Update MXNet examples with Gluon training APIs

[eric-haibin-lin]

A large number of examples in the mxnet official repo is using the Module APIs for training. Since the Module APIs will be removed in mxnet 2.0 in favor of the easier-to-use Gluon APIs, it will be great to get help from the community to update/rewrite these examples using the Gluon APIs.

Below is a list of examples using Module APIs

• [ ] autoencoder / variational autoencoder
• [ ] baysian-method
• [ ] capsnet (capsule network)
• [ ] cnn_chinese_text_classification
• [ ] ctc (connectionist temporal classification for speech)
• [ ] deep-embedded-clustering
• [ ] dsd (dense-sparse-dense training)
• [ ] fcn-xs
• [ ] gan
• [ ] model-parallel
• [ ] module
• [ ] multivariate-time-series
• [ ] named-entity-recognition
• [ ] nce-loss
• [ ] neural-style
• [ ] rcnn
• [ ] numpy-ops
• [ ] reinforcement-learning
• [ ] rnn
• [ ] sparse
• [ ] speech-recognition
• [ ] svrg-module
• [ ] vae-gan

[szha]

Going forward, we should have properly maintained examples that are checked by CI and rerun frequently to make sure they are function. We should probably decide on the proper infrastructure for doing that first.

Also, since the toolkits are better equipped to maintain examples for each domain, it might be a good idea to having them in the toolkits instead.

[pengzhao-intel]

Going forward, we should have properly maintained examples that are checked by CI and rerun frequently to make sure they are function. We should probably decide on the proper infrastructure for doing that first.

Also, since the toolkits are better equipped to maintain examples for each domain, it might be a good idea to having them in the toolkits instead.

Toolkits distributed in different repos and it's hard to get by newer. I think the basic examples are still very helpful to illustrate the FWK's usage and capability.

[szha]

I created the repository for examples here: https://github.com/apache/incubator-mxnet-examples