After the release of generative AI models that can handle text content, audio, and translation, Meta has now moved its attention to coding. The company has introduced Code Llama, a version of its large language model that is made specifically for coding. The main function of this model is to take text prompts and then turn them into code in the same way as ChatGPT and DALL-E2 would do if asked to write some paragraphs or generate an image, respectively. They have released Code Llama on GitHub next to a research paper that offers a deeper dive into the code-specific generative AI tool. It extends the Llama 2 LLM.
Meta won’t be the first company to introduce generative AI smarts into the world of coding. Like, Microsoft recently incorporated the Chat GPT-4 capabilities into the GitHub Copilot system for the assistance of its developers. Similarly, In April 2023, Google also pushed advanced reasoning and mathematical capabilities for its Bard AI, which allows it to help with programming tasks like code generation. Meta is helping its users to generate original code and also fix their existing code With Code Llama.
Meta has clearly said that 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. The most interesting fact is that the prompts given by its users can either be in natural language or code segments. Depending on the nature of the prompt, Code LLama will generate a new code for you. It will act like some sort of auto-complete engine (like GitHub Copilot), and help its users in resolving any glitches. Code Llama also supports many of the most common programming languages, like C++, Python, Javascript, and PHP.
The Meta company also claims that Code Llama is ahead of its competitors in popular coding benchmarks such as HumanEval. They are releasing Code Llama in three size ranges of 7 billion parameters, 13 billion parameters, and 34 billion parameters. Its lower-end model will be more useful for less demanding tasks, whereas its higher-end model demands more hardware needs with greater capabilities.
For instance, a basic 7B tier can be run on a machine with a single GPU for low-latency tasks like code completion. The 13B model offers slightly more powerful fill-in-the-middle (FIM) capabilities, while the 34B variant is more useful for the experts looking for advanced code assistance with heavy code generation, block insertion, and debugging as they have the hardware to handle it.
Moreover, Meta has created two other variants of Code Llama based on the coding environment. Its Python targets Python, which is one of the most effective and widely used programming languages for AI. The other version is Code Llama — Instruct, which is better suited for natural language prompts and for non-experts looking to generate code. Although it is not perfect, still Meta claims that it will deliver safer responses compared to competitors.
Overall, Meta has definitely unleashed a gem as Code Llama is a versatile tool that has revolutionized the coding industry by completing both unfinished codes, and troubleshooting errors, and can also provide explanations about code like a pro. It supports across many programming languages, from Python to C++, Java to PHP, Typescript to C#, and even Bash. It is like a superhero for developers and has made coding smoother, faster, and more accessible. If you are an experienced programmer or just starting your coding journey, Code Llama has got your back.
