SantaCoder

lastmod 2023-05-10created 2023-05-101 edits

A code generation model trained with 1.1 billion (1.1B) parameters

Features

• A multilingual language model trained on Python, Java, and Javascript code
• Reportedly outperforms Facebook's InCoder (6.7B) and Salesforce's CodeGen-Multi (2.7B) in LTR generation and infilling
• Uses part of The-Stack v1.1 (6TB) dataset previously released by BigCode

See also

• StarCoder

Favorite site

• Huggingface - santacoder