Overview

gorilla-llm/gorilla-7b-th-delta-v0 is a 7 billion parameter language model developed by Gorilla LLM (UC Berkeley). It is an open-source API caller, fine-tuned from LLaMA weights, designed to enable Large Language Models (LLMs) to interact with external tools via API calls. The model focuses on translating natural language queries into accurate API invocations, aiming to reduce hallucination in this process.

Key Capabilities

• API Invocation: Translates natural language into semantically and syntactically correct API calls.
• Reduced Hallucination: Specifically engineered to minimize errors when generating API calls.
• Torch Hub API Integration: The gorilla-7b-th-delta-v0 variant is specifically fine-tuned for reliable use with Torch Hub APIs.
• Retriever-Aware Training: Can be trained using a novel retriever-aware pipeline, alongside standard fine-tuning.

Use Cases

• Tool Use for LLMs: Ideal for applications where LLMs need to interact with external systems or databases through APIs.
• Automated API Generation: Useful for generating API calls from user prompts, such as "I want to generate an image from text."
• Expanding API Ecosystem: Part of a broader project to expand the largest API store and teach LLMs how to write them, leveraging the APIBench dataset.

For more details, refer to the project website and the arXiv paper.