项目作者: DeadAt0m

项目描述 :
JupyterLab IDE + ML environment docker container
高级语言: Shell
项目地址: git://github.com/DeadAt0m/JupyterLabIDE-docker.git
创建时间: 2020-02-15T12:29:48Z
项目社区:https://github.com/DeadAt0m/JupyterLabIDE-docker

开源协议:

下载


JupyterLab IDE container:

  1. - based on jupyter docker stacks(https://github.com/jupyter/docker-stacks) and can be usable inside JupyterHub or standalone.
  2. - includes all necessary(and little more) ML, CV and NeuroImaging libraries including(PyTorch, TensorFlow, Pyro, OpencCV, MNE, PyWavelets, etc.).
  3. - carefully selected extensions for JupyterLab turning it to powerfull IDE.

Requirements:

  1. - Docker(>=19.03.6, API>=1.40)
  2. - NVIDIA Container Toolkit(https://github.com/NVIDIA/nvidia-docker)

Getting started:

  1. 1. Edit "env" file.
  2. 2. Run ./build_jupyter_kernel.sh and wait(~1 hour, depends on your network speed).
  3. 3. Run ./run_jupyter.sh to lauch standalone instance of Jupyter Lab.

Description:

  1. JupyterLab container builds in two steps:
  2. 1. Building of base image(via Dockerfile.KernelBase): Installs all reqiured system libs(and much more, I should reduce the apps list in future), launch settings(taking from jupyter docker-stacks) and miniconda. It bases on nvidia/cuda:${CUDA_VERSION}-cudnn${CUDNN_VERSION}-devel-ubuntu18.04 image. Docker image size after build ~ 7Gb.
  3. 2. Building of main image(via Dockerfile.DLkernel): Install conda enviroment(located in ./resources/conda-env-py3.yml) and JupyterLab extensions(located inside docker file). So, you can change what you want before the build.
  4. You can controll build procedure (if something will go wrong) by editing build_jupyter_kernel.sh(just comment corresponding line).
  5. run_jupyter.sh - is example of launch of JupyterLab standalone. For using it with JupyterHub refer to docker-stacks instructions(by link above).