blocks|key|2368186|text|最近，我创建了一个用于绘制NN体系结构和导出SVG的工具，名为神经网络-SVG。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2368187|📷|atomic|2368188|entityMap|0|LINK|mutability|MUTABLE|url|http://alexlenail.me/NN-SVG/index.html|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/f96kw.jpg|imageAlt|2|https://i.stack.imgur.com/OpS8l.png^0|V|8|0|0|0|1|1|0|0|1|2^^$0|@$1|2|3|4|5|6|7|W|8|@]|9|@$A|X|B|Y|1|Z]]|C|$]]|$1|D|3|E|5|F|7|10|8|@]|9|@$A|11|B|12|1|13]]|C|$]]|$1|G|3|E|5|F|7|14|8|@]|9|@$A|15|B|16|1|17]]|C|$]]]|H|$I|$5|J|K|L|C|$M|N]]|O|$5|P|K|Q|C|$R|S|T|-4]]|U|$5|P|K|Q|C|$R|V|T|-4]]]]

I recently created a tool for drawing NN architectures and exporting SVG, called <a href="http://alexlenail.me/NN-SVG/index.html" rel="noreferrer">NN-SVG</a>
<a href="https://i.stack.imgur.com/f96kw.jpg" rel="noreferrer"><img src="https://i.stack.imgur.com/f96kw.jpg" alt="enter image description here" /></a>
<a href="https://i.stack.imgur.com/OpS8l.png" rel="noreferrer"><img src="https://i.stack.imgur.com/OpS8l.png" alt="second screenshot" /></a>

blocks|key|2367091|text|Tensorflow，Keras，MXNet，PyTorch|type|header-two|depth|inlineStyleRanges|entityRanges|data|2367092|如果神经网络给出的是一个Tensorflow图，那么您可以用TensorBoard可视化此图。|unstyled|offset|length|2367093|以下是MNIST+CNN的样子：|2367094|📷|atomic|2367095|您可以自己添加名称/作用域(如“退出”、"softmax“、"fc1”、"conv1“、"conv2")。|2367096|Interpretation|2367097|以下只是关于左图的内容。我忽略了右边的四个小图。|2367098|每个框都是一个具有可学习参数的层。为了推论，信息流从下到上。椭圆是不包含学习参数的层。|2367099|盒子的颜色没有意义。|2367100|我不确定虚线小框的价值(“渐变”、“亚当”、“保存”)。|entityMap|0|LINK|mutability|MUTABLE|url|https://www.tensorflow.org/guide/graph_viz|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/zJHpV.png|imageAlt^0|0|T|H|0|0|0|0|1|1|0|0|0|0|0|0^^$0|@$1|2|3|4|5|6|7|1A|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|1B|8|@]|9|@$E|1C|F|1D|1|1E]]|A|$]]|$1|G|3|H|5|D|7|1F|8|@]|9|@]|A|$]]|$1|I|3|J|5|K|7|1G|8|@]|9|@$E|1H|F|1I|1|1J]]|A|$]]|$1|L|3|M|5|D|7|1K|8|@]|9|@]|A|$]]|$1|N|3|O|5|6|7|1L|8|@]|9|@]|A|$]]|$1|P|3|Q|5|D|7|1M|8|@]|9|@]|A|$]]|$1|R|3|S|5|D|7|1N|8|@]|9|@]|A|$]]|$1|T|3|U|5|D|7|1O|8|@]|9|@]|A|$]]|$1|V|3|W|5|D|7|1P|8|@]|9|@]|A|$]]]|X|$Y|$5|Z|10|11|A|$12|13]]|14|$5|15|10|16|A|$17|18|19|-4]]]]

<h2>Tensorflow, Keras, MXNet, PyTorch</h2>

If the neural network is given as a Tensorflow graph, then you can <a href="https://www.tensorflow.org/guide/graph_viz" rel="noreferrer">visualize this graph with TensorBoard</a>.

Here is how the MNIST CNN looks like:

<a href="https://i.stack.imgur.com/zJHpV.png" rel="noreferrer"><img src="https://i.stack.imgur.com/zJHpV.png" alt="enter image description here"></a>

You can add names / scopes (like "dropout", "softmax", "fc1", "conv1", "conv2") yourself.

<h3>Interpretation</h3>

The following is only about the left graph. I ignore the 4 small graphs on the right half.

Each box is a layer with parameters that can be learned. For inference, information flows from bottom to the top. Ellipses are layers which do not contain learned parameters.

The color of the boxes does not have a meaning.

I'm not sure of the value of the dashed small boxes ("gradients", "Adam", "save").

blocks|key|1225324|text|有一个名为内特恩的开源项目|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|1225325|Netron是神经网络、深度学习和机器学习模型的观察者。Netron支持ONNX+(.onnx，.pb)、Keras+(.h5，.keras)、CoreML+(.mlmodel)和TensorFlow+Lite+(.tflite)。Netron对Caffe+(.caffemodel)、Caffe2+(predict_net.pb)、MXNet+(-symbol.json)、TensorFlow.js+(model.json，.pb)和TensorFlow+(.pb，.meta)有实验支持。|blockquote|1225326|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/lutzroeder/Netron|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/R7PT1.png|imageAlt^0|5|3|0|0|0|0|1|1^^$0|@$1|2|3|4|5|6|7|W|8|@]|9|@$A|X|B|Y|1|Z]]|C|$]]|$1|D|3|E|5|F|7|10|8|@]|9|@]|C|$]]|$1|G|3|H|5|I|7|11|8|@]|9|@$A|12|B|13|1|14]]|C|$]]]|J|$K|$5|L|M|N|C|$O|P]]|Q|$5|R|M|S|C|$T|U|V|-4]]]]

