NVIDIA-NeMo / NVIDIA-NeMo/Curator
[Ray] Use PyTorch `DataLoader` for model inference
Open
@sarahyurick is already working on this.
Since Aug 4, 2025.
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 328
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 30
Description
As a follow-up to https://github.com/NVIDIA-NeMo/Curator/pull/753, we should look into using the PyTorch DataLoader class for enhanced performance at the model inference stage.
This works by allowing the next batch(es) to be loaded into memory asynchronously while running inference on the current batch.
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.
Assessment
This issue has not been assessed yet.