmoid(unittest.TestCase):
def setUp(self):
self.model = LearnerModel(categories=10,
use_sigmoid=True,
learning_rate=1.5,
logger=logging.getLogger("test"))
self.model.build_graph()
self.model.initialize()
print("Testing sigmoid functionality")
def test_sigmoid(self):
print(self.model.predict(np.zeros([1, 2048, 32])))
self.model.train(np.zeros([800, 2048, 32]), np.zeros([800, 10]))
print(self.model.predict(np.zeros([1, 2048, 32])))
def tearDown(self):
self.model.close()
tf.reset_default_graph()
def train(flags: argparse.Namespace, configuration: configparser.ConfigParser):
global IS_RUNNING
operation_logger = logging.getLogger("operations")
driver = configuration.get("rmdb", "engine")
host = configuration.get("rmdb", "host")
port = configuration.get("rmdb", "port")
username = configuration.get("rmdb", "username")
password = configuration.get("rmdb", "password")
database = configuration.get("rmdb", "database")
session = sessionmaker()
engine = create_engine(
f"{driver}://{username}:{password}@{host}:{port}/{database}")
session.configure(bind=engine)
sess = session()
database = RelationalDatabase(sess=sess,
logger=logging.getLogger("database"))
database.get_all_category()
model = LearnerModel(categories=database.genre_size() +
database.category_size(),
logger=logging.getLogger("learner_model"))
session_file = configuration.get("recommender", "session_file")
session_configuration = configparser.ConfigParser()
session_configuration.read_file(session_file)
if session_configuration.get("session", "id"):
batch_manager = DatabaseBatchManager(sess,
logging.getLogger("manager"))
else:
operation_logger.error(
"error with getting the session id. create id with batch operation"
)
return
saver = FileSaver("results/default.map", logging.getLogger("file_saver"))
collector = SpotifyCollector(configuration.get("spotify", "id"),
configuration.get("spotify", "secret"))
if configuration.get(
"recommender", "cache_directory") and not os.path.exists(
configuration.get("recommender", "cache_directory")):
os.makedirs(configuration.get("recommender", "cache_directory"))
tempdir = configuration.get("recommender",
"cache_directory") if configuration.get(
"recommender", "cache_directory") else ""
training_config = TrainingConfiguration(
False, flags.batch_size, tempdir, 32,
session_configuration.get("session", "id"))
download_config = DownloadConfiguration(tempdir, flags.batch_size)
trainer = Trainer(training_config, model, batch_manager, collector,
saver, database,
configuration.get("recommender", "session_file"))
downloader = Downloader(collector, download_config)
deleter = DeleteRunner(tempdir)
downloader.run()
deleter.run()
trainer.run()
print("Press [Enter] to save the model at the snapshot.")
while IS_RUNNING:
input()
IS_RUNNING = False
def setUp(self):
self.model = LearnerModel(categories=10,
logger=logging.getLogger("test"))
self.model.build_graph()
self.model.initialize()
class TestPredictor(unittest.TestCase):
def setUp(self):
self.model = LearnerModel(categories=10,
learning_rate=0.01,
logger=logging.getLogger("test"),
beta=0.005)
self.model.build_graph()
self.model.initialize()
def test_train(self):
"""Test whether the model can be trained"""
for _ in range(1, 20):
self.model.train(np.zeros([1, 2048, 32]), np.zeros([1, 10]))
def test_flush(self):
"""Test whether the graph can be run with a zero array"""
result = self.model.predict(np.zeros([1, 2048, 32]))
print(result)
def test_predict(self):
"""Test whether the graph can be run with a zero array"""
print()
print("Testing the prediction:")
print("Ensure that the second result is smaller than the first")
result = self.model.predict(np.zeros([1, 2048, 32]))
print(result)
for _ in range(1, 20):
self.model.train(np.zeros([10, 2048, 32]), np.zeros([10, 10]))
result = self.model.predict(np.zeros([1, 2048, 32]))
print(result)
def tearDown(self):
self.model.close()
tf.reset_default_graph()
def setUp(self):
self.model = LearnerModel(categories=10,
logger=logging.getLogger("test"))