[FR] Engine with data echoing/minibatch persistency/BA
Open
Nobody has claimed this yet.
enhancement
help wanted
- Dominant language
- Python
- Stars
- 4.8k
- Forks
- 726
- Avg merge
- 5d 21h
- Merged PRs (30d)
- 5
Description
Idea is to propose 2 things:
- a supervised trainer with minibatch persistency [3] (configurable K = same batch reusage)
- a handler that can be used inside user's
update_fnof the trainer and implementing- minibatch persistency [3] or
- Batch Augmentation [1]
Refs:
[1] https://arxiv.org/abs/1901.09335
[2] https://arxiv.org/abs/1907.05550
[3] https://arxiv.org/abs/1806.07353
cc @ericspod
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the trainer's user-supplied update_fn and the three referenced papers, especially the minibatch persistency and batch augmentation definitions. Clarify whether the goal is a supervised trainer, an update_fn handler, or both, and specify configurable K and expected behavior before implementation. Done should mean the behavior for minibatch reuse and batch augmentation is agreed and implemented.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100