back to home

oumi-ai / oumi

Easily fine-tune, evaluate and deploy gpt-oss, Qwen3, DeepSeek-R1, or any open source LLM / VLM!

8,861 stars
700 forks
10 issues
PythonJupyter NotebookShell
Chat with CodebaseArchitecture ScanSecurity AuditExplain Codebase

AI Architecture Analysis

This repository is indexed by RepoMind. By analyzing oumi-ai/oumi in our AI interface, you can instantly generate complete architecture diagrams, visualize control flows, and perform automated security audits across the entire codebase.

Our Agentic Context Augmented Generation (Agentic CAG) engine loads full source files into context, avoiding the fragmentation of traditional RAG systems. Ask questions about the architecture, dependencies, or specific features to see it in action.

Click here to launch the interactive analysis workspace

Embed this Badge

Showcase RepoMind's analysis directly in your repository's README.

[![Analyzed by RepoMind](https://img.shields.io/badge/Analyzed%20by-RepoMind-4F46E5?style=for-the-badge)](https://repomind-ai.vercel.app/repo/oumi-ai/oumi)
Preview:Analyzed by RepoMind

Repository Summary (README)

Preview

Oumi Logo

Documentation Blog Twitter Discord PyPI version License Tests GPU Tests GitHub Repo stars Code style: black pre-commit About

Everything you need to build state-of-the-art foundation models, end-to-end

<p align="center"> <a href="https://trendshift.io/repositories/12865"> <img alt="GitHub trending" src="https://trendshift.io/api/badge/repositories/12865" /> </a> </p>

🔥 News

🔎 About

Oumi is a fully open-source platform that streamlines the entire lifecycle of foundation models - from data preparation and training to evaluation and deployment. Whether you're developing on a laptop, launching large scale experiments on a cluster, or deploying models in production, Oumi provides the tools and workflows you need.

With Oumi, you can:

  • 🚀 Train and fine-tune models from 10M to 405B parameters using state-of-the-art techniques (SFT, LoRA, QLoRA, GRPO, and more)
  • 🤖 Work with both text and multimodal models (Llama, DeepSeek, Qwen, Phi, and others)
  • 🔄 Synthesize and curate training data with LLM judges
  • ⚡️ Deploy models efficiently with popular inference engines (vLLM, SGLang)
  • 📊 Evaluate models comprehensively across standard benchmarks
  • 🌎 Run anywhere - from laptops to clusters to clouds (AWS, Azure, GCP, Lambda, and more)
  • 🔌 Integrate with both open models and commercial APIs (OpenAI, Anthropic, Vertex AI, Together, Parasail, ...)

All with one consistent API, production-grade reliability, and all the flexibility you need for research.

Learn more at oumi.ai, or jump right in with the quickstart guide.

🚀 Getting Started

NotebookTry in ColabGoal
🎯 Getting Started: A Tour<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - A Tour.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>Quick tour of core features: training, evaluation, inference, and job management
🔧 Model Finetuning Guide<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - Finetuning Tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>End-to-end guide to LoRA tuning with data prep, training, and evaluation
📚 Model Distillation<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - Distill a Large Model.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>Guide to distilling large models into smaller, efficient ones
📋 Model Evaluation<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - Evaluation with Oumi.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>Comprehensive model evaluation using Oumi's evaluation framework
☁️ Remote Training<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - Running Jobs Remotely.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>Launch and monitor training jobs on cloud (AWS, Azure, GCP, Lambda, etc.) platforms
📈 LLM-as-a-Judge<a target="_blank" href="https://colab.research.google.com/github/oumi-ai/oumi/blob/main/notebooks/Oumi - Simple Judge.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>Filter and curate training data with built-in judges

🔧 Usage

Installation

Choose the installation method that works best for you:

<details open> <summary><b>Using pip (Recommended)</b></summary> 
# Basic installation
uv pip install oumi

# With GPU support
uv pip install 'oumi[gpu]'

# Latest development version
uv pip install git+https://github.com/oumi-ai/oumi.git

Don't have uv? Install it or use pip instead.

