def test_set_eval_mode(self, mock_eval, mock_call):
""" Make sure that evaluation is done in evaluation mode. """
mock_mgr = MagicMock()
mock_mgr.attach_mock(mock_eval, 'eval')
mock_mgr.attach_mock(mock_call, 'call')
evaluator = Evaluator()
evaluator.evaluate(self.seq2seq, self.dataset)
expected_calls = [call.eval()] + \
self.dataset.num_batches(evaluator.batch_size) * [call.call(ANY, ANY, volatile=ANY)]
self.assertEquals(expected_calls, mock_mgr.mock_calls)
def test_set_eval_mode(self, mock_eval, mock_call):
""" Make sure that evaluation.txt is done in evaluation.txt mode. """
mock_mgr = MagicMock()
mock_mgr.attach_mock(mock_eval, 'eval')
mock_mgr.attach_mock(mock_call, 'call')
evaluator = Evaluator(batch_size=64)
with patch('seq2seq.evaluator.evaluator.torch.stack', return_value=None), \
patch('seq2seq.loss.NLLLoss.eval_batch', return_value=None):
evaluator.evaluate(self.seq2seq, self.dataset)
num_batches = int(math.ceil(len(self.dataset) / evaluator.batch_size))
expected_calls = [call.eval()] + num_batches * [call.call(ANY, ANY, ANY)]
self.assertEquals(expected_calls, mock_mgr.mock_calls)
def test_set_eval_mode(self, mock_eval, mock_call):
""" Make sure that evaluation is done in evaluation mode. """
mock_mgr = MagicMock()
mock_mgr.attach_mock(mock_eval, 'eval')
mock_mgr.attach_mock(mock_call, 'call')
evaluator = Evaluator()
with patch('machine.evaluator.evaluator.torch.stack', return_value=None), \
patch('machine.metrics.WordAccuracy.eval_batch', return_value=None), \
patch('machine.metrics.WordAccuracy.eval_batch', return_value=None), \
patch('machine.loss.NLLLoss.eval_batch', return_value=None):
evaluator.evaluate(self.seq2seq, self.data_iterator,
trainer.get_batch_data)
num_batches = len(self.data_iterator)
expected_calls = [call.eval()] + num_batches * \
[call.call(ANY, ANY, ANY)]
self.assertEqual(expected_calls, mock_mgr.mock_calls)
param.data.uniform_(-0.08, 0.08)
@patch.object(Seq2seq, '__call__', return_value=(
[], None, dict(inputs=[], length=[10] * 64, sequence=MagicMock())))
@patch.object(Seq2seq, 'eval')
def test_set_eval_mode(self, mock_eval, mock_call):
""" Make sure that evaluation is done in evaluation mode. """
mock_mgr = MagicMock()
mock_mgr.attach_mock(mock_eval, 'eval')
mock_mgr.attach_mock(mock_call, 'call')
evaluator = Evaluator(batch_size=64)
<<<<<<< HEAD
with patch('seq2seq.evaluator.evaluator.torch.stack', return_value=None), \
patch('seq2seq.metrics.WordAccuracy.eval_batch', return_value=None), \
patch('seq2seq.loss.NLLLoss.eval_batch', return_value=None):
evaluator.evaluate(self.seq2seq, self.dataset, trainer.get_batch_data)
=======
with patch('machine.evaluator.evaluator.torch.stack', return_value=None), \
patch('machine.metrics.WordAccuracy.eval_batch', return_value=None), \
patch('machine.metrics.WordAccuracy.eval_batch', return_value=None), \
patch('machine.loss.NLLLoss.eval_batch', return_value=None):
evaluator.evaluate(self.seq2seq, self.dataset,
trainer.get_batch_data)
>>>>>>> upstream/master
num_batches = int(math.ceil(len(self.dataset) / evaluator.batch_size))
expected_calls = [call.eval()] + num_batches * \
[call.call(ANY, ANY, ANY)]
self.assertEquals(expected_calls, mock_mgr.mock_calls)
def editor(vim):
mockeditor = Editor(vim)
assert vim.mock_calls == [call.eval("has('nvim')")]
vim.reset_mock() # Clear above constructor vim calls from call list
return mockeditor