NVIDIA-NeMo / NVIDIA-NeMo/Curator

tutorials/nemo-retriever-synthetic-data-generation: ETA for Training Data Generation Example

Open
#989 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

community-request enhancement waiting-on-maintainers
Dominant language
Python
Stars
1.8k
Forks
328
Avg merge
4d 5h
Merged PRs (30d)
30

Description

https://github.com/NVIDIA-NeMo/Curator/tree/v0.9.0/tutorials/nemo-retriever-synthetic-data-generation

This version supports the generation of evaluation datasets, creating synthetic benchmark datasets compatible with commonly used evaluation frameworks such as BEIR. Synthetic training dataset generation will be supported in an upcoming version.


Hello @vinay-raman

Could you please help clarify the following:

  • What are the key differences between synthetic datasets intended for evaluation and those intended for training in the context of embedding-model fine-tuning?
  • For the upcoming synthetic training dataset feature, what format do you plan to produce (e.g., positive/negative pairs, triplets with hard negatives, qrels), and will example notebooks or tutorials be provided?
  • Is there an estimated timeline for when the synthetic training dataset generation example will be available?

Thank you very much for your guidance and support.

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 with the linked tutorials/nemo-retriever-synthetic-data-generation example at the v0.9.0 tag and compare its evaluation-dataset scope with the requested training-data use case. The issue is complete only when the project provides or documents the training-data format, example materials, and availability timeline; no implementation entry point is specified.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
data, machine-learning
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.