付俊伟-AIGC 浪潮下 WebNN 的演进与实践.pdf

1、演讲人：付俊伟胡宁馨,英特尔首席工程师,W3C Web Neural Network(WebNN)标准的起草和主要编辑者,Chromium committer and Chromium WebNN 组件的主要拥有者张敏,Intel WebNN 团队的技术经理,Chromium and ONNX Runtime WebNN EP 的开发者,WebNN developer preview的作者付俊伟,英特尔高级软件工程师,Chromium committer and Chromium WebNN的基础架构设计和Chromium Shape Detection API 主要开发者目 录01WebNN

2、出现的背景02WebNN的架构设计03如何使用WebNN04WebNN的性能对比https:/microsoft.github.io/webnn-developer-preview/WebNN Execution Provider of ONNX Runtime Web with GPU acceleration from DirectML.Running on Intel CoreUltra 7 processor 155H with integrated ArcGPU.Stable DiffusionA cat under the snowText EncoderText EncoderI

3、mage GenerationUnetUnetStepStep1 1Image DecoderUnetUnetStepStep2 2UnetUnetStepStep3 3UnetUnetStepStep4 4WebNN OperationmatMulgathersigmoidsoftmaxDirectMLGEMMGATHERLOGISTICSOFTMAXTFLiteBATH_MATMULGATHERACTIVATION_SIGMOIDACTIVATION_SOFTMAXCoreMLmatmulgather_along_axissigmoidsoftmaxCPUGPUNPU系统ML APIsWe

4、b Browser(e.g.,Chrome/Edge)框架运用场景WebNNJavaScript Runtime(e.g.,Electron/Node.js)Noise SuppressionImageClassificationBackgroundSegmentationTensorFlow.jsONNXRuntimeWebMediaPipe WebNatural Language硬件CoreMLDirectMLWeb API Web引擎Transformers.jsWebAssemblyWebGPUObject DetectionTFLiteOther ML OS APIsWindows

5、Studio EffectsAPI extensionsComputational Graph(Web)conv2daddreluinputoutputfilterbiastmptmpcompilecomputeInput Buffers(CPU/GPU)MLGraphBuilderMLGraphBuilderMLContextMLContextMLGraphMLGraphOutput Buffers(CPU/GPU)device type:cpu/gpu/npupower preference:high-perf/low-powerbuildcreateCompiled Graph(Nati

6、ve)Fused conv2dinputoutputWebNN为Web带来了神经网络的统一抽象Other Web APIWebNN API Call flowDataflowWeb ApplicationJS ML FrameworksGPUNPUCPUWebNNMojo ClientDirectMLBackendWebNN Mojo ServerBNNS/MPSCoreMLMCDMApps/FrameworksHardwareChromiumNative ML APIsOS DriversRenderer ProcessIPCmacOSWindowsMLContextMLGraphBuild