cudnn Backend API注意事项
2024-10-21 11:30:22
一、在包含多个节点的图中,不支持in-place node。(如果图只包含一个节点,支持in-place node)
Note that graphs with more than one operation node do not support in-place operations (that is, where any of the input UIDs matches any of the output UIDs). Such in-place operations are considered cyclic in later graph analysis and deemed unsupported. In-place operations are supported for single-node graphs.
二、op的存放顺序无关紧要,因为cudnn可以推理出图的拓扑结构。
Also note that the operation descriptors can be created and passed into cuDNN in any order, as the tensor UIDs are enough to determine the dependencies in the graph.
几类融合算子的限制
https://docs.nvidia.com/deeplearning/cudnn/developer-guide/index.html#op-specific-contraints-runtime-fusion-engine
最新文章
- LoadRunner ---协议分析
- hdu 5115 区间dp ***
- Ant编译环境
- 转载MSDN 在ASP.NET 中执行 URL 重写
- linux内存管理解析1----linux物理,线性内存布局及页表的初始化
- 查看编译后的calss文件编译jdk版本
- Web前端开发面试题赋答案
- C# 移动无标题栏窗体的几种方法
- [转]oracle系统表v$session、v$sql字段说明
- POJ2135:Farm Tour
- 宁撞金钟一下,不打破鼓三千,IT人要有志气,要进就进大的好的公司
- Linux网络基本网络配置方法介绍
- redis学习(八)——redis应用场景
- 学习笔记TF039:TensorBoard
- 使用TestServer测试ASP.NET Core API
- linux添加磁盘空间
- SQL语句导致性能问题
- 关于Class.getResource和ClassLoader.getResource的路径问题(转)
- SDUT3926 kmp
- rest_framework之魔法类