Open daurer opened 4 months ago
Hi, I have finally checked it, and it seems to be working well now. The old problem seems to no longer exist. The new exposed parameter smooth_gradient_method
is a very good asset. I will keep testing it in the long term, but all seems good to me. Thank you!
Nevertheless, I have only tested with ML_pycuda
since I have not yet been successful in installing CuPy
in order to try ML_cupy
, but I assume it should work as well. I have the same installation problem in the 2 HPC clusters I have access to.
can re-use FFTGaussianSmoothingKernel implemented for ML_pycuda and ML_cupy.