Meta has recently been active in the field of artificial intelligence, and after the new artificial intelligence model Llama 2 and SeamlessM4T to translate phrases in different languages (including Farsi), now Code Llama has made it available to the public; A machine learning system that can coding and provide explanations about the generated codes.
Just like Copilot and CodeWhisperer, as well as other open source artificial intelligence models such as StableCode and StarCoder, according to TechCrunch, Code Llama can not only write code, but also complete code written for various programming languages, including Python and C++, and solve problems. fix them
Meta tells TechCrunch:
“At META we believe that AI models, and especially large language models for coding, will benefit most from an open approach, in terms of innovation and safety. “When these models become publicly available, they can facilitate the development of new technologies that improve people’s lives, identify problems and fix vulnerabilities.”
Code Llama, which comes in several different versions, such as a Python-optimized version and an instruction-specific version, is based on the Llama 2 text generation model, which Meta made publicly available earlier this month. While coding can be done with the help of Llama 2, its performance is certainly not comparable to the code provided by the Copilot model.
Although Meta trained Code Llama using the same dataset used to train Llama 2, Code Llama had more time to learn the relationships between code and natural language than its parent model. However, the model is specifically designed for coding, and the tech giant has noted that its coding model may sometimes provide “inappropriate” answers.
RCO NEWS