ForgeHLS
收藏资源简介:
ForgeHLS是一个大规模、开源的高层次综合数据集,包含JSON格式的数据,位于code/forgehls_dataset_json/目录中。完整数据集及相关工件可在huggingface/ForgeHLS上获取,由Hugging Face团队提供官方托管支持,以容纳大规模基准测试
ForgeHLS is a large-scale, open-source high-level synthesis dataset. Its data is formatted in JSON and stored in the code/forgehls_dataset_json/ directory. The full dataset and associated artifacts are available at huggingface/ForgeHLS, which is officially hosted and supported by the Hugging Face team to accommodate large-scale benchmarking tasks.
ForgeHLS 数据集概述
数据集简介
ForgeHLS 是一个大规模、开源的高层次综合(HLS)数据集,专为支持 HLS 相关研究而设计。
数据集内容
- 数据格式:JSON 格式
- 存储位置:
- GitHub:
source_code/forgehls_dataset_json/data_of_designs_forgehls_with_strategy_formatted.json.gz - Hugging Face:https://huggingface.co/datasets/zedongpeng/forgehls
- GitHub:
数据集结构
数据集包含以下组成部分:
- 高层次综合设计数据
- 策略格式化信息
- 质量结果(QoR)数据
应用场景
- 质量结果预测(QoR Prediction)
- 自动编译指示插入(Automatic Pragma Insertion)
技术依赖
- 工具要求:
- Vitis HLS
- LLVM 和 Clang(版本需一致)
- Python 包依赖:
- openai, anytree, pandas, matplotlib, notebook, seaborn, scikit-learn, scikit-optimize, tiktoken
引用信息
bibtex @misc{peng2025forgehls, title={ForgeHLS: A Large-Scale, Open-Source Dataset for High-Level Synthesis}, author={Zedong Peng and Zeju Li and Mingzhe Gao and Qiang Xu and Chen Zhang and Jieru Zhao}, year={2025}, eprint={2507.03255}, archivePrefix={arXiv}, primaryClass={cs.AR} }
相关资源
- 论文地址:https://arxiv.org/abs/2507.03255
- 团队文档:https://zeju.gitbook.io/lcm-team
- 数据集文档:
source_code/forgehls_dataset_json/README.md




