henryzlo/deepdock repository overview

⁠CUDA-enabled Deep Learning Image

Jupyter notebook server (port 8888) for deep learning. Contains:

• Caffe
• Theano
• TensorFlow
• MXnet
• Keras
• Lasagne
• Scikit-learn
• Jupyter

The github repo and docker image are both henryzlo/deepdock

⁠Requirements

You will need CUDA 7.5 installed and a GPU. Then you need to attach the GPU to the image with the flags:

--device /dev/nvidia0 --device /dev/nvidia-uvm --/device /dev/nvidiactl

⁠Cheatsheet

Usually I have a data and workspace folder that I like to attach. This can be done using the flags:

-v `pwd`/workspace:/root/workspace -v `pwd`/data:/root/data

Putting it all together:

docker run -it -P --device /dev/nvidia0 --device /dev/nvidia-uvm --device /dev/nvidiactl -v `pwd`/workspace:/root/workspace -v `pwd`/data:/root/data henryzlo/deepdock

If you want a shell in the image, run the command above, then use:

docker exec -it <image-name> bash

Where <image-name> can be obtained via docker ps.

To connect to the notebook server, point browser to <IP>:8888, where <IP> can be obtained via docker inspect <image-name> | grep IPAddress

Default password is abc123ak47