I am trying to slice the output of dask.array.argwhere(), but the result of this has unknown chunk sizes. The suggested solution, calling compute_chunk_sizes(), returns the error

TypeError: tuple indices must be integers or slices, not tuple

And the error points towards the line within Array.compute_chunk_sizes(self)

x._chunks = tuple(tuple(int(chunk) for chunk in chunks) for chunks in compute(tuple(c))[0]

Is there a way to fix this?

I tried dask.array.nonzero() too, but with the same result.