Apache Mxnet 简明教程

pip install mxnet

$ sudo docker pull mxnet/python:gpu

$ sudo docker images

$ sudo docker pull mxnet/python:1.3.0_cpu_mkl
$ sudo docker images

tar xvzf cudnn-9.2-linux-x64-v7.1
sudo cp -P cuda/include/cudnn.h /usr/local/cuda/include
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
sudo ldconfig

sudo apt-get update
sudo apt-get install -y build-essential git ninja-build ccache libopenblas-dev
libopencv-dev cmake

git clone –recursive https://github.com/apache/incubator-mxnet.git mxnet
cd mxnet
cp config/linux_gpu.cmake #for build with CUDA

rm -rf build
mkdir -p build && cd build
cmake -GNinja ..
cmake --build .

cmake -DCMAKE_BUILD_TYPE=Debug -GNinja ..

cmake --build . --parallel N

pip install mxnet

pip install mxnet-mkl

$ sudo docker pull mxnet/python

$ sudo docker images

$ sudo docker pull mxnet/python:1.3.0_cpu_mkl
$ sudo docker images

sudo apt-get update

sudo apt-get install -y build-essential git ninja-build ccache libopenblas-dev libopencv-dev cmake

git clone –recursive https://github.com/apache/incubator-mxnet.git mxnet

cd mxnet
cp config/linux.cmake config.cmake

rm -rf build
mkdir -p build && cd build
cmake -GNinja ..
cmake --build .

cmake -DCMAKE_BUILD_TYPE=Debug -GNinja ..

cmake --build . --parallel N

xcode-select –-install #Install OS X Developer Tools

/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)" #Install Homebrew

brew install cmake ninja ccache opencv # Install dependencies

git clone –-recursive https://github.com/apache/incubator-mxnet.git mxnet

cd mxnet
cp config/linux.cmake config.cmake

rm -rf build
mkdir -p build && cd build
cmake -GNinja ..
cmake --build .

cmake -DCMAKE_BUILD_TYPE=Debug -GNinja ..

cmake --build . --parallel N

pip install mxnet

$ docker pull mxnet/python

$ docker images

$ docker pull mxnet/python:1.3.0_cpu_mkl
$ docker images

xcode-select –-install #Install OS X Developer Tools
/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)" #Install Homebrew
brew install cmake ninja ccache opencv # Install dependencies

git clone –-recursive https://github.com/apache/incubator-mxnet.git mxnet

cd mxnet

cp config/linux.cmake config.cmake

rm -rf build
mkdir -p build && cd build
cmake -GNinja ..
cmake --build .

cmake -DCMAKE_BUILD_TYPE=Debug -GNinja ..

cmake --build . --parallel N

pip install mxnet-cu92

"C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Auxiliary\Build\vcvars64.bat" -vcvars_ver=14.11

cd C:\

git clone https://github.com/apache/incubator-mxnet.git --recursive

mkdir C:\incubator-mxnet\build
cd C:\incubator-mxnet\build

cmake -G "Visual Studio 15 2017 Win64" -T cuda=9.2,host=x64 -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_LIST=Common -DCUDA_TOOLSET=9.2 -DCUDNN_INCLUDE=C:\cuda\include -DCUDNN_LIBRARY=C:\cuda\lib\x64\cudnn.lib "C:\incubator-mxnet"

msbuild mxnet.sln /p:Configuration=Release;Platform=x64 /maxcpucount

cd C:\
git clone https://github.com/apache/incubator-mxnet.git --recursive

msbuild mxnet.sln /p:Configuration=Release;Platform=x64 /maxcpucount

pip install mxnet

pip install mxnet-mkl

$ docker pull mxnet/python

$ docker images

$ docker pull mxnet/python:1.3.0_cpu_mkl
$ docker images

sudo apt-get update
sudo apt-get -y install git cmake ninja-build build-essential g++-4.9 c++-4.9 liblapack*
libblas* libopencv*
libopenblas* python3-dev python-dev virtualenv

git clone https://github.com/apache/incubator-mxnet.git --recursive

cd incubator-mxnet

mkdir -p build && cd build
cmake \
-DUSE_SSE=OFF \
-DUSE_CUDA=OFF \
-DUSE_OPENCV=ON \
-DUSE_OPENMP=ON \
-DUSE_MKL_IF_AVAILABLE=OFF \
-DUSE_SIGNAL_HANDLER=ON \

-DCMAKE_BUILD_TYPE=Release \
-GNinja ..
ninja -j$(nproc)

cd python
pip install --upgrade pip
pip install -e .

ci/docker/runtime_functions.sh build_wheel python/ $(realpath build)

sudo apt update
sudo apt -y install \
   build-essential \
   git \
   graphviz \
   libatlas-base-dev \
   libopencv-dev \
   python-pip
sudo pip install --upgrade \
   pip \
   setuptools
sudo pip install \
   graphviz==0.8.4 \
   jupyter \
   numpy==1.15.2

git clone --recursive https://github.com/apache/incubator-mxnet.git mxnet

export PATH=/usr/local/cuda/bin:$PATH
export MXNET_HOME=$HOME/mxnet/
export PYTHONPATH=$MXNET_HOME/python:$PYTHONPATH

source .profile

nvcc --version

sudo rm /usr/local/cuda
sudo ln -s /usr/local/cuda-10.0 /usr/local/cuda

$MXNET_HOME/ci/build.py -p jetson

cp $MXNET_HOME/make/crosscompile.jetson.mk config.mk

MSHADOW_CFLAGS += -DMSHADOW_USE_PASCAL=1

cd $MXNET_HOME
make -j $(nproc)

cd $MXNET_HOME/python
sudo pip install -e .

import mxnet
mxnet.__version__

pip install --upgrade mxnet gluoncv

pip install --upgrade mxnet gluonnlp

pip install gluonts

pip install gluoncv --pre
pip install mxnet-mkl --pre --upgrade
pip install mxnet-cuXXmkl --pre –upgrade # if cuda XX is installed

pip install git+https://github.com/THUFutureLab/gluon-face.git@master

pip install gluonfr

pip install --upgrade insightface

pip install keras-mxnet

pip install mxnet-cu90

pip install mxnet

pip install mxfusion

virtual bool InferShape(mxnet::ShapeVector *in_shape,
mxnet::ShapeVector *out_shape,
mxnet::ShapeVector *aux_shape) const = 0;

virtual std::vector<ResourceRequest> ForwardResource(
   const mxnet::ShapeVector &in_shape) const;
virtual std::vector<ResourceRequest> BackwardResource(
   const mxnet::ShapeVector &in_shape) const;

void FullyConnectedForward(TBlob weight, TBlob in_data, TBlob out_data);
void FullyConnectedBackward(TBlob weight, TBlob in_data, TBlob out_grad, TBlob in_grad);
void PoolingForward(TBlob in_data, TBlob out_data);
void PoolingBackward(TBlob in_data, TBlob out_data, TBlob out_grad, TBlob in_grad);

virtual Operator* CreateOperator(Context ctx) const = 0;

class ConvolutionOp {
   public:
      void Forward( ... ) { ... }
      void Backward( ... ) { ... }
};
class ConvolutionOpProperty : public OperatorProperty {
   public:
      Operator* CreateOperator(Context ctx) const {
         return new ConvolutionOp;
      }
};

#include <dmlc/parameter.h>
struct ConvolutionParam : public dmlc::Parameter<ConvolutionParam> {
   mxnet::TShape kernel, stride, pad;
   uint32_t num_filter, num_group, workspace;
   bool no_bias;
};

class ConvolutionOp {
   public:
      ConvolutionOp(ConvolutionParam p): param_(p) {}
      void Forward( ... ) { ... }
      void Backward( ... ) { ... }
   private:
      ConvolutionParam param_;
};
class ConvolutionOpProperty : public OperatorProperty {
   public:
      void Init(const vector<pair<string, string>& kwargs) {
         // initialize param_ using kwargs
      }
      Operator* CreateOperator(Context ctx) const {
         return new ConvolutionOp(param_);
      }
   private:
      ConvolutionParam param_;
};

DMLC_REGISTER_PARAMETER(ConvolutionParam);
MXNET_REGISTER_OP_PROPERTY(Convolution, ConvolutionOpProperty);

struct EnvArguments {
   real_t scalar; // scalar argument, if enabled
   std::vector<std::pair<std::string, std::string> > kwargs; // keyword arguments
   std::vector<Resource> resource; // pointer to the resources requested
};

struct ResourceRequest {
   enum Type { // Resource type, indicating what the pointer type is
      kRandom, // mshadow::Random<xpu> object
      kTempSpace // A dynamic temp space that can be arbitrary size
   };
   Type type; // type of resources
};

auto tmp_space_res = env.resources[0].get_space(some_shape, some_stream);
auto rand_res = env.resources[0].get_random(some_stream);

namespace mshadow_op {
   struct smooth_l1_loss {
      // a is x, b is sigma2
      MSHADOW_XINLINE static real_t Map(real_t a, real_t b) {
         if (a > 1.0f / b) {
            return a - 0.5f / b;
         } else if (a < -1.0f / b) {
            return -a - 0.5f / b;
         } else {
            return 0.5f * a * a * b;
         }
      }
   };
}

import mxnet as mx
tensor_cpu = mx.nd.zeros((100,), ctx=mx.cpu())
tensor_gpu= mx.nd.zeros((100,), ctx=mx.gpu(0))

import mxnet as mx
x = mx.sym.Variable(“X”)
y = mx.sym.Variable(“Y”)
z = (x+y)
m = z/100

kv = mxnet.kvstore.create(‘dist_sync’)

trainer = gluon.Trainer(net.collect_params(), optimizer='sgd',
   optimizer_params={'learning_rate': opt.lr,
      'wd': opt.wd,
      'momentum': opt.momentum,
      'multi_precision': True},
      kvstore=kv,
   update_on_kvstore=True)

from mxnet import autograd
import mxnet as mx
from mxnet.gluon.nn import HybridSequential, Dense
from mxnet.gluon.loss import L2Loss

N_net = HybridSequential()
N_net.add(Dense(units=3))
N_net.add(Dense(units=1))
N_net.initialize()

loss_function = L2Loss()

x = mx.nd.array([[0.5, 0.9]])
y = mx.nd.array([[1.5]])

with autograd.record():
   y_hat = N_net(x)
   loss = loss_function(y_hat, y)

loss.backward()

N_net[0].weight.grad()

[[-0.00470527 -0.00846948]
[-0.03640365 -0.06552657]
[ 0.00800354 0.01440637]]
<NDArray 3x2 @cpu(0)>

from mxnet import autograd
import mxnet as mx
from mxnet.gluon.nn import HybridSequential, Dense
from mxnet.gluon.loss import L2Loss
N_net = HybridSequential()
N_net.add(Dense(units=3))
N_net.add(Dense(units=1))
N_net.initialize()
loss_function = L2Loss()
x = mx.nd.array([[0.5, 0.9]])
y = mx.nd.array([[1.5]])
with autograd.record():
y_hat = N_net(x)
loss = loss_function(y_hat, y)
loss.backward()
N_net[0].weight.grad()

import mxnet as mx
from mxnet import nd

x = nd.array([1,2,3,4,5,6,7,8,9,10])
print(x)

[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.]
<NDArray 10 @cpu(0)>

y = nd.array([[1,2,3,4,5,6,7,8,9,10], [1,2,3,4,5,6,7,8,9,10], [1,2,3,4,5,6,7,8,9,10]])
print(y)

[[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.]
[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.]
[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.]]
<NDArray 3x10 @cpu(0)>

x = nd.empty((3, 4))
print(x)
x = nd.full((3,4), 8)
print(x)

[[0.000e+00 0.000e+00 0.000e+00 0.000e+00]
 [0.000e+00 0.000e+00 2.887e-42 0.000e+00]
 [0.000e+00 0.000e+00 0.000e+00 0.000e+00]]
<NDArray 3x4 @cpu(0)>

[[8. 8. 8. 8.]
 [8. 8. 8. 8.]
 [8. 8. 8. 8.]]
<NDArray 3x4 @cpu(0)>

x = nd.zeros((3, 8))
print(x)

[[0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0.]]
<NDArray 3x8 @cpu(0)>

x = nd.ones((3, 8))
print(x)

[[1. 1. 1. 1. 1. 1. 1. 1.]
   [1. 1. 1. 1. 1. 1. 1. 1.]
   [1. 1. 1. 1. 1. 1. 1. 1.]]
<NDArray 3x8 @cpu(0)>

y = nd.random_normal(0, 1, shape=(3, 4))
print(y)

[[ 1.2673576 -2.0345826 -0.32537818 -1.4583491 ]
 [-0.11176403 1.3606371 -0.7889914 -0.17639421]
 [-0.2532185 -0.42614475 -0.12548696 1.4022992 ]]
<NDArray 3x4 @cpu(0)>

y.shape

(3, 4)

y.size

12

y.dtype

numpy.float32

import mxnet as mx
from mxnet import nd
x = nd.ones((3, 5))
y = nd.random_normal(0, 1, shape=(3, 5))
print('x=', x)
print('y=', y)
x = x + y
print('x = x + y, x=', x)

x=
[[1. 1. 1. 1. 1.]
[1. 1. 1. 1. 1.]
[1. 1. 1. 1. 1.]]
<NDArray 3x5 @cpu(0)>
y=
[[-1.0554522 -1.3118273 -0.14674698 0.641493 -0.73820823]
[ 2.031364 0.5932667 0.10228804 1.179526 -0.5444829 ]
[-0.34249446 1.1086396 1.2756858 -1.8332436 -0.5289873 ]]
<NDArray 3x5 @cpu(0)>
x = x + y, x=
[[-0.05545223 -0.3118273 0.853253 1.6414931 0.26179177]
[ 3.031364 1.5932667 1.102288 2.1795259 0.4555171 ]
[ 0.6575055 2.1086397 2.2756858 -0.8332436 0.4710127 ]]
<NDArray 3x5 @cpu(0)>

x = nd.array([1, 2, 3, 4])
y = nd.array([2, 2, 2, 1])
x * y

[2. 4. 6. 4.]
<NDArray 4 @cpu(0)>

nd.exp(x)

[ 2.7182817 7.389056 20.085537 54.59815 ]
<NDArray 4 @cpu(0)>

nd.dot(x, y.T)

[16.]
<NDArray 1 @cpu(0)>

from __future__ import print_function
import mxnet as mx
from mxnet import gluon, nd, autograd
from mxnet.ndarray import NDArray
from mxnet.gluon import HybridBlock
import numpy as np

class LossBuffer(object):
   """
   Simple buffer for storing loss value
   """

   def __init__(self):
      self._loss = None

   def new_loss(self, loss):
      ret = self._loss
      self._loss = loss
      return ret

      @property
      def loss(self):
         return self._loss

net = gluon.nn.Dense(10)
ce = gluon.loss.SoftmaxCELoss()
net.initialize()
data = nd.random.uniform(shape=(1024, 100))
label = nd.array(np.random.randint(0, 10, (1024,)), dtype='int32')
train_dataset = gluon.data.ArrayDataset(data, label)
train_data = gluon.data.DataLoader(train_dataset, batch_size=128, shuffle=True, num_workers=2)
trainer = gluon.Trainer(net.collect_params(), optimizer='sgd')
loss_buffer = LossBuffer()
for data, label in train_data:
   with autograd.record():
      out = net(data)
      # This call saves new loss and returns previous loss
      prev_loss = loss_buffer.new_loss(ce(out, label))
   loss_buffer.loss.backward()
   trainer.step(data.shape[0])
   if prev_loss is not None:
      print("Loss: {}".format(np.mean(prev_loss.asnumpy())))

Loss: 2.3373236656188965
Loss: 2.3656985759735107
Loss: 2.3613128662109375
Loss: 2.3197104930877686
Loss: 2.3054862022399902
Loss: 2.329197406768799
Loss: 2.318927526473999

from mxnet import nd
from mxnet.gluon import nn
x = nd.random.uniform(shape=(2, 20))
N_net = nn.Sequential()
N_net.add(nn.Dense(256, activation='relu'))
N_net.add(nn.Dense(10))
N_net.initialize()
N_net(x)

[[ 0.09543004 0.04614332 -0.00286655 -0.07790346 -0.05130241 0.02942038
0.08696645 -0.0190793 -0.04122177 0.05088576]
[ 0.0769287 0.03099706 0.00856576 -0.044672 -0.06926838 0.09132431
0.06786592 -0.06187843 -0.03436674 0.04234696]]
<NDArray 2x10 @cpu(0)>

from mxnet import nd
from mxnet.gluon import nn
class MySequential(nn.Block):
   def __init__(self, **kwargs):
      super(MySequential, self).__init__(**kwargs)

   def add(self, block):
      self._children[block.name] = block
   def forward(self, x):
   for block in self._children.values():
      x = block(x)
   return x
x = nd.random.uniform(shape=(2, 20))
N_net = MySequential()
N_net.add(nn.Dense(256, activation
='relu'))
N_net.add(nn.Dense(10))
N_net.initialize()
N_net(x)

[[ 0.09543004 0.04614332 -0.00286655 -0.07790346 -0.05130241 0.02942038
0.08696645 -0.0190793 -0.04122177 0.05088576]
[ 0.0769287 0.03099706 0.00856576 -0.044672 -0.06926838 0.09132431
0.06786592 -0.06187843 -0.03436674 0.04234696]]
<NDArray 2x10 @cpu(0)>

from __future__ import print_function
import mxnet as mx
from mxnet import nd, gluon, autograd
from mxnet.gluon.nn import Dense
mx.random.seed(1)
class NormalizationLayer(gluon.Block):
   def __init__(self):
      super(NormalizationLayer, self).__init__()

   def forward(self, x):
      return (x - nd.min(x)) / (nd.max(x) - nd.min(x))
x = nd.random.uniform(shape=(2, 20))
N_net = NormalizationLayer()
N_net.initialize()
N_net(x)

[[0.5216355 0.03835821 0.02284337 0.5945146 0.17334817 0.69329053
0.7782702 1. 0.5508242 0. 0.07058554 0.3677264
0.4366546 0.44362497 0.7192635 0.37616986 0.6728799 0.7032008

 0.46907538 0.63514024]
[0.9157533 0.7667402 0.08980197   0.03593295 0.16176797 0.27679572
 0.07331014 0.3905285 0.6513384 0.02713427 0.05523694 0.12147208
 0.45582628 0.8139887 0.91629887 0.36665893 0.07873632 0.78268915
 0.63404864 0.46638715]]
 <NDArray 2x20 @cpu(0)>

class NormalizationHybridLayer(gluon.HybridBlock):
   def __init__(self):
      super(NormalizationHybridLayer, self).__init__()

   def hybrid_forward(self, F, x):
      return F.broadcast_div(F.broadcast_sub(x, F.min(x)), (F.broadcast_sub(F.max(x), F.min(x))))

layer_hybd = NormalizationHybridLayer()
layer_hybd(nd.array([1, 2, 3, 4, 5, 6], ctx=mx.cpu()))

[0. 0.2 0.4 0.6 0.8 1. ]
<NDArray 6 @cpu(0)>

net = gluon.nn.HybridSequential()
with net.name_scope():
net.add(Dense(5))
net.add(NormalizationHybridLayer())
net.add(Dense(1))
net.initialize(mx.init.Xavier(magnitude=2.24))
net.hybridize()
input = nd.random_uniform(low=-10, high=10, shape=(10, 2))
net(input)

[[-1.1272651]
 [-1.2299833]
 [-1.0662932]
 [-1.1805027]
 [-1.3382034]
 [-1.2081106]
 [-1.1263978]
 [-1.2524893]

 [-1.1044774]

 [-1.316593 ]]
<NDArray 10x1 @cpu(0)>

class NormalizationHybridLayer(gluon.HybridBlock):
   def __init__(self, hidden_units, scales):
      super(NormalizationHybridLayer, self).__init__()
      with self.name_scope():
      self.weights = self.params.get('weights',
      shape=(hidden_units, 0),
      allow_deferred_init=True)
      self.scales = self.params.get('scales',
         shape=scales.shape,
         init=mx.init.Constant(scales.asnumpy()),
         differentiable=False)
      def hybrid_forward(self, F, x, weights, scales):
         normalized_data = F.broadcast_div(F.broadcast_sub(x, F.min(x)),
         (F.broadcast_sub(F.max(x), F.min(x))))
         weighted_data = F.FullyConnected(normalized_data, weights, num_hidden=self.weights.shape[0], no_bias=True)
         scaled_data = F.broadcast_mul(scales, weighted_data)
return scaled_data

import mxnet as mx
kv = mx.kv.create('local') # create a local KVStore.
shape = (3,3)
kv.init(3, mx.nd.ones(shape)*2)
a = mx.nd.zeros(shape)
kv.pull(3, out = a)
print(a.asnumpy())

[[2. 2. 2.]
[2. 2. 2.]
[2. 2. 2.]]

kv.push(3, mx.nd.ones(shape)*8)
kv.pull(3, out = a)
print(a.asnumpy())

[[8. 8. 8.]
 [8. 8. 8.]
 [8. 8. 8.]]

contexts = [mx.cpu(i) for i in range(4)]
b = [mx.nd.ones(shape, ctx) for ctx in contexts]
kv.push(3, b)
kv.pull(3, out = a)
print(a.asnumpy())

[[4. 4. 4.]
 [4. 4. 4.]
 [4. 4. 4.]]

def update(key, input, stored):
   print("update on key: %d" % key)

   stored += input * 2
kv.set_updater(update)
kv.pull(3, out=a)
print(a.asnumpy())

[[4. 4. 4.]
 [4. 4. 4.]
 [4. 4. 4.]]

kv.push(3, mx.nd.ones(shape))
kv.pull(3, out=a)
print(a.asnumpy())

update on key: 3
[[6. 6. 6.]
 [6. 6. 6.]
 [6. 6. 6.]]

b = [mx.nd.ones(shape, ctx) for ctx in contexts]
kv.pull(3, out = b)
print(b[1].asnumpy())

[[6. 6. 6.]
 [6. 6. 6.]
 [6. 6. 6.]]

import mxnet as mx
kv = mx.kv.create('local')
shape = (3,3)
kv.init(3, mx.nd.ones(shape)*2)
a = mx.nd.zeros(shape)
kv.pull(3, out = a)
print(a.asnumpy())
kv.push(3, mx.nd.ones(shape)*8)
kv.pull(3, out = a) # pull out the value
print(a.asnumpy())
contexts = [mx.cpu(i) for i in range(4)]
b = [mx.nd.ones(shape, ctx) for ctx in contexts]
kv.push(3, b)
kv.pull(3, out = a)
print(a.asnumpy())
def update(key, input, stored):
   print("update on key: %d" % key)
   stored += input * 2
kv._set_updater(update)
kv.pull(3, out=a)
print(a.asnumpy())
kv.push(3, mx.nd.ones(shape))
kv.pull(3, out=a)
print(a.asnumpy())
b = [mx.nd.ones(shape, ctx) for ctx in contexts]
kv.pull(3, out = b)
print(b[1].asnumpy())

keys = [5, 7, 9]
kv.init(keys, [mx.nd.ones(shape)]*len(keys))
kv.push(keys, [mx.nd.ones(shape)]*len(keys))
b = [mx.nd.zeros(shape)]*len(keys)
kv.pull(keys, out = b)
print(b[1].asnumpy())

update on key: 5
update on key: 7
update on key: 9
[[3. 3. 3.]
 [3. 3. 3.]
 [3. 3. 3.]]

b = [[mx.nd.ones(shape, ctx) for ctx in contexts]] * len(keys)
kv.push(keys, b)
kv.pull(keys, out = b)
print(b[1][1].asnumpy())

update on key: 5
update on key: 7
update on key: 9
[[11. 11. 11.]
 [11. 11. 11.]
 [11. 11. 11.]]

import mxnet as mx
user = mx.symbol.Variable('user')
item = mx.symbol.Variable('item')
score = mx.symbol.Variable('score')

# Set the dummy dimensions
k = 64
max_user = 100
max_item = 50

# The user feature lookup
user = mx.symbol.Embedding(data = user, input_dim = max_user, output_dim = k)

# The item feature lookup
item = mx.symbol.Embedding(data = item, input_dim = max_item, output_dim = k)

# predict by the inner product and then do sum
N_net = user * item
N_net = mx.symbol.sum_axis(data = N_net, axis = 1)
N_net = mx.symbol.Flatten(data = N_net)

# Defining the loss layer
N_net = mx.symbol.LinearRegressionOutput(data = N_net, label = score)

# Visualize the network
mx.viz.plot_network(N_net)

import mxnet as mx
from mxnet import nd
data = nd.array([[[[2, 5, 3, 6],
   [1, 8, 7, 9],
   [0, 4, 1, 8],
   [2, 0, 3, 4]]]])
affine_matrix = nd.array([[2, 0, 0],
   [0, 2, 0]])

affine_matrix = nd.reshape(affine_matrix, shape=(1, 6))

grid = nd.GridGenerator(data=affine_matrix, transform_type='affine', target_shape=(4, 4))

output = nd.BilinearSampler(data, grid)

[[[[0. 0. 0. 0. ]
   [0. 4.0000005 6.25 0. ]
   [0. 1.5 4. 0. ]
   [0. 0. 0. 0. ]]]]
<NDArray 1x1x4x4 @cpu(0)>

import mxnet as mx
from mxnet import nd
data = nd.array([[[[2, 5, 3, 6],
   [1, 8, 7, 9],
   [0, 4, 1, 8],
   [2, 0, 3, 4]]]])
warp_matrix = nd.array([[[[1, 1, 1, 1],
   [1, 1, 1, 1],
   [1, 1, 1, 1],
   [1, 1, 1, 1]],
   [[0, 0, 0, 0],
   [0, 0, 0, 0],
   [0, 0, 0, 0],
   [0, 0, 0, 0]]]])
grid = nd.GridGenerator(data=warp_matrix, transform_type='warp')
output = nd.BilinearSampler(data, grid)

[[[[5. 3. 6. 0.]
[8. 7. 9. 0.]
[4. 1. 8. 0.]
[0. 3. 4. 0.]]]]
<NDArray 1x1x4x4 @cpu(0)>

nd.cast(nd.array([300, 10.1, 15.4, -1, -2]), dtype='uint8')

[ 44 10 15 255 254]
<NDArray 5 @cpu(0)>

import mxnet as mx
from mxnet import nd
trueclass = mx.nd.array([2])
samples, exp_count_true, exp_count_sample = mx.nd.contrib.rand_zipfian(trueclass, 3, 4)
samples

[0 0 1]
<NDArray 3 @cpu(0)>

exp_count_true

[0.53624076]
<NDArray 1 @cpu(0)>

exp_count_sample

[1.29202967 1.29202967 0.75578891]
<NDArray 3 @cpu(0)>

cond = lambda i, s: i <= 7
func = lambda i, s: ([i + s], [i + 1, s + i])
loop_var = (mx.nd.array([0], dtype="int64"), mx.nd.array([1], dtype="int64"))
outputs, states = mx.nd.contrib.while_loop(cond, func, loop_vars, max_iterations=10)
outputs

[
[[       1]
 [      2]
 [      4]
 [      7]
 [     11]
 [     16]
 [     22]
 [     29]
 [3152434450384]
 [     257]]
<NDArray 10x1 @cpu(0)>]

States

[
[8]
<NDArray 1 @cpu(0)>,
[29]
<NDArray 1 @cpu(0)>]

import numpy as np
img = mx.nd.random.uniform(0, 255, (4, 2, 3)).astype(dtype=np.uint8)
mx.nd.image.to_tensor(img)

[[[0.972549 0.5058824 ]
   [0.6039216 0.01960784]
   [0.28235295 0.35686275]
   [0.11764706 0.8784314 ]]

[[0.8745098 0.9764706 ]
   [0.4509804 0.03529412]
   [0.9764706 0.29411766]
   [0.6862745 0.4117647 ]]

[[0.46666667 0.05490196]
   [0.7372549 0.4392157 ]
   [0.11764706 0.47843137]
   [0.31764707 0.91764706]]]
<NDArray 3x4x2 @cpu(0)>

img = mx.nd.random.uniform(0, 255, (2, 4, 2, 3)).astype(dtype=np.uint8)

mx.nd.image.to_tensor(img)

[[[[0.0627451 0.5647059 ]
[0.2627451 0.9137255 ]
[0.57254905 0.27450982]
[0.6666667 0.64705884]]
[[0.21568628 0.5647059 ]
[0.5058824 0.09019608]
[0.08235294 0.31764707]
[0.8392157 0.7137255 ]]
[[0.6901961 0.8627451 ]
[0.52156866 0.91764706]
[0.9254902 0.00784314]
[0.12941177 0.8392157 ]]]
[[[0.28627452 0.39607844]
[0.01960784 0.36862746]
[0.6745098 0.7019608 ]
[0.9607843 0.7529412 ]]
[[0.2627451 0.58431375]
[0.16470589 0.00392157]
[0.5686275 0.73333335]
[0.43137255 0.57254905]]
[[0.18039216 0.54901963]
[0.827451 0.14509805]
[0.26666668 0.28627452]
[0.24705882 0.39607844]]]]
<NDArgt;ray 2x3x4x2 @cpu(0)>

img = mx.nd.random.uniform(0, 1, (3, 4, 2))

mx.nd.image.normalize(img, mean=(0, 1, 2), std=(3, 2, 1))

[[[ 0.29391178 0.3218054 ]
[ 0.23084386 0.19615503]
[ 0.24175143 0.21988946]
[ 0.16710812 0.1777354 ]]
[[-0.02195817 -0.3847335 ]
[-0.17800489 -0.30256534]
[-0.28807247 -0.19059572]
[-0.19680339 -0.26256624]]
[[-1.9808068 -1.5298678 ]
[-1.6984252 -1.2839255 ]
[-1.3398265 -1.712009 ]
[-1.7099224 -1.6165378 ]]]
<NDArray 3x4x2 @cpu(0)>

img = mx.nd.random.uniform(0, 1, (2, 3, 4, 2))

mx.nd.image.normalize(img, mean=(0, 1, 2), std=(3, 2, 1))

[[[[ 2.0600514e-01 2.4972327e-01]
[ 1.4292289e-01 2.9281738e-01]
[ 4.5158025e-02 3.4287784e-02]
[ 9.9427439e-02 3.0791296e-02]]
[[-2.1501756e-01 -3.2297665e-01]
[-2.0456362e-01 -2.2409186e-01]
[-2.1283737e-01 -4.8318747e-01]
[-1.7339960e-01 -1.5519112e-02]]
[[-1.3478968e+00 -1.6790028e+00]
[-1.5685816e+00 -1.7787373e+00]
[-1.1034534e+00 -1.8587360e+00]
[-1.6324382e+00 -1.9027401e+00]]]
[[[ 1.4528830e-01 3.2801408e-01]
[ 2.9730779e-01 8.6780310e-02]
[ 2.6873133e-01 1.7900752e-01]
[ 2.3462953e-01 1.4930873e-01]]
[[-4.4988656e-01 -4.5021546e-01]
[-4.0258706e-02 -3.2384416e-01]
[-1.4287934e-01 -2.6537544e-01]
[-5.7649612e-04 -7.9429924e-02]]
[[-1.8505517e+00 -1.0953522e+00]
[-1.1318740e+00 -1.9624406e+00]
[-1.8375070e+00 -1.4916846e+00]
[-1.3844404e+00 -1.8331525e+00]]]]
<NDArray 2x3x4x2 @cpu(0)>

mx.nd.random.uniform(0, 1)

[0.12381998]
<NDArray 1 @cpu(0)>

mx.nd.random.uniform(-1, 1, shape=(2,))

[0.558102 0.69601643]
<NDArray 2 @cpu(0)>

low = mx.nd.array([1,2,3])
high = mx.nd.array([2,3,4])
mx.nd.random.uniform(low, high, shape=2)

[[1.8649333 1.8073189]
 [2.4113967 2.5691009]
 [3.1399727 3.4071832]]
<NDArray 3x2 @cpu(0)>

mx.nd.random.generalized_negative_binomial(10, 0.5)

[1.]
<NDArray 1 @cpu(0)>

mx.nd.random.generalized_negative_binomial(10, 0.5, shape=(2,))

[16. 23.]
<NDArray 2 @cpu(0)>

mu = mx.nd.array([1,2,3])
alpha = mx.nd.array([0.2,0.4,0.6])
mx.nd.random.generalized_negative_binomial(mu, alpha, shape=2)

[[0. 0.]
 [4. 1.]
 [9. 3.]]
<NDArray 3x2 @cpu(0)>

mx.nd.zeros((1,2), mx.cpu(), stype='csr')

<CSRNDArray 1x2 @cpu(0)>

mx.nd.zeros((1,2), mx.cpu(), 'float16', stype='row_sparse').asnumpy()

array([[0., 0.]], dtype=float16)

x = mx.nd.zeros((2,3))
y = mx.nd.ones((1,4))
mx.nd.save('list', [x,y])
mx.nd.save('dict', {'x':x, 'y':y})
mx.nd.load('list')

[
[[0. 0. 0.]
[0. 0. 0.]]
<NDArray 2x3 @cpu(0)>,
[[1. 1. 1. 1.]]
<NDArray 1x4 @cpu(0)>]

mx.nd.load('my_dict')

{'x':
[[0. 0. 0.]
[0. 0. 0.]]
<NDArray 2x3 @cpu(0)>, 'y':
[[1. 1. 1. 1.]]
<NDArray 1x4 @cpu(0)>}

from mxnet.gluon import Block, nn
class Model(Block):
   def __init__(self, **kwargs):
      super(Model, self).__init__(**kwargs)
      # use name_scope to give child Blocks appropriate names.
      with self.name_scope():
         self.dense0 = nn.Dense(20)
         self.dense1 = nn.Dense(20)
   def forward(self, x):

      x = mx.nd.relu(self.dense0(x))
      return mx.nd.relu(self.dense1(x))

model = Model()
model.initialize(ctx=mx.cpu(0))
model(mx.nd.zeros((5, 5), ctx=mx.cpu(0)))

[[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]
<NDArray 5x20 @cpu(0)*gt;

import mxnet as mx
from mxnet.gluon import HybridBlock, nn


class Model(HybridBlock):
   def __init__(self, **kwargs):
      super(Model, self).__init__(**kwargs)
      # use name_scope to give child Blocks appropriate names.
      with self.name_scope():
         self.dense0 = nn.Dense(20)
         self.dense1 = nn.Dense(20)

   def forward(self, x):
      x = nd.relu(self.dense0(x))
      return nd.relu(self.dense1(x))
model = Model()
model.initialize(ctx=mx.cpu(0))

model.hybridize()
model(mx.nd.zeros((5, 5), ctx=mx.cpu(0)))

[[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]
<NDArray 5x20 @cpu(0)>

layer = mx.gluon.rnn.GRU(100, 3)
layer.initialize()
input_seq = mx.nd.random.uniform(shape=(5, 3, 10))
out_seq = layer(input_seq)
h0 = mx.nd.random.uniform(shape=(3, 3, 100))
out_seq, hn = layer(input_seq, h0)
out_seq

[[[ 1.50152072e-01 5.19012511e-01 1.02390535e-01 ... 4.35803324e-01
1.30406499e-01 3.30152437e-02]
[ 2.91542172e-01 1.02243155e-01 1.73325196e-01 ... 5.65296151e-02
1.76546033e-02 1.66693389e-01]
[ 2.22257316e-01 3.76294643e-01 2.11277917e-01 ... 2.28903517e-01
3.43954474e-01 1.52770668e-01]]


[[ 1.40634328e-01 2.93247789e-01 5.50393537e-02 ... 2.30207980e-01
6.61415309e-02 2.70989928e-02]
[ 1.11081995e-01 7.20834285e-02 1.08342394e-01 ... 2.28330195e-02
6.79589901e-03 1.25501186e-01]
[ 1.15944080e-01 2.41565228e-01 1.18612610e-01 ... 1.14908054e-01
1.61080107e-01 1.15969211e-01]]
………………………….

hn

[[[-6.08105101e-02 3.86217088e-02   6.64453954e-03 8.18805695e-02
3.85607071e-02 -1.36945639e-02 7.45836645e-03 -5.46515081e-03
9.49622393e-02 6.39371723e-02 -6.37890724e-03 3.82240303e-02
9.11015049e-02 -2.01375950e-02 -7.29381144e-02 6.93765879e-02
2.71829776e-02 -6.64435029e-02 -8.45306814e-02 -1.03075653e-01
6.72040805e-02 -7.06537142e-02 -3.93818803e-02 5.16211614e-03
-4.79770005e-02 1.10734522e-01 1.56721435e-02 -6.93409378e-03
1.16915874e-01 -7.95962065e-02 -3.06530762e-02 8.42394680e-02
7.60370195e-02 2.17055440e-01 9.85361822e-03 1.16660878e-01
4.08297703e-02 1.24978097e-02 8.25245082e-02 2.28673983e-02
-7.88266212e-02 -8.04114193e-02 9.28791538e-02 -5.70827350e-03
-4.46166918e-02 -6.41122833e-02 1.80885363e-02 -2.37745279e-03
4.37298454e-02 1.28888980e-01 -3.07202265e-02 2.50503756e-02
4.00907174e-02 3.37077095e-03 -1.78839862e-02 8.90695080e-02
6.30150884e-02 1.11416787e-01 2.12221760e-02 -1.13236710e-01
5.39616570e-02 7.80710578e-02 -2.28817668e-02 1.92073174e-02
………………………….

layer = mx.gluon.rnn.LSTM(100, 3)
layer.initialize()

input_seq = mx.nd.random.uniform(shape=(5, 3, 10))
out_seq = layer(input_seq)
h0 = mx.nd.random.uniform(shape=(3, 3, 100))
c0 = mx.nd.random.uniform(shape=(3, 3, 100))
out_seq, hn = layer(input_seq,[h0,c0])
out_seq

[[[ 9.00025964e-02 3.96071747e-02 1.83841765e-01 ... 3.95872220e-02
1.25569820e-01 2.15555862e-01]
[ 1.55962542e-01 -3.10300849e-02 1.76772922e-01 ... 1.92474753e-01
2.30574399e-01 2.81707942e-02]
[ 7.83204585e-02 6.53361529e-03 1.27262697e-01 ... 9.97719541e-02
1.28254429e-01 7.55299702e-02]]
[[ 4.41036932e-02 1.35250352e-02 9.87644792e-02 ... 5.89378644e-03
5.23949116e-02 1.00922674e-01]
[ 8.59075040e-02 -1.67027581e-02 9.69351009e-02 ... 1.17763653e-01
9.71239135e-02 2.25218050e-02]
[ 4.34580036e-02 7.62207608e-04 6.37005866e-02 ... 6.14888743e-02
5.96345589e-02 4.72368896e-02]]
……………

hn

[
[[[ 2.21408084e-02 1.42750628e-02 9.53067932e-03 -1.22849066e-02
1.78788435e-02 5.99269159e-02 5.65306023e-02 6.42553642e-02
6.56616641e-03 9.80876666e-03 -1.15729487e-02 5.98640442e-02
-7.21173314e-03 -2.78371759e-02 -1.90690923e-02 2.21447181e-02
8.38765781e-03 -1.38521893e-02 -9.06938594e-03 1.21346042e-02
6.06449470e-02 -3.77471633e-02 5.65885007e-02 6.63008019e-02
-7.34188128e-03 6.46054149e-02 3.19911093e-02 4.11194898e-02
4.43960279e-02 4.92892228e-02 1.74766723e-02 3.40303481e-02
-5.23341820e-03 2.68163737e-02 -9.43402853e-03 -4.11836170e-02
1.55221792e-02 -5.05655073e-02 4.24557598e-03 -3.40388380e-02
……………………

weight = mx.gluon.Parameter('weight', shape=(2, 2))
weight.initialize(ctx=mx.cpu(0))
weight.data()

[[-0.0256899 0.06511251]
[-0.00243821 -0.00123186]]
<NDArray 2x2 @cpu(0)>

weight.grad()

[[0. 0.]
[0. 0.]]
<NDArray 2x2 @cpu(0)>

weight.initialize(ctx=[mx.gpu(0), mx.gpu(1)])
weight.data(mx.gpu(0))

[[-0.00873779 -0.02834515]
 [ 0.05484822 -0.06206018]]
<NDArray 2x2 @gpu(0)>

weight.data(mx.gpu(1))

[[-0.00873779 -0.02834515]
 [ 0.05484822 -0.06206018]]
<NDArray 2x2 @gpu(1)>

import mxnet as mx
from mxnet.gluon import data
sampler = mx.gluon.data.SequentialSampler(15)
batch_sampler = mx.gluon.data.BatchSampler(sampler, 4, 'keep')
list(batch_sampler)

[[0, 1, 2, 3], [4, 5, 6, 7], [8, 9, 10, 11], [12, 13, 14]]

# written to text file *.lst

0 0 root/cat/0001.jpg
1 0 root/cat/xxxa.jpg
2 0 root/cat/yyyb.jpg
3 1 root/dog/123.jpg
4 1 root/dog/023.jpg
5 1 root/dog/wwww.jpg

# A pure list, each item is a list [imagelabel: float or list of float, imgpath]

[[0, root/cat/0001.jpg]
[0, root/cat/xxxa.jpg]
[0, root/cat/yyyb.jpg]
[1, root/dog/123.jpg]
[1, root/dog/023.jpg]
[1, root/dog/wwww.jpg]]

class sigmoid(mx.autograd.Function):
   def forward(self, x):
      y = 1 / (1 + mx.nd.exp(-x))
      self.save_for_backward(y)
      return y

   def backward(self, dy):
      y, = self.saved_tensors
      return dy * y * (1-y)

func = sigmoid()
x = mx.nd.random.uniform(shape=(10,))
x.attach_grad()
with mx.autograd.record():
m = func(x)
m.backward()
dx_grad = x.grad.asnumpy()
dx_grad

array([0.21458015, 0.21291625, 0.23330082, 0.2361367 , 0.23086983,
0.24060014, 0.20326573, 0.21093895, 0.24968489, 0.24301809],
dtype=float32)

x = mx.nd.ones((2,))
x.attach_grad()
with mx.autograd.record():
z = mx.nd.elemwise_add(mx.nd.exp(x), x)
dx_grad = mx.autograd.grad(z, [x], create_graph=True)
dx_grad

[
[3.7182817 3.7182817]
<NDArray 2 @cpu(0)>]

with mx.autograd.record():
y = model(x)
with mx.autograd.predict_mode():
y = sampling(y)
backward([y])

init = mx.init.Normal(0.8)
init.dumps()

'["normal", {"sigma": 0.8}]'

init = mx.init.Xavier(factor_type="in", magnitude=2.45)
init.dumps()

'["xavier", {"rnd_type": "uniform", "factor_type": "in", "magnitude": 2.45}]'

init = mx.initializer.Mixed(['bias', '.*'], [mx.init.Zero(),
mx.init.Uniform(0.1)])
module.init_params(init)

for dictionary in module.get_params():
for key in dictionary:
print(key)
print(dictionary[key].asnumpy())

fullyconnected1_weight
[[ 0.0097627 0.01856892 0.04303787]]
fullyconnected1_bias
[ 0.]

import mxnet as mx
mx.sym.power(3, 5)

243

x = mx.sym.Variable('x')
y = mx.sym.Variable('y')
z = mx.sym.power(x, 3)
z.eval(x=mx.nd.array([1,2]))[0].asnumpy()

array([1., 8.], dtype=float32)

z = mx.sym.power(4, y)
z.eval(y=mx.nd.array([2,3]))[0].asnumpy()

array([16., 64.], dtype=float32)

z = mx.sym.power(x, y)
z.eval(x=mx.nd.array([4,5]), y=mx.nd.array([2,3]))[0].asnumpy()

array([ 16., 125.], dtype=float32)

input_data = mx.nd.array([[2., 0.9, -0.5, 4., 8.], [4., -.7, 9., 2., 0.9]])
soft_max_act = mx.nd.softmax(input_data)
print (soft_max_act.asnumpy())

[[2.4258138e-03 8.0748333e-04 1.9912292e-04 1.7924475e-02 9.7864312e-01]
[6.6843745e-03 6.0796250e-05 9.9204916e-01 9.0463174e-04 3.0112563e-04]]

import mxnet as mx
true_cls = mx.sym.Variable('true_cls')
samples, exp_count_true, exp_count_sample = mx.sym.contrib.rand_zipfian(true_cls, 5, 6)
samples.eval(true_cls=mx.nd.array([3]))[0].asnumpy()

array([4, 0, 2, 1, 5], dtype=int64)

exp_count_true.eval(true_cls=mx.nd.array([3]))[0].asnumpy()

array([0.57336551])

exp_count_sample.eval(true_cls=mx.nd.array([3]))[0].asnumpy()

array([1.78103594, 0.46847373, 1.04183923, 0.57336551, 1.04183923])

cond = lambda i, s: i <= 7
func = lambda i, s: ([i + s], [i + 1, s + i])
loop_vars = (mx.sym.var('i'), mx.sym.var('s'))
outputs, states = mx.sym.contrib.while_loop(cond, func, loop_vars, max_iterations=10)
print(outputs)

[<Symbol _while_loop0>]

Print(States)

[<Symbol _while_loop0>, <Symbol _while_loop0>]

import mxnet as mx
a = mx.nd.zeros((6,3))
b = mx.nd.array([[1,2,3],[4,5,6],[7,8,9]])
index = mx.nd.array([0,4,2])
mx.nd.contrib.index_copy(a, index, b)

[[1. 2. 3.]
[0. 0. 0.]
[7. 8. 9.]
[0. 0. 0.]
[4. 5. 6.]
[0. 0. 0.]]
<NDArray 6x3 @cpu(0)>

import numpy as np

img = mx.sym.random.uniform(0, 255, (4, 2, 3)).astype(dtype=np.uint8)

mx.sym.image.to_tensor(img)

<Symbol to_tensor4>

img = mx.sym.random.uniform(0, 255, (2, 4, 2, 3)).astype(dtype=np.uint8)

mx.sym.image.to_tensor(img)

<Symbol to_tensor5>

img = mx.sym.random.uniform(0, 1, (3, 4, 2))

mx.sym.image.normalize(img, mean=(0, 1, 2), std=(3, 2, 1))

<Symbol normalize0>

img = mx.sym.random.uniform(0, 1, (2, 3, 4, 2))

mx.sym.image.normalize(img, mean=(0, 1, 2), std=(3, 2, 1))

<Symbol normalize1>

data = mx.nd.array([[0, 1, 2], [3, 4, 5], [6, 7, 8],[9,10,11]])
x = mx.sym.Variable('x')
y = mx.sym.random.shuffle(x)
y.eval(x=data)

[
[[ 9. 10. 11.]
[ 0. 1. 2.]
[ 6. 7. 8.]
[ 3. 4. 5.]]
<NDArray 4x3 @cpu(0)>]

y.eval(x=data)

[
[[ 6. 7. 8.]
[ 0. 1. 2.]
[ 3. 4. 5.]
[ 9. 10. 11.]]
<NDArray 4x3 @cpu(0)>]

mx.sym.random.generalized_negative_binomial(10, 0.1)

<Symbol _random_generalized_negative_binomial0>

input_dim = 4
output_dim = 5

/* Here every row in weight matrix y represents a word. So, y = (w0,w1,w2,w3)
y = [[ 0., 1., 2., 3., 4.],
[ 5., 6., 7., 8., 9.],
[ 10., 11., 12., 13., 14.],
[ 15., 16., 17., 18., 19.]]
/* Here input array x represents n-grams(2-gram). So, x = [(w1,w3), (w0,w2)]
x = [[ 1., 3.],
[ 0., 2.]]
/* Now, Mapped input x to its vector representation y.
Embedding(x, y, 4, 5) = [[[ 5., 6., 7., 8., 9.],
[ 15., 16., 17., 18., 19.]],
[[ 0., 1., 2., 3., 4.],
[ 10., 11., 12., 13., 14.]]]

import mxnet as mx
input_data = mx.symbol.Variable('input_data')
f_connected1 = mx.symbol.FullyConnected(data, name='f_connected1', num_hidden=128)
activation_1 = mx.symbol.Activation(f_connected1, name='relu1', act_type="relu")
f_connected2 = mx.symbol.FullyConnected(activation_1, name = 'f_connected2', num_hidden = 64)
activation_2 = mx.symbol.Activation(f_connected2, name='relu2',
act_type="relu")
f_connected3 = mx.symbol.FullyConnected(activation_2, name='fc3', num_hidden=10)
out = mx.symbol.SoftmaxOutput(f_connected3, name = 'softmax')
mod = mx.mod.Module(out)
print(out)

<Symbol softmax>

print(mod)

<mxnet.module.module.Module object at 0x00000123A9892F28>

import mxnet as mx
from collections import namedtuple
Batch = namedtuple('Batch', ['data'])
data = mx.sym.Variable('data')
out = data * 2
mod = mx.mod.Module(symbol=out, label_names=None)
mod.bind(data_shapes=[('data', (1, 10))])
mod.init_params()
data1 = [mx.nd.ones((1, 10))]
mod.forward(Batch(data1))
print (mod.get_outputs()[0].asnumpy())

[[2. 2. 2. 2. 2. 2. 2. 2. 2. 2.]]

data2 = [mx.nd.ones((3, 5))]

mod.forward(Batch(data2))
print (mod.get_outputs()[0].asnumpy())

[[2. 2. 2. 2. 2.]
[2. 2. 2. 2. 2.]
[2. 2. 2. 2. 2.]]