The output pickle object is a little bit of a mess in terms of data types, mixing int, uint, float32, float64 etc.. This makes it very hard to convert to a numpy array with different dtypes (e.g., numpy.void where each column might be a different dtype). Forcing everything as numpy.float64 is fine, but leads to extra space allocation/necessity.
Be more aware when writing the results.
The output pickle object is a little bit of a mess in terms of data types, mixing int, uint, float32, float64 etc.. This makes it very hard to convert to a numpy array with different dtypes (e.g., numpy.void where each column might be a different dtype). Forcing everything as numpy.float64 is fine, but leads to extra space allocation/necessity.
Be more aware when writing the results.