LSTMでstatefulにした際のbatch_input

LSTMで時系列予測をする際に、stateful=Trueを指定し、batch_input_shapeを指定したのですが下記の様なエラーになってしまい、解決方法がわかりません。

Keras Documentation
KerasのステートレスLSTMとステートフルLSTMの違いについて
を参考にしました。

解決策や原因が分かる方いらっしゃいましたら、アドバイスをお願いします。

python
1batch_input_shape=(batch_size, train_steps, np_data.shape[-1])
2batch_input_shape
3>>>(64, 123, 25)
4
5model = Sequential()
6model.add(layers.LSTM(hl1,
7                      recurrent_dropout=recurrent_dropout1,
8                      return_sequences=True,
9                      stateful=True,
10                      batch_input_shape=batch_input_shape))
11
12model.add(layers.LSTM(hl2,
13                      recurrent_dropout=recurrent_dropout2,
14                      return_sequences=False,
15                      stateful=True))
16
17model.add(layers.Dense(1))

error
1---------------------------------------------------------------------------
2RuntimeError                              Traceback (most recent call last)
3<ipython-input-243-a1221c150bf4> in <module>
4      4                       return_sequences=True,
5      5                       stateful=True,
6----> 6                       batch_input_shape=batch_input_shape))
7      7 
8      8 model.add(layers.LSTM(hl2,
9
10/usr/local/lib/python3.6/dist-packages/keras/engine/sequential.py in add(self, layer)
11    164                     # and create the node connecting the current layer
12    165                     # to the input layer we just created.
13--> 166                     layer(x)
14    167                     set_inputs = True
15    168             else:
16
17/usr/local/lib/python3.6/dist-packages/keras/layers/recurrent.py in __call__(self, inputs, initial_state, constants, **kwargs)
18    539 
19    540         if initial_state is None and constants is None:
20--> 541             return super(RNN, self).__call__(inputs, **kwargs)
21    542 
22    543         # If any of `initial_state` or `constants` are specified and are Keras
23
24/usr/local/lib/python3.6/dist-packages/keras/engine/base_layer.py in __call__(self, inputs, **kwargs)
25    487             # Actually call the layer,
26    488             # collecting output(s), mask(s), and shape(s).
27--> 489             output = self.call(inputs, **kwargs)
28    490             output_mask = self.compute_mask(inputs, previous_mask)
29    491 
30
31/usr/local/lib/python3.6/dist-packages/keras/layers/recurrent.py in call(self, inputs, mask, training, initial_state)
32   2245                                       mask=mask,
33   2246                                       training=training,
34-> 2247                                       initial_state=initial_state)
35   2248 
36   2249     @property
37
38/usr/local/lib/python3.6/dist-packages/keras/layers/recurrent.py in call(self, inputs, mask, training, initial_state, constants)
39    680                                              mask=mask,
40    681                                              unroll=self.unroll,
41--> 682                                              input_length=timesteps)
42    683         if self.stateful:
43    684             updates = []
44
45/usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py in rnn(step_function, inputs, initial_states, go_backwards, mask, constants, unroll, input_length)
46   3101         constants=constants,
47   3102         unroll=unroll,
48-> 3103         input_length=input_length)
49   3104     reachable = tf_utils.get_reachable_from_inputs([learning_phase()],
50   3105                                                    targets=[last_output])
51
52/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/backend.py in rnn(step_function, inputs, initial_states, go_backwards, mask, constants, unroll, input_length, time_major, zero_output_for_mask)
53   3929     # the value is discarded.
54   3930     output_time_zero, _ = step_function(
55-> 3931         input_time_zero, tuple(initial_states) + tuple(constants))
56   3932     output_ta = tuple(
57   3933         tensor_array_ops.TensorArray(
58
59/usr/local/lib/python3.6/dist-packages/keras/layers/recurrent.py in step(inputs, states)
60    671         else:
61    672             def step(inputs, states):
62--> 673                 return self.cell.call(inputs, states, **kwargs)
63    674 
64    675         last_output, outputs, states = K.rnn(step,
65
66/usr/local/lib/python3.6/dist-packages/keras/layers/recurrent.py in call(self, inputs, states, training)
67   2035             z = K.dot(inputs, self.kernel)
68   2036             if 0. < self.recurrent_dropout < 1.:
69-> 2037                 h_tm1 *= rec_dp_mask[0]
70   2038             z += K.dot(h_tm1, self.recurrent_kernel)
71   2039             if self.use_bias:
72
73/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/resource_variable_ops.py in __imul__(self, unused_other)
74   1227 
75   1228   def __imul__(self, unused_other):
76-> 1229     raise RuntimeError("Variable *= value not supported. Use "
77   1230                        "`var.assign(var * value)` to modify the variable or "
78   1231                        "`var = var * value` to get a new Tensor object.")
79
80RuntimeError: Variable *= value not supported. Use `var.assign(var * value)` to modify the variable or `var = var * value` to get a new Tensor object.
81