- N/A
QuantilesSketch
now ignores NaNs in input values and weights. Previously, NaNs would lead to incorrect quantiles calculation.- Fixes a bug when
MisraGriesSketch
would discard excessive number of elements duringAddValues
andCompress
and output fewer elements than requested. - Depends on
tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,<3
. - Depends on
tensorflow-serving-api>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,<3
.
- N/A
- N/A