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

Dmatrix refactor stage 1 (#3301)

Browse Source 
* Use sparse page as singular CSR matrix representation

* Simplify dmatrix methods

* Reduce statefullness of batch iterators

* BREAKING CHANGE: Remove prob_buffer_row parameter. Users are instead recommended to sample their dataset as a preprocessing step before using XGBoost.
This commit is contained in:
Rory Mitchell
2018-06-07 10:25:58 +12:00
committed by GitHub
parent 286dccb8e8
commit a96039141a
47 changed files with 650 additions and 1036 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/cpp/tree/test_gpu_hist.cu
View File

@@ -26,10 +26,10 @@ TEST(gpu_hist_experimental, TestSparseShard) {
TrainParam p;
p.max_depth = 6;
dmlc::DataIter<RowBatch>* iter = dmat->RowIterator();
dmlc::DataIter<SparsePage>* iter = dmat->RowIterator();
iter->BeforeFirst();
CHECK(iter->Next());
const RowBatch& batch = iter->Value();
const SparsePage& batch = iter->Value();
DeviceShard shard(0, 0, 0, rows, hmat.row_ptr.back(), p);
shard.Init(hmat, batch);
CHECK(!iter->Next());
@@ -67,10 +67,10 @@ TEST(gpu_hist_experimental, TestDenseShard) {
TrainParam p;
p.max_depth = 6;
dmlc::DataIter<RowBatch>* iter = dmat->RowIterator();
dmlc::DataIter<SparsePage>* iter = dmat->RowIterator();
iter->BeforeFirst();
CHECK(iter->Next());
const RowBatch& batch = iter->Value();
const SparsePage& batch = iter->Value();
DeviceShard shard(0, 0, 0, rows, hmat.row_ptr.back(), p);
shard.Init(hmat, batch);