我基于tensorflow LSTM类定义了一个已更改的LSTM单元，尝试在其上添加关注机，如下所示：
class MyLSTMCell(tf.nn.rnn_cell.LSTMCell):
def __init__(self, num_units, input_size, grid_size,
use_peepholes=True, cell_clip=None,
initializer=tf.random_normal_initializer(stddev=0.01), num_proj=None, proj_clip=None,
num_unit_shards=None, num_proj_shards=None,
forget_bias=1.0, batch_size=32, state_is_tuple=False,
activation=None, reuse=None, name=None, with_att=True):
super(tf.nn.rnn_cell.LSTMCell, self).__init__(_reuse=reuse, name=name)
# self.input_spec = tf.nn.rnn_cell.base_layer.InputSpec(ndim=2)
self._batch_size = batch_size
self._num_units = num_units
self._input_size = input_size
self._grid_size = grid_size
self._use_peepholes = use_peepholes
self._cell_clip = cell_clip
self._initializer = initializer
self._num_proj = num_proj
self._proj_clip = proj_clip
self._num_unit_shards = num_unit_shards
self._num_proj_shards = num_proj_shards
self._forget_bias = forget_bias
self._state_is_tuple = state_is_tuple
self._activation = activation or tf.tanh
self._with_att = with_att
if num_proj:
self._state_size = (
tf.nn.rnn_cell.LSTMStateTuple(num_units, num_proj)
if state_is_tuple else num_units + num_proj)
self._output_size = num_proj
else:
self._state_size = (
tf.nn.rnn_cell.LSTMStateTuple(num_units, num_units)
if state_is_tuple else 2 * num_units)
self._output_size = num_units
@property
def state_size(self):
return self._state_size
@property
def output_size(self):
return self._output_size
def build(self, inputs_shape):
if inputs_shape[2].value is None:
raise ValueError("Expected inputs.shape[-1] to be known, saw shape: %s"
% inputs_shape)
input_depth = inputs_shape[2].value
h_depth = self._num_units if self._num_proj is None else self._num_proj
maybe_partitioner = (
tf.nn.rnn_cell.partitioned_variables.fixed_size_partitioner(self._num_unit_shards)
if self._num_unit_shards is not None
else None)
self._kernel = self.add_variable(
"kernal",
shape=[input_depth + h_depth, 4 * self._num_units],
initializer=self._initializer,
partitioner=maybe_partitioner)
self._bias = self.add_variable(
"bias",
shape=[4 * self._num_units],
initializer=tf.zeros_initializer(dtype=self.dtype))
if self._use_peepholes:
self._w_f_diag = self.add_variable("w_f_diag", shape=[self._num_units],
initializer=self._initializer)
self._w_i_diag = self.add_variable("w_i_diag", shape=[self._num_units],
initializer=self._initializer)
self._w_o_diag = self.add_variable("w_o_diag", shape=[self._num_units],
initializer=self._initializer)
if self._num_proj is not None:
maybe_proj_partitioner = (
tf.nn.rnn_cell.partitioned_variables.fixed_size_partitioner(self._num_proj_shards)
if self._num_proj_shards is not None
else None)
self._proj_kernel = self.add_variable(
"projection/%s" % "kernal",
shape=[self._num_units, self._num_proj],
initializer=self._initializer,
partitioner=maybe_proj_partitioner)
self.built = True
def call(self, inputs, state):
print(state.shape)
num_proj = self._num_units if self._num_proj is None else self._num_proj
sigmoid = tf.sigmoid
if self._state_is_tuple:
(c_prev, h_prev) = state
else:
c_prev = tf.slice(state, [0, 0], [-1, self._num_units])
h_prev = tf.slice(state, [0, self._num_units], [-1, num_proj])
print(inputs.shape)
input_size = inputs.get_shape().with_rank(3)[2]
if input_size.value is None:
raise ValueError("Could not infer input size from inputs.get_shape()[-1]")
Kx = tf.layers.dense(inputs, self._input_size, use_bias=False)
print('begin')
print(Kx.shape)
Uh = tf.layers.dense(h_prev, self._input_size, use_bias=True)
print(Uh.shape)
e = tf.tanh(Uh + Kx)
print(e.shape)
e = tf.layers.dense(e, self._grid_size)
print(e.shape)
alpha = tf.nn.softmax(e, axis=2)
print(alpha.shape)
z = tf.reduce_mean(tf.matmul(alpha, inputs), axis=1)
print(z.shape)
# i = input_gate, j = new_input, f = forget_gate, o = output_gate
lstm_matrix = tf.matmul(
tf.concat([z, h_prev], 1), self._kernel)
lstm_matrix = tf.nn.bias_add(lstm_matrix, self._bias)
i, g, f, o = tf.split(
value=lstm_matrix, num_or_size_splits=4, axis=1)
# Diagonal connections
if self._use_peepholes:
c = (sigmoid(f + self._forget_bias + self._w_f_diag * c_prev) * c_prev +
sigmoid(i + self._w_i_diag * c_prev) * self._activation(g))
else:
c = (sigmoid(f + self._forget_bias) * c_prev + sigmoid(i) *
self._activation(g))
if self._cell_clip is not None:
# pylint: disable=invalid-unary-operand-type
c = tf.nn.rnn_cell_impl.clip_ops.clip_by_value(c, -self._cell_clip, self._cell_clip)
# pylint: enable=invalid-unary-operand-type
if self._use_peepholes:
h = sigmoid(o + self._w_o_diag * c) * self._activation(c)
else:
h = sigmoid(o) * self._activation(c)
if self._num_proj is not None:
h = tf.matmul(h, self._proj_kernel)
if self._proj_clip is not None:
# pylint: disable=invalid-unary-operand-type
h = tf.nn.rnn_cell_impl.clip_ops.clip_by_value(h, -self._proj_clip, self._proj_clip)
# pylint: enable=invalid-unary-operand-type
new_state = (tf.nn.rnn_cell.LSTMStateTuple(c, h) if self._state_is_tuple else
tf.concat([c, h], 1))
return h, new_state
我想将第一个状态初始化为：
c0 = tf.layers.dense(tf.reduce_mean(feature, axis=1), num_units, activation=tf.tanh)
h0 = tf.layers.dense(tf.reduce_mean(feature, axis=1), num_units, activation=tf.tanh)
init_state = tf.nn.rnn_cell.LSTMStateTuple(c0,h0)
但是我得到一个错误：Tensor在急切执行时对象不是不可迭代的，我真的很注意state_size，不知道为什么，有人可以帮助我吗？谢谢！