There is an open source project called <a href="https://github.com/lutzroeder/Netron" rel="noreferrer">Netron</a>
<blockquote>
Netron is a viewer for neural network, deep learning and machine learning models.
Netron supports ONNX (.onnx, .pb), Keras (.h5, .keras), CoreML (.mlmodel) and TensorFlow Lite (.tflite). Netron has experimental support for Caffe (.caffemodel), Caffe2 (predict_net.pb), MXNet (-symbol.json), TensorFlow.js (model.json, .pb) and TensorFlow (.pb, .meta).
</blockquote>
<a href="https://i.stack.imgur.com/R7PT1.png" rel="noreferrer"><img src="https://i.stack.imgur.com/R7PT1.png" alt="enter image description here" /></a>

blocks|key|2367299|text|在Caffe中，您可以使用首席执行官/拖拽绘制NetParameter原型缓冲区：|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2367300|📷|atomic|2367301|在Matlab中，您可以使用视图(净额)。|2367302|2367303|Keras.js：|2367304|2367305|另外，请参见有人能推荐一个网络架构可视化工具吗？(Reddit/Reddit.Reddit.)。|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/BVLC/caffe/blob/master/python/caffe/draw.py|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/5Z1Cb.png|imageAlt|2|http://www.mathworks.com/help/nnet/ref/view.html|3|https://i.stack.imgur.com/rPpfa.png|4|https://transcranial.github.io/keras-js/#/inception-v3|5|https://i.stack.imgur.com/vEfTv.png|6|https://redd.it/kmfx2d^0|D|8|0|0|0|1|1|0|E|6|2|0|0|1|3|0|0|8|4|0|0|1|5|0|6|15|6^^$0|@$1|2|3|4|5|6|7|1B|8|@]|9|@$A|1C|B|1D|1|1E]]|C|$]]|$1|D|3|E|5|F|7|1F|8|@]|9|@$A|1G|B|1H|1|1I]]|C|$]]|$1|G|3|H|5|6|7|1J|8|@]|9|@$A|1K|B|1L|1|1M]]|C|$]]|$1|I|3|E|5|F|7|1N|8|@]|9|@$A|1O|B|1P|1|1Q]]|C|$]]|$1|J|3|K|5|6|7|1R|8|@]|9|@$A|1S|B|1T|1|1U]]|C|$]]|$1|L|3|E|5|F|7|1V|8|@]|9|@$A|1W|B|1X|1|1Y]]|C|$]]|$1|M|3|N|5|6|7|1Z|8|@]|9|@$A|20|B|21|1|22]]|C|$]]]|O|$P|$5|Q|R|S|C|$T|U]]|V|$5|W|R|X|C|$Y|Z|10|-4]]|11|$5|Q|R|S|C|$T|12]]|13|$5|W|R|X|C|$Y|14|10|-4]]|15|$5|Q|R|S|C|$T|16]]|17|$5|W|R|X|C|$Y|18|10|-4]]|19|$5|Q|R|S|C|$T|1A]]]]

In Caffe you can use <a href="https://github.com/BVLC/caffe/blob/master/python/caffe/draw.py" rel="nofollow noreferrer">caffe/draw.py</a> to draw the NetParameter protobuffer:
<a href="https://i.stack.imgur.com/5Z1Cb.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/5Z1Cb.png" alt="enter image description here" /></a>
In Matlab, you can use <a href="http://www.mathworks.com/help/nnet/ref/view.html" rel="nofollow noreferrer">view(net)</a>
<a href="https://i.stack.imgur.com/rPpfa.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/rPpfa.png" alt="enter image description here" /></a>
<a href="https://transcranial.github.io/keras-js/#/inception-v3" rel="nofollow noreferrer">Keras.js</a>:
<a href="https://i.stack.imgur.com/vEfTv.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/vEfTv.png" alt="enter image description here" /></a>
Also, see <a href="https://redd.it/kmfx2d" rel="nofollow noreferrer">Can anyone recommend a Network Architecture visualization tool? (Reddit/self.MachineLearning)</a>.

blocks|key|1225250|text|我将使用角化-序贯性添加ASCII可视化(免责声明:我是作者)。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|1225251|CIFAR-10+(来自本教程)的一个小型网络是：|1225252|+++++++OPERATION+++++++++++DATA+DIMENSIONS+++WEIGHTS(N)+++WEIGHTS(%25)

+++++++++++Input+++#####+++++32+++32++++3
++++++++++Conv2D++++\%7C/++-------------------+++++++896+++++2.1%25
++++++++++++relu+++#####+++++30+++30+++32
++++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
+++++++++++++++++++#####+++++15+++15+++32
++++++++++Conv2D++++\%7C/++-------------------+++++18496++++43.6%25
++++++++++++relu+++#####+++++13+++13+++64
++++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
+++++++++++++++++++#####++++++6++++6+++64
+++++++++Flatten+++%7C%7C%7C%7C%7C+-------------------+++++++++0+++++0.0%25
+++++++++++++++++++#####++++++++2304
+++++++++++Dense+++XXXXX+-------------------+++++23050++++54.3%25
+++++++++softmax+++#####++++++++++10|code-block|syntax|javascript|1225253|对于VGG16来说，它应该是：|1225254|+++++++OPERATION+++++++++++DATA+DIMENSIONS+++WEIGHTS(N)+++WEIGHTS(%25)

++++++++++Input+++#####++++++3++224++224
+++++InputLayer+++++%7C+++-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####++++++3++224++224
++Convolution2D++++\%7C/++-------------------++++++1792+++++0.0%25
+++++++++++relu+++#####+++++64++224++224
++Convolution2D++++\%7C/++-------------------+++++36928+++++0.0%25
+++++++++++relu+++#####+++++64++224++224
+++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####+++++64++112++112
++Convolution2D++++\%7C/++-------------------+++++73856+++++0.1%25
+++++++++++relu+++#####++++128++112++112
++Convolution2D++++\%7C/++-------------------++++147584+++++0.1%25
+++++++++++relu+++#####++++128++112++112
+++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####++++128+++56+++56
++Convolution2D++++\%7C/++-------------------++++295168+++++0.2%25
+++++++++++relu+++#####++++256+++56+++56
++Convolution2D++++\%7C/++-------------------++++590080+++++0.4%25
+++++++++++relu+++#####++++256+++56+++56
++Convolution2D++++\%7C/++-------------------++++590080+++++0.4%25
+++++++++++relu+++#####++++256+++56+++56
+++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####++++256+++28+++28
++Convolution2D++++\%7C/++-------------------+++1180160+++++0.9%25
+++++++++++relu+++#####++++512+++28+++28
++Convolution2D++++\%7C/++-------------------+++2359808+++++1.7%25
+++++++++++relu+++#####++++512+++28+++28
++Convolution2D++++\%7C/++-------------------+++2359808+++++1.7%25
+++++++++++relu+++#####++++512+++28+++28
+++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####++++512+++14+++14
++Convolution2D++++\%7C/++-------------------+++2359808+++++1.7%25
+++++++++++relu+++#####++++512+++14+++14
++Convolution2D++++\%7C/++-------------------+++2359808+++++1.7%25
+++++++++++relu+++#####++++512+++14+++14
++Convolution2D++++\%7C/++-------------------+++2359808+++++1.7%25
+++++++++++relu+++#####++++512+++14+++14
+++MaxPooling2D+++Y+max+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####++++512++++7++++7
++++++++Flatten+++%7C%7C%7C%7C%7C+-------------------+++++++++0+++++0.0%25
++++++++++++++++++#####+++++++25088
++++++++++Dense+++XXXXX+-------------------+102764544++++74.3%25
+++++++++++relu+++#####++++++++4096
++++++++++Dense+++XXXXX+-------------------++16781312++++12.1%25
+++++++++++relu+++#####++++++++4096
++++++++++Dense+++XXXXX+-------------------+++4097000+++++3.0%25
++++++++softmax+++#####++++++++1000|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/stared/keras-sequential-ascii|1|https://blog.deepsense.ai/deep-learning-hands-on-image-classification/^0|4|6|0|0|C|3|1|0|0|0^^$0|@$1|2|3|4|5|6|7|X|8|@]|9|@$A|Y|B|Z|1|10]]|C|$]]|$1|D|3|E|5|6|7|11|8|@]|9|@$A|12|B|13|1|14]]|C|$]]|$1|F|3|G|5|H|7|15|8|@]|9|@]|C|$I|J]]|$1|K|3|L|5|6|7|16|8|@]|9|@]|C|$]]|$1|M|3|N|5|H|7|17|8|@]|9|@]|C|$I|J]]]|O|$P|$5|Q|R|S|C|$T|U]]|V|$5|Q|R|S|C|$T|W]]]]

