Skip to content

Latest commit

 

History

History
135 lines (85 loc) · 5.31 KB

nvidia-docker-run-digits.md

File metadata and controls

135 lines (85 loc) · 5.31 KB

Use nvidia-docker-run-digits

https://github.com/NVIDIA/nvidia-docker

Follow the guide(NVIDIA Container Runtime for Docker) to install

It works for me follow the steps

Prerequisites

The QUICKSTART in https://github.com/NVIDIA/nvidia-docker show some prerequisites

  1. GNU/Linux x86_64 with kernel version > 3.10
  2. Docker CE >= 1.12
  3. NVIDIA GPU with Architecture > Fermi (2.1)
  4. NVIDIA drivers ~= 361.93 (untested on older versions)
  5. CUDA

NOTE Your driver version might limit your CUDA capabilities (see CUDA requirements)

Install

QUICKSTART

Make sure you have installed the NVIDIA driver and a supported version of Docker for your distribution (see prerequisites).

If you have a custom /etc/docker/daemon.json, the nvidia-docker2 package might override it.

Ubuntu 14.04/16.04/18.04, Debian Jessie/Stretch

# If you have nvidia-docker 1.0 installed: we need to remove it and all existing GPU containers
docker volume ls -q -f driver=nvidia-docker | xargs -r -I{} -n1 docker ps -q -a -f volume={} | xargs -r docker rm -f
sudo apt-get purge -y nvidia-docker

# Add the package repositories
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \
  sudo apt-key add -
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | \
  sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update

# Install nvidia-docker2 and reload the Docker daemon configuration
sudo apt-get install -y nvidia-docker2
sudo pkill -SIGHUP dockerd

# Test nvidia-smi with the latest official CUDA image
docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi

Use it

Basic Usage

nvidia-docker registers a new container runtime to the Docker daemon. You must select the nvidia runtime when using docker run:

docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi

Bellow is the infomation about my enviroments.

$ sudo docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi
Unable to find image 'nvidia/cuda:9.0-base' locally
9.0-base: Pulling from nvidia/cuda
18d680d61657: Pull complete
0addb6fece63: Pull complete
78e58219b215: Pull complete
eb6959a66df2: Pull complete
6ef1ff668c93: Pull complete
f5f8f0544aa2: Pull complete
3d28d96eb352: Pull complete
Digest: sha256:764039ce9ff2cfb44d646fde6930099493334bb743e5b4f089d820de023c5d9a
Status: Downloaded newer image for nvidia/cuda:9.0-base
Wed Nov 21 03:37:25 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.72   Driver Version: 410.72   CUDA Version: 10.0 |
|-------------------------------+----------------------+----------------------+
| GPU  NamePersistence-M        | Bus-IdDisp.A         | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap| Memory-Usage         | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro K620 Off          | 00000000:03:00.0  On |  N/A                 |
| 34%   41CP8 1W /  30W         | 52MiB /  2000MiB     |  1%  Default         |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:   GPU Memory                                                     |
|  GPU   PID   Type   Process name Usage                                      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Advance Usage

There is a very good project in github to guide you use Nvidia DIGITS in docker to train a Caffe model.

https://github.com/humphd/have-fun-with-machine-learning

It has a very sweet README to lead you train a model step-by-step.

BUT

It just show use CPU-only docker image to do that which is very slowly. So if you have NVIDIA GPU has capablity of speed up. You'd like to try levelage the GPU with nvidia-docker.

Here is the example.

Start the nvidia-docker images.

$sudo nvidia-docker run --name digits5 -d -p 9090:5000 -v /DATA/Gits/have-fun-with-machine-learning:/data/repo nvidia/digits:5.0

You must note that we still use the data(image data for training model) got from https://github.com/humphd/have-fun-with-machine-learning .

Other things about do the training model is same as description in https://github.com/humphd/have-fun-with-machine-learning

As my experience to train a same model with same data, the nvidia-docker w/ gpu may be 30times fast than docker w/o gpu.

ISSUE

  1. no response when do "apt install/update" This may cause by your wrong DNS.

The default network of docker is bridge mode. You may need to set DNS in your container as your real network. This can be done by modify the /etc/resolv.conf in the container. You just need to add a line like "nameserver [your DNS server IP]" at the end of /etc/resolv.conf

Reference

https://github.com/NVIDIA/nvidia-docker  
https://github.com/nvidia/nvidia-docker/wiki/Installation-(version-2.0)