I am trying simulate US liver images and corresponding labels using the SimConfig2.iws. I encountered following error. Could you please help me solve it. Used Academia Version 3.11.0
(Link to SimConfig2.iws )
Thanks
I am trying simulate US liver images and corresponding labels using the SimConfig2.iws. I encountered following error. Could you please help me solve it. Used Academia Version 3.11.0
(Link to SimConfig2.iws )
Thanks
Welcome to the forum Maitri!
I just requested access to the google drive link you provided to access the workspace file.
In case it is not already in the goodle drive share, could you also provide the full log-file?
You can export it via “Help → Export Logs”
Thanks!
Thanks for the access to the workspace file.
In the workspace by itself I can’t find a reason for it not working.
Could you please provide a log export for a failed run (as described above)?
This should help with determining the source of the problem.
Thank you for verifying. I am still getting the same error. I uploaded error logs along with other two iws and imf file pair, the logs are from them. I am trying to generate liver US and labels, is there any alternate files or method to do so?
Any help is much appreciated.
Thank you.
Checking the log, the reported errors are from the OpenCL environment. They are
CL_MEM_OBJECT_ALLOCATION_FAILURE
and
CL_OUT_OF_HOST_MEMORY
I also see from the logs, that on start of the ImFusionSuite, a large portion of the GPU memory was already in use:
Memory: 1859 MB of 11 GB available
Is it possible that there was another process running at the same time, that was using a lot of GPU memoery and RAM, for example a pytorch training process?
If so, you could try again when theses are not running.
Another option could be to lower the number of frames to be simulated, but that might make it difficult to cover your area of interest with a smooth sweep motion.
I am already simulating only 50 frames.
The baffling part is the simulations happen in my local laptop which has Imfusion version 2.47.0
Unfortunately my laptop does not have GPU capabilities.
Could you also take look at images generated from the same iws file in my laptop? I will upload them to drive.
Thanks for a quick reply.
Regards,
Maitri
I tried running the again and error remains the same. Though intitally 10GB of GPU is available later there is still failures. Could you please specify what is the minimum GPU required to run a simulation of 50 sweeps?
Thanks and Regards,
Maitri Bhat
Which error code appears in that situation? In the first log there was -5 which means CL_MEM_OBJECT_ALLOCATION_FAILURE
(so likely insufficient GPU memory) and -6, which is the CL_OUT_OF_HOST_MEMORY
. Can you verify that there is also enough (host / CPU) RAM free?
Regarding the simulation on your laptop: The hybrid simulation required an OpenCL device even in version 2.47.0. That suggests that there is some OpenCL implementation present on the laptop, even if it’s just a CPU based one.
The required GPU and CPU memory is tricky to estimate, as it depends on a number of factors. But in a setting with 10GB GPU memory available, there should not be any memory issues.
I noticed that the workspace file you shared is relatively old. Could you try running the simulation example from our example repository?