Using the DevBox

Accessing the DevBox

Connect to the DevBox by SSH over the internal network. It is on:

• 192.168.2.202
• Username: researcher
• Password hint: littleamount

It has a /bulk samba share that can be mounted from the RDP. Follow the instructions on Help:Access RDP Server.

Specification

Our DIGITS DevBox, affectionately named after Lois McMaster Bujold's fifth God, has a XEON e5-2620v3 processor, 256GB of DDR4 RAM, two GPUs - one Titan RTX and one Titan Xp - with room for two more, a 500GB SSD hard drive (mounting /), and an 8TB RAID5 array bcached with a 512GB m.2 drive (mounting the /bulk share, which is available over samba). It runs Ubuntu 18.04, CUDA 10.0, cuDNN 7.6.1, Anaconda3-2019.03, python 3.7, tensorflow 1.13, digits 6, and other useful machine learning tools/libraries.

Working in Tensorflow

After you've connected to the box as researcher, you should be in /home/researcher. If not, cd there. Then load the virtual environment:

source ./venv/bin/activate

Test tensorflow:

python -c "import tensorflow as tf; tf.enable_eager_execution(); print(tf.reduce_sum(tf.random_normal([1000, 1000])))"

Sample CNN Projects

There are some nice Tensorflow CNN tutorials online:

• https://www.tensorflow.org/tutorials/estimators/cnn
• https://www.tensorflow.org/alpha/tutorials/images/intro_to_cnns
• https://www.tensorflow.org/tutorials/images/deep_cnn

Some of which use standard datasets:

• https://medium.com/tensorflow/introducing-tensorflow-datasets-c7f01f7e19f3

The code for MNIST.py is in E:\projects\tensorflow.

Word2Vec

There's a nice Word2Vec guide with code and data here: https://www.tensorflow.org/tutorials/representation/word2vec

DIGITS

DIGITS runs in docker containers. See https://docs.nvidia.com/deeplearning/digits/digits-release-notes/running-digits.html

As root run:

docker run --runtime=nvidia --name digits -d -p 5000:5000 nvidia/digits

Then browse to http://192.168.2.202:5000

Notes

Docker won't run if you left a container with the same name running (active or not). List docker containers:

docker ps -a

Stop a container:

docker stop containername

Prune them (-a is all, -f is no prompt)!

docker system prune -af

If you do a full prune, then docker will download a new image for nvidia/digits:latest