imcomking/bi_deeplearning repository overview

⁠* Host System : Ubuntu 14.04

⁠Announcement for CUDA 7.5

You need to install cuda 7.5 [ NVIDIA-SMI 367.44 Driver Version: 367.44 ] and CuDNN v5.1 on your host machine first. See gpu driver 367.44 --> http://www.nvidia.com/download/driverResults.aspx/106780/en-us⁠ See cuda 7.5 --> https://developer.nvidia.com/cuda-downloads⁠ See cudnn v5.1 --> https://developer.nvidia.com/cudnn⁠

⁠Announcement for CUDA 8

• There is updated version of image for the cuda 8 user. Please see the Tags. (But there was one report of not working of Caffe. However the other library works well.)

You need to install cuda 8 [ NVIDIA-SMI 375.26 Driver Version: 375.26 ] and CuDNN v5.1 on your host machine first.

⁠* General Usage

⁠# create container with gpu device:

docker run -it --device /dev/nvidiactl --device /dev/nvidia-uvm --device /dev/nvidia0 --dns=8.8.8.8 -p 22 -p 6006 -p 8888 imcomking/bi_deeplearning

⁠if you get an error message [ "/dev/nvidia-uvm": lstat /dev/nvidia-uvm: no such file or directory ], firstly try to run [ nvidia-smi ] on your host machine. Because "/dev/nvidia-uvm" is created sometimes when nvidia-smi is firstly called. Otherwise try these commands on your host machine.

• sudo apt-get install nvidia-modprobe
• sudo modprobe nvidia-uvm
• sudo mknod -m 666 /dev/nvidia-uvm c 250 0

⁠if you get an error message [ failed call to cuInit: CUDA_ERROR_UNKNOWN ], try these commands on your host machine. (I don't know the reason, but if you install tensorflow on you host machine, the container can use the CUDA correctly)

• wget https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow-0.10.0-cp27-none-linux_x86_64.whl⁠
• sudo pip install --upgrade tensorflow-0.10.0-cp27-none-linux_x86_64.whl
• export PATH=/usr/local/cuda/bin:$PATH
• export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

⁠# create container without gpu device (if you have some problem on above method):

docker run -it --dns=8.8.8.8 -p 22 -p 6006 -p 8888 imcomking/bi_deeplearning

⁠- Setting IPython notebook[3.2.1]

cd /root python setup_nbserver.py (enter a powerful password) screen -S ipy ipython notebook --profile=nbserver (ctrl + AD)

⁠- Setting ssh/sftp

passwd (enter a powerful password) wget https://raw.githubusercontent.com/bi-lab/deeplearning_tutorial/master/sshd_config⁠ -O /etc/ssh/sshd_config

screen -S ssh /usr/sbin/sshd -D (ctrl + AD)

• ssh/sftp user info id : root

• If this '/usr/sbin/sshd -D' command doesn't work, please restart the container.

⁠* Core Library List

Cuda 7.5 CuDNN v5.1 TensorFlow r0.10.0 Theano latest(2016.9.19) Torch 7 latest(2016.9.19) Scikit-learn latest(2016.9.19) Keras latest(2016.9.19) Caffe latest(2016.9.19) (path: /opt/caffe) Bazel 0.3.1 lateset(2016.9.19) Lasagne latest(2016.9.19) JSAnimation latest(2016.9.19) IPython Notebook 3.2.1 SSH/SFTP

⁠* Apt-get installed item list

⁠#for Tensorflow. this is come from the official tensorflow docker file

build-essential
curl
git
libfreetype6-dev
libpng12-dev
libzmq3-dev
pkg-config
python-dev
python-numpy
python-pip
software-properties-common
swig
zip
zlib1g-dev \

⁠#for Caffe. this is come from the official caffe dockerfile

cmake
wget
libatlas-base-dev
libboost-all-dev
libgflags-dev
libgoogle-glog-dev
libhdf5-serial-dev
libleveldb-dev
liblmdb-dev
libopencv-dev
libprotobuf-dev
libsnappy-dev
protobuf-compiler \
graphviz \

⁠#for etc

python-scipy
python-nose
python-setuptools
python-h5py \ python-matplotlib
python-yaml
libopenblas-dev
screen \ vim \ unzip
libatlas-dev
libhdf5-dev
libatlas3gf-base

⁠* pip installed item list

pip future euclid redis pandas Request six mpld3 pydotplus