def test_reproduce(self):
# disable all logging
logging.disable(logging.CRITICAL)
set_reproducibility(0)
trainer = Trainer(self.args, tf_log=False)
train_set, dev_set = trainer.load_dataset()
trainer._init_model()
expected_losses = [
483.9374983910236, 396.7885422831847, 313.4529746738469,
412.6149972063738, 294.88370214788677, 276.9205821059003,
242.83269801561872, 225.63258253138852, 223.06413635684396,
209.42066629479157, 182.57185706984563, 199.2475491835321,
203.06883826238152, 166.84528868322133, 177.92200707610107,
163.95598015036012, 143.41510641020656, 164.14947102979863,
157.85085816150064, 145.01048472122704
]
losses = self._test_train(trainer, train_set)
# print(losses)
self.assertEqual(losses, expected_losses)
expected_valid_loss = -19.518319110144216
valid_loss = self._test_valid(trainer, dev_set)
trainer.save_model()
# print(valid_loss)
self.assertEqual(valid_loss, expected_valid_loss)
expected_hypos = [
[
Hypothesis(value=[
'Examines', 'the', 'undistorted', 'gray', 'input', 'image',
'for', 'squares', '<con>', '<con>', '.', '.', '.'
],
score=-9.169617064004049),
Hypothesis(value=[
'Examines', 'the', 'undistorted', 'gray', 'input', 'image',
'for', 'squares', '<con>', '<con>', '.', '.'
],
score=-8.4981576175081)
],
[
Hypothesis(value=[
'Save', 'basic', 'clusters', '<con>', '.', '.', '.', '.',
'.'
],
score=-4.295478831250667),
Hypothesis(value=[
'Save', 'basic', 'clusters', '<con>', '.', '.', '.', '.',
'.', '.'
],
score=-4.809735736699733)
],
[
Hypothesis(value=[
'Configure', 'a', 'ssl', '<con>', 'Config', 'for', 'the',
'server', 'using', 'the', 'legacy', 'legacy',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration', 'configuration',
'configuration', 'configuration'
],
score=-35.71404207185598)
]
]
hypos = self._test_infer()
self.assertEqual(hypos, expected_hypos)
