Groq has recently released two innovative open source models for tool usage: Llama-3-Groq-70B-Tool Use and Llama-3-Groq-8B-Tool UseThese models were developed in collaboration with Glaive and are designed to improve tool usage and function calling capabilities in ai.
He Tool usage model Llama-3-Groq-70B is the top performing model on the Berkeley Function Calling Leaderboard (BFCL), outperforming all other open source and proprietary models. Achieving an impressive overall accuracy of 90.76% has set a new benchmark in the field. Similarly, the Llama-3-Groq-8B-Tool-Use model has also demonstrated remarkable performance with an overall accuracy of 89.06%, securing the third position on the BFCL. These models are now available on the GroqCloud Developer Hub and Hugging Face under the same permissive style license as the original Llama-3 models.
The development of these models involved a meticulous training approach that combined thorough fine-tuning and direct preference optimization (DPO). Notably, no user data was used in the training process; instead, the models were trained using ethically generated data. This approach ensures that the models perform highly and adhere to ethical standards in ai development. The training process also included a thorough contamination analysis using the LMSYS method. This resulted in a low contamination rate of only 5.6% for the SFT data and 1.3% for the DPO data, indicating minimal overfitting on the evaluation parameter.
In addition to their specialized tool usage capabilities, Llama-3 Groq tool usage models are recommended to be used in a hybrid approach with general-purpose language models. This strategy involves implementing a routing system that analyzes incoming user queries to determine the most appropriate model for each request. For queries involving function calls, API interactions, or structured data manipulation, Llama-3 Groq tool usage models are used. For general knowledge, open-ended conversations, or tasks not specifically related to tool usage, a general-purpose language model such as the unmodified Llama-3 70B is recommended. This approach ensures that each query is handled by the most appropriate model, maximizing the overall performance and capabilities of the ai system.
Both Llama-3-Groq-70B-Tool-Use and Llama-3-Groq-8B-Tool-Use are available for preview via the Groq API, with the model IDs llama3-groq-70b-8192-tool-use-preview and llama3-groq-8b-8192-tool-use-preview, respectively. Groq encourages the community to start developing and experimenting with these models via the GroqCloud Developer Center, paving the way for future innovations in ai tool use.
In conclusion, Groq introduced the Llama-3-Groq-Tool-Use models with their cutting-edge performance and permissive licensing. These models are poised to make a substantial impact on ai research and development. Groq’s commitment to ethical ai development and its collaborative approach with the community underscore the company’s leadership in the field.
Asif Razzaq is the CEO of Marktechpost Media Inc. As a visionary engineer and entrepreneur, Asif is committed to harnessing the potential of ai for social good. His most recent initiative is the launch of an ai media platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is technically sound and easily understandable to a wide audience. The platform has over 2 million monthly views, illustrating its popularity among the public.