Weaviate Gorilla: LLaMA 7B Fine-tuned for GraphQL APIs

AI Impact Summary

The team has developed WeaviateGraphQLGorilla, a fine-tuned LLaMA 7B model designed to translate natural language commands into Weaviate GraphQL Search APIs. This allows developers to interact with Weaviate’s powerful vector search capabilities without needing to manually construct GraphQL queries. The model was trained on a synthetic dataset of 2,300 examples generated using GPT-3.5-16K, leveraging techniques like Self-Instruct prompting and Retrieval-Aware Training to optimize the model’s performance and reduce training costs.