I would add ASCII visualizations using <a href="https://github.com/stared/keras-sequential-ascii" rel="noreferrer">keras-sequential-ascii</a> (disclaimer: I am the author).

A small network for CIFAR-10 (from <a href="https://blog.deepsense.ai/deep-learning-hands-on-image-classification/" rel="noreferrer">this tutorial</a>) would be:

<pre><code> OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)

 Input ##### 32 32 3
 Conv2D \|/ ------------------- 896 2.1%
 relu ##### 30 30 32
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 15 15 32
 Conv2D \|/ ------------------- 18496 43.6%
 relu ##### 13 13 64
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 6 6 64
 Flatten ||||| ------------------- 0 0.0%
 ##### 2304
 Dense XXXXX ------------------- 23050 54.3%
 softmax ##### 10
</code></pre>

For VGG16 it would be:

<pre><code> OPERATION DATA DIMENSIONS WEIGHTS(N) WEIGHTS(%)

 Input ##### 3 224 224
 InputLayer | ------------------- 0 0.0%
 ##### 3 224 224
 Convolution2D \|/ ------------------- 1792 0.0%
 relu ##### 64 224 224
 Convolution2D \|/ ------------------- 36928 0.0%
 relu ##### 64 224 224
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 64 112 112
 Convolution2D \|/ ------------------- 73856 0.1%
 relu ##### 128 112 112
 Convolution2D \|/ ------------------- 147584 0.1%
 relu ##### 128 112 112
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 128 56 56
 Convolution2D \|/ ------------------- 295168 0.2%
 relu ##### 256 56 56
 Convolution2D \|/ ------------------- 590080 0.4%
 relu ##### 256 56 56
 Convolution2D \|/ ------------------- 590080 0.4%
 relu ##### 256 56 56
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 256 28 28
 Convolution2D \|/ ------------------- 1180160 0.9%
 relu ##### 512 28 28
 Convolution2D \|/ ------------------- 2359808 1.7%
 relu ##### 512 28 28
 Convolution2D \|/ ------------------- 2359808 1.7%
 relu ##### 512 28 28
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 512 14 14
 Convolution2D \|/ ------------------- 2359808 1.7%
 relu ##### 512 14 14
 Convolution2D \|/ ------------------- 2359808 1.7%
 relu ##### 512 14 14
 Convolution2D \|/ ------------------- 2359808 1.7%
 relu ##### 512 14 14
 MaxPooling2D Y max ------------------- 0 0.0%
 ##### 512 7 7
 Flatten ||||| ------------------- 0 0.0%
 ##### 25088
 Dense XXXXX ------------------- 102764544 74.3%
 relu ##### 4096
 Dense XXXXX ------------------- 16781312 12.1%
 relu ##### 4096
 Dense XXXXX ------------------- 4097000 3.0%
 softmax ##### 1000
</code></pre>

blocks|key|1224903|text|Keras|type|header-two|depth|inlineStyleRanges|entityRanges|data|1224904|keras.utils.vis_utils模块提供实用函数来绘制Keras模型(使用graphviz)|unstyled|offset|length|1224905|下面是一个网络模型，第一个隐层有50个神经元，期望104个输入变量。|1224906|plot_model(model,+to_file='model.png',+show_shapes=True,+show_layer_names=True)|code-block|syntax|javascript|1224907|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|https://keras.io/visualization/|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/o17GY.png|imageAlt^0|0|0|N|0|0|0|0|0|1|1^^$0|@$1|2|3|4|5|6|7|13|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|14|8|@]|9|@$E|15|F|16|1|17]]|A|$]]|$1|G|3|H|5|D|7|18|8|@]|9|@]|A|$]]|$1|I|3|J|5|K|7|19|8|@]|9|@]|A|$L|M]]|$1|N|3|O|5|P|7|1A|8|@]|9|@$E|1B|F|1C|1|1D]]|A|$]]]|Q|$R|$5|S|T|U|A|$V|W]]|X|$5|Y|T|Z|A|$10|11|12|-4]]]]

Keras

The <a href="https://keras.io/visualization/" rel="noreferrer">keras.utils.vis_utils module</a> provides utility functions to plot a Keras model (using graphviz)

The following shows a network model that the first hidden layer has 50 neurons and expects 104 input variables.

