def test_load(self):
vocab = self.vocab
seq = ["i", "like", "python"]
vocab.add_sequence(seq)
file_name = "vocab_file"
vocab.save(file_name)
loaded_vocab = Vocabulary.load(file_name)
os.remove(file_name)
self.assertEqual(vocab, loaded_vocab)
def load(cls, path):
"""
Loads a Checkpoint object that was previously saved to disk.
Args:
path (str): path to the checkpoint subdirectory
Returns:
checkpoint (Checkpoint): checkpoint object with fields copied from those stored on disk
"""
print("Loading checkpoints from {}".format(path))
resume_checkpoint = torch.load(
os.path.join(path, cls.TRAINER_STATE_NAME))
model = torch.load(os.path.join(path, cls.MODEL_NAME))
input_vocab = Vocabulary.load(os.path.join(path, cls.INPUT_VOCAB_FILE))
output_vocab = Vocabulary.load(
os.path.join(path, cls.OUTPUT_VOCAB_FILE))
return Checkpoint(model=model,
input_vocab=input_vocab,
output_vocab=output_vocab,
optimizer_state_dict=resume_checkpoint['optimizer'],
epoch=resume_checkpoint['epoch'],
step=resume_checkpoint['step'],
path=path)
def test_load(self):
vocab = self.vocab
seq = ["i", "like", "python"]
vocab.add_sequence(seq)
pickle_file = "vocab_pickle"
input_vocab_pickle = pickle.dumps(vocab)
with open(pickle_file, "wb") as f:
f.write(input_vocab_pickle)
with open(pickle_file, "rb") as f:
pickled_vocab = pickle.load(f)
loaded_vocab = Vocabulary.load(pickle_file)
self.assertEqual(pickled_vocab, loaded_vocab)