CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation

Abstract

Pre-trained models for Natural Languages (NL) like BERT and GPT have been recently shown to transfer well to Programming Languages (PL) and largely benefit a broad set of code-related tasks. Despite their success, most current methods either rely on an encoder-only (or decoder-only) pre-training that is suboptimal for generation (resp. understanding) tasks or process the code snippet in the same way as NL, neglecting the special characteristics of PL such as token types. We present CodeT5, a unified pre-trained encoder-decoder Transformer model that better leverages the code semantics conveyed from the developer-assigned identifiers. Our model employs a unified framework to seamlessly support both code understanding and generation tasks and allows for multi-task learning. Besides, we propose a novel identifier-aware pre-training task that enables the model to distinguish which code tokens are identifiers and to recover them when they are masked. Furthermore, we propose to exploit the user-written code comments with a bimodal dual generation task for better NL-PL alignment. Comprehensive experiments show that CodeT5 significantly outperforms prior methods on understanding tasks such as code defect detection and clone detection, and generation tasks across various directions including PL-NL, NL-PL, and PL-PL. Further analysis reveals that our model can better capture semantic information from code. Our code and pre-trained models are released at https: //github.com/salesforce/CodeT5.

Investigating Explainability of Generative AI for Code through Scenario-based Design

“…• Code translation, in which a generative model translates source code from one language (e.g., Java) to another (e.g., Python). This task has been an important benchmark for technical work in GenAI for code [61] and has gained extensive attention from both industry and academia [3,23,30,88,99]. Such technologies can significantly reduce the cost and expertise barriers for code modernization work, in which a legacy codebase is ported to a modern programming language.…”

GAP-Gen: Guided Automatic Python Code Generation

“…In this work, we present GAP-Gen, a method to improve automatic Python source code generation from natural language description. Our GAP-Gen is fine-tuning of the pre-trained T5-English (Raffel et al, 2020a) and CodeT5 (Wang et al, 2021) language models that employ Syntax-Flow and Variable-Flow as guidance and has shown on being able to understand the relationship between natural language description and Python code from syntactic and semantic level of the Python code.…”

“…More related to our work, (Husain et al, 2019;Clement et al, 2020) explore pre-training methodologies for learning better structural and syntactical information for automatic code generation. Moreover, (Wang et al, 2021;Guo et al, 2021) incorporates Variable-Flows and identifier information into their pre-training process for better code generation performance.…”

“…It can be effectively applied for maintaining the naming semantics of the code during the code generation process. (Wang et al, 2021;Guo et al, 2021) use Variable-Flow during their pre-training process and achieve good performances on programming language relevant tasks. In their works, they extract function variables names as Variable-Flow which is integrated into their pre-training process for improving langauage models' capability on understanding the code semantic structure.…”

CodeRetriever: Unimodal and Bimodal Contrastive Learning

“…However, existing code pre-training approaches directly adopt (masked) language modeling as the training objective which targets on learning to predict (masked) tokens in a given code context (Feng et al, 2020;Guo et al, 2021;Wang et al, 2021b). However, this token-based approach generally results in poor code semantic representations due to two reasons.…”