Question

该怎么做？

:18+GCC_BUILD=false
:19+GLIBC_BUILD=false
:20+KERNEL_BUILD=false
:21+POSTGRESQL_BUILD=false
:22+PYTHON_BUILD=true
:23+REINSTALL_PACKAGES=false
:24+RUBY_BUILD=true
:25+SEGMENTED_DOWNLOAD=false
:26+UBUNTU_RSYNC=false
:29+echo false
:30+echo false
:31+echo false
:32+echo false
:33+echo true
:34+echo false
:35+echo true
:36+echo false
:37+echo false
:39+ARTIFACTORY=myregistry
:42+[[ false == \t\r\u\e ]]
:47+[[ false == \t\r\u\e ]]
:52+[[ false == \t\r\u\e ]]
:57+[[ false == \t\r\u\e ]]
:62+[[ true == \t\r\u\e ]]
:63+echo 'Running PYTHON_BUILD'
Running PYTHON_BUILD
running: docker run -i --name OTQ0YzJjZTU4YWQy myregistry/image
:64+run_container python-build
::6+date +%s
::6+sha256sum
::6+base64
::6+head -c 16
:6+CONTAINER_NAME=OTQ0YzJjZTU4YWQy
:7+echo 'running: docker run -i --name OTQ0YzJjZTU4YWQy myregistry/image'
:8+docker run -i --name OTQ0YzJjZTU4YWQy myregistry/image

Done with container python-build

:9+echo 'removing: container OTQ0YzJjZTU4YWQy from python-build'
:10+docker container rm OTQ0YzJjZTU4YWQy
:11+echo 'removing:  image python-build'
:12+docker image rm myregistry/image
:13+echo -e '\nDone with container python-build\n\n\n'
:14+return 0

因此nn中的所有张量都相同，只有输入训练分支不同？

在纯张量流中，您可以使用

nn = get_networks()
A = nn(X_input)
B = nn(X_other_input)
C = A + B
model = ...

并仔细命名您的图层。

但是基本上您可以首先构建nn，因为您不能将非调用层传递给调用层！

例如：

tf.variable_scope('something', reuse=tf.AUTO_REUSE):
       define stuff here

更新：

我已经通过创建未编译的模型作为子网来完成此任务。然后可以将该“模型”传递给其他网络创建功能。例如，如果您有一个要求解的泛函方程，则可以使用网络对函数进行近似，然后将网络传递给本身就是网络的函数。

Answer 1

这取决于您要如何重用它，但是其想法是保存初始化后的图层，并在以后多次使用。

import tensorflow as tf
import tensorflow.keras as keras
import tensorflow.keras.layers as layers
import numpy as np

layers = {}

def net1(input):
    layers["l1"] = keras.layers.Dense(10)
    layers["l2"] = keras.layers.Dense(10)

    return layers["l1"](layers["l2"](keras.layers.Flatten()(input)))

def net2(input):
    return layers["l1"](layers["l2"](keras.layers.Flatten()(input)))

input1 = keras.layers.Input((2, 2))
input2 = keras.layers.Input((2, 2))

model1 = keras.Model(inputs=input1, outputs=net1(input1)) 
model1.compile(loss=keras.losses.mean_squared_error, optimizer=keras.optimizers.Adam())

model2 = keras.Model(inputs=input2, outputs=net2(input2)) 
model2.compile(loss=keras.losses.mean_squared_error, optimizer=keras.optimizers.Adam())

x = np.random.randint(0, 100, (50, 2, 2))
m1 = model1.predict(x)
m2 = model2.predict(x)
print(x[0])
print(m1[0])
print(m2[0])

输出相同：

[ 10.114908  -13.074531   -8.671929  -59.03201    55.389366    1.3610549
 -38.051434    8.355987    7.5310936 -27.717983 ]
[ 10.114908  -13.074531   -8.671929  -59.03201    55.389366    1.3610549
 -38.051434    8.355987    7.5310936 -27.717983 ]

如何在网络的两个输入分支之间构造和重用网络？

1 个答案: