Container objects

Besides constructing neural networks through creating a subclass of Block, Gluon provides containers to facilitate the construction.

Example

Import modules

>>> from mxnet import np, npx
>>> from mxnet.gluon import nn
>>> npx.set_np()

Construct a MLP

>>> net = nn.Sequential()
>>> net.add(nn.Dense(10, activation='relu'),
...         nn.Dense(2))

Initialize the network and run forward

>>> net.initialize()
>>> x = np.random.uniform(size=(4, 20))
>>> print(net(x))
[[-0.01270512  0.01225275]
 [-0.01303939  0.01191816]
 [-0.01165176  0.00252366]
 [-0.02191715  0.01045103]]

nn.Sequential([prefix, params])	Stacks Blocks sequentially.
nn.HybridSequential([prefix, params])	Stacks HybridBlocks sequentially.

contrib.nn.Concurrent([axis, prefix, params])	Lays Block s concurrently.
contrib.nn.HybridConcurrent([axis, prefix, …])	Lays HybridBlock s concurrently.