According to Jeremy’s post: (assume you have gpu with cuda, conda and git installed)

1. clone the git repo

git clone https://github.com/fastai/fastai.git

2. open your anaconda prompt and set up the conda env fastai

cd fastai
conda env update

3. type this to activate the environment every time
   in your anaconda prompt

activate fastai

also fine with

conda activate fastai

then you will see the prefix of each prompt turns to (fastai)

4. install ipython widgets into jupyter

jupyter nbextension enable --py widgetsnbextension --sys-prefix

5. before you replace the simulink (which was prepared for linux system) to fastai library, to remove it you requires admin.

deactivate

and safely close the prompt,
and then run an admin prompt by right clicking.

6. remove the simulink and create one fit for windows

cd path\to\fastai
cd courses\ml1
del fastai
mklink fastai ..\..\fastai
cd ..\..

7. all done!