hkust-nlp/drkernel-validation-data
收藏Hugging Face2026-02-06 更新2026-02-07 收录
下载链接：
https://hf-mirror.com/datasets/hkust-nlp/drkernel-validation-data
下载链接
链接失效反馈官方服务：
资源简介：
DR.Kernel验证数据集（KernelBench Level 2）是一个用于内核生成模型验证和评估的数据集，来源于KernelBench Level 2任务。当前版本包含100个任务，每个任务包括数据源标签、评估提示（单用户消息）、参考元数据（包括参考PyTorch实现）、能力标签（内核优化）和问题元数据（名称和问题ID）。数据集结构清晰，提供了详细的使用方法和统计信息，适用于DR.Kernel模型的评估和基准测试。
The DR.Kernel Validation Dataset (KernelBench Level 2) is a dataset designed for the validation and evaluation of kernel generation models, sourced from KernelBench Level 2 tasks. The current release contains 100 tasks, each including a data source tag, evaluation prompt (single user message in this release), reference metadata (with ground truth PyTorch implementation), ability tag (kernel_optimization), and problem metadata (name and problem_id). The dataset features a clear structure, detailed usage instructions, and statistical information, making it suitable for DR.Kernel model evaluation and benchmarking.
提供机构：
hkust-nlp
