CSv5.0.5 GPU memory allocation error

2D Classification
1

Hi,

I’m facing an issue after upgrading from v4.7.1 to v5.0.5 (nvidia:610.57.04, CUDA:13.3, 18x6GB RTX4000).

I’m running the same 2D classification job with 20million particles (90pix box, 300 classes, rest parameters default) in v5.0.5 which I ran successfully in v4.7.1. However, newer version the job is failing with the following “memory allocation error”. This seems kind of odd to me since the older version was doing fantastic with same set of parameters.

“/home/emPackages/CS/cryosparc_worker/.pixi/envs/worker/lib/python3.12/site-packages/numba/cuda/cudadrv/driver.py”, line 408, in _check_cuda_python_error
raise CudaAPIError(retcode, msg)
numba.cuda.cudadrv.driver.CudaAPIError: [2] Call to cuMemAlloc results in CUDA_ERROR_OUT_OF_MEMORY

Please advice if I’m missing something. If not then I should probably roll back to previous version.

Thanks

2

@diffracteD Does an adjustment of the Plotting sort method parameter CUDA_ERROR_OUT_OF_MEMORY error for 2D classification after upgrading to v5 - #2 by wtempel help in your case?