<pre><code>plot_model(model, to_file='model.png', show_shapes=True, show_layer_names=True)
</code></pre>

<a href="https://i.stack.imgur.com/o17GY.png" rel="noreferrer"><img src="https://i.stack.imgur.com/o17GY.png" alt="enter image description here"></a>

blocks|key|2367769|text|这里还有另一种方式--+网网，它使用墨维兹，深受蒂亚戈·马丁斯(ThiagoG.Martins)+这个职位的启发。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2367770|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/martisak/dotnets|1|http://www.graphviz.org/|2|https://tgmstat.wordpress.com/2013/06/12/draw-neural-network-diagrams-graphviz/|3|IMAGE|IMMUTABLE|imageUrl|https://github.com/martisak/dotnets/raw/master/test.png|imageAlt^0|C|2|0|I|3|1|1D|4|2|0|0|1|3^^$0|@$1|2|3|4|5|6|7|X|8|@]|9|@$A|Y|B|Z|1|10]|$A|11|B|12|1|13]|$A|14|B|15|1|16]]|C|$]]|$1|D|3|E|5|F|7|17|8|@]|9|@$A|18|B|19|1|1A]]|C|$]]]|G|$H|$5|I|J|K|C|$L|M]]|N|$5|I|J|K|C|$L|O]]|P|$5|I|J|K|C|$L|Q]]|R|$5|S|J|T|C|$U|V|W|-4]]]]

Here is yet another way - <a href="https://github.com/martisak/dotnets" rel="noreferrer">dotnets</a>, using <a href="http://www.graphviz.org/" rel="noreferrer">Graphviz</a>, heavily inspired by <a href="https://tgmstat.wordpress.com/2013/06/12/draw-neural-network-diagrams-graphviz/" rel="noreferrer">this post</a> by Thiago G. Martins.

<img src="https://github.com/martisak/dotnets/raw/master/test.png" alt="dotnets example">

blocks|key|2368469|text|我一直在研究一种拖放式神经网络可视化器(还有更多)。下面是一个类似于LeNet的体系结构的可视化示例。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|2368470|📷|atomic|offset|length|2368471|带有扇出和扇入的模型也很容易建模。您可以访问https://math.mit.edu/ennui/的网站|2368472|开源实现可在https://github.com/martinjm97/ENNUI上使用.|entityMap|0|IMAGE|mutability|IMMUTABLE|imageUrl|https://i.stack.imgur.com/pRLeG.png|imageAlt|1|LINK|MUTABLE|url|https://math.mit.edu/ennui/|2|https://github.com/martinjm97/ENNUI^0|0|0|1|0|0|M|R|1|0|6|Z|2^^$0|@$1|2|3|4|5|6|7|Z|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|10|8|@]|9|@$E|11|F|12|1|13]]|A|$]]|$1|G|3|H|5|6|7|14|8|@]|9|@$E|15|F|16|1|17]]|A|$]]|$1|I|3|J|5|6|7|18|8|@]|9|@$E|19|F|1A|1|1B]]|A|$]]]|K|$L|$5|M|N|O|A|$P|Q|R|-4]]|S|$5|T|N|U|A|$V|W]]|X|$5|T|N|U|A|$V|Y]]]]

I've been working on a drag-and-drop neural network visualizer (and more). Here's an example of a visualization for a LeNet-like architecture. <a href="https://i.stack.imgur.com/pRLeG.png" rel="noreferrer"><img src="https://i.stack.imgur.com/pRLeG.png" alt="A visualization of a LeNet-like architecture"></a>
Models with fan-out and fan-in are also quite easily modeled. You can visit the website at <a href="https://math.mit.edu/ennui/" rel="noreferrer">https://math.mit.edu/ennui/</a>

The open-source implementation is available at <a href="https://github.com/martinjm97/ENNUI" rel="noreferrer">https://github.com/martinjm97/ENNUI</a>.

blocks|key|1225111|text|package+conx可以通过激活net.picture()函数来可视化网络，从而生成SVG、PNG或PIL映像，如下所示：|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|1225112|📷|atomic|1225113|Conx建立在Keras上，可以在Keras的模型中读取。每个银行的颜色图可以改变，它可以显示所有的银行类型。|1225114|更多信息可在以下网站找到：http://conx.readthedocs.io/en/latest/|entityMap|0|IMAGE|mutability|IMMUTABLE|imageUrl|https://i.stack.imgur.com/nhHjO.png|imageAlt|1|LINK|MUTABLE|url|http://conx.readthedocs.io/en/latest/^0|8|4|I|D|0|0|1|0|0|0|D|11|1^^$0|@$1|2|3|4|5|6|7|Z|8|@$9|10|A|11|B|C]|$9|12|A|13|B|C]]|D|@]|E|$]]|$1|F|3|G|5|H|7|14|8|@]|D|@$9|15|A|16|1|17]]|E|$]]|$1|I|3|J|5|6|7|18|8|@]|D|@]|E|$]]|$1|K|3|L|5|6|7|19|8|@]|D|@$9|1A|A|1B|1|1C]]|E|$]]]|M|$N|$5|O|P|Q|E|$R|S|T|-4]]|U|$5|V|P|W|E|$X|Y]]]]

The Python package <code>conx</code> can visualize networks with activations with the function <code>net.picture()</code> to produce SVG, PNG, or PIL Images like this:

<a href="https://i.stack.imgur.com/nhHjO.png" rel="noreferrer"><img src="https://i.stack.imgur.com/nhHjO.png" alt="enter image description here"></a>

Conx is built on Keras, and can read in Keras' models. The colormap at each bank can be changed, and it can show all bank types.

More information can be found at: <a href="http://conx.readthedocs.io/en/latest/" rel="noreferrer">http://conx.readthedocs.io/en/latest/</a>

blocks|key|1225911|text|PlotNeuralNet+LaTex+tool|type|header-two|depth|inlineStyleRanges|entityRanges|data|1225912|这个解决方案不是自动生成的(您需要自己构建图形)，但是PlotNeuralNet+github回购允许您直接从LaTex构建映像，而且效果很好！例如，请参见自述文件中的以下图像：|unstyled|offset|length|1225913|📷|atomic|1225914|或者我的例子：|1225915|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/HarisIqbal88/PlotNeuralNet|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/U7JAf.png|imageAlt|2|https://i.stack.imgur.com/9pCUZ.png^0|0|R|M|0|0|0|1|1|0|0|0|1|2^^$0|@$1|2|3|4|5|6|7|11|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|12|8|@]|9|@$E|13|F|14|1|15]]|A|$]]|$1|G|3|H|5|I|7|16|8|@]|9|@$E|17|F|18|1|19]]|A|$]]|$1|J|3|K|5|D|7|1A|8|@]|9|@]|A|$]]|$1|L|3|H|5|I|7|1B|8|@]|9|@$E|1C|F|1D|1|1E]]|A|$]]]|M|$N|$5|O|P|Q|A|$R|S]]|T|$5|U|P|V|A|$W|X|Y|-4]]|Z|$5|U|P|V|A|$W|10|Y|-4]]]]

