Closed chrisbarber closed 6 years ago
looks good to me on axon.
@NileGraddis thanks for testing
@NileGraddis just to double check-- does running your conn bridge tests verify that your additional OR {"field": "experiment", "op": "=", "value": seed_idx} clause does not interfere with this?
Yes. In some cases the r values have changed on the order of 10^-{4,5,6}, but I think that might be unavoidable - it could be caused by running sums in different orders for instance, which is the sort of thing we don't really have detailed control over when deferring to a backend like dask.
This isn't using dask, just ndarrays. The parallelize decorator works on xarray but the chunks all end up being views on to the original ndarrays.
code looks good - testing on axon shortly