Overview

Gorilla-7b-tf-delta-v0 is a 7 billion parameter auto-regressive language model, fine-tuned from LLaMA, developed by Gorilla LLM (UC Berkeley). Its core innovation lies in its ability to connect Large Language Models (LLMs) with a vast array of APIs, allowing them to act as tool users. The model can translate natural language queries into semantically and syntactically correct API calls, addressing a common challenge of hallucination in API invocation.

Key Capabilities

• API Invocation: Proficiently invokes over 1,600 APIs, with a specific focus on TensorFlow Hub APIs.
• Reduced Hallucination: Designed to minimize errors and hallucinations when generating API calls.
• Natural Language to API: Converts natural language instructions into executable API code.
• Retriever-Aware Training: Can be trained using a novel retriever-aware pipeline, alongside standard fine-tuning.

Good For

• Tool Use in LLMs: Ideal for applications requiring LLMs to interact with external tools and services via APIs.
• Code Generation (APIs): Generating accurate API calls based on user prompts.
• Expanding LLM Functionality: Enabling LLMs to perform tasks beyond their inherent knowledge by leveraging external APIs. The project also maintains APIBench, a large collection of APIs for training.