While large language models (LLMs) have been widely applied to code generation, they struggle with generating entire deep learning projects, which are characterized by complex structures, longer functions, and stronger reliance on domain knowledge than general-purpose code. An open-domain LLM often lacks coherent contextual guidance and domain expertise for specific projects, making it challenging to produce complete code that fully meets user requirements. In this paper, we propose a novel planning-guided code generation method, DLCodeGen, tailored for generating deep learning projects. DLCodeGen predicts a structured solution plan, offering global guidance for LLMs to generate the project. The generated plan is then leveraged to retrieve semantically analogous code samples and subsequently abstract a code template. To effectively integrate these multiple retrieval-augmented techniques, a comparative learning mechanism is designed to generate the final code. We validate the effectiveness of our approach on a dataset we build for deep learning code generation. Experimental results demonstrate that DLCodeGen outperforms other baselines, achieving improvements of 9.7% in CodeBLEU and 3.6% in human evaluation metrics.

You can accelerate the simulation of your algorithms in Simulink by using different execution environments. By using support packages, you can also generate and deploy C/C and CUDA code on target hardware. Simulate and generate code for deep learning models in Simulink using MATLAB function blocks. Simulate and generate code for deep learning models in Simulink using library blocks. This example shows how to develop a CUDA application from a Simulink model that performs lane and vehicle detection using convolutional neural networks (CNN).