我正在尝试使用this页来构建变体自动编码器。这是我的模型的摘要：
Model: "model_44"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_32 (InputLayer) (None, 784) 0
__________________________________________________________________________________________________
dense_70 (Dense) (None, 32) 25120 input_32[0][0]
__________________________________________________________________________________________________
dense_71 (Dense) (None, 2) 66 dense_70[0][0]
__________________________________________________________________________________________________
dense_72 (Dense) (None, 2) 66 dense_70[0][0]
__________________________________________________________________________________________________
lambda_16 (Lambda) (None, 2) 0 dense_71[0][0]
dense_72[0][0]
__________________________________________________________________________________________________
dense_73 (Dense) (None, 32) 96 lambda_16[0][0]
__________________________________________________________________________________________________
dense_74 (Dense) (None, 784) 25872 dense_73[0][0]
==================================================================================================
Total params: 51,220
Trainable params: 51,220
Non-trainable params: 0
但是，每当我开始训练时，在第一个时期，我都会遇到以下错误：
InvalidArgumentError：不兼容的形状：[96]与[128] [[node 渐变_8 /损失_17 /密集_74_损失/ vae_loss / weighted_loss / mul_grad / Mul_1 （定义为 /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3009） ]] [操作：__ inference_keras_scratch_graph_49801]
函数调用堆栈：keras_scratch_graph
我的代码与链接完全相同。你能帮我吗？