@liviocali generated new npz LUT files with the right number of optical channels, addressing the issue of missing channels described in #223. I did a run with a single file and passed it through larndsim_validation.py. The output is here, and below is the plot of the light signals across the channels of the 2x2, which look much better in comparison to the MR5 beta1 validation plots:
@liviocali mentioned that these LUT files also feature a change in the object types, from [('vis','f4'), ('t0','f4'),('time_dist','f4',(int(nbins_time),))] to [('vis','f4'), ('t0','f2'),('time_dist','u2',(int(nbins_time),))], in order to reduce the memory usage when loading the arrays.
A quick check on the NERSC Jupyter platform shows that the memory used when loading the array from the old LUT file is ~12GB, whereas with the new version of the file it is ~2.5 GB. The change of the object types, along with the elimination of the null entries in the old file which reduces the array shape from (64, 128, 32, 120) to (64, 128, 32, 48), seem to help with the memory usage.
@liviocali generated new npz LUT files with the right number of optical channels, addressing the issue of missing channels described in #223. I did a run with a single file and passed it through larndsim_validation.py. The output is here, and below is the plot of the light signals across the channels of the 2x2, which look much better in comparison to the MR5 beta1 validation plots:
@liviocali mentioned that these LUT files also feature a change in the object types, from
[('vis','f4'), ('t0','f4'),('time_dist','f4',(int(nbins_time),))]to[('vis','f4'), ('t0','f2'),('time_dist','u2',(int(nbins_time),))], in order to reduce the memory usage when loading the arrays.A quick check on the NERSC Jupyter platform shows that the memory used when loading the array from the old LUT file is ~12GB, whereas with the new version of the file it is ~2.5 GB. The change of the object types, along with the elimination of the null entries in the old file which reduces the array shape from
(64, 128, 32, 120)to(64, 128, 32, 48), seem to help with the memory usage.