Deep Learning System Lab1 Notebooks

This repo contains various notebooks that introduces basic usages and training of deep learning models and applying pretrained state-of-the-art models.

How to use

The python notebooks are written in Jupyter.

Setup

• We can run and modify these notebooks if both mxnet and jupyter are installed. Here is an example script to install all these packages on Ubuntu.

• Intructions to setup on AWS instances to run the notebooks:

  1. Create an AWS account, and apply the student credit from AWS Education.

  2. Launch a g2.2xlarge or p2.2xlarge instance. The AMI Ids that you can use:

     AWS Region Name	Region	AMI Id
     US West (Oregon)	us-west-2	ami-dfb13ebf
     US East (N. Virginia)	us-east-1	ami-e7c96af1
     EU (Ireland)	eu-west-1	ami-6e5d6808

     Remember to open the TCP port 22 and 8888 in the security group. You can modify the security group after you create the instance. Find the Instances page at the EC2 Dashboard, and click the security group of the instance you created, which is at the last column. In the Inbound tab, add the SSH rule and customized TCP Rule with port 8888 from anywhere.

  3. Once launch is succeed, setup the following variable with proper value

      export HOSTNAME=ec2-107-22-159-132.compute-1.amazonaws.com
      export PERM=~/Downloads/my.pem
     

  4. Now we should be able to ssh to the machine by

      chmod 400 $PERM
      ssh -i $PERM -L 8888:localhost:8888 ec2-user@HOSTNAME
     

     Here we forward the EC2 machine's 8888 port into localhost.

  5. Clone this repo on the EC2 machine and run jupyter

      sudo yum install -y graphviz
      git clone https://github.com/dlsys-course/lab1.git
      cd lab1
      jupyter notebook
     

  6. Wait until jupyter creates the notebook and copy the URL it outputs to your brower. The URL should look like: http://localhost:8888/?token=5b870699ac133c42c56400de91f66256b89ab211ee38c7e0