When I use DatasetFxNormalizer in Iterator, for each iteration the dataset is not the original one but a modified version, so that the normalizer always normalizes on a normalized one!!!
I added ds.copy() in the Normalizer but this solution may be unoptimal for memory usage.
When I use
DatasetFxNormalizer
inIterator
, for each iteration the dataset is not the original one but a modified version, so that the normalizer always normalizes on a normalized one!!!I added
ds.copy()
in theNormalizer
but this solution may be unoptimal for memory usage.