flammenai/Llama3.1-Flammades-70B
Llama3.1-Flammades-70B is a 70 billion parameter language model developed by flammenai, fine-tuned from nbeerbower/Llama3.1-Gutenberg-Doppel-70B. This model utilizes ORPO tuning over 3 epochs, leveraging datasets like flammenai/Date-DPO-NoAsterisks and jondurbin/truthy-dpo-v0.1. It features a 32768 token context length and is optimized for general language understanding and generation tasks, demonstrating an average score of 35.74 on the Open LLM Leaderboard.
Loading preview...
Llama3.1-Flammades-70B Overview
Llama3.1-Flammades-70B is a 70 billion parameter language model developed by flammenai, built upon the nbeerbower/Llama3.1-Gutenberg-Doppel-70B base model. It has been fine-tuned using the ORPO method over three epochs, leveraging two distinct datasets: flammenai/Date-DPO-NoAsterisks and jondurbin/truthy-dpo-v0.1. This specific tuning approach aims to enhance its performance across various language tasks.
Key Capabilities & Performance
This model demonstrates a solid foundation in general language understanding and generation, as indicated by its evaluation on the Open LLM Leaderboard. Notable performance metrics include:
- Average Score: 35.74
- IFEval (0-Shot): 70.58
- BBH (3-Shot): 52.55
- MMLU-PRO (5-shot): 41.69
While excelling in certain areas like instruction following (IFEval), it also provides a baseline for complex reasoning and knowledge-based tasks. Its 32768 token context length allows for processing longer inputs and generating more coherent, extended responses.
When to Use This Model
Llama3.1-Flammades-70B is suitable for developers seeking a robust 70B parameter model for a wide range of applications, particularly those benefiting from its ORPO-tuned characteristics. It can be a strong candidate for tasks requiring general conversational abilities, content generation, and understanding complex instructions, especially where the specific DPO datasets used in its training align with the desired output style or domain.