活性化関数sigmoidのエラー

#前提・実現したいこと
Variational Autoencoderを使った画像の異常検知　後編　（塩尻MLもくもく会＃7）
上記URLのサイト様のコードを参考にVAEとオリジナルデータセットを用いて異常検知を行いたいと考えております。

#エラーメッセージ

Python
1-------------------------------------------------------------------------
2InvalidArgumentError                    Traceback (most recent call last)
3<ipython-input-40-8e19e3bc72e7> in <module>
4     62 vae.fit(x_train_b,
5     63         epochs=epochs,
6---> 64         batch_size=batch_size)
7     65         #validation_data=(x_test, None))
8     66 vae.save_weights('vae_mlp_mnist.h5')
9
10~/.pyenv/versions/3.6.5/lib/python3.6/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, validation_freq, max_queue_size, workers, use_multiprocessing, **kwargs)
11   1237                                         steps_per_epoch=steps_per_epoch,
12   1238                                         validation_steps=validation_steps,
13-> 1239                                         validation_freq=validation_freq)
14   1240 
15   1241     def evaluate(self,
16
17~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/keras/engine/training_arrays.py in fit_loop(model, fit_function, fit_inputs, out_labels, batch_size, epochs, verbose, callbacks, val_function, val_inputs, shuffle, initial_epoch, steps_per_epoch, validation_steps, validation_freq)
18    194                     ins_batch[i] = ins_batch[i].toarray()
19    195 
20--> 196                 outs = fit_function(ins_batch)
21    197                 outs = to_list(outs)
22    198                 for l, o in zip(out_labels, outs):
23
24~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/keras/backend.py in __call__(self, inputs)
25   3790         value = math_ops.cast(value, tensor.dtype)
26   3791       converted_inputs.append(value)
27-> 3792     outputs = self._graph_fn(*converted_inputs)
28   3793 
29   3794     # EagerTensor.numpy() will often make a copy to ensure memory safety.
30
31~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/eager/function.py in __call__(self, *args, **kwargs)
32   1603       TypeError: For invalid positional/keyword argument combinations.
33   1604     """
34-> 1605     return self._call_impl(args, kwargs)
35   1606 
36   1607   def _call_impl(self, args, kwargs, cancellation_manager=None):
37
38~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/eager/function.py in _call_impl(self, args, kwargs, cancellation_manager)
39   1643       raise TypeError("Keyword arguments {} unknown. Expected {}.".format(
40   1644           list(kwargs.keys()), list(self._arg_keywords)))
41-> 1645     return self._call_flat(args, self.captured_inputs, cancellation_manager)
42   1646 
43   1647   def _filtered_call(self, args, kwargs):
44
45~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/eager/function.py in _call_flat(self, args, captured_inputs, cancellation_manager)
46   1744       # No tape is watching; skip to running the function.
47   1745       return self._build_call_outputs(self._inference_function.call(
48-> 1746           ctx, args, cancellation_manager=cancellation_manager))
49   1747     forward_backward = self._select_forward_and_backward_functions(
50   1748         args,
51
52~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/eager/function.py in call(self, ctx, args, cancellation_manager)
53    596               inputs=args,
54    597               attrs=attrs,
55--> 598               ctx=ctx)
56    599         else:
57    600           outputs = execute.execute_with_cancellation(
58
59~/.pyenv/versions/3.6.5/lib/python3.6/site-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
60     58     ctx.ensure_initialized()
61     59     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
62---> 60                                         inputs, attrs, num_outputs)
63     61   except core._NotOkStatusException as e:
64     62     if name is not None:
65
66InvalidArgumentError:  Incompatible shapes: [960000] vs. [3200]
67	 [[node gradients_5/Sub_511_grad/BroadcastGradientArgs (defined at /Users/user_name/.pyenv/versions/3.6.5/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3009) ]] [Op:__inference_keras_scratch_graph_54577]
68
69Function call stack:
70keras_scratch_graph