<h1>PlotNeuralNet LaTex tool</h1>
This solution is not automatically generated (you need to construct the graph by yourself) but <a href="https://github.com/HarisIqbal88/PlotNeuralNet" rel="noreferrer">the PlotNeuralNet github repo</a> allows you to build images directly from LaTex, and the result is great ! See for example the image below from the README :
<a href="https://i.stack.imgur.com/U7JAf.png" rel="noreferrer"><img src="https://i.stack.imgur.com/U7JAf.png" alt="Example of NN draw with " /></a>
or my example :
<a href="https://i.stack.imgur.com/9pCUZ.png" rel="noreferrer"><img src="https://i.stack.imgur.com/9pCUZ.png" alt="My own example, without the operations legend (sorry)" /></a>

blocks|key|1224431|text|在R中，nnet不附带绘图函数，而是提供了这里代码。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|1224432|或者，您也可以使用名为neuralnet的最近的和IMHO更好的包，它具有一个plot.neuralnet函数，因此您可以这样做：|1224433|data(infert,+package="datasets")
plot(neuralnet(case~parity%2Binduced%2Bspontaneous,+infert))|code-block|syntax|javascript|1224434|📷|atomic|1224435|neuralnet没有nnet那么多的使用，因为nnet更老，并且随r-cran一起提供。但是neuralnet有更多的训练算法，包括弹性反向传播，即使在Tensorflow这样的软件包中也缺乏，并且对超参数选择更加健壮，而且总体上有更多的特性。|entityMap|0|LINK|mutability|MUTABLE|url|https://beckmw.wordpress.com/2013/03/04/visualizing-neural-networks-from-the-nnet-package/|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/yyftd.png|imageAlt|2|https://en.wikipedia.org/wiki/Rprop^0|4|4|L|2|0|0|B|9|13|E|0|0|0|1|1|0|0|9|B|4|O|4|1B|9|1V|6|2^^$0|@$1|2|3|4|5|6|7|16|8|@$9|17|A|18|B|C]]|D|@$9|19|A|1A|1|1B]]|E|$]]|$1|F|3|G|5|6|7|1C|8|@$9|1D|A|1E|B|C]|$9|1F|A|1G|B|C]]|D|@]|E|$]]|$1|H|3|I|5|J|7|1H|8|@]|D|@]|E|$K|L]]|$1|M|3|N|5|O|7|1I|8|@]|D|@$9|1J|A|1K|1|1L]]|E|$]]|$1|P|3|Q|5|6|7|1M|8|@$9|1N|A|1O|B|C]|$9|1P|A|1Q|B|C]|$9|1R|A|1S|B|C]|$9|1T|A|1U|B|C]]|D|@$9|1V|A|1W|1|1X]]|E|$]]]|R|$S|$5|T|U|V|E|$W|X]]|Y|$5|Z|U|10|E|$11|12|13|-4]]|14|$5|T|U|V|E|$W|15]]]]

In R, <code>nnet</code> does not come with a plot function, but code for that is provided <a href="https://beckmw.wordpress.com/2013/03/04/visualizing-neural-networks-from-the-nnet-package/" rel="noreferrer">here</a>.

Alternatively, you can use the more recent and IMHO better package called <code>neuralnet</code> which features a <code>plot.neuralnet</code> function, so you can just do:

<pre><code>data(infert, package="datasets")
plot(neuralnet(case~parity+induced+spontaneous, infert))
</code></pre>

<a href="https://i.stack.imgur.com/yyftd.png" rel="noreferrer"><img src="https://i.stack.imgur.com/yyftd.png" alt="neuralnet"></a>

<code>neuralnet</code> is not used as much as <code>nnet</code> because <code>nnet</code> is much older and is shipped with r-cran. But <code>neuralnet</code> has more training algorithms, including <a href="https://en.wikipedia.org/wiki/Rprop" rel="noreferrer">resilient backpropagation</a> which is lacking even in packages like Tensorflow, and is much more robust to hyperparameter choices, and has more features overall.

blocks|key|1224771|text|在神经网络可视化方面有一些新的替代方法。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|1224772|请参阅以下文章：|1224773|惊人的‘人工智能大脑扫描’揭示了机器在学习新技能时所看到的东西|offset|length|1224774|在人工智能的“大脑”里-机器学习是什么样子的？|1224775|这些方法更倾向于可视化神经网络操作，然而，神经网络体系结构在结果图上也是可见的。|1224776|示例：|1224777|📷|atomic|1224778|1224779|1224780|entityMap|0|LINK|mutability|MUTABLE|url|http://www.wired.co.uk/article/ai-machine-learning-brain-scan|1|https://www.graphcore.ai/posts/what-does-machine-learning-look-like|2|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/qd5rF.jpg|imageAlt|3|https://i.stack.imgur.com/q80fj.jpg|4|https://i.stack.imgur.com/LQQuG.jpg|5|https://i.stack.imgur.com/t1xZk.jpg^0|0|0|0|V|0|0|0|N|1|0|0|0|0|1|2|0|0|1|3|0|0|1|4|0|0|1|5^^$0|@$1|2|3|4|5|6|7|1E|8|@]|9|@]|A|$]]|$1|B|3|C|5|6|7|1F|8|@]|9|@]|A|$]]|$1|D|3|E|5|6|7|1G|8|@]|9|@$F|1H|G|1I|1|1J]]|A|$]]|$1|H|3|I|5|6|7|1K|8|@]|9|@$F|1L|G|1M|1|1N]]|A|$]]|$1|J|3|K|5|6|7|1O|8|@]|9|@]|A|$]]|$1|L|3|M|5|6|7|1P|8|@]|9|@]|A|$]]|$1|N|3|O|5|P|7|1Q|8|@]|9|@$F|1R|G|1S|1|1T]]|A|$]]|$1|Q|3|O|5|P|7|1U|8|@]|9|@$F|1V|G|1W|1|1X]]|A|$]]|$1|R|3|O|5|P|7|1Y|8|@]|9|@$F|1Z|G|20|1|21]]|A|$]]|$1|S|3|O|5|P|7|22|8|@]|9|@$F|23|G|24|1|25]]|A|$]]]|T|$U|$5|V|W|X|A|$Y|Z]]|10|$5|V|W|X|A|$Y|11]]|12|$5|13|W|14|A|$15|16|17|-4]]|18|$5|13|W|14|A|$15|19|17|-4]]|1A|$5|13|W|14|A|$15|1B|17|-4]]|1C|$5|13|W|14|A|$15|1D|17|-4]]]]

There are some novel alternative efforts on neural network visualization.

