Skip to content
This repository has been archived by the owner on Dec 23, 2020. It is now read-only.
/ gcp-ubuntu-gpu Public archive

This is a quick shell script file to set up with NVIDIA GPUs on Google Cloud Platform VM.

License

Notifications You must be signed in to change notification settings

GzuPark/gcp-ubuntu-gpu

Repository files navigation

[한국어 가이드]

Quick Set Up NVIDIA GPUs with GCP VM

  • OS: Ubuntu 16.04
  • GPU: NVIDIA K80, P100, V100
    • CUDA: 9.0.176
    • NCCL: 2.2.13
    • CUDNN: 7.1.4.18
  • Python: 3.5.2
    • TensorFlow: 1.12.0
    • PyTorch: 1.0.0
    • Other ML libraries
  • Programs:
    • Anaconda: 5.3.1, Python: 3.7.1 (is not default)
    • nvidia-docker: 2.0.0

How to use

  • Run gcp_setup_ubuntu1604.sh
  • Before you run this file, please check your OS Ubuntu 16.04.
  • You can modify this code if you want to use other OS or CUDA version.
wget https://raw.githubusercontent.com/GzuPark/gcp-ubuntu-gpu/master/gcp_setup_ubuntu1604.sh
bash gcp_setup_ubuntu1604.sh

Etc

Anaconda

  • If you want to change from default python interpreter to Anaconda:
sudo bash
ln -s -f /usr/bin/anaconda3/bin/python3 /usr/bin/python
ln -s -f /usr/bin/anaconda3/bin/pip /usr/bin/pip
ln -s -f /usr/bin/anaconda3/bin/jupyter /usr/bin/jupyter

Jupyter

  • First, run this command:
jupyter notebook --allow-root
  • And then, check your EXTERNAL_IP of a instance, enter like below on your web-browser (ex: 10.100.0.8)
10.100.0.8:8888

Issues

Please comment at the Issues tap.

About

This is a quick shell script file to set up with NVIDIA GPUs on Google Cloud Platform VM.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published