README.md

metadata

language:
  - en
license: llama3
pipeline_tag: text-generation
tags:
  - facebook
  - meta
  - pytorch
  - llama
  - llama-3
  - groq
  - tool-use
  - function-calling

Llama-3-70B-Tool-Use

This is the 70B parameter version of the Llama 3 Groq Tool Use model, specifically designed for advanced tool use and function calling tasks.

Model Details

• Model Type: Causal language model fine-tuned for tool use
• Language(s): English
• License: Meta Llama 3 Community License
• Model Architecture: Optimized transformer
• Training Approach: Full fine-tuning and Direct Preference Optimization (DPO) on Llama 3 70B base model
• Input: Text
• Output: Text, with enhanced capabilities for tool use and function calling

Performance

• Berkeley Function Calling Leaderboard (BFCL) Score: 90.76% overall accuracy
• This score represents the best performance among all open-source 70B LLMs on the BFCL

Usage and Limitations

This model is designed for research and development in tool use and function calling scenarios. It excels at tasks involving API interactions, structured data manipulation, and complex tool use. However, users should note:

• For general knowledge or open-ended tasks, a general-purpose language model may be more suitable
• The model may still produce inaccurate or biased content in some cases
• Users are responsible for implementing appropriate safety measures for their specific use case

Note the model is quite sensitive to the temperature and top_p sampling configuration. Start at temperature=0.5, top_p=0.65 and move up or down as needed.

Ethical Considerations

While fine-tuned for tool use, this model inherits the ethical considerations of the base Llama 3 model. Use responsibly and implement additional safeguards as needed for your application.

Availability

The model is available through:

• Groq API console
• Hugging Face

For full details on responsible use, ethical considerations, and latest benchmarks, please refer to the official Llama 3 documentation and the Groq model card.