NVIDIA-NeMo / NVIDIA-NeMo/RL

[TRAIN-5] Example configs are too fragmented; no canonical config per model family

Open
#2,721 2 comments 0 reactions 1 assignee Claimed by @sharonyu-115 View on GitHub
Documentation Feature
Dominant language
Python
Stars
2k
Forks
561
Avg merge
4d 5h
Merged PRs (30d)
145

Description

## Problem

The example configs consist of many near-identical recipe files that differ only in minor parallelism settings. There is no canonical config per model family. This makes it difficult to know which config to use as a starting point.

## Suggested fix

Reorganize configs around one canonical recipe per model family (e.g., one for Qwen3.5-0.8B, one for Qwen3.5-9B). Express parallelism choices as documented override examples rather than duplicated config files.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.