PidResidualDiagnostics API【免费下载链接】mat-chem-sim-pred面向工业领域聚焦计算仿真、预测两大核心场景构建面向流程工业机理数据双轮驱动的领域计算层推动AI for Science在材料化学领域的深度应用。项目地址: https://gitcode.com/cann/mat-chem-sim-predextern C uint64_t aclnnPidResidualDiagnosticsGetWorkspaceSize( int64_t batch, int64_t sample_count, int64_t max_lag); extern C int32_t aclnnPidResidualDiagnostics( void* actual, void* predicted, void* metrics, void* autocorr, int64_t batch, int64_t sample_count, int64_t max_lag, void* workspace, uint64_t workspace_size, void* stream);参数参数说明actualDevice 输入shape[batch, sample_count]float32实测过程输出predictedDevice 输入shape[batch, sample_count]float32模型预测输出metricsDevice 输出shape[batch, 8]float32autocorrDevice 输出shape[batch, max_lag]float32batch回路数量sample_count每条回路采样点数必须大于 1max_lag自相关最大滞后阶数范围为1 max_lag sample_countworkspaceDevice workspaceworkspace_sizeworkspace 大小必须不小于aclnnPidResidualDiagnosticsGetWorkspaceSize返回值streamACL stream输出指标顺序mean_residual,std_residual,mae,rmse,max_abs_residual,fit_percent,durbin_watson,ljung_box_qautocorr[b, lag - 1]对应第lag阶残差自相关lag从 1 到max_lag。返回值返回值说明ACL_SUCCESSlaunch 成功ACL_ERROR_INVALID_PARAM输入指针为空、shape 参数非法或 workspace 不足约束当前原型仅支持 float32。输入和输出均为连续 ND 布局。actual/predicted/metrics/autocorr/workspace必须位于 Device 内存。max_lag越大计算量按B * N * max_lag增长。【免费下载链接】mat-chem-sim-pred面向工业领域聚焦计算仿真、预测两大核心场景构建面向流程工业机理数据双轮驱动的领域计算层推动AI for Science在材料化学领域的深度应用。项目地址: https://gitcode.com/cann/mat-chem-sim-pred创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考