Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix CPU bin compression with categorical data. (#8809)

Browse Source 
* Fix CPU bin compression with categorical data.

* The bug causes the maximum category to be lesser than 256 or the maximum number of bins when
the input data is dense.
This commit is contained in:
Jiaming Yuan
2023-02-16 04:20:34 +08:00
committed by GitHub
parent cce4af4acf
commit c0afdb6786
6 changed files with 49 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/data/gradient_index_format.cc
View File

@@ -35,7 +35,7 @@ class GHistIndexRawFormat : public SparsePageFormat<GHistIndexMatrix> {
if (!fi->Read(&page->hit_count)) {
return false;
}
if (!fi->Read(&page->max_num_bins)) {
if (!fi->Read(&page->max_numeric_bins_per_feat)) {
return false;
}
if (!fi->Read(&page->base_rowid)) {
@@ -76,8 +76,8 @@ class GHistIndexRawFormat : public SparsePageFormat<GHistIndexMatrix> {
page.hit_count.size() * sizeof(decltype(page.hit_count)::value_type) +
sizeof(uint64_t);
// max_bins, base row, is_dense
fo->Write(page.max_num_bins);
bytes += sizeof(page.max_num_bins);
fo->Write(page.max_numeric_bins_per_feat);
bytes += sizeof(page.max_numeric_bins_per_feat);
fo->Write(page.base_rowid);
bytes += sizeof(page.base_rowid);
fo->Write(page.IsDense());