DeepLearning network compiling for embedded code error

Question

Yang Liu 2023-12-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2057924-deeplearning-network-compiling-for-embedded-code-error

评论： Yang Liu 2023-12-28

I have trained a DL network and save as a .mat file (attached as "Paper3_LSTM_Based_Network2.mat"). The DL network simulates correctly under the Simulink environment (attached as "temp3.slx").

But when I tried to generate the C++ code through Embedded Code, it revealed:

Internal error while creating code interface description file: codeInfo.mat. Aborting code generation.

Caused by:

Cannot set ActualArgs property. Number of elements specified does not match number of elements in the Prototype's Arguments property.

Can someone help me understand this error information? Any solution I can apply with?

Note: The DL network contains the SeqFolding and SeqUnfolding, thus I have to rely on the ARM Compute Library.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ganesh 2023-12-18

编辑：Ganesh 2023-12-18

Hi @Yang Liu

May I know which version of MATLAB you are using?

Code generation seems to work fine on MATLAB R2023b.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Ganesh 2023-12-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2057924-deeplearning-network-compiling-for-embedded-code-error#answer_1376202

Hi @Yang Liu,

I understand that you are facing an issue with Code Generation using "Embedded Coder" when you are using a trained DL Network. The error you are facing is a known issue with r2021a. Kindly refer to the following link for more information on the bug:

https://in.mathworks.com/support/bugreports/2637595

I would suggest you to upgrade to a later version of MATLAB, which can be found in the following link:

https://in.mathworks.com/downloads/message/error_page/unlicensed?release=R2023b

Hope this helps!

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Yang Liu 2023-12-28

Hi, Ganesh,

Thanks a lot for you kind reply! I'm using Matlab 2022b.

I just solved this problem yesterday, by entering "which codeInfo.mat -all" in the command window, and wiping out the paths revealed below. Now, my model can be generated with source code.

Thank you very much!

Yang Liu

请先登录，再进行评论。