fastai-docker

Docker Container supporting the PyTorch based fast.ai library and Jupyter notebooks

This docker is designed for development not deployment. During build no code is copied to the container, instead it mounts the current repository directory as /local.

CPU

A CPU only version specifically for use on non-GPU backend servers, MacOS or Windows desktops.

cd cpu

build:

make build

launch jupyter:

make run

bash

make bash

GPU

For linux machines. Requires a GPU, CUDA and nvidia docker. Currently this image is build assuming CUDA 10.1. To work with other versions, modify the xx.x in this first line of the GPU Dockerfile: FROM nvidia/cuda:xx.x-cudnn7-devel-ubuntu18.04

cd gpu

build:

make build

launch jupyter:

make run

bash

make bash