Closed hihunjin closed 3 years ago
Here S3 is for remote storage of the trained models. It can be substituted with any of the remote storage. DVC supports the following ones: Remote storages. Based on your requirement you can change the dockerfile. Once it is done, the Github Action should contain necessary credentials to fetch the model from remote location.
Oh sorry for the misunderstanding.
In this Deployment section, you said, I can use on-prem. Is there any guidance for this?
If I understand correctly, you want to store the models in your machine instead of S3 and want github actions to read the model from the machine. If your local machine is hosted somewhere means, you can use the SSH
remote config in DVC. And in github actions you can do the same by configuring the credentials. If the local machine is a personal system that is not hosted anywhere means you can use the local remote
config in DVC. For this, I am quite sure about configuring Github Action.
Mhm That is not what I want to do.
I want to use my own cpu, ram machines (let's say on-prems) when model is running. Storage is also in my machine.
You can use local remote for that in dvc. Configuring github action would be challenging task to do. Since it needs to access your machine to download the model and create the docker image. You can try a different hook instead of on: [push]
hook. There are pre-commit hooks that will run before pushing to github. That will create the Docker image directly from the local model.
I see. Thanks a lot.
Then how can I configure my github action to give a permission to access my machine? is it possible? If there's any reference, please share it.
That you need to explore. I haven't worked on it. Either try working on it (or) you can create the docker image yourself or using some script (makefile) and then continue with other steps.
I have my own local machine and I'd like to substitute what AWS S3 bucket does into my machine.
From my thought, the followings are the steps
Would you give me any advice/article to read?