GPU打印机器学习层无

Question

我正在运行以下代码，对GPU 使用tensorflow

import numpy as np
import keras
from keras.models import Model, Sequential
from keras import layers
from keras.layers import BatchNormalization, Dropout, Activation, Flatten, Dense, Reshape, Conv2DTranspose, Conv2D
input_img = keras.Input(shape=(144, 48, 1))
x = layers.Conv2D(64, 3, activation="relu", strides=2,padding='same')(input_img)
print(x.shape)
x = layers.Conv2D(256, 3, activation="relu", strides=2, padding='same')(x)
print(x.shape)
x = layers.Conv2D(128, 2, activation="relu", strides=2,padding='same')(x)
print(x.shape)
encoded = layers.Conv2D(64, 2, activation="relu", strides=2,padding='same')(x)
print(encoded.shape)
encoder = Model(input_img, encoded)

encoded_input= keras.Input(shape=(9,3,64))
x = Conv2DTranspose(64, 3, activation="relu", strides=2,padding='same')(encoded_input)
print('transpose', x.shape)
x = Conv2DTranspose(32, 3, activation="relu", strides=2,padding='same')(x)
print('transpose',x.shape)
x = Conv2DTranspose(16, 3, activation="relu", strides=2,padding='same')(x)
print('transpose',x.shape)
decoded = Conv2DTranspose(1, 3, activation="relu", strides=2,padding='same')(x)
print('transpose',decoded.shape)
decoder = Model(encoded_input, decoded)

但当我打印每一层的形状时，我得到了以下输出。conv2d转置层显示"无"。

(None, 72, 24, 64)
(None, 36, 12, 256)
(None, 18, 6, 128)
(None, 9, 3, 64)
(transpose, None, None, None, 64)
(transpose, None, None, None, 32)
(transpose, None, None, None, 16)
(transpose, None, None, None, 1)

反过来，我从运行模型中得到了错误的结果。在CPU上，我没有无问题

Answer 1

我已经解决了这个问题。基本上，我改变了

layers.Conv2D

至

Conv2D