</details> <details> <summary><b>Using Docker</b></summary> 
# Pull the latest image
docker pull ghcr.io/oumi-ai/oumi:latest

# Run oumi commands
docker run --gpus all -it ghcr.io/oumi-ai/oumi:latest oumi --help

# Train with a mounted config
docker run --gpus all -v $(pwd):/workspace -it ghcr.io/oumi-ai/oumi:latest \
    oumi train --config /workspace/my_config.yaml
</details> <details> <summary><b>Quick Install Script (Experimental)</b></summary>

Try Oumi without setting up a Python environment. This installs Oumi in an isolated environment:

curl -LsSf https://oumi.ai/install.sh | bash
</details>

For more advanced installation options, see the installation guide.

Oumi CLI

You can quickly use the oumi command to train, evaluate, and infer models using one of the existing recipes:

# Training
oumi train -c configs/recipes/smollm/sft/135m/quickstart_train.yaml

# Evaluation
oumi evaluate -c configs/recipes/smollm/evaluation/135m/quickstart_eval.yaml

# Inference
oumi infer -c configs/recipes/smollm/inference/135m_infer.yaml --interactive

For more advanced options, see the training, evaluation, inference, and llm-as-a-judge guides.

Running Jobs Remotely

You can run jobs remotely on cloud platforms (AWS, Azure, GCP, Lambda, etc.) using the oumi launch command:

# GCP
oumi launch up -c configs/recipes/smollm/sft/135m/quickstart_gcp_job.yaml

# AWS
oumi launch up -c configs/recipes/smollm/sft/135m/quickstart_gcp_job.yaml --resources.cloud aws

# Azure
oumi launch up -c configs/recipes/smollm/sft/135m/quickstart_gcp_job.yaml --resources.cloud azure

# Lambda
oumi launch up -c configs/recipes/smollm/sft/135m/quickstart_gcp_job.yaml --resources.cloud lambda

Note: Oumi is in <ins>beta</ins> and under active development. The core features are stable, but some advanced features might change as the platform improves.

💻 Why use Oumi?

If you need a comprehensive platform for training, evaluating, or deploying models, Oumi is a great choice.

Here are some of the key features that make Oumi stand out:

  • 🔧 Zero Boilerplate: Get started in minutes with ready-to-use recipes for popular models and workflows. No need to write training loops or data pipelines.
  • 🏢 Enterprise-Grade: Built and validated by teams training models at scale
  • 🎯 Research Ready: Perfect for ML research with easily reproducible experiments, and flexible interfaces for customizing each component.
  • 🌐 Broad Model Support: Works with most popular model architectures - from tiny models to the largest ones, text-only to multimodal.
  • 🚀 SOTA Performance: Native support for distributed training techniques (FSDP, DeepSpeed, DDP) and optimized inference engines (vLLM, SGLang).
  • 🤝 Community First: 100% open source with an active community. No vendor lock-in, no strings attached.

📚 Examples & Recipes

Explore the growing collection of ready-to-use configurations for state-of-the-art models and training workflows:

Note: These configurations are not an exhaustive list of what's supported, simply examples to get you started. You can find a more exhaustive list of supported models, and datasets (supervised fine-tuning, pre-training, preference tuning, and vision-language finetuning) in the oumi documentation.

Qwen Family

ModelExample Configurations
Qwen3-Next 80B A3BLoRA • Inference • Inference (Instruct) • Evaluation
Qwen3 30B A3BLoRA • Inference • Evaluation
Qwen3 32BLoRA • Inference • Evaluation
Qwen3 14BLoRA • Inference • Evaluation
Qwen3 8BFFT • Inference • Evaluation
Qwen3 4BFFT • Inference • Evaluation
Qwen3 1.7BFFT • Inference • Evaluation
Qwen3 0.6BFFT • Inference • Evaluation
QwQ 32BFFT • LoRA • QLoRA • Inference • Evaluation
Qwen2.5-VL 3BSFT • LoRA• Inference (vLLM) • Inference
Qwen2-VL 2BSFT • LoRA • Inference (vLLM) • Inference (SGLang) • Inference • Evaluation

🐋 DeepSeek R1 Family

ModelExample Configurations
DeepSeek R1 671BInference (Together AI)
Distilled Llama 8BFFT • LoRA • QLoRA • Inference • Evaluation
Distilled Llama 70BFFT • LoRA • QLoRA • Inference • Evaluation
Distilled Qwen 1.5BFFT • LoRA • Inference • Evaluation
Distilled Qwen 32BLoRA • Inference • Evaluation

🦙 Llama Family

ModelExample Configurations
Llama 4 Scout Instruct 17BFFT • LoRA • QLoRA • Inference (vLLM) • Inference • Inference (Together.ai)
Llama 4 Scout 17BFFT
Llama 3.1 8BFFT • LoRA • QLoRA • Pre-training • Inference (vLLM) • Inference • Evaluation
Llama 3.1 70BFFT • LoRA • QLoRA • Inference • Evaluation
Llama 3.1 405BFFT • LoRA • QLoRA
Llama 3.2 1BFFT • LoRA • QLoRA • Inference (vLLM) • Inference (SGLang) • Inference • Evaluation
Llama 3.2 3BFFT • LoRA • QLoRA • Inference (vLLM) • Inference (SGLang) • Inference • Evaluation
Llama 3.3 70BFFT • LoRA • QLoRA • Inference (vLLM) • Inference • Evaluation
Llama 3.2 Vision 11BSFT • Inference (vLLM) • Inference (SGLang) • Evaluation

🦅 Falcon family

ModelExample Configurations
Falcon-H1FFT • Inference • Evaluation
Falcon-E (BitNet)FFT • DPO • Evaluation

💎 Gemma 3 Family

ModelExample Configurations
Gemma 3 4B InstructFFT • Inference • Evaluation
Gemma 3 12B InstructLoRA • Inference • Evaluation
Gemma 3 27B InstructLoRA • Inference • Evaluation

🦉 OLMo 3 Family

ModelExample Configurations
OLMo 3 7B InstructFFT • Inference • Evaluation
OLMo 3 32B InstructLoRA • Inference • Evaluation

🎨 Vision Models

ModelExample Configurations
Llama 3.2 Vision 11BSFT • LoRA • Inference (vLLM) • Inference (SGLang) • Evaluation
LLaVA 7BSFT • Inference (vLLM) • Inference
Phi3 Vision 4.2BSFT • LoRA • Inference (vLLM)
Phi4 Vision 5.6BSFT • LoRA • Inference (vLLM) • Inference
Qwen2-VL 2BSFT • LoRA • Inference (vLLM) • Inference (SGLang) • Inference • Evaluation
Qwen3-VL 2BInference
Qwen3-VL 4BInference
Qwen3-VL 8BInference
Qwen2.5-VL 3BSFT • LoRA• Inference (vLLM) • Inference
SmolVLM-Instruct 2BSFT • LoRA

🔍 Even more options

This section lists all the language models that can be used with Oumi. Thanks to the integration with the 🤗 Transformers library, you can easily use any of these models for training, evaluation, or inference.

Models prefixed with a checkmark (✅) have been thoroughly tested and validated by the Oumi community, with ready-to-use recipes available in the configs/recipes directory.

<details> <summary>📋 Click to see more supported models</summary>

Instruct Models

ModelSizePaperHF HubLicenseOpen 1
✅ SmolLM-Instruct135M/360M/1.7BBlogHubApache 2.0✅
✅ DeepSeek R1 Family1.5B/8B/32B/70B/671BBlogHubMIT❌
✅ Llama 3.1 Instruct8B/70B/405BPaperHubLicense❌
✅ Llama 3.2 Instruct1B/3BPaperHubLicense❌
✅ Llama 3.3 Instruct70BPaperHubLicense❌
✅ Phi-3.5-Instruct4B/14BPaperHubLicense❌
✅ Qwen30.6B-32BPaperHubLicense❌
Qwen2.5-Instruct0.5B-70BPaperHubLicense❌
OLMo 2 Instruct7BPaperHubApache 2.0✅
✅ OLMo 3 Instruct7B/32BPaperHubApache 2.0✅
MPT-Instruct7BBlogHubApache 2.0✅
Command R35B/104BBlogHubLicense❌
Granite-3.1-Instruct2B/8BPaperHubApache 2.0❌
Gemma 2 Instruct2B/9BBlogHubLicense❌
✅ Gemma 3 Instruct4B/12B/27BBlogHubLicense❌
DBRX-Instruct130B MoEBlogHubApache 2.0❌
Falcon-Instruct7B/40BPaperHubApache 2.0❌
✅ Llama 4 Scout Instruct17B (Activated) 109B (Total)PaperHubLicense❌
✅ Llama 4 Maverick Instruct17B (Activated) 400B (Total)PaperHubLicense❌

Vision-Language Models

ModelSizePaperHF HubLicenseOpen
✅ Llama 3.2 Vision11BPaperHubLicense❌
✅ LLaVA-1.57BPaperHubLicense❌
✅ Phi-3 Vision4.2BPaperHubLicense❌
✅ BLIP-23.6BPaperHubMIT❌
✅ Qwen2-VL2BBlogHubLicense❌
✅ Qwen3-VL2B/4B/8BBlogHubLicense❌
✅ SmolVLM-Instruct2BBlogHubApache 2.0✅

Base Models

ModelSizePaperHF HubLicenseOpen
✅ SmolLM2135M/360M/1.7BBlogHubApache 2.0✅
✅ Llama 3.21B/3BPaperHubLicense❌
✅ Llama 3.18B/70B/405BPaperHubLicense❌
✅ GPT-2124M-1.5BPaperHubMIT✅
DeepSeek V27B/13BBlogHubLicense❌
Gemma22B/9BBlogHubLicense❌
GPT-J6BBlogHubApache 2.0✅
GPT-NeoX20BPaperHubApache 2.0✅
Mistral7BPaperHubApache 2.0❌
Mixtral8x7B/8x22BBlogHubApache 2.0❌
MPT7BBlogHubApache 2.0✅
OLMo1B/7BPaperHubApache 2.0✅
✅ Llama 4 Scout17B (Activated) 109B (Total)PaperHubLicense❌

Reasoning Models

ModelSizePaperHF HubLicenseOpen
✅ gpt-oss20B/120BPaperHubApache 2.0❌
✅ Qwen30.6B-32BPaperHubLicense❌
✅ Qwen3-Next80B-A3BBlogHubLicense❌
Qwen QwQ32BBlogHubLicense❌

Code Models

ModelSizePaperHF HubLicenseOpen
✅ Qwen2.5 Coder0.5B-32BBlogHubLicense❌
DeepSeek Coder1.3B-33BPaperHubLicense❌
StarCoder 23B/7B/15BPaperHubLicense✅

Math Models

ModelSizePaperHF HubLicenseOpen
DeepSeek Math7BPaperHubLicense❌
</details>

📖 Documentation

To learn more about all the platform's capabilities, see the Oumi documentation.

🤝 Join the Community

Oumi is a community-first effort. Whether you are a developer, a researcher, or a non-technical user, all contributions are very welcome!

  • To contribute to the oumi repository, please check the CONTRIBUTING.md for guidance on how to contribute to send your first Pull Request.
  • Make sure to join our Discord community to get help, share your experiences, and contribute to the project!
  • If you are interested in joining one of the community's open-science efforts, check out our open collaboration page.

🙏 Acknowledgements

Oumi makes use of several libraries and tools from the open-source community. We would like to acknowledge and deeply thank the contributors of these projects! ✨ 🌟 💫

📝 Citation

If you find Oumi useful in your research, please consider citing it:

@software{oumi2025,
  author = {Oumi Community},
  title = {Oumi: an Open, End-to-end Platform for Building Large Foundation Models},
  month = {January},
  year = {2025},
  url = {https://github.com/oumi-ai/oumi}
}

📜 License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

Footnotes

  1. Open models are defined as models with fully open weights, training code, and data, and a permissive license. See Open Source Definitions for more information. ↩