因为用官网的是官网的代码，所有没有贴全

createInferenceSession(modelPath) {
    return new Promise((resolve, reject) => {
      this.session = wx.createInferenceSession({
        model: modelPath,
        /* 0: Lowest precision e.g., LS16 + A16 + Winograd A16 + approx. math
           1: Lower precision e.g., LS16 + A16 + Winograd off + approx. math
           2: Modest precision e.g., LS16 + A32 + Winograd A32 + approx. math
           3: Higher precision e.g., LS32 + A32 + Winograd A32 + approx. math
           4: Highest precision e.g., LS32 + A32 + Winograd A32 + precise math
           Higher precision always require longer time to run session
        */
        precisionLevel : 0,
        allowNPU : false, // wheather use NPU for inference, only useful for IOS
        allowQuantize: false, // wheather generate quantize model
      });
      // 监听error事件
      this.session.onError((error) => {
        console.log("onError")
        console.error(error);
        reject(error);
      });
      this.session.onLoad(() => {
console.log("onload")
        this.ready = true;
        resolve();
      });
      this.session.offError(()=>{
        console.log("offErr")
      });
      this.session.offLoad(()=>{
        console.log("offLoad")
      });
    })
  }
真机调试可以执行到onLoad,体验版和线上都无法执行

onLoad函数在除了 真机调试 意外的环境不执行

存到本机路径：${wx.env.USER_DATA_PATH}/mobilenetv2-12.onnx

调试显示：file already exist at: http://usr/mobilenetv2-12.onnx