Please see these articles:

<a href="http://www.wired.co.uk/article/ai-machine-learning-brain-scan" rel="nofollow noreferrer">Stunning 'AI brain scans' reveal what machines see as they learn new skills</a>

<a href="https://www.graphcore.ai/posts/what-does-machine-learning-look-like" rel="nofollow noreferrer">Inside an AI 'brain' - What does machine learning look like?</a>

These approaches are more oriented towards visualizing neural network operation, however, NN architecture is also somewhat visible on the resulting diagrams.

Examples:

<a href="https://i.stack.imgur.com/qd5rF.jpg" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/qd5rF.jpg" alt="enter image description here"></a> 

<a href="https://i.stack.imgur.com/q80fj.jpg" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/q80fj.jpg" alt="enter image description here"></a>

<a href="https://i.stack.imgur.com/LQQuG.jpg" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/LQQuG.jpg" alt="enter image description here"></a>

<a href="https://i.stack.imgur.com/t1xZk.jpg" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/t1xZk.jpg" alt="enter image description here"></a>

blocks|key|1224927|text|您可以阅读流行的论文基于深度可视化的神经网络理解，其中讨论了卷积网的可视化。其实施情况不仅显示每一层，而且还描述了本文中深入讨论的激活、权值、解卷积和许多其他内容。它的代码在caffe'中。有趣的是，你可以用你自己的训练过的模型来代替它。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|entityMap|0|LINK|mutability|MUTABLE|url|http://yosinski.com/media/papers/Yosinski__2015__ICML_DL__Understanding_Neural_Networks_Through_Deep_Visualization__.pdf|1|https://github.com/yosinski/deep-visualization-toolbox^0|2F|6|A|E|0|12|5|1^^$0|@$1|2|3|4|5|6|7|O|8|@$9|P|A|Q|B|C]]|D|@$9|R|A|S|1|T]|$9|U|A|V|1|W]]|E|$]]]|F|$G|$5|H|I|J|E|$K|L]]|M|$5|H|I|J|E|$K|N]]]]

You can read the popular paper <a href="http://yosinski.com/media/papers/Yosinski__2015__ICML_DL__Understanding_Neural_Networks_Through_Deep_Visualization__.pdf" rel="noreferrer">Understanding Neural Networks Through Deep Visualization</a> which discusses visualization of convolutional nets. <a href="https://github.com/yosinski/deep-visualization-toolbox" rel="noreferrer">Its implementation</a> not only displays each layer but also depicts the activations, weights, deconvolutions and many other things that are deeply discussed in the paper. It's code is in <code>caffe'</code>. The interesting part is that you can replace the pre-trained model with your own.

blocks|key|2368340|text|Tensorspace-JS是一个非常棒的网络体系结构三维可视化工具：|type|unstyled|depth|inlineStyleRanges|entityRanges|data|2368341|📷|atomic|offset|length|2368342|https://tensorspace.org/|2368343|这里有一篇关于如何编写程序的很好的文章：|2368344|https://medium.freecodecamp.org/tensorspace-js-a-way-to-3d-visualize-neural-networks-in-browsers-2c0afd7648a8|entityMap|0|IMAGE|mutability|IMMUTABLE|imageUrl|https://i.stack.imgur.com/ekF5v.png|imageAlt|1|LINK|MUTABLE|url|2^0|0|0|1|0|0|0|O|1|0|0|0|31|2^^$0|@$1|2|3|4|5|6|7|Z|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|10|8|@]|9|@$E|11|F|12|1|13]]|A|$]]|$1|G|3|H|5|6|7|14|8|@]|9|@$E|15|F|16|1|17]]|A|$]]|$1|I|3|J|5|6|7|18|8|@]|9|@]|A|$]]|$1|K|3|L|5|6|7|19|8|@]|9|@$E|1A|F|1B|1|1C]]|A|$]]]|M|$N|$5|O|P|Q|A|$R|S|T|-4]]|U|$5|V|P|W|A|$X|H]]|Y|$5|V|P|W|A|$X|L]]]]

Tensorspace-JS is a fantastic tool for 3d visualization of network architecture:

<a href="https://i.stack.imgur.com/ekF5v.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/ekF5v.png" alt="enter image description here"></a>

<a href="https://tensorspace.org/" rel="nofollow noreferrer">https://tensorspace.org/</a>

and here is a nice post about how to write a program:

<a href="https://medium.freecodecamp.org/tensorspace-js-a-way-to-3d-visualize-neural-networks-in-browsers-2c0afd7648a8" rel="nofollow noreferrer">https://medium.freecodecamp.org/tensorspace-js-a-way-to-3d-visualize-neural-networks-in-browsers-2c0afd7648a8</a>

blocks|key|1225761|text|Tensorflow+/+Keras+/+Python|type|header-two|depth|inlineStyleRanges|entityRanges|data|1225762|我编写了一个名为visualkeras的小型python包，它允许您从keras模型直接生成体系结构。|unstyled|1225763|通过pip+install+visualkeras安装|offset|length|style|CODE|1225764|然后简单的说：|1225765|import+visualkeras
visualkeras.layered_view(<model>)|code-block|syntax|javascript|1225766|有很多选择来调整它，我正在做更多的可视化工作。此外，始终打开PRs或特性请求。|1225767|下面是VGG16的样子：|1225768|📷|atomic|entityMap|0|IMAGE|mutability|IMMUTABLE|imageUrl|https://i.stack.imgur.com/sITF5.png|imageAlt^0|0|0|2|N|0|0|0|0|0|0|1|0^^$0|@$1|2|3|4|5|6|7|16|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|17|8|@]|9|@]|A|$]]|$1|E|3|F|5|D|7|18|8|@$G|19|H|1A|I|J]]|9|@]|A|$]]|$1|K|3|L|5|D|7|1B|8|@]|9|@]|A|$]]|$1|M|3|N|5|O|7|1C|8|@]|9|@]|A|$P|Q]]|$1|R|3|S|5|D|7|1D|8|@]|9|@]|A|$]]|$1|T|3|U|5|D|7|1E|8|@]|9|@]|A|$]]|$1|V|3|W|5|X|7|1F|8|@]|9|@$G|1G|H|1H|1|1I]]|A|$]]]|Y|$Z|$5|10|11|12|A|$13|14|15|-4]]]]

