ba28f13e3c6bbeb8521feef63af72557ac08781b,opennmt/layers/transformer.py,MultiHeadAttention,map_v1_weights,#MultiHeadAttention#Any#,145

Before Change

          lambda w: np.split(w, num_splits, axis=0 if w.ndim == 1 else 1)[index],
          weights[key])

    m = {}
    if "conv1d_2" not in weights:  // Case self-attention.
      m.update(self.linear_queries.map_v1_weights(_partial_weights("conv1d", 3, 0)))
      m.update(self.linear_keys.map_v1_weights(_partial_weights("conv1d", 3, 1)))
      m.update(self.linear_values.map_v1_weights(_partial_weights("conv1d", 3, 2)))
      m.update(self.linear_output.map_v1_weights(weights["conv1d_1"]))
    else:
      m.update(self.linear_queries.map_v1_weights(weights["conv1d"]))
      m.update(self.linear_keys.map_v1_weights(_partial_weights("conv1d_1", 2, 0)))
      m.update(self.linear_values.map_v1_weights(_partial_weights("conv1d_1", 2, 1)))
      m.update(self.linear_output.map_v1_weights(weights["conv1d_2"]))
    return m

  def call(self, inputs, memory=None, mask=None, cache=None, training=None):  // pylint: disable=arguments-differ
    Runs the layer.

After Change

          lambda w: np.split(w, num_splits, axis=0 if w.ndim == 1 else 1)[index],
          weights[key])

    m = []
    if "conv1d_2" not in weights:  // Case self-attention.
      m += self.linear_queries.map_v1_weights(_partial_weights("conv1d", 3, 0))
      m += self.linear_keys.map_v1_weights(_partial_weights("conv1d", 3, 1))
      m += self.linear_values.map_v1_weights(_partial_weights("conv1d", 3, 2))
      m += self.linear_output.map_v1_weights(weights["conv1d_1"])
    else:
      m += self.linear_queries.map_v1_weights(weights["conv1d"])
      m += self.linear_keys.map_v1_weights(_partial_weights("conv1d_1", 2, 0))
      m += self.linear_values.map_v1_weights(_partial_weights("conv1d_1", 2, 1))
      m += self.linear_output.map_v1_weights(weights["conv1d_2"])
    return m

  def call(self, inputs, memory=None, mask=None, cache=None, training=None):  // pylint: disable=arguments-differ
    Runs the layer.

Instances