cuda returns wrong results?

[CoderJackLiu]

I try this model in pure c++ , while it runs fine in cpu , while I try gpu cuda ,it returns wrong result. why?

the result is like the following:

GPU Result:

CPU Result:

[brainstormyyf]

I try this model in pure c++ , while it runs fine in cpu , while I try gpu cuda ,it returns wrong result. why?

the result is like the following:

GPU Result:

CPU Result:

May I ask you how you implemented it in pure c++?