Meet Gorilla : The Language Model That Can Interact with Over 1,600 APIs and Outperforms GPT-4 for API calls
Gorilla is an open-source project, created by Shishir Patil, a PhD Student in ML systems at UC Berkeley.
It’s based on Falcon and MPT, two powerful language models that have been fine-tuned for API integration and usage.
Gorilla can parse the Abstract Syntax Tree (AST) when writing code, resulting in semantically and syntactically correct API invocations. This means that it can significantly reduce hallucinations and incorrect syntax, which are common problems for other language models, such as GPT-4.
Comments
No affiliation.
From their announcement post:
Meet Gorilla : The Language Model That Can Interact with Over 1,600 APIs and Outperforms GPT-4 for API calls
Gorilla is an open-source project, created by Shishir Patil, a PhD Student in ML systems at UC Berkeley.
It’s based on Falcon and MPT, two powerful language models that have been fine-tuned for API integration and usage.
Gorilla can parse the Abstract Syntax Tree (AST) when writing code, resulting in semantically and syntactically correct API invocations. This means that it can significantly reduce hallucinations and incorrect syntax, which are common problems for other language models, such as GPT-4.
Apache 2.0 license.