You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We could implement this for step > 1 by making the output chunks be 1/step of the size of the input in the indexed dimension. This would ensure that the output chunk doesn't read from a potentially large number of input chunks if step was large (100 in this case), which could be slow and unreliable. We could follow with a rechunk to restore the original chunk size. I think the hardest bit would be doing the indexing arithmetic.
Are you blocked on this or did you find a workaround @TomNicholas?
Also, note there's a test covering this case (it currently just checks for NotImplementedError).
The text was updated successfully, but these errors were encountered: