Closed Liu-Delin closed 3 years ago
This PR is a bug fix for this one: https://github.com/ShifuML/shifu/pull/732.
We changed segment column's name from name_1 to name_seg1 in stats step. But we didn't change the logic in norm step, so we will find empty segment data in normalized data.
name_1
name_seg1
stats
norm
For example, the header is:
target_column|column_1|column_2|column_1_seg1|column_2_seg1
We will get data like below (with the bug):
1|0.5|0.1|||1.0
But we want to get data like:
1|0.5|0.1|0.2|-0.1|1.0
Therefore, I changed the logic in norm step and any other step which will use BasicUpdater.
BasicUpdater
I manually tested it with large data. It generated right data. For unit test, I didn't find related UT for BasicUpdater, I think I can add it later.
Description
This PR is a bug fix for this one: https://github.com/ShifuML/shifu/pull/732.
We changed segment column's name from
name_1
toname_seg1
instats
step. But we didn't change the logic innorm
step, so we will find empty segment data in normalized data.For example, the header is:
We will get data like below (with the bug):
But we want to get data like:
Therefore, I changed the logic in
norm
step and any other step which will useBasicUpdater
.Tests
I manually tested it with large data. It generated right data. For unit test, I didn't find related UT for
BasicUpdater
, I think I can add it later.