[mindspore] [Lite end-to-side training reasoning] mindspore lit runs the lenet training example code according to the instructions and reports an error

Mindspore lite

【 Operation steps & Problem phenomenon 】

1、 Compile first according to the instructions Mindspore lite,train-linux-x64 as well as train-android-aarch64 package

Model training and validation

The sample code is in MindSpore Source code Under the mindspore/lite/examples/train_lenet Catalog . Local cloning MindSpore After the source , Get into mindspore/lite/examples/train_lenet Catalog , After executing the following order , Script will be exported lenet_tod.mindir Model , And then use it converter Tools will MindIR The model is converted to MindSpore Lite Recognizable lenet_tod.ms Model ; Last , take lenet_tod.ms Model file 、MNIST Data sets and MindSpore Lite Training runtime The package is pushed to Andorid On the device , Executive Training .

bash prepare_and_run.sh -D /PATH/MNIST_Data -t arm64

Report errors

【 Screenshot information 】

This problem cannot be found clang++ compiler , First of all, confirm whether your environment is installed clang, Secondly, whether the environment variable is set .

At present MS Lite Support is gcc 7.3 And above , Suggest using gcc.