Tensorflow 转换到特定长度

我正在建立一个组织，面临着一个愚蠢的问题。 我的特征是1D和[174,1]作为形状，当鉴别器工作时，我有一些问题要在生成器中将我的潜在变量上采样到174,1。 我修改了mnist的一些代码，它使用了conv2dtranspose，所以可能有直接用于1d信号的东西

z_size = 100
output_size=(174, 1)
n_filters=128
n_blocks=2
size_factor = 2**n_blocks
hidden_size = (output_size[0]//size_factor)
model = tf.keras.Sequential()
model.add(Input(shape=z_size,))
model.add(Dense(units=n_filters*hidden_size))
model.add(BatchNormalization())
model.add(LeakyReLU())
model.add(Reshape((hidden_size, n_filters)))
# now we upsample the feature space to
model.add(Conv1DTranspose(filters=n_filters,
                          kernel_size=4,
                          strides=1,
                          padding='same',
                          use_bias=False
                          ))
model.add(BatchNormalization())
model.add(LeakyReLU())

nf = n_filters
for i in range(n_blocks):
    nf = nf // 2
    model.add(Conv1DTranspose(filters=nf,
                              kernel_size=5,
                              strides=2,
                              padding='same',
                              use_bias=False
                              ))
    model.add(BatchNormalization())
    model.add(LeakyReLU())

model.add(Conv1DTranspose(filters=output_size[1],
                          kernel_size=5,
                          strides=1,
                          padding='same',
                          use_bias=False,
                          activation='tanh')
                          )
model.summary()

但是尺寸就差了

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 5504)              555904    
_________________________________________________________________
batch_normalization (BatchNo (None, 5504)              22016     
_________________________________________________________________
leaky_re_lu (LeakyReLU)      (None, 5504)              0         
_________________________________________________________________
reshape (Reshape)            (None, 43, 128)           0         
_________________________________________________________________
conv1d_transpose (Conv1DTran (None, 43, 128)           65536     
_________________________________________________________________
batch_normalization_1 (Batch (None, 43, 128)           512       
_________________________________________________________________
leaky_re_lu_1 (LeakyReLU)    (None, 43, 128)           0         
_________________________________________________________________
conv1d_transpose_1 (Conv1DTr (None, 86, 64)            40960     
_________________________________________________________________
batch_normalization_2 (Batch (None, 86, 64)            256       
_________________________________________________________________
leaky_re_lu_2 (LeakyReLU)    (None, 86, 64)            0         
_________________________________________________________________
conv1d_transpose_2 (Conv1DTr (None, 172, 32)           10240     
_________________________________________________________________
batch_normalization_3 (Batch (None, 172, 32)           128       
_________________________________________________________________
leaky_re_lu_3 (LeakyReLU)    (None, 172, 32)           0         
_________________________________________________________________
conv1d_transpose_3 (Conv1DTr (None, 172, 1)            160       
=================================================================
Total params: 695,712
Trainable params: 684,256
Non-trainable params: 11,456
 _________________________________________________________________

这源于隐藏的大小=（输出大小[0]//大小系数），它是174//4->43.5四舍五入到43，因此当上采样到172时。 那么我应该手动计算正确的大小吗？或者只进行一次上采样