def test_loss_init_WITH_NON_LOSS(self):
<<<<<<< HEAD
self.assertRaises(ValueError, lambda: Loss("name", "shortname", "decoder_outputs", "decoder_targets", "loss"))
def test_loss_backward_WITH_NO_LOSS(self):
loss = Loss("name", "shortname", "decoder_output", "decoder_output", torch.nn.NLLLoss())
=======
self.assertRaises(ValueError, lambda: Loss("name", "shortname",
"decoder_outputs", "decoder_targets", "loss"))
def test_loss_backward_WITH_NO_LOSS(self):
loss = Loss("name", "shortname", "decoder_output",
"decoder_output", torch.nn.NLLLoss())
>>>>>>> upstream/master
self.assertRaises(ValueError, lambda: loss.backward())
def test_nllloss_init(self):
loss = NLLLoss()
self.assertEqual(loss.name, NLLLoss._NAME)
self.assertEqual(loss.log_name, NLLLoss._SHORTNAME)
<<<<<<< HEAD
self.assertTrue(type(loss.criterion) is torch.nn.NLLLoss)
=======
self.assertTrue(isinstance(loss.criterion, torch.nn.NLLLoss))
>>>>>>> upstream/master
def test_nllloss(self):
loss = NLLLoss()
pytorch_loss = 0
pytorch_criterion = torch.nn.NLLLoss()
def test_loss_backward_WITH_NO_LOSS(self):
loss = Loss("name", "shortname", "decoder_output", "decoder_output",
torch.nn.NLLLoss())
self.assertRaises(ValueError, lambda: loss.backward())