Closed leonardcser closed 1 year ago
thanks @leonardcser for the effort!
Modularise code into functions
I'm not sure how far we want to go with this, tbh. On one hand this is the exactly right way to do things from the software eng perspective, but on the other hand even the need for these functions means that code is quite complicated for a DS project (should we try to simplify it?), and it makes it farther away from a quite regular (for good or bad) notebook-style flow ... what do you think @daavoo @dberenbaum ?
what do you think @daavoo @dberenbaum ?
I think the src
in this P.R. looks more like the state I would expect for people using a dvc.yaml
with isolated scripts per stage.
and it makes it farther away from a quite regular (for good or bad) notebook-style flow
I think the motivation is not good enough for trying to make artificially "messy" scripts.
We could have an actual notebook in the project like in example-get-started-experiments and acknowledge that you are expected to clean up and refactor the code when wanting to use dvc.yaml
, like we acknowledge in https://dvc.org/doc/start/experiments/experiment-pipelines#stepping-up-and-out-of-the-notebook
means that code is quite complicated for a DS project (should we try to simplify it?),
Which parts look complicated? It doesn't look complicated for a DS project to me.
The reason for the PR is to make the source code clearer for the
example-get-started
while keeping it as simple as possible. The actual code was left unchanged.Here are my changes for a more understandable starter source code: