Open nvaytet opened 1 month ago
Note that the docs are also quite long to build, maybe something similar can be done there, although that may be more tricky if users download notebooks directly from the docs.
This will be easy to do for our own files, but not for the ISIS ones, I think?
I think we are loading hdf5 scipp format files that we created from the ISIS files? If so, we can do the same selecting a vertical strip of pixels? (maybe we want to keep a cross (+) of pixels so we can also compute Qx, Qy?
Maybe simply keeping every 2nd or 4th pixel would get us far enough?
Running the unit tests is taking a long time (circa 10 minutes on CI). We should try to reduce this.
This is most probably due to running workflows on full loki/sans2d data, which contains many pixels, leading to expensive memory allocations. We should reduce the data to just a few pixels. We could for example extract a vertical line or slab of pixels, that would still have a discernable scattering pattern, but would be much less expensive. Note that in the process, we would need a new detector bank shape specific for the tests.