The code that's used for the encoder network is as follows:

# Encoder
input_layer <- 
         layer_input(shape = c(28,28,1)) 
encoder <-  input_layer %>% 
         layer_conv_2d(filters = 32, 
                       kernel_size = c(3,3), 
                       activation = 'relu', 
                       padding = 'same') %>%   
         layer_max_pooling_2d(pool_size = c(2,2),
                              padding = 'same') %>% 
         layer_conv_2d(filters = 32, 
                       kernel_size = c(3,3), 
                       activation = 'relu', 
                       padding = 'same') %>%  
         layer_max_pooling_2d(pool_size = c(2,2), 
                              padding = 'same') 
summary(encoder)
OutputTensor("max_pooling2d_6/MaxPool:0", shape=(?, 7, 7, 32), dtype=float32)

Here, the input layer is specified to be 28 x 28 x 1 in size. We use two convolution layers with 32 filters each and a rectifier linear unit as the activation function. Each convolution layer is followed by pooling layers. After the first pooling layer, the height and width change to 14 x 14, and after the second pooling layer, this changes to 7 x 7. The output of the encoder network in this example has 7 x 7 x 32 dimensions.