InvalidArgumentError, когда я использую model.fit ()

Question

Я пытаюсь создать пользовательский слой в keras / tenorflow. Цель состоит в том, чтобы расширить его как RNN. Проблема возникает во время обучения, где, по-видимому, имеется несовместимость с выходной формой и набором обучающих данных.

X_train.shape
(100, 5)
X_target.shape
(100, 5)

Это пользовательский слой:

class MyLayer(Layer):

    def __init__(self, output_dim, **kwargs):
        self.output_dim = output_dim
        super(MyLayer, self).__init__(**kwargs)

    def build(self, input_shape):
        print(self.output_dim)
        print(input_shape)
        # Create a trainable weight variable for this layer.

        self.kernel = self.add_weight(name='kernel', 
                                  shape=(self.output_dim,   input_shape[1]), 
                                  initializer='uniform',
                                  trainable=True)
        super(MyLayer, self).build(input_shape)  # Be sure to call this at the end

    def call(self, x):
        print('shape x: ')
        print(x.shape)
        print('shape kernel: ')
        print(self.kernel.shape)
        matrix = tf.transpose(self.kernel)
        print('matrix')
        print(matrix.shape)
        prod = K.dot(x, matrix)
        print('after product')
        print(prod.shape)

        return prod

    def compute_output_shape(self, input_shape):
        print('Compute output shape')
        print(input_shape)
        print(self.output_dim)
        return (input_shape[0], self.output_dim)

model = Sequential()
model.add(MyLayer(5, batch_input_shape=(100, 5)))
model.compile(optimizer='adam', loss='mse')
# fit model
model.fit(X_train, X_target)

tf.keras.__version__ = '2.1.6-tf'

Поскольку я передаю x_training с формой (100,5), я ожидаю, что, умножив его на матрицу (5,5), я получу матрицу (100, 5), которая имеет ту же форму, что и цель, и поэтому я смог бы тренировать в этом случае матрицу веса 5x5. Вместо этого я получаю это:

InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-77-4dee23ead957> in <module>()
      6 model.compile(optimizer='adam', loss='mse')
      7 # fit model
----> 8 model.fit(X_train, X_target)#, epochs=300, verbose=0)

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, **kwargs)
   1037                                         initial_epoch=initial_epoch,
   1038                                         steps_per_epoch=steps_per_epoch,
-> 1039                                         validation_steps=validation_steps)
   1040 
   1041     def evaluate(self, x=None, y=None,

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/keras/engine/training_arrays.py in fit_loop(model, f, ins, out_labels, batch_size, epochs, verbose, callbacks, val_f, val_ins, shuffle, callback_metrics, initial_epoch, steps_per_epoch, validation_steps)
    197                     ins_batch[i] = ins_batch[i].toarray()
    198 
--> 199                 outs = f(ins_batch)
    200                 outs = to_list(outs)
    201                 for l, o in zip(out_labels, outs):

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/keras/backend/tensorflow_backend.py in __call__(self, inputs)
   2713                 return self._legacy_call(inputs)
   2714 
-> 2715             return self._call(inputs)
   2716         else:
   2717             if py_any(is_tensor(x) for x in inputs):

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/keras/backend/tensorflow_backend.py in _call(self, inputs)
   2673             fetched = self._callable_fn(*array_vals, run_metadata=self.run_metadata)
   2674         else:
-> 2675             fetched = self._callable_fn(*array_vals)
   2676         return fetched[:len(self.outputs)]
   2677 

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/tensorflow/python/client/session.py in __call__(self, *args, **kwargs)
   1380           ret = tf_session.TF_SessionRunCallable(
   1381               self._session._session, self._handle, args, status,
-> 1382               run_metadata_ptr)
   1383         if run_metadata:
   1384           proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

~/anaconda3/envs/ldsa/lib/python3.5/site-packages/tensorflow/python/framework/errors_impl.py in __exit__(self, type_arg, value_arg, traceback_arg)
    517             None, None,
    518             compat.as_text(c_api.TF_Message(self.status.status)),
--> 519             c_api.TF_GetCode(self.status.status))
    520     # Delete the underlying status object from memory otherwise it stays alive
    521     # as there is a reference to status from this from the traceback due to

InvalidArgumentError: Incompatible shapes: [100,5] vs. [32,5]
     [[Node: training_12/Adam/gradients/loss_17/my_layer_19_loss/sub_grad/BroadcastGradientArgs = BroadcastGradientArgs[T=DT_INT32, _class=["loc:@training_12/Adam/gradients/loss_17/my_layer_19_loss/sub_grad/Reshape"], _device="/job:localhost/replica:0/task:0/device:CPU:0"](training_12/Adam/gradients/loss_17/my_layer_19_loss/sub_grad/Shape, training_12/Adam/gradients/loss_17/my_layer_19_loss/sub_grad/Shape_1)]]

Меня удивляет следующее: несовместимые формы: [100,5] против [32,5], откуда взялись эти 32?