<h1>Tensorflow / Keras / Python</h1>
I wrote a small python package called visualkeras that allows you to directly generate the architecture from your keras model.
Install via <code>pip install visualkeras</code>
And then it's as simple as:
<pre><code>import visualkeras
visualkeras.layered_view(&lt;model&gt;)
</code></pre>
There are lots of options to tweak it and I am working on more visualizations. Also, always open for PRs or feature requests.
Here's what VGG16 looks like:
<a href="https://i.stack.imgur.com/sITF5.png" rel="noreferrer"><img src="https://i.stack.imgur.com/sITF5.png" alt="VGG16 architecture" /></a>

blocks|key|2367489|text|从本质上讲，论文不是很好，但对于向不太了解神经网络的人展示他们的拓扑结构是非常有用的。这个库(Neataptic)允许你可视化你的网络：|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2367490|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/wagenaartje/neataptic|1|IMAGE|IMMUTABLE|imageUrl|https://camo.githubusercontent.com/6e033f6c37b57562bbab24f6579bc0fdcbd30544/68747470733a2f2f692e6779617a6f2e636f6d2f35333235636139323137646263613331353161383931373339353438613031642e706e67|imageAlt^0|19|C|0|0|0|1|1^^$0|@$1|2|3|4|5|6|7|T|8|@]|9|@$A|U|B|V|1|W]]|C|$]]|$1|D|3|E|5|F|7|X|8|@]|9|@$A|Y|B|Z|1|10]]|C|$]]]|G|$H|$5|I|J|K|C|$L|M]]|N|$5|O|J|P|C|$Q|R|S|-4]]]]

Not per se nifty for papers, but very useful for showing people who don't know a lot of about neural networks what their topology may look like. This <a href="https://github.com/wagenaartje/neataptic" rel="nofollow noreferrer">Javascript library (Neataptic)</a> lets you visualise your network:

<a href="https://camo.githubusercontent.com/6e033f6c37b57562bbab24f6579bc0fdcbd30544/68747470733a2f2f692e6779617a6f2e636f6d2f35333235636139323137646263613331353161383931373339353438613031642e706e67" rel="nofollow noreferrer"><img src="https://camo.githubusercontent.com/6e033f6c37b57562bbab24f6579bc0fdcbd30544/68747470733a2f2f692e6779617a6f2e636f6d2f35333235636139323137646263613331353161383931373339353438613031642e706e67" alt="enter image description here"></a>

blocks|key|1225529|text|网镜是我日常使用的Caffe模型工具。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|1225530|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|http://dgschwend.github.io/netscope/quickstart.html|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/VVDsg.png|imageAlt^0|0|2|0|0|0|1|1^^$0|@$1|2|3|4|5|6|7|T|8|@]|9|@$A|U|B|V|1|W]]|C|$]]|$1|D|3|E|5|F|7|X|8|@]|9|@$A|Y|B|Z|1|10]]|C|$]]]|G|$H|$5|I|J|K|C|$L|M]]|N|$5|O|J|P|C|$Q|R|S|-4]]]]

<a href="http://dgschwend.github.io/netscope/quickstart.html" rel="nofollow noreferrer">Netscope</a> is my everyday tool for Caffe models.

<a href="https://i.stack.imgur.com/VVDsg.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/VVDsg.png" alt="enter image description here"></a>

blocks|key|2368745|text|我找到了一个很棒的网站。您只需上传您的h5模型，然后您将得到一个美丽的可视化在几秒钟内。看看这个！|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2368746|📷|atomic|entityMap|0|LINK|mutability|MUTABLE|url|https://netron.app/|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/SRHPm.png|imageAlt^0|18|5|0|0|0|1|1^^$0|@$1|2|3|4|5|6|7|T|8|@]|9|@$A|U|B|V|1|W]]|C|$]]|$1|D|3|E|5|F|7|X|8|@]|9|@$A|Y|B|Z|1|10]]|C|$]]]|G|$H|$5|I|J|K|C|$L|M]]|N|$5|O|J|P|C|$Q|R|S|-4]]]]

I have found one amazing website. You just need to upload your h5 model, Then you will get a beautiful visualization within a few seconds. <a href="https://netron.app/" rel="nofollow noreferrer">Check it out!</a>
<a href="https://i.stack.imgur.com/SRHPm.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/SRHPm.png" alt="enter image description here" /></a>

blocks|key|2368576|text|您可以使用eiffel2，您可以使用pip安装它：|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|2368577|python+-m+pip+install+eiffel2|code-block|syntax|javascript|2368578|只需从builder导入eiffel，并提供网络中每层神经元的列表作为输入。|2368579|示例：|2368580|from+eiffel2+import+builder

builder([1,+10,+10,+5,+5,+2,+1])
#+or+the+following+if+you+want+to+have+a+dark+theme
builder([1,+10,+10,+5,+5,+2,+1],+bmode="night")|2368581|输出：|2368582|📷|atomic|2368583|2368584|要了解更多关于eiffel2的信息，请访问Github存储库：|2368585|https://github.com/Ale9806/Eiffel2/blob/master/README.md|entityMap|0|IMAGE|mutability|IMMUTABLE|imageUrl|https://i.stack.imgur.com/5v9k0.png|imageAlt|1|https://i.stack.imgur.com/GbRqP.png|2|LINK|MUTABLE|url^0|5|7|I|3|0|0|3|7|C|6|0|0|0|0|0|1|0|0|0|1|1|0|0|0|1K|2^^$0|@$1|2|3|4|5|6|7|1E|8|@$9|1F|A|1G|B|C]|$9|1H|A|1I|B|C]]|D|@]|E|$]]|$1|F|3|G|5|H|7|1J|8|@]|D|@]|E|$I|J]]|$1|K|3|L|5|6|7|1K|8|@$9|1L|A|1M|B|C]|$9|1N|A|1O|B|C]]|D|@]|E|$]]|$1|M|3|N|5|6|7|1P|8|@]|D|@]|E|$]]|$1|O|3|P|5|H|7|1Q|8|@]|D|@]|E|$I|J]]|$1|Q|3|R|5|6|7|1R|8|@]|D|@]|E|$]]|$1|S|3|T|5|U|7|1S|8|@]|D|@$9|1T|A|1U|1|1V]]|E|$]]|$1|V|3|T|5|U|7|1W|8|@]|D|@$9|1X|A|1Y|1|1Z]]|E|$]]|$1|W|3|X|5|6|7|20|8|@]|D|@]|E|$]]|$1|Y|3|Z|5|6|7|21|8|@]|D|@$9|22|A|23|1|24]]|E|$]]]|10|$11|$5|12|13|14|E|$15|16|17|-4]]|18|$5|12|13|14|E|$15|19|17|-4]]|1A|$5|1B|13|1C|E|$1D|Z]]]]

You can use <code>eiffel2</code>, which you can install using <code>pip</code>:

<pre><code>python -m pip install eiffel2
</code></pre>

Just import <code>builder</code> from <code>eiffel</code> and provide a list of neurons per layer in your network as an input.

