Starstruk Logo MarkStarstruk

v0.1.6

hiyouga/LlamaFactoryv0.1.6Aug 11, 2023by hiyouga

AI Summary

Adds Direct Preference Optimization (DPO) training and expands model support to Qwen and XVERSE.

Key Highlights

  • DPO training integration from TRL library
  • Support for Qwen-7B, XVERSE-13B, and ChatGLM2-6B
  • Safe ChatML template for Qwen-7B-Chat

New Features

  • DPO training implementation
  • Qwen-7B/XVERSE-13B/ChatGLM2-6B fine-tuning
  • ChatML template support
  • Improved Web UI

Full Release Notes

- Adapt **[DPO training](https://arxiv.org/abs/2305.18290)** from the [TRL](https://github.com/huggingface/trl) library
- Support fine-tuning the Qwen-7B, Qwen-7B-Chat, XVERSE-13B, and ChatGLM2-6B models
- Implement the "safe" [ChatML template](https://github.com/openai/openai-python/blob/main/chatml.md) for Qwen-7B-Chat
- Better Web UI
- Pretty readme by @codemayq #382 
- New features: #395 #451 
- Fix InternLM-7B inference #312 
- Fix bugs: #351 #354 #361 #376 #408 #417 #420 #423 #426