iterative / vscode-dvc

Machine learning experiment tracking and data versioning with DVC extension for VS Code
https://marketplace.visualstudio.com/items?itemName=Iterative.dvc
Apache License 2.0
193 stars 29 forks source link

Feature enhacement : pipelines in DVC VS code extension #4242

Open SoyGema opened 1 year ago

SoyGema commented 1 year ago

Motivation

Proposal

I'm sure that you can improve this one , and not fully convinced myself, but I shall put maybe a diagram or animation about the relationship in between the 3 building blocks

https://github.com/iterative/vscode-dvc/assets/24204714/b26ca3a5-1e31-4760-a928-0df3c1ecf81e

Thanks for the code snippets and stages structure proposal for dvc.yaml file in #4233 Related to #4194

### Tasks
mattseddon commented 1 year ago

@SoyGema would a button that could generate the DAG be useful from this screen?

I think if we were going to use an animation like the one shown in the screen grab we would need to swap the positions of params.yaml and load_dataset.py. I think we would also want to stick with a train example as that is the DVCLive onboarding path that we've been focused on up until this point.

SoyGema commented 1 year ago

Hey @mattseddon ! Thanks for the attention and the quick response 🙏

1. Would a button that could generate the DAG be useful from this screen? Yes, it would, once the pipeline is configured. In fact, it's pretty coherent with the DVC done configuration scenario moment in the UI . Once the DAG might be created, would you please consider reinforcing that it is done or give some stamina to the user for running the experiment ?

2. Swaps. Makes sense to me, especially given that the params.yml file will be the file the user will Iterate from. I would consider automating the params.yml creation as well, or even creating it when the user is focusing on the dvc.yamlfile. Please note that the animation is an idea, and you might want to test another approaches ... If so, it would be amazing that it still distills the essence : improve high-level understanding for the user about pipelines and the relationship between code , dvc.yaml, and params.yaml file without leaving the IDE.

See you in the Open! 🛰️

mattseddon commented 1 year ago

We will get to this but we don't have capacity right now 🙏🏻. Thank you @SoyGema.

dberenbaum commented 1 year ago

We should consider adding a visual like this to the dvc pipeline docs. If we can include hydra or other advanced config as well, that would be amazing.