StarCoder: LLM for Responsible Code Development | | | | Turtles AI

StarCoder: LLM for Responsible Code Development
  #OpenAI and #ServiceNow have joined forces to form #BigCode, an #open #scientific #collaboration aimed at developing #LLM for #code in a #responsible manner. They have introduced #StarCoder and #StarCoderBase, #LLMs for code trained on permissively licensed data from #GitHub and other sources. The models outperform existing open code models and match or surpass closed models such as OpenAI's code-cushman-001. With a context length of over 8,000 tokens, the StarCoder models can process more input than any other open language model. They can be used for autocomplete, code modification, and explaining code snippets in natural language. StarCoder is publicly available under an improved version of the OpenRAIL license, which simplifies integration into products. The model also acts as a technical assistant, able to answer programming-related requests. The training data for the models consists only of permissively licensed code, and personal identifiable information has been removed. See the project page here for more info.