Example:

<pre class="lang-py prettyprint-override"><code>from eiffel2 import builder

builder([1, 10, 10, 5, 5, 2, 1])
# or the following if you want to have a dark theme
builder([1, 10, 10, 5, 5, 2, 1], bmode="night")
</code></pre>

Output:

<a href="https://i.stack.imgur.com/5v9k0.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/5v9k0.png" alt="Normal output"></a>

<a href="https://i.stack.imgur.com/GbRqP.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/GbRqP.png" alt="output with bmode=&quot;night&quot;"></a>

To see more about eiffel2 visit the Github repository: 

<a href="https://github.com/Ale9806/Eiffel2/blob/master/README.md" rel="nofollow noreferrer">https://github.com/Ale9806/Eiffel2/blob/master/README.md</a>

blocks|key|2368809|text|对于PyTorch的解决方案，我要添加TorchView。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|2368810|它很简单，就像：|2368811|from+torchview+import+draw_graph

model+=+MLP()
batch_size+=+2
#+device='meta'+->+no+memory+is+consumed+for+visualization
model_graph+=+draw_graph(model,+input_size=(batch_size,+128),+device='meta')
model_graph.visual_graph|code-block|syntax|javascript|2368812|产生的结果：|2368813|📷|atomic|2368814|它也有许多定制选项。|entityMap|0|LINK|mutability|MUTABLE|url|https://github.com/mert-kurttutan/torchview|1|IMAGE|IMMUTABLE|imageUrl|https://i.stack.imgur.com/L0PxJ.png|imageAlt^0|J|9|0|0|0|0|0|0|1|1|0^^$0|@$1|2|3|4|5|6|7|14|8|@]|9|@$A|15|B|16|1|17]]|C|$]]|$1|D|3|E|5|6|7|18|8|@]|9|@]|C|$]]|$1|F|3|G|5|H|7|19|8|@]|9|@]|C|$I|J]]|$1|K|3|L|5|6|7|1A|8|@]|9|@]|C|$]]|$1|M|3|N|5|O|7|1B|8|@]|9|@$A|1C|B|1D|1|1E]]|C|$]]|$1|P|3|Q|5|6|7|1F|8|@]|9|@]|C|$]]]|R|$S|$5|T|U|V|C|$W|X]]|Y|$5|Z|U|10|C|$11|12|13|-4]]]]

For a solution for PyTorch I'd add <a href="https://github.com/mert-kurttutan/torchview" rel="nofollow noreferrer">TorchView</a>.
It is as easy as:
<pre class="lang-py prettyprint-override"><code>from torchview import draw_graph

model = MLP()
batch_size = 2
# device='meta' -&gt; no memory is consumed for visualization
model_graph = draw_graph(model, input_size=(batch_size, 128), device='meta')
model_graph.visual_graph
</code></pre>
Which yields:
<a href="https://i.stack.imgur.com/L0PxJ.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/L0PxJ.png" alt="enter image description here" /></a>
It has many customization options as well.

When writing a paper / making a presentation about a topic which is about neural networks, one usually visualizes the networks architecture.

What are good / simple ways to visualize common architectures automatically?

How do you visualize neural network architectures?

翻译质量差，导致语言生硬或混乱。

没有提供实际的解决方法或示例。

解答不清晰，无法理解或解决问题。

页面排版不美观，阅读体验差。

文章

问答

视频

教程

学习中心

腾讯云实验室

直播

竞赛

腾讯云代码分析专区

腾讯iOA零信任安全管理系统专区

腾讯云架构师技术同盟交流圈

腾讯云数据库专区

腾讯云智能顾问专区

腾讯云原生专区

腾讯混元专区

腾讯云TCE专区

腾讯云Lighthouse专区

腾讯云HAI专区

腾讯云Edgeone专区

腾讯云存储专区

腾讯云智能专区

腾讯轻联专区 

腾讯云开发专区

TAPD专区

腾讯轻量云游戏服专区

腾讯云最具价值专家

腾讯云架构师技术同盟

腾讯云创作之星

腾讯云开发者先锋

腾讯云AI代码助手

云原生构建

TAPD 敏捷项目管理

Cloud Studio

SDK中心

API中心

命令行工具

功能1上新10个字符

功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符功能2描述100个字符。

功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符功能2上新100个字符。

功能5描述100个字符功能5描述100个字符功能5描述100个字符功能5描述100个字符功能5描述100个字符功能5描述100个字符

功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符功能5上新100个字符

功能4上新

文章&问答评论现已支持表情

全新交互，全新视觉，新增快捷键、悬浮工具栏、高亮块等功能并同时优化现有功能，全面提升创作效率和体验

社区富文本编辑器全新改版！诚邀体验～ 

精选全网热门MCP server，让你的AI更好用 🚀

💥开发者 MCP广场重磅上线！

涵盖代码开发、场景应用、自动测试全流程，助你从零构建专属AI助手

一站式MCP教程库，解锁AI应用新玩法

聚焦“写作效率、视觉美观与运行性能”三方面进行全面升级，为您提供更高效、稳定的创作环境

社区富文本&Markdown编辑器全新改版上线，欢迎大家体验!

诚挚邀请您参与本次调研，分享您的真实使用感受与建议。您的反馈至关重要，感谢您的支持与参与！

社区新版编辑器体验调研

当你写一篇论文/做一个关于神经网络的专题报告时，人们通常会看到网络体系结构。有什么好的/简单的方法可以自动地可视化公共架构？

问如何可视化神经网络体系结构？
EN

回答 20

Data Science用户

Data Science用户

Tensorflow，Keras，MXNet，PyTorch

Interpretation

Data Science用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问如何可视化神经网络体系结构？EN

回答 20

Data Science用户

Data Science用户

Tensorflow，Keras，MXNet，PyTorch

Interpretation

Data Science用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问如何可视化神经网络体系结构？
EN