Python keras子类模型的多输出
下面是一个重现问题的小示例: 问题1:尽管我调用了Python keras子类模型的多输出,python,tensorflow,keras,deep-learning,Python,Tensorflow,Keras,Deep Learning,下面是一个重现问题的小示例: 问题1:尽管我调用了build()函数,summary()没有显示输出形状。 问题2:尽管我通过调用build()修复了模型的输入维度,但模型仍然接受任何其他形状 import os from tensorflow.keras.models import Model from tensorflow.keras.layers import Conv2D, MaxPool2D, BatchNormalization, ReLU import numpy as np
build()
函数,summary()
没有显示输出形状。问题2:尽管我通过调用
build()
修复了模型的输入维度,但模型仍然接受任何其他形状
import os
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Conv2D, MaxPool2D, BatchNormalization, ReLU
import numpy as np
class SubModel(Model):
def __init__(self, image_shape=(512, 512, 3), name='SubModel'):
# - front_end -
super().__init__(name=name)
self.image_shape = image_shape
# block1
self.block1_conv1 = Conv2D(filters=64, kernel_size=3,
activation=None, padding='same',
name='block1_conv1')
self.build(input_shape=(None, *(self.image_shape)))
def call(self, x):
x = self.block1_conv1(x)
return x
model = SubModel()
model.summary()
x= np.random.normal(size=(1, 256, 256, 3))
y= model(x)
print(y.shape)