chattg1 / oreilly-mxnet-workshop

Lab resources for Oreilly online Apache MXNet workshop

O'Reilly Apache MXNet Workshop Lab Materials

• Workshop Name: Applied Deep Learning for coders with Apache MXNet - Hands-on deep learning in Computer Vision and Natural Language Processing

• Workshop link: https://www.safaribooksonline.com/live-training/courses/applied-deep-learning-for-coders-with-apache-mxnet/0636920232841/

Lab Exercises

1. Lab 1 - Basics of NDArray (Tensors) - Fundamental Datastructure in Deep Learning
2. Lab 2 - Fashion MNIST - Train your first Neural Network in Gluon - Multi Layer Perceptron(MLP)
3. Lab 3 - Facial Emotion Recognition in Gluon - Intuition behind Convolutional Neural Networks(CNN)
4. Lab 4 - Text to Emoji Prediction in Gluon - Intuition behind Recurrent Neural Networks(RNN)

Prerequisites / Installation

(Option 1) On Ubuntu / Mac

Install Anaconda and Setup Conda Environment

1. Install Anaconda - https://docs.continuum.io/anaconda/install/mac-os.html
2. Set up Conda Environment for MXNet development

Execute the below commands for the first time, when you are setting up your machine:

# Download this repository with lab exercises and resources
$ git clone  https://github.com/sandeep-krishnamurthy/oreilly-mxnet-workshop

# Create a conda environment with name "mxnet_dev"
$ conda create -n mxnet_dev python=3 numpy jupyter

# Activate
$ source activate mxnet_dev

# Install MXNet

(mxnet_dev) $ pip install mxnet-mkl # for CPU machines
(mxnet_dev) $ pip install mxnet-cu92 # for GPU machines with CUDA 9.2

# Dependencies for CNN Lab exercise
(mxnet_dev) $ pip install Pillow # For image processing
(mxnet_dev) $ pip install graphviz # For MXNet network visualization
(mxnet_dev) $ pip install matplotlib # For plotting training graphs

# Dependencies for RNN Lab exercise
(mxnet_dev) $ pip install emoji
(mxnet_dev) $ pip install gluonnlp
(mxnet_dev) $ pip install spacy -U --quiet
(mxnet_dev) $ python -m spacy download en

# Dependencies for model serving with MXNet Model Server
(mxnet_dev) $ pip install mxnet-model-server
(mxnet_dev) $ pip install scikit-image
(mxnet_dev) $ pip install opencv-python

# Deactivate the environment
(mxnet_dev) $ source deactivate

Execute, below command whenever you want to work with these Lab exercises: NOTE: Make sure you have cloned (downloaded) this repository and you are in the downloaded directory (oreilly-mxnet-workshop)

# Activate the mxnet_dev conda environment we have prepared
$ source activate mxnet_dev

# Start Jupyter Notebook
(mxnet_dev) $ jupyter notebook

# When you are done. Deactivate the conda environment
(mxnet_dev) $ source deactivate

(Option 2) Use cloud - AWS / Google Cloud

1. On AWS, you can use pre-configured Deep Learning AMIs, which comes with all frameworks and libraries pre-installed in Conda environments. You just need to launch an EC2 instance with Deep Learning AMI and then get started with coding! See here for Instructions

2. On Google Cloud, you can use pre-configured Deep Learning images. Similar to AWS Deep Learning AMI, all the deep learning frameworks and required library dependencies are pre-installed and you can just get started with coding! See here for Instructions

Resources

1. Apache MXNet (Incubating) - http://mxnet.incubator.apache.org/
2. Learning Deep Learning with Gluon - https://gluon.mxnet.io/
3. (Highly Recommended) Dive into Deep Learning with Gluon - http://d2l.ai/