-
Workshop Name: Applied Deep Learning for coders with Apache MXNet - Hands-on deep learning in Computer Vision and Natural Language Processing
-
Workshop link:
- Lab 1 - Basics of NDArray (Tensors) - Fundamental Datastructure in Deep Learning
- Lab 2 - Fashion MNIST - Train your first Neural Network in Gluon - Multi Layer Perceptron(MLP)
- Lab 3 - Facial Emotion Recognition in Gluon - Intuition behind Convolutional Neural Networks(CNN)
- Lab 4 - Text to Emoji Prediction in Gluon - Intuition behind Recurrent Neural Networks(RNN)
- Install Anaconda - https://docs.continuum.io/anaconda/install/mac-os.html OR https://docs.continuum.io/anaconda/install/linux/
- 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
-
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
-
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
- Apache MXNet (Incubating) - http://mxnet.incubator.apache.org/
- Learning Deep Learning with Gluon - https://gluon.mxnet.io/
- (Highly Recommended) Dive into Deep Learning with Gluon - http://d2l.ai/