ram-lexsi/safetune-testrun-unlearn_simdpo
ram-lexsi/safetune-testrun-unlearn_simdpo is a 1.5 billion parameter causal language model built from Qwen/Qwen2.5-1.5B-Instruct. Developed by Lexsi Labs using the SafeTune library, this model was created with the SimDPOTrainer method. It is specifically designed to explore and demonstrate LLM safety methods, including unlearning and hardening techniques. This model serves as an artifact for evaluating and understanding safety-focused fine-tuning processes.
Loading preview...
Model Overview
ram-lexsi/safetune-testrun-unlearn_simdpo is a 1.5 billion parameter language model developed by Lexsi Labs. It is built upon the Qwen/Qwen2.5-1.5B-Instruct base model and was fine-tuned using the SimDPOTrainer method.
Key Characteristics
- Origin: Developed as part of the SafeTune library, which focuses on LLM safety methods.
- Purpose: This model is an artifact from a test run, specifically designed to demonstrate and evaluate unlearning techniques within the SafeTune framework.
- Base Model: Utilizes the Qwen2.5-1.5B-Instruct architecture, providing a robust foundation for safety-oriented experimentation.
Intended Use
This model is primarily intended for researchers and developers interested in:
- Exploring and understanding LLM safety methods, particularly unlearning and hardening.
- Evaluating the effectiveness of the SafeTune library and its various techniques.
- Studying the impact of SimDPOTrainer on model behavior and safety properties.