我正在尝试在张量流中构建矢量量化的变分自动编码器。当我建立模型并开始训练时，会发生一个非常神秘的错误。
我已遍历代码以查找错误，但找不到。该错误对查找原因非常不利。下面是模型代码的一部分
def call(self, inputs):
z = self.encode(inputs)
vq_output = self.vector_quantizer(z)
reconstructed_image = self.decode(vq_output['quantize'])
return reconstructed_image, vq_output
def encode(self, image):
encoded = Conv2D(int(self._num_hidden/2), kernel_size=(4,4), strides=(2,2), activation='relu', name='enc1')(image)
encoded = Conv2D(self._num_hidden, kernel_size=(4,4), strides=(2,2), activation='relu', name='enc2')(encoded)
encoded = Conv2D(self._num_hidden, kernel_size=(3,3), strides=(1,1), name='enc2')(encoded)
encoded = residual_stack(encoded, self._num_hidden, self._num_residual_layers, self._num_residual_hiddens)
return Conv2D(self._embeding_dim, kernel_size=(1,1) ,strides=(1,1), name='enc3')(encoded)
def decode(self, encoded_input):
decoded = Conv2D(self._decoder_num_hidden, kernel_size=(3,3), strides=(1,1), name='dec1')(encoded_input)
decoded = residual_stack(decoded, self._decoder_num_hidden, self._decoder_num_residual_layers, self._decoder_num_residual_hiddens)
decoded = Conv2DTranspose(int(self._decoder_num_hidden/2), kernel_size=(4,4), strides=(2,2), activation='relu', name='dec2')(decoded)
return Conv2DTranspose(3, kernel_size=(4,4), strides=(2,2), name='dec3')(decoded)
发生以下错误：
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-52-90a0a1fb38f6> in <module>()
15
16 for batch in train_batches:
---> 17 _loss, _perplexity = train(vq_vae, batch['images'], optimizer)
18 loss.update_state(_loss)
19 perplexity.update_state(_perplexity)
14 frames
<ipython-input-45-0025b05fe5c9> in train(model, inputs, optimizer, outputs)
1 def train(model, inputs, optimizer, outputs=None):
2 with tf.GradientTape() as tape:
----> 3 loss, perplexity = compute_loss(model, inputs)
4 gradients = tape.gradient(loss, model.trainable_variables)
5 optimizer.apply_gradients(zip(gradients, model.trainable_variables))
<ipython-input-44-242959fe043f> in compute_loss(model, inputs, outputs)
1 def compute_loss(model, inputs, outputs=None):
----> 2 recon, vq_outputs = model(inputs)
3 recon_loss = tf.reduce_mean((recon - inputs)**2) / data_variance
4 loss = recon_loss + vq_outputs['loss']
5 return loss, vq_outputs['perplexity']
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py in __call__(self, inputs, *args, **kwargs)
677 with base_layer_utils.autocast_context_manager(
678 input_list, self._mixed_precision_policy.should_cast_variables):
--> 679 outputs = self.call(inputs, *args, **kwargs)
680 self._handle_activity_regularization(inputs, outputs)
681 self._set_mask_metadata(inputs, outputs, previous_mask)
<ipython-input-43-f28d8aaad600> in call(self, inputs)
40 vq_output = self.vector_quantizer(z)
41
---> 42 reconstructed_image = self.decode(vq_output['quantize'])
43
44 return reconstructed_image, vq_output
<ipython-input-43-f28d8aaad600> in decode(self, encoded_input)
55
56 def decode(self, encoded_input):
---> 57 decoded = Conv2D(self._decoder_num_hidden, kernel_size=(3,3), strides=(1,1), name='dec1')(encoded_input)
58
59 decoded = residual_stack(decoded, self._decoder_num_hidden, self._decoder_num_residual_layers, self._decoder_num_residual_hiddens)
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py in __call__(self, inputs, *args, **kwargs)
674 # Eager execution on data tensors.
675 with backend.name_scope(self._name_scope()):
--> 676 self._maybe_build(inputs)
677 with base_layer_utils.autocast_context_manager(
678 input_list, self._mixed_precision_policy.should_cast_variables):
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py in _maybe_build(self, inputs)
1879 # operations.
1880 with tf_utils.maybe_init_scope(self):
-> 1881 self.build(input_shapes)
1882 # We must set self.built since user defined build functions are not
1883 # constrained to set self.built.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/layers/convolutional.py in build(self, input_shape)
163 constraint=self.kernel_constraint,
164 trainable=True,
--> 165 dtype=self.dtype)
166 if self.use_bias:
167 self.bias = self.add_weight(
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py in add_weight(self, name, shape, dtype, initializer, regularizer, trainable, constraint, partitioner, use_resource, synchronization, aggregation, **kwargs)
382 collections=collections_arg,
383 synchronization=synchronization,
--> 384 aggregation=aggregation)
385 backend.track_variable(variable)
386
/usr/local/lib/python3.6/dist-packages/tensorflow/python/training/tracking/base.py in _add_variable_with_custom_getter(self, name, shape, dtype, initializer, getter, overwrite, **kwargs_for_getter)
661 dtype=dtype,
662 initializer=initializer,
--> 663 **kwargs_for_getter)
664
665 # If we set an initializer and the variable processed it, tracking will not
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer_utils.py in make_variable(name, shape, dtype, initializer, trainable, caching_device, validate_shape, constraint, use_resource, collections, synchronization, aggregation, partitioner)
140 # TODO(apassos,rohanj) figure out how to remove collections from here so we
141 # can remove the V1.
--> 142 variable_shape = tensor_shape.TensorShape(shape)
143 return tf_variables.VariableV1(
144 initial_value=init_val,
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py in __init__(self, dims)
772 else:
773 # Got a list of dimensions
--> 774 self._dims = [as_dimension(d) for d in dims_iter]
775
776 @property
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py in <listcomp>(.0)
772 else:
773 # Got a list of dimensions
--> 774 self._dims = [as_dimension(d) for d in dims_iter]
775
776 @property
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py in as_dimension(value)
714 return value
715 else:
--> 716 return Dimension(value)
717
718
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py in __init__(self, value)
183 raise TypeError("Cannot convert %s to Dimension" % value)
184 else:
--> 185 self._value = int(value)
186 if (not isinstance(value, compat.bytes_or_text_types) and
187 self._value != value):
TypeError: int() argument must be a string, a bytes-like object or a number, not 'tuple'
通过查看此错误，我无法理解我在哪里犯了错误。