MIOpen Kernels

该数据集选取了MIOpen的多个GPU内核作为候选，旨在验证所提出神经架构的有效性，重点关注不同卷积配置的最优性能参数。此外，数据集包含了各种卷积配置的最优性能参数，并且所提出的方法在预测这些参数方面实现了超过90%的准确率。考虑到可能存在的参数总量极为庞大，任务之一是预测GPU计算内核的最优调整参数。

This dataset selects multiple GPU kernels from MIOpen as candidates, aiming to validate the effectiveness of the proposed neural architecture, with a focus on the optimal performance parameters for various convolution configurations. Furthermore, the dataset includes optimal performance parameters for various convolution configurations, and the proposed method has achieved an accuracy of over 90% when predicting these parameters. Given the potentially extremely large total number of parameters, one of the tasks is to predict the optimal tuning parameters for GPU computing kernels.