Profiling is not working in Nvidia Nsight Deeplearning Designer

Hai @gxie,
I’m using NVIDIA Nsight Deep Learning Designer for a project activity. After creating a custom ONNX model, whenever I try to profile it, I always get the message “process terminated.” I’ve tested both the CPU and CUDA execution providers, but the result is the same in both cases. The tool is installed on my Windows PC.

Hi,

The error message has the clue:“Input channel C is not equal to kernel channels*group“. Note, this is an error message emitted by OnnxRuntime that the model itself has a problem.

Did you notice any error message from NDLD’s type checker? Is it possible for you to share the model?

Thanks,

GXie

Hi @gxie

Thanks for your quick replay, the below are the models attached
In the below dummy model also i am getting the same issue…

AuthExt.zip (521 Bytes)

AuthExt.zip (521 Bytes)

The attachment only has a .cfg file. Can you please provide the .onnx file?

Hi @gxie

Thanks for your quick replay, the below are the models attached

AuthExt.zip (835 Bytes)

Thanks! I tried this test model on my Windows machine, and the profiling works well with both ORT CPU EP and ORT CUDA EP.

Does it work for you with the CPU EP?

What GPU do you have and what driver are you using? CUDA does require a GPU and proper driver to work.

Hi,

I have CUDA installed in my machine(Because i am training my ML models in this existing machine)

The below is the status when i run it in the CPU exection provider

Same issue…

nvlog_config.txt (269 Bytes)

We need more logging messages to understand the problem. You can following these steps to obtain the logs:

1)Download and put the attached file into the same folder where the NDLD binary (nsight-dl.exe) is

  1. Repeat what you were doing before (load the ONNX model and profile)

  2. Find a log file named “nvlog_output.txt“ in the same location above and share with us.

Hi @gxie,
I Have already installed the nsight-dl.exe in my windows machine …


This is my folder tree , i have placed the nvlog_config.txt in the same location..
Is this correct

No.

Try to find where nsight-dl.exe is, usually its path is something like this: “C:\Program Files\NVIDIA Corporation\Nsight DL Designer 2025.3.0\host\windows-desktop-dl-x64“

Hi @gxie,

Thanks for the quick replay
I have placed the *.txt file but ,the output file mentioned is not formed after the launch of profile with the current .onnx model

PLEASE FIND THE VIDEO DATA

video_data.zip (10.3 MB)

Tried with both the execution providers

Sorry about the trouble. “nvlog_config.txt” needs to be renamed as “nvlog.config”. When you do this in File Explorer, make sure you have the file extensions visible so that you can rename it correctly.

Hi @gxie ,
I have attached the result video

video_data.zip (11.9 MB)

Can you either run nsight-dl.exe with administrator privilege, or change the path text (“nvlog_output.txt“) in nvlog.config to an absolute path that you have the write permission, e.g. D:\ONNX\nvlog_output.txt?

Hi @gxie

Now i am able to generate the logs, Please find the attached below

nvlog_output.txt (9.3 KB)

Thank you for providing the log. Unfortunately, it does not capture any helpful messages about the problem.

For the next step, could you try launching the profiler directly from a command window and share its outputs?

  • Navigate to C:\Users\USER_NAME\AppData\Local\NVIDIA Corporation\NVIDIA Nsight Deep Learning Designer\2025.3.25220.1113\target\windows-desktop-dl-x64 (where USER_NAME should be replaced by his windows username)
  • And launch from a command line DLDesignerWorker.exe -v profile-onnx -e cpu --generate-random-inputs "Path_to_onnx_file" "Output_profiler_path.nv-dld-report" preferably using the same onnx path and profiler output path that was provided in the UI.

Hi @gxie,

C:\Users\M S I\AppData\Local\NVIDIA Corporation\NVIDIA Nsight Deep Learning Designer\2025.3.25220.1113 (build 36368985) (public-release)\target\windows-desktop-dl-x64>DLDesignerRemoteHelper.exe -v profile-onnx -e cpu --generate-random-inputs “D:\ONNX\AuthExt\tesr.onnx” “D:\ONNX\AuthExt\Output_profiler_path.nv-dld-report”
A subcommand is required
Run with --help for more information.

I am getting this while running could you please check any argument is missing

It is DLDesignerWorker.exe that you should execute. It looks you were trying to execute DLDesignerRemoteHelper.exe.

Hi @gxie,

Sorry for the minor mistake, I have attached the log below

Thank you for the screen capture. It looks the profiler ran to completion. Could you check if the profiler report was generated correctly at “D:\ONNX\AuthExt\Output_profiler_path.nv-dld-report“. If the file exists, try opening it in NDLD.