b46fa00aa4256914240cb154b910bebfaf1693d6,tensorflow/python/keras/layers/preprocessing/index_lookup.py,IndexLookup,call,#IndexLookup#Any#,379

Before Change

  def call(self, inputs):
    if self._key_dtype == dtypes.int64 and inputs.dtype == dtypes.int32:
      inputs = math_ops.cast(inputs, dtypes.int64)
    return self._table_handler.lookup(inputs)

  def _use_v1_apis(self):
    return False

After Change

    self.set_vocabulary(updates[_VOCAB_NAME])

  def call(self, inputs):
    if not self.max_tokens:
      raise ValueError("You must set the layer"s vocabulary before calling it. "
                       "Either pass a `vocabulary` argument to the layer, or "
                       "call `layer.adapt(dataset)` with some sample data.")
    if self._key_dtype == dtypes.int64 and inputs.dtype == dtypes.int32:
      inputs = math_ops.cast(inputs, dtypes.int64)
    lookup_result = self._table_handler.lookup(inputs)
    if self.output_mode == INT:
      return lookup_result

    binary_output = (self.output_mode == BINARY)
    if self.sparse:
      return category_encoding.sparse_bincount(
          lookup_result, self.max_tokens, binary_output)
    else:
      return category_encoding.dense_bincount(
          lookup_result, self.max_tokens, binary_output)

  def _use_v1_apis(self):
    return False

Instances