IMHO if the goal is to facilitate understanding then the setup phase is particularly important as it is a barrier to entry.
There are numerous solutions to overcome that, from containerization, e.g Docker or Podman, to dedicated AI ones e.g cog or HuggingFace with a gradio interface, to finally a Jupyter Notebook that could itself be part of an online setup, e.g Gitpod.
I believe providing a Dockerfile with instructions on how to do the most basic steps, e.g training or a single inference, would help newcomers who can then focus on the actual core of the project.
IMHO if the goal is to facilitate understanding then the setup phase is particularly important as it is a barrier to entry.
There are numerous solutions to overcome that, from containerization, e.g Docker or Podman, to dedicated AI ones e.g cog or HuggingFace with a gradio interface, to finally a Jupyter Notebook that could itself be part of an online setup, e.g Gitpod.
I believe providing a
Dockerfilewith instructions on how to do the most basic steps, e.g training or a single inference, would help newcomers who can then focus on the actual core of the project.