Model Overview

argilla/Llama-3.2-1B-Instruct-APIGen-FC-v0.1 is a 1 billion parameter instruction-tuned model, fine-tuned by argilla from the meta-llama/Llama-3.2-1B-Instruct base model. Its primary specialization is function calling, allowing it to interpret user queries and generate appropriate structured calls to external tools or APIs.

Key Capabilities

• Function Calling: Designed to parse natural language into structured function calls, including arguments.
• Versatile Call Handling: Supports various function calling patterns:
  • Simple: Single function call based on a query.
  • Parallel: Multiple independent function calls from one query.
  • Multiple: Chained or dependent function calls.
  • Multi-turn: Maintains context for sequential function calls in a conversation.
• Robust Error Handling: Can identify when no suitable tools are available or when required parameters are missing for a function call.

Training Details

The model was fine-tuned using the TRL library on the argilla-warehouse/apigen-synth-trl dataset, which is derived from argilla/Synth-APIGen-v0.1. The training process was efficient, completing in 13 minutes on a node with 8xH100 GPUs.

Good for

• API Integration: Automating interactions with external APIs through natural language.
• Tool Use: Developing agents that can leverage a suite of tools to fulfill user requests.
• Structured Output Generation: Generating JSON-formatted tool calls from conversational input.