-
-
Notifications
You must be signed in to change notification settings - Fork 16
Dataset/dataloader APIΒ #174
Copy link
Copy link
Open
Labels
Type: Improvement πPerformance improvement not introducing a new feature or requiring a major refactorPerformance improvement not introducing a new feature or requiring a major refactorType: New Feature βIntroduction of a completely new addition to the codebaseIntroduction of a completely new addition to the codebase
Milestone
Description
Feature Description
Add helper class that to make easier data loading, batching, shuffling, etc.
I.e. dataset and/or dataloader class API that would help user to prepare data for training.
Use cases:
- Easify and standardize data preparation on the client side, across all libraries.
- Supply dataset into
Job.train(..., {data: dataset, ...})and update Job.train to work with the dataset API.
Design: TBD
Is your feature request related to a problem?
Standardize the way data is loaded into training loop across workers.
What alternatives have you considered?
Discussed with FL team.
Additional Context
See also #172
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
Type: Improvement πPerformance improvement not introducing a new feature or requiring a major refactorPerformance improvement not introducing a new feature or requiring a major refactorType: New Feature βIntroduction of a completely new addition to the codebaseIntroduction of a completely new addition to the codebase