def setUp(self):
session.init("test_cross_entropy")
self.softmax_loss = SoftmaxCrossEntropyLoss()
self.y_list = [i % 5 for i in range(100)]
self.predict_list = [np.array([random.random() for i in range(5)]) for j in range(100)]
self.y = session.parallelize(self.y_list, include_key=False, partition=16)
self.predict = session.parallelize(self.predict_list, include_key=False, partition=16)
class TestSoftmaxCrossEntropyLoss(unittest.TestCase):
def setUp(self):
session.init("test_cross_entropy")
self.softmax_loss = SoftmaxCrossEntropyLoss()
self.y_list = [i % 5 for i in range(100)]
self.predict_list = [
np.array([random.random() for i in range(5)]) for j in range(100)
]
self.y = session.parallelize(self.y_list,
include_key=False,
partition=16)
self.predict = session.parallelize(self.predict_list,
include_key=False,
partition=16)
def test_predict(self):
for i in range(10):
list = [random.random() for j in range(5)]
pred_arr = np.asarray(list, dtype='float64')
mx = pred_arr.max()
predict = np.exp(pred_arr - mx) / sum(np.exp(pred_arr - mx))
softmaxloss_predict = self.softmax_loss.predict(pred_arr)
self.assertTrue(
np.fabs(predict -
softmaxloss_predict).all() < consts.FLOAT_ZERO)
def test_compute_grad(self):
for i in range(10):
pred = np.asarray([random.random() for j in range(5)],
dtype="float64")
label = random.randint(0, 4)
softmaxloss_grad = self.softmax_loss.compute_grad(label, pred)
grad = pred.copy()
grad[label] -= 1
self.assertTrue(
np.fabs(grad - softmaxloss_grad).all() < consts.FLOAT_ZERO)
def test_compute_hess(self):
for i in range(10):
pred = np.asarray([random.random() for j in range(5)],
dtype='float64')
label = random.randint(0, 4)
softmaxloss_hess = self.softmax_loss.compute_hess(label, pred)
hess = pred * (1 - pred)
self.assertTrue(
np.fabs(hess - softmaxloss_hess).all() < consts.FLOAT_ZERO)
def test_compute_loss(self):
softmax_loss = self.softmax_loss.compute_loss(self.y, self.predict)
loss = sum(-np.log(pred[yi])
for yi, pred in zip(self.y_list, self.predict_list)) / len(
self.y_list)
self.assertTrue(np.fabs(softmax_loss - loss) < consts.FLOAT_ZERO)
def tearDown(self):
session.stop()
def set_loss(self, objective_param):
loss_type = objective_param.objective
params = objective_param.params
LOGGER.info("set objective, objective is {}".format(loss_type))
if self.task_type == consts.CLASSIFICATION:
if loss_type == "cross_entropy":
if self.num_classes == 2:
self.loss = SigmoidBinaryCrossEntropyLoss()
else:
self.loss = SoftmaxCrossEntropyLoss()
else:
raise NotImplementedError("objective %s not supported yet" %
(loss_type))
elif self.task_type == consts.REGRESSION:
if loss_type == "lse":
self.loss = LeastSquaredErrorLoss()
elif loss_type == "lae":
self.loss = LeastAbsoluteErrorLoss()
elif loss_type == "huber":
self.loss = HuberLoss(params[0])
elif loss_type == "fair":
self.loss = FairLoss(params[0])
elif loss_type == "tweedie":
self.loss = TweedieLoss(params[0])
elif loss_type == "log_cosh":
self.loss = LogCoshLoss()
else:
raise NotImplementedError("objective %s not supported yet" %
(loss_type))
else:
raise NotImplementedError("objective %s not supported yet" %
(loss_type))
def set_loss(self, loss_type):
LOGGER.info("set loss, loss type is {}".format(loss_type))
if self.task_type == "classification":
if loss_type == "cross_entropy":
if self.num_classes == 2:
self.loss = SigmoidBinaryCrossEntropyLoss()
else:
self.loss = SoftmaxCrossEntropyLoss()
else:
raise NotImplementedError("Loss type %s not supported yet" %
(self.loss_type))
else:
raise NotImplementedError("Loss type %s not supported yet" %
(self.loss_type))