accuracy of cpp example is a constant value when training, no matter how many epochs trained!

[BackT0TheFuture]

cpp example : lenet_with_mxdataiter(gpu and cpu) mlp_cpu mlp_gpu

I've tested examples on three PCs (win8.1 64bit && win2012 64bit && centos7 64bit ) ,
roughly it's same, just mlp_gpu works correctly.

Environment info Operating System && Compiler

win8.1 64bit && win2012 64bit ===> vs2017 cmake
centos7 64bit ===> gcc

Package used: cpp

MXNet version: installed from source

MXNet commit hash (git rev-parse HEAD):

SHA-1: 6d71577

Error Message:

accuracy is a constant value , no matter how many epochs trained
please refer to attached images.

Minimum reproducible example

no modification just run after compiled

Steps to reproduce

no modification just run after compilederror.

What have you tried to solve it?

run example on other PCs

[BackT0TheFuture]

@piiswrong @mli

[BackT0TheFuture]

nobody cares about cpp ?

[eric-haibin-lin]

ping @lx75249

[szha]

@apache/mxnet-committers: This issue has been inactive for the past 90 days. It has no label and needs triage.

For general "how-to" questions, our user forum (and Chinese version) is a good place to get help.

[chsin]

@goodtogood if you pull master and rebuild, do you have the same issue? Apparently, the fix was merged: #8551. I'm curious bc I've just built master, but have the same problem with lenet.cpp