def load_from_file(cls, vocab_type: VocabType, file, special_words: SpecialVocabWordsType) -> 'Vocab':
special_words_as_unique_list = common.get_unique_list(special_words.__dict__.values())
# Notice: From historical reasons, a saved vocab doesn't include special words,
# so they should be added upon loading.
word_to_index_wo_specials = pickle.load(file)
index_to_word_wo_specials = pickle.load(file)
size_wo_specials = pickle.load(file)
assert len(index_to_word_wo_specials) == len(word_to_index_wo_specials) == size_wo_specials
min_word_idx_wo_specials = min(index_to_word_wo_specials.keys())
if min_word_idx_wo_specials != len(special_words_as_unique_list):
raise ValueError(
"Error while attempting to load vocabulary `{vocab_type}` from file `{file_path}`. "
"The stored vocabulary has minimum word index {min_word_idx}, "
"while expecting minimum word index to be {nr_special_words} "
"because having to use {nr_special_words} special words, which are: {special_words}. "
"Please check the parameter `config.SEPARATE_OOV_AND_PAD`.".format(
vocab_type=vocab_type, file_path=file.name, min_word_idx=min_word_idx_wo_specials,
nr_special_words=len(special_words_as_unique_list), special_words=special_words))
vocab = cls(vocab_type, [], special_words)
vocab.word_to_index = {**word_to_index_wo_specials,
**{word: idx for idx, word in enumerate(special_words_as_unique_list)}}
vocab.index_to_word = {**index_to_word_wo_specials,
**{idx: word for idx, word in enumerate(special_words_as_unique_list)}}
vocab.size = size_wo_specials + len(special_words_as_unique_list)
return vocab
def save_to_file(self, file):
# Notice: From historical reasons, a saved vocab doesn't include special words.
special_words_as_unique_list = common.get_unique_list(self.special_words.__dict__.values())
nr_special_words = len(special_words_as_unique_list)
word_to_index_wo_specials = {word: idx for word, idx in self.word_to_index.items() if idx >= nr_special_words}
index_to_word_wo_specials = {idx: word for idx, word in self.index_to_word.items() if idx >= nr_special_words}
size_wo_specials = self.size - nr_special_words
pickle.dump(word_to_index_wo_specials, file)
pickle.dump(index_to_word_wo_specials, file)
pickle.dump(size_wo_specials, file)
def __init__(self,
vocab_type: VocabType,
words: Iterable[str],
special_words: Optional[SpecialVocabWordsType] = None):
if special_words is None:
special_words = Namespace()
self.vocab_type = vocab_type
self.word_to_index: Dict[str, int] = {}
self.index_to_word: Dict[int, str] = {}
self._word_to_index_lookup_table = None
self._index_to_word_lookup_table = None
self.special_words: SpecialVocabWordsType = special_words
for index, word in enumerate(
chain(common.get_unique_list(special_words.__dict__.values()),
words)):
self.word_to_index[word] = index
self.index_to_word[index] = word
self.size = len(self.word_to_index)
