GE/CANN融合模式传递示例

Fusion Pass Examples【免费下载链接】geGEGraph Engine是面向昇腾的图编译器和执行器提供了计算图优化、多流并行、内存复用和模型下沉等技术手段加速模型执行效率减少模型内存占用。 GE 提供对 PyTorch、TensorFlow 前端的友好接入能力并同时支持 onnx、pb 等主流模型格式的解析与编译。项目地址: https://gitcode.com/cann/geThis directory provides examples for implementing custom fusion passes by inheriting GE-provided classes and overriding their methods:ExampleExample LinkMatMulAdd fused to GEMM custom pass exampleREADMEMatMulAdd fused to GEMM custom pass example (Python version)READMEcapture tensor feature usageREADMEcapture tensor feature usage (Python version)READMEPatternMatcherConfig feature usageREADMEPatternMatcherConfig feature usage (Python version)READMECustom pass example for deleting add zero operationREADMECustom pass example for deleting add zero operation (Python version, demonstratingpatternexpression syntax)READMECustom pass example for custom operatorsREADMECustom pass example for custom operators (Python version)READMECustom pass example for splitting grouped convolutionREADMECustom pass example for splitting grouped convolution (Python version)READMECustom pass example for flattening BatchMatMul to MatMulREADMEDevelopment GuideRecommended to first read mechanism description, then choose language guide:Fusion Pattern Pass MechanismPython Fusion Pass Development GuideC Fusion Pass Development Guide【免费下载链接】geGEGraph Engine是面向昇腾的图编译器和执行器提供了计算图优化、多流并行、内存复用和模型下沉等技术手段加速模型执行效率减少模型内存占用。 GE 提供对 PyTorch、TensorFlow 前端的友好接入能力并同时支持 onnx、pb 等主流模型格式的解析与编译。项目地址: https://gitcode.com/cann/ge创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考