lmms-finetune
A unified code repository for fine-tuning large multimodal models
- Provide a unified fine-tuning framework to simplify the integration and fine-tuning process
- Support multiple fine-tuning strategies such as full fine-tuning, lora, q-lora, etc
- Maintain the simplicity of the code repository for easy understanding and modification
- Supports multiple types of LMMs, including single image models, multi image/interleaved image models, and video models
- Provide detailed documentation and examples to help users quickly get started
- Flexible code repository, supporting customization and quick experimentation
Product Details
LMMS finetune is a unified code repository designed to simplify the fine-tuning process of large multimodal models (LMMs). It provides a structured framework that allows users to easily integrate the latest LMMs and make fine-tuning, supporting full fine-tuning and strategies such as Lora. The code library design is simple and lightweight, easy to understand and modify, and supports multiple models including LLaVA-1.5, Phi-3-Vision, Qwen VL Chat, LLaVA NeXT Interleaf, and LLaVA NeXT Video.
