Skills
Nemo-Skills is a project designed to enhance the capabilities of large language models.
- Type
- Agent Skill
- Open source
- Yes
- GitHub Stars
- ★ 1.0k
- Source
- skill-github
- Repository
- github.com/NVIDIA-NeMo/Skills
Overview
Nemo-Skills provides a suite of pipelines to improve large language models (LLMs). It supports all requirements from synthetic data generation and model training to evaluation on various benchmarks. Development can begin on a local workstation and seamlessly scale to large-scale Slurm clusters with a single line of code change. The platform offers flexible LLM inference, model evaluation, and model training capabilities, making it ideal for developers aiming to boost LLM performance.
Capabilities
- ▪Seamlessly switch between API providers and local servers
- ▪Support multiple model hosting options
- ▪Scale SDG tasks to thousands of GPUs
- ▪Multi-domain model evaluation
- ▪Parallelized evaluation
- ▪Train models using NeMo-RL or verl
Use cases
Setup
Refer to the official documentation for installation and configuration: https://github.com/NVIDIA-NeMo/Skills
This information was compiled by AI from public sources and may contain inaccuracies — please refer to the source.
FAQ
How do I get started with Nemo-Skills?
Please visit the GitHub repository and follow the installation and configuration steps in the README.
Which model hosting options does Nemo-Skills support?
Supports TensorRT-LLM, vLLM, sglang, and Megatron.
Related skills
gemini-cli
Gemini CLI is an open-source AI agent that brings Gemini's powerful capabilities directly into the terminal.
bioSkills
A skill set for AI coding agents performing bioinformatics tasks.
planning-with-files
File-based persistent planning skill, ideal for AI coding agents and long-running tasks.
scientific-agent-skills
Transform any AI agent into a scientific assistant with 147 ready-to-use research skills.
susi_alexa_skill
A skill that enables question-and-answer interactions between Alexa and Susi AI.
claude-context
Provides code search capabilities for Claude Code, turning the entire codebase into context.