用合成数据提升硬件设计模型泛化能力，效果显著。
Iceberg: Enhancing HLS Modeling with Synthetic Data
- 通过大模型生成程序和弱标签扩展合成数据
- 少样本下几何平均精度提升86.4%
- 适合芯片设计自动化研究者使用
基于深度学习的高层次综合（HLS）预测模型常难以泛化。本文通过在合成数据上预训练，提出Iceberg合成数据增强方法，扩展大型语言模型生成的代码及未见设计配置的弱标签。弱标签生成与上下文内模型架构结合，实现从真实与邻近标签中进行元学习。在六个真实应用中少样本适配时，几何平均建模准确率提升86.4%；在两个不同测试数据集上，离线设计空间探索（DSE）性能分别提升2.47倍和1.12倍。代码已开源：https://github.com/UCLA-VAST/iceberg
原文摘要 · Abstract (English)
Deep learning-based prediction models for High-Level Synthesis (HLS) of hardware designs often struggle to generalize. In this paper, we study how to close the generalizability gap of these models through pretraining on synthetic data and introduce Iceberg, a synthetic data augmentation approach that expands both large language model (LLM)-generated programs and weak labels of unseen design configurations. Our weak label generation method is integrated with an in-context model architecture, enabling meta-learning from actual and proximate labels. Iceberg improves the geometric mean modeling accuracy by $86.4\%$ when adapt to six real-world applications with few-shot examples and achieves a $2.47\times$ and a $1.12\times$ better offline DSE performance when adapting to two different test datasets. Our open-sourced code is here: https://github.com/UCLA-VAST/iceberg
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。