1 个仓库
Numerical representations of source code using specialized vocabularies tailored for programming syntax.
Distinct from Source Code Tokenizers: Focuses on the numerical embedding/encoding for LLMs rather than the lexical analysis of a compiler tokenizer.
Explore 1 awesome GitHub repository matching programming languages & runtimes · Code Vocabulary Encodings. Refine with filters or upvote what's useful.
CodeGen 是一个经过训练的大语言模型与程序合成模型,旨在生成功能性源代码。它利用神经网络架构,根据自然语言描述或部分代码片段合成可执行代码。 该模型通过预测并填充程序中缺失的代码部分,实现了自动化程序合成与 AI 辅助编码。它将自然语言描述转换为功能性编程逻辑,以自动化样板代码与业务逻辑的创建。
Converts source code into numerical representations using a specialized vocabulary optimized for programming language syntax.