You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Defining the workflows implicitly with dvc run is very convenient. However, defining the workflows explicitly adds a possibility for decoupling the workflows from specific repositories. Perhaps it is even possible to have a library of workflow templates. Pachyderm specifies the workflows in a json format, which is probably a good starting point.
The text was updated successfully, but these errors were encountered:
On second thought, the json format might not be a good idea. Luigi, which I'm not very familiar with, defines the dependency graph in Python. This is obviously more powerful and it is easier to make workflow templates in code.
That's funny. Right now @efiop is working on yaml file for dvc run. It is going to be an alternative to implicit dvc run python script.py .... We are going to support both.
Using Python as a format is probably not the best choice for DVC since DVC is language agnostic. Also, we preferred yaml over json because yaml is more human-readable (comments, aliases). While we keep json format for internal machine generated status files.
Defining the workflows implicitly with
dvc run
is very convenient. However, defining the workflows explicitly adds a possibility for decoupling the workflows from specific repositories. Perhaps it is even possible to have a library of workflow templates. Pachyderm specifies the workflows in a json format, which is probably a good starting point.The text was updated successfully, but these errors were encountered: