Public Repository

Last pushed: 2 years ago
Short Description
This is Caffe compiled for GPU + Anaconda + Jupyter
Full Description

On AWS g2.2xlarge instance

  1. Choose public AMI: ami-2cbf3e44
  2. Once you ssh onto the instance run:
    cd /usr/local/cuda/samples/1_Utilities/deviceQuery
    make
    ./deviceQuery
    
    Check that everything is OK-ish
  3. install docker:
    sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-keys 36A1D7869245C8950F966E92D8576A8BA88D21E9
    sudo sh -c "echo deb https://get.docker.com/ubuntu docker main > /etc/apt/sources.list.d/docker.list"
    sudo apt-get update
    sudo apt-get install lxc-docker
    
    (Copied from tleyden)
  4. run
    DOCKER_NVIDIA_DEVICES="--device /dev/nvidia0:/dev/nvidia0 --device /dev/nvidiactl:/dev/nvidiactl --device /dev/nvidia-uvm:/dev/nvidia-uvm"
    sudo docker run -ti $DOCKER_NVIDIA_DEVICES mjaskowski/caffe-gpu /bin/bash
    
  5. Now you are inside the docker, you can check that everything is ok by running:
    cd /opt/caffe
    make test && make runtest
    

Why care?

This container is similar to and heavily inspired by tleyden5iwx/caffe-gpu-master but there you have support built on top of pip, whereas here whole anaconda is available with jupyter.

You can run jupyter server like this:

sudo docker run -d -p 0.0.0.0:8888:8888 $DOCKER_NVIDIA_DEVICES mjaskowski/caffe-gpu sh -c "ipython notebook --ip=* --no-browser"

Then go to YOUR_AMAZON_PUBLIC_DNS:8888 and browse to /opt/caffe/examples and play happily!

Note! The jupyter notebook is not password protected!

Docker Pull Command
Owner
mjaskowski

Comments (0)