-
Notifications
You must be signed in to change notification settings - Fork 82
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ability to save pipelines using pickle #1956
Comments
The issue at hand is that Options:
|
@chukarsten suggested looking at defining However that does still leave us with another challenge. Currently we define the pipeline classes used in automl at runtime, and as @angela97lin said above, the original issue was that those definitions weren't being found/saved during the pickling, and that issue still needs a solution. |
We did some digging this morning and listed options. |
Next steps: open a spike PR, see what difficulties we encounter with updating our pipeline API. |
Background
The pipeline
save
andload
methods currently usecloudpickle
instead of pythonpickle
.We closed #1400 a couple months ago. But this led to a bug (#1912) where our code was setting class attributes which apply across all instances of our template classes.
Plan
The short-term resolution for #1912 is to revert the change for #1400, so that
get_pipelines
returns a pipeline instance which is notpickle
-able using pythonpickle
but also does not have the buggy side-effect behavior when it comes to the class attributes likename
andcomponent_graph
Once that's done, this issue tracks figuring out how we can modify our approach to support python
pickle
of our pipelines, and filing issues to execute that plan.The text was updated successfully, but these errors were encountered: