v0.1.11

oumi-ai/oumiv0.1.11Apr 6, 2025by oelachqar

AI Summary

Adds Llama 4 support, LoRA training for Phi3/Phi4/Qwen2.5-VL, introduces MLflow integration, and adds HallOumi inference notebooks.

Key Highlights

  • Integration of Llama 4 (Scout and Maverick).
  • LoRA training for Phi3, Phi4, and Qwen2.5-VL models.
  • MLflow integration for experiment tracking.
  • Dedicated HallOumi inference notebooks.

New Features

  • Llama 4 configurations.
  • Vision-Language Model (VLM) LoRA configs.
  • MLflow integration for experiment tracking.
  • HallOumi inference notebooks (generative and classifier).
  • CLI alias functionality enhancements.

Full Release Notes

# Oumi v0.1.11 Release Notes 🚀

## Key Highlights

### Model Support 🤖
- Integrated Llama 4 (Scout and Maverick variants) with complete workflow configs 🦙
- Added LoRA training for Phi3, Phi4, and Qwen2.5-VL multimodal models 🖼️

### Developer Experience 💻
- Introduced MLflow integration for experiment tracking 📝
- Enhanced CLI with convenient alias functionality ⌨️

### HallOumi Framework 🧠
- Added examples for Halloumi
- Added dedicated inference notebooks for both generative and classifier approaches 📓

Welcome to our new contributors @hommayushi3 and @gabrielaugz! 👋

For details, see the [[full changelog](https://github.com/oumi-ai/oumi/compare/v0.1.10...v0.1.11)](https://github.com/oumi-ai/oumi/compare/v0.1.10...v0.1.11).