#コード

Python
1from __future__ import absolute_import
2from __future__ import division
3from __future__ import print_function
4
5from keras.layers import Lambda, Input, Dense, Reshape
6from keras.models import Model
7from keras.losses import mse
8from keras.utils import plot_model
9from keras import backend as K
10from keras.layers import BatchNormalization, Activation, Flatten
11from keras.layers.convolutional import Conv2DTranspose, Conv2D
12
13import numpy as np
14import matplotlib.pyplot as plt
15import matplotlib.colors as colors
16import os
17from sklearn import metrics
18
19#〜文字数制限のため省略〜
20
21# reparameterization trick
22# instead of sampling from Q(z|X), sample eps = N(0,I)
23# z = z_mean + sqrt(var)*eps
24def sampling(args):
25    z_mean, z_log_var = args
26    batch = K.shape(z_mean)[0]
27    dim = K.int_shape(z_mean)[1]
28    # by default, random_normal has mean=0 and std=1.0
29    epsilon = K.random_normal(shape=(batch, dim))
30    return z_mean + K.exp(0.5 * z_log_var) * epsilon
31
32# dataset
33from bcn_dataset import BCN_Dataset2
34(x_train, y_train), (x_test, y_test) = BCN_Dataset2.create_bcn()
35
36x_train = x_train.reshape(x_train.shape[0], 224, 224, 3)
37x_test = x_test.reshape(x_test.shape[0], 224, 224, 3)
38
39x_train = x_train.astype('float32') / 255
40x_test = x_test.astype('float32') / 255
41
42x_train_b = []
43x_train_shape = x_train.shape
44
45for i in range(len(x_train)):
46  if y_train[i] == 1:#スニーカーは7
47    temp = x_train[i,:,:,:]
48    x_train_b.append(temp.reshape((x_train_shape[1],x_train_shape[2],x_train_shape[3])))
49
50x_train_b = np.array(x_train_b)
51x_train_b = cut_img(x_train_b, 50)
52print("train data:",len(x_train_b))
53
54x_test_b = []
55x_test_n = []
56for i in range(len(x_test)):
57  if y_test[i] == 1:#スニーカーは7
58    temp = x_test[i,:,:,:]
59    x_test_b.append(temp.reshape((x_train_shape[1],x_train_shape[2],x_train_shape[3])))
60
61  if y_test[i] == 2:
62    temp = x_test[i,:,:,:]
63    x_test_n.append(temp.reshape((x_train_shape[1],x_train_shape[2],x_train_shape[3])))
64
65x_test_b = np.array(x_test_b)
66x_test_n = np.array(x_test_n)
67
68# network parameters
69input_shape=(80, 80, 3)
70batch_size = 128
71latent_dim = 2
72epochs = 10
73Nc = 16
74
75# build encoder model
76inputs = Input(shape=input_shape, name='encoder_input')
77x = Conv2D(Nc, kernel_size=2, strides=2)(inputs)
78x = BatchNormalization()(x)
79x = Activation('relu')(x)
80x = Conv2D(2*Nc, kernel_size=2, strides=2)(x)
81x = BatchNormalization()(x)
82x = Activation('relu')(x)
83x = Flatten()(x)
84
85z_mean = Dense(latent_dim, name='z_mean')(x)
86z_log_var = Dense(latent_dim, name='z_log_var')(x)
87z = Lambda(sampling, output_shape=(latent_dim,), name='z')([z_mean, z_log_var])
88
89encoder = Model(inputs, [z_mean, z_log_var, z], name='encoder')
90#encoder.summary()
91
92# build decoder model
93latent_inputs = Input(shape=(latent_dim,), name='z_sampling')
94x = Dense(2*2)(latent_inputs)
95x = BatchNormalization()(x)
96x = Activation('relu')(x)
97x = Reshape((2,2,1))(x)
98x = Conv2DTranspose(2*Nc, kernel_size=2, strides=2, padding='same')(x)
99x = BatchNormalization()(x)
100x = Activation('relu')(x)
101x = Conv2DTranspose(Nc, kernel_size=2, strides=2, padding='same')(x)
102x = BatchNormalization()(x)
103x = Activation('relu')(x)
104
105x1 = Conv2DTranspose(1, kernel_size=4, padding='same')(x)
106x1 = BatchNormalization()(x1)
107out1 = Activation('sigmoid')(x1)#out.shape=(n,8,8,1)
108
109x2 = Conv2DTranspose(1, kernel_size=4, padding='same')(x)
110x2 = BatchNormalization()(x2)
111out2 = Activation('sigmoid')(x2)#out.shape=(n,8,8,1)
112
113decoder = Model(latent_inputs, [out1, out2], name='decoder')
114#decoder.summary()
115
116# build VAE model
117outputs_mu, outputs_sigma_2 = decoder(encoder(inputs)[2])
118vae = Model(inputs, [outputs_mu, outputs_sigma_2], name='vae_mlp')
119
120# VAE loss
121m_vae_loss = (K.flatten(inputs) - K.flatten(outputs_mu))**2 / K.flatten(outputs_sigma_2)
122m_vae_loss = 0.5 * K.sum(m_vae_loss)
123
124a_vae_loss = K.log(2 * 3.14 * K.flatten(outputs_sigma_2))
125a_vae_loss = 0.5 * K.sum(a_vae_loss)
126
127kl_loss = 1 + z_log_var - K.square(z_mean) - K.exp(z_log_var)
128kl_loss = K.sum(kl_loss, axis=-1)
129kl_loss *= -0.5
130
131vae_loss = K.mean(kl_loss + m_vae_loss + a_vae_loss)
132vae.add_loss(vae_loss)
133vae.compile(optimizer='adam')
134
135# train the autoencoder
136vae.fit(x_train_b,
137        epochs=epochs,
138        batch_size=batch_size)
139        #validation_data=(x_test, None))
140vae.save_weights('vae_mlp_mnist.h5')

#試していること
サイト様のコードでは、サイズが(28,28)の白黒のmnistデータセットを使用しているため、(8,8,1)となっておりました。しかし私のデータセットはカラー、かつ画像のサイズが(224,224)であるため値を(80,80,3)に変更してコードの実行を行っておりましたところ、上記のようなエラーが発生してしまいました。コードの下のほうにある# build decoder modelより下の#out.shape=(n,8,8,1)に解決策があるのではないか、もしかするとsigmoid関数ではうまくいかないのか？などの考察をしておりますが、なかなか解決せず頭を抱えております。何かご助言をいただけたら幸いです。

#補足
PCはmacOS Catalina バージョン10.15.5
Pythonのバージョンは3.6.5です
Jupiter notebookを使用しています
文字数制限の関係上、コードの関係ないと思われる関数を一部、省略させていただいております。