Skip to content
This repository was archived by the owner on Jun 4, 2025. It is now read-only.

Conversation

@natuan
Copy link

@natuan natuan commented Oct 6, 2021

When a model is trained with modifiers such as QAT or layer pruning, we save the entire recipe containing them and apply it again when the model is loaded (before applying a new recipe such as pruning on top of it). This change is to limit the saving to only modifiers that change the architecture of the models.

@natuan natuan requested review from a team, bfineran, markurtz and spacemanidol October 6, 2021 06:29
Copy link

@spacemanidol spacemanidol left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good

Copy link
Member

@markurtz markurtz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good for now, I do think there's more we can do here to generalize, though, so let's sync and continue iterating

@markurtz markurtz merged commit aa5128f into master Oct 8, 2021
KSGulin pushed a commit that referenced this pull request Mar 9, 2022
* Only save modifiers that change the model

* Only save arch changing modifiers with the model
@dbogunowicz dbogunowicz deleted the simplify_saved_recipes branch December 5, 2023 10:28
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants