meta-pytorch / meta-pytorch/data

DataPipes for Error Handling and Retry

Open
#962 1 comment 5 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

feature
Dominant language
Python
Stars
1.3k
Forks
179
Avg merge
6d 1h
Merged PRs (30d)
2

Description

🚀 The feature

Provide a mechanism to catch exception raised by a previous DataPipe and retry. These can be related but separate DataPipes. Non-DataPipes implementation should also be considered.

Motivation, pitch

Currently, there is no standard mechanism to handle exception from a source DataPipe. One example of such issues is raised on the PyTorch forum.

Alternatives

Given how iterators are implemented, it is unclear to me if this is feasible as a DataPipe implementation. We may have to modify the iterator wrapper within _hook_iterator.py (inside PyTorch core) to make this feature possible.

We can also consider modifying specific DataPipes where custom error handling and retries are frequently needed .

Additional context

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by examining the DataPipe iterator behavior and the iterator wrapper in _hook_iterator.py. Determine whether exception catching and retry can be supported as related DataPipes or requires changes in PyTorch core, while considering non-DataPipes implementations. Done requires a defined approach for handling exceptions from a source DataPipe and retrying it.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
data
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.