tensor2tensor/data_generators/multinli.py [86:111]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  @property
  def is_generate_per_split(self):
    return True

  @property
  def dataset_splits(self):
    return [{
        "split": problem.DatasetSplit.TRAIN,
        "shards": 100,
    }, {
        "split": problem.DatasetSplit.EVAL,
        "shards": 1,
    }]

  @property
  def approx_vocab_size(self):
    return 2**15

  @property
  def num_classes(self):
    return 3

  def class_labels(self, data_dir):
    del data_dir
    # Note this binary classification is different from usual MNLI.
    return ["contradiction", "entailment", "neutral"]
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



tensor2tensor/data_generators/stanford_nli.py [43:68]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  @property
  def is_generate_per_split(self):
    return True

  @property
  def dataset_splits(self):
    return [{
        "split": problem.DatasetSplit.TRAIN,
        "shards": 100,
    }, {
        "split": problem.DatasetSplit.EVAL,
        "shards": 1,
    }]

  @property
  def approx_vocab_size(self):
    return 2**15

  @property
  def num_classes(self):
    return 3

  def class_labels(self, data_dir):
    del data_dir
    # Note this binary classification is different from usual SNLI.
    return ["contradiction", "entailment", "neutral"]
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -