# NVIDIA-NeMo/Skills

A project to improve skills of large language models

Repository: https://github.com/NVIDIA-NeMo/Skills
Canonical: https://ross.abutalabs.com/products/nvidia-nemo-skills
Homepage: https://nvidia-nemo.github.io/Skills/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T11:54:00+00:00

## Health v2 (maintenance only)
Score: 70/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 35, longevity 66
- inputs: {"age_days": 934, "days_push": 7, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1031, forks 198 (observed 2026-08-28T04:03:18.089524+00:00)

## What it is
Nemo-Skills is a collection of Python pipelines for improving the skills of large language models, covering synthetic data generation, model training, and evaluation on a wide range of benchmarks. It scales seamlessly from a local workstation to large Slurm clusters with a one-line configuration change.

## Use cases
- generate synthetic training data for LLMs at scale
- evaluate LLMs on math, code, and science benchmarks
- run LLM inference on a Slurm cluster with vLLM or TensorRT-LLM
- fine-tune and train large language models
- benchmark models on AIME, SWE-bench, GPQA and other evals
- scale data generation jobs from one GPU to thousands of GPUs

## When to choose
- you need end-to-end LLM development pipelines from data generation to training to evaluation
- you run workloads on Slurm HPC clusters and want scalable LLM jobs
- you want to evaluate models across many diverse benchmarks with parallelization
- you need to switch between API providers and self-hosted inference servers easily

## When to avoid
- you only need simple single-GPU inference without cluster orchestration
- you want lightweight benchmark harnesses outside the NVIDIA ecosystem
- you need benchmark rollouts specifically, since those are moving to NeMo-Gym

## Facets
- artifact type: framework
- maturity: active
- function: llm-inference, llm-training, machine-learning, benchmarking, data-generation, etl
- domain: large-language-models, machine-learning, artificial-intelligence, gpu-computing, developer-tools
- platform: python, cli
- tags: slurm, synthetic-data-generation, model-evaluation, nemo, vllm, tensorrt-llm, pipelines, gpu, linux, docker

## Member repositories
- NVIDIA-NeMo/Skills (main) score 70

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:18.089524+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:07:24.724691+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA-NeMo/Skills (fetched 2026-08-28T04:03:18.089524+00:00, sha d9f5bd92dfee)
  - homepage: https://nvidia-nemo.github.io/Skills/ (fetched 2026-08-29T13:06:54.410899+00:00, sha e69048a44ed3)
- Data as of 2026-08-30T08:39:29.467469+00:00.