@MarcYin suggested to use Alpine as a first image, rather than Ubuntu to simplify image creation. I have also noticed that most of the storage in the Ubuntu install is there for latex used by pandoc in turn used by nbconvert, so I have created a new Dockerfile that extends the Alpine base with the Jupyter stacks.
This should result in a simpler image that is based off a standard small image, that should be directly available from CN docker mirrors.
@MarcYin suggested to use Alpine as a first image, rather than Ubuntu to simplify image creation. I have also noticed that most of the storage in the Ubuntu install is there for latex used by pandoc in turn used by nbconvert, so I have created a new Dockerfile that extends the Alpine base with the Jupyter stacks.
This should result in a simpler image that is based off a standard small image, that should be directly available from CN docker mirrors.