Skip to content
This repository has been archived by the owner on Oct 9, 2023. It is now read-only.

Simplify InputTransform hooks #924

Closed
ethanwharris opened this issue Nov 3, 2021 · 0 comments · Fixed by #1010
Closed

Simplify InputTransform hooks #924

ethanwharris opened this issue Nov 3, 2021 · 0 comments · Fixed by #1010
Assignees
@tchaton
Labels
Data Pipeline V2 Refactor (Functional)

Comments

@ethanwharris
Copy link
Collaborator

The pre_tensor_transform, to_tensor_transform, post_tensor_transform hook flow is confusing and not always honoured by our tasks. This would simplify the InputTransform hooks to be just per_sample_transform and per_batch_transform on or off device (so 4 hooks in total).

It might be possible to do this and still be backwards compatible by just converting the pre, to, and post hooks into a single sequential, but we may just want to have an error instead.
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees

@tchaton tchaton

Labels
Data Pipeline V2 Refactor (Functional)
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

DataPipeline Refactor V2: Simplify Input Transform Lightning-Universe/lightning-flash
2 participants
@ethanwharris @tchaton