Error in from kcRampPathSampler2

[herma686]

I successfully installed the package and dependencies in a similar Windows environment. However, despite much effort, I am getting an error (below) stemming from one of the MEX files (kcRampPathSampler2) that was compiled as part of the process. It's at line 560 of kcRampPathSampler2.cu (in
the code/CUDAlib/src folder).

Here is the error:

Simulating from the ramping model (500 total trials, 5 coherence levels)...
done.
Current GPU device: 0
Changed to GPU device: 0
Device reset.
Starting Ramping MCMC sampler...
Error using kcRampPathSampler2
Error random number generating seed (201).

Error in fitRampingModel (line 187)
kcRampPathSampler2(gpu_lambdaN,gpu_auxThresholdN,gpu_y,gpu_trIndex,gpu_trBetaIndex,RampSamples.betas(ss-1,:),RampSamples.w2s(ss-1),RampSamples.l_0(ss-1),RampSamples.gammas(ss-1),timeSeries.delta_
Error in runModelComparison (line 86)
[RampFit, RampSamples,LatentDataHandler] = fitRampingModel(timeSeries,params);

Error in exampleScript (line 55)
[DICs, StepFit, RampFit, StepModelComp, RampModelComp] = runModelComparison( timeSeries,
params, resultsFiles, samplesFiles);

[latimerk]

Hi,
With the 201 error code, my best guess is that it can't find one of the CUDA libraries (such as cudadevrt) or it's loading a different version of the library than the one used for compiling.
Can you confirm that the paths are pointing to the correct directories?
This MATLAB tool might help: https://www.mathworks.com/help/gpucoder/ref/gpucodersetup-app.html
The Windows environment variables might need checking too: https://docs.nvidia.com/gameworks/content/developertools/desktop/environment_variables.htm

During early development with the version of MATLAB I had at the time, I found MATLAB had it's own copy of some of the CUDA libraries stored in one of its subfolders and that caused linking problems. If that's the case here, it may be interfering. I was able to get around it by moving MATLAB's copies to a different folder (keeping them around so I could undo this change if needed).

As stated in the documentation, this code was released for Ubuntu and for a now old version of CUDA and MATLAB. We hope that you can get it to work in Windows. If not and if you do not have access to a Ubuntu machine, you could try Ubuntu in Windows through the WSL.