LightGBM icon indicating copy to clipboard operation
LightGBM copied to clipboard

Fix calculation of number of bins in FindGroup

Open CVPaul opened this issue 1 year ago • 3 comments

Fixes #3339.

The calculation method for num_total_bin used during Exclusive Feature Bundling

LightGBM/src/io/dataset.cpp

Line 134 in 665c473

(bin_mappers[fidx]->GetDefaultBin() == 0 ? -1 : 0); doesn't align completely with the way num_total_bin is calculated during the creation of a FeatureGroup LightGBM/include/LightGBM/feature_group.h

Line 68 in 665c473

if (bin_mappers_[i]->GetMostFreqBin() == 0) { As a result, the max_bin_per_group (=256) is working during Bundling, but it is not working when creating the FeatureGroup. When I replaced the GetDefaultBin() at dataset.cpp#L134 with GetMostFreqBin(), the issue was resolved.

CVPaul avatar Aug 04 '23 08:08 CVPaul

@microsoft-github-policy-service agree

CVPaul avatar Aug 04 '23 08:08 CVPaul

@shiyu1994 could you please edit the title of this to make it more informative in the release notes? I would, but I don't really understand the effect of this PR.

jameslamb avatar Aug 15 '23 02:08 jameslamb

this would be great to have in any release soon, please :pray:

Borda avatar Feb 07 '24 16:02 Borda

this would be great to have in any release soon, please 🙏

I've just come across this issue. No release with this solution yet??

fillipemadureira avatar Apr 26 '24 21:04 fillipemadureira

I also meet same problem. When release this version?

karta282950 avatar Apr 29 '24 12:04 karta282950

Unfortunately, we are currently battling CI issues (cf. https://github.com/microsoft/LightGBM/pull/6425). Once we got our GPU runners working again, we are planning to prepare another release -- this will likely happen in the early part of May.

borchero avatar Apr 29 '24 12:04 borchero