rnn_cell._linear

源码

def _linear(args, output_size, bias, bias_start=0.0, scope=None):
  """Linear map: sum_i(args[i] * W[i]), where W[i] is a variable.

  Args:
    args: a 2D Tensor or a list of 2D, batch x n, Tensors.
    output_size: int, second dimension of W[i].
    bias: boolean, whether to add a bias term or not.
    bias_start: starting value to initialize the bias; 0 by default.
    scope: (optional) Variable scope to create parameters in.

  Returns:
    A 2D Tensor with shape [batch x output_size] equal to
    sum_i(args[i] * W[i]), where W[i]s are newly created matrices.

  Raises:
    ValueError: if some of the arguments has unspecified or wrong shape.
  """
  if args is None or (nest.is_sequence(args) and not args):
    raise ValueError("`args` must be specified")
  if not nest.is_sequence(args):
    args = [args]

  # Calculate the total size of arguments on dimension 1.
  total_arg_size = 0
  shapes = [a.get_shape() for a in args]
  for shape in shapes:
    if shape.ndims != 2:
      raise ValueError("linear is expecting 2D arguments: %s" % shapes)
    if shape[1].value is None:
      raise ValueError("linear expects shape[1] to be provided for shape %s, "
                       "but saw %d" % (shape, shape[1]))
    else:
      total_arg_size += shape[1].value

  dtype = [a.dtype for a in args][0]

  # Now the computation.
  scope = vs.get_variable_scope()
  with vs.variable_scope(scope) as outer_scope:
    weights = vs.get_variable(
        "weights", [total_arg_size, output_size], dtype=dtype)
    if len(args) == 1:
      res = math_ops.matmul(args[0], weights)
    else:
      res = math_ops.matmul(array_ops.concat(1, args), weights)
    if not bias:
      return res
    with vs.variable_scope(outer_scope) as inner_scope:
      inner_scope.set_partitioner(None)
      biases = vs.get_variable(
          "biases", [output_size],
          dtype=dtype,
          initializer=init_ops.constant_initializer(bias_start, dtype=dtype))
  return nn_ops.bias_add(res, biases)

每次调用linear函数时不会创建一个新的矩阵，因为tf.get_variable(…) 只会在scope 没有设置为重用变量时才会创建新的矩阵。例如：

with tf.variable_scope("foo"):
      v = tf.get_variable("v", [1])  # v.name == "foo/v:0"
      w = tf.get_variable("w", [1])  # w.name == "foo/w:0"
  with tf.variable_scope("foo", reuse=True)
      v1 = tf.get_variable("v")  # The same as v above.

linear函数测试：

def testLinear(self):
    with self.test_session() as sess:
      with tf.variable_scope("root", initializer=tf.constant_initializer(1.0)):
        x = tf.zeros([1, 2])
        l = tf.nn.rnn_cell._linear([x], 2, False)
        sess.run([tf.initialize_all_variables()])
        res = sess.run([l], {x.name: np.array([[1., 2.]])})
        self.assertAllClose(res[0], [[3.0, 3.0]])
        # Checks prevent you from accidentally creating a shared function.
        with self.assertRaises(ValueError):
          l1 = tf.nn.rnn_cell._linear([x], 2, False)
        # But you can create a new one in a new scope and share the variables.
        with tf.variable_scope("l1") as new_scope:
          l1 = tf.nn.rnn_cell._linear([x], 2, False)
        with tf.variable_scope(new_scope, reuse=True):
          tf.nn.rnn_cell._linear([l1], 2, False)
        self.assertEqual(len(tf.trainable_variables()), 2)