张量流构建中的输入大小

Question

我正在查看文档中的示例： https://www.tensorflow.org/tutorials/generative/autoencoder

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import tensorflow as tf

from sklearn.metrics import accuracy_score, precision_score, recall_score
from sklearn.model_selection import train_test_split
from tensorflow.keras import layers, losses
from tensorflow.keras.datasets import fashion_mnist
from tensorflow.keras.models import Model


class Autoencoder(Model):
  def __init__(self, latent_dim, shape):
    super(Autoencoder, self).__init__()
    self.latent_dim = latent_dim
    self.shape = shape
    self.encoder = tf.keras.Sequential([
      layers.Flatten(),  # flattens input WITHOUT affecting batch size
      layers.Dense(latent_dim, activation='relu'),
    ])
    self.decoder = tf.keras.Sequential([
      layers.Dense(tf.math.reduce_prod(shape).numpy(),
                   activation='sigmoid'),
      layers.Reshape(shape)
    ])

  def call(self, x):
    encoded = self.encoder(x)
    decoded = self.decoder(encoded)
    return decoded


# (x_train, _), (x_test, _) = fashion_mnist.load_data()

x_train = x_train.astype('float32') / 255.
x_test = x_test.astype('float32') / 255.

print(x_train.shape)  # (60000, 28, 28)
print(x_test.shape)   # (10000, 28, 28)

shape = x_test.shape[1:]
latent_dim = 64
autoencoder = Autoencoder(latent_dim, shape)

autoencoder.compile(optimizer='adam', loss=losses.MeanSquaredError())
# autoencoder.build(input_shape=x_train.shape)  # nope
# autoencoder.build(input_shape=(None, 28, 28))  # nope
# autoencoder.summary()

我想构建模型来获得摘要。 input_shape 应该是什么？输入尺寸为 28x28 的图像。

我希望这能起作用：

autoencoder.build(input_shape=(None, 28, 28))

非常感谢！

Answer 1

谢谢您的建议。

重新审视这个例子：

autoencoder.build(input_shape=x_train.shape)
autoencoder.summary()

产量：

_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 sequential (Sequential)     (60000, 64)               50240     
                                                                 
 sequential_1 (Sequential)   (60000, 28, 28)           50960     
                                                                 
=================================================================
Total params: 101,200

以下：

autoencoder.build(input_shape=(None, 28, 28))
autoencoder.summary()

产量：

_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 sequential_2 (Sequential)   (None, 64)                50240     
                                                                 
 sequential_3 (Sequential)   (None, 28, 28)            50960     
                                                                 
=================================================================
Total params: 101,200
Trainable params: 101,200
Non-trainable params: 0
_________________________________________________________________

在这两种情况下，第一层的参数数量都是 28 x 28 x 64 + 64 = 50240，正如预期的那样。

感谢您的提示！