Overview
Llama-3-8B-Function-Calling-xLAM is an 8 billion parameter language model developed by ermiaazarkhalili, fine-tuned from Meta's Llama-3-8B-Instruct. Its primary focus is on function calling, achieved through Supervised Fine-Tuning (SFT) using LoRA (Low-Rank Adaptation) with 4-bit quantization. The model was trained on the Salesforce/xlam-function-calling-60k dataset, specifically designed for function calling tasks.
Key Capabilities
- Function Calling Optimization: Specifically fine-tuned to understand and generate function calls based on user prompts.
- Efficient Training: Utilizes LoRA with 4-bit NF4 quantization for efficient adaptation of the base model.
- Inference Flexibility: Available in multiple formats, including GGUF quantizations, for deployment on various hardware, including CPU.
- Base Model Strength: Benefits from the robust architecture and pre-training of the Meta-Llama-3-8B-Instruct model.
Good For
- Research: Ideal for studying language model fine-tuning techniques, particularly for function calling.
- Prototyping Conversational AI: Suitable for developing and testing conversational agents that require tool use or function invocation.
- Educational Purposes: Can be used for learning about SFT, LoRA, and function calling implementations in LLMs.
- Personal Projects: Applicable for various personal AI projects where function calling is a core requirement.
Limitations
- Context Length: Fine-tuned with a 2,048 token context limit.
- Language: Primarily trained on English data.
- Safety: Not extensively safety-tuned; requires additional guardrails for production use.
- Knowledge Cutoff: Inherits the knowledge cutoff of its base model.