Support one to many mapping?
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.6k
- Forks
- 206
- PR merge metrics
- No merged PRs in 30d
Description
Hi,
I have a Mosaic ML streaming dataset. Each sample in my dataset has shape (Time, N, 3).
What the model needs for training are pairs ((N, 3) , (N, 3)). To do this I extract the pairs using the time dimension. I have a function extract() which returns an iterator of pairs from a single (Time, N, 3) input.
How would I combine this with the StreamingDataset / StreamingDataloader? I want to be able to do the extraction() as part of the data processing pipeline, so it can happen in different worker threads, and do not block the main thread which is busy training.
Is this supported / how should one go about implementing this?
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 by reading the StreamingDataset and StreamingDataloader entry points and compare them with the issue's extract() iterator shape. Determine whether worker-side one-to-many expansion is supported; done means a documented implementation path or focused feature specification explaining how pairs are emitted without blocking training.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- data-engineering, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100