Meta is joining other fellow tech giants in creating its own AI-assisted coding tool. The tool, Code Llama, generates code based on text prompts. It can also help with other development tasks, like code completion and debugging.
Meta believes that Code Llama will make development workflows faster and lower the bar to entry for new developers.
“Code Llama is state-of-the-art for publicly available LLMs on coding tasks. It has the potential to make workflows faster and more efficient for developers and lower the barrier to entry for people who are learning to code. Code Llama has the potential to be used as a productivity and educational tool to help programmers write more robust, well-documented software,” Meta wrote in a blog post.
Code Llama is based on Llama 2, which is a large language model focused on code generation. In order to create Code Llama, the company did additional training on code-specific datasets.
Meta is releasing Code Llama under a community license in order to follow an open approach to AI innovation. It is using the same license it used for Llama 2.
The tool already supports most popular programming languages, including Python, C++, Java, PHP, TypeScript, C#, and Bash, among others.
The company has also released two specialized versions of Code Llama. Code Llama – Python is a variation that was further trained on 100B tokens of Python code. Code Llama – Instruct is fine-tuned for natural language instruction.
“The model is fed a natural language instruction input and the expected output. This makes it better at understanding what people expect out of their prompts. We recommend using Code Llama – Instruct variants whenever using Code Llama for code generation since Code Llama – Instruct has been fine-tuned to generate helpful and safe answers in natural language,” Meta wrote.