def test_train_and_eval(self, secondary_loss):
data_dir = tf.compat.v1.test.get_temp_dir()
data_file = os.path.join(data_dir, "libvsvm.txt")
if tf.io.gfile.exists(data_file):
tf.io.gfile.remove(data_file)
with open(data_file, "wt") as writer:
writer.write(LIBSVM_DATA)
output_dir = os.path.join(data_dir, secondary_loss or "")
with flagsaver.flagsaver(
train_path=data_file,
vali_path=data_file,
test_path=data_file,
output_dir=output_dir,
loss="pairwise_logistic_loss",
secondary_loss=secondary_loss,
num_train_steps=10,
list_size=10,
group_size=2,
num_features=100):
tf_ranking_libsvm.train_and_eval()
if tf.io.gfile.exists(output_dir):
tf.io.gfile.rmtree(output_dir)
def test_train_and_eval(self):
data_dir = tf.test.get_temp_dir()
data_file = os.path.join(data_dir, "libvsvm.txt")
if tf.gfile.Exists(data_file):
tf.gfile.Remove(data_file)
with open(data_file, "wt") as writer:
writer.write(LIBSVM_DATA)
self._data_file = data_file
self._output_dir = data_dir
with flagsaver.flagsaver(train_path=self._data_file,
vali_path=self._data_file,
test_path=self._data_file,
output_dir=self._output_dir,
num_train_steps=10,
list_size=10,
group_size=2,
num_features=100):
tf_ranking_libsvm.train_and_eval()
if tf.gfile.Exists(self._output_dir):
tf.gfile.DeleteRecursively(self._output_dir)