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

Fix GPU bugs (#3051)

Browse Source 
* Change uint to unsigned int

* Fix no root predictions bug

* Remove redundant splitting due to numerical instability
This commit is contained in:
Rory Mitchell 2018-01-23 13:14:15 +13:00 committed by GitHub
parent 8b2f4e2d39
commit f87802f00c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 16 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/objective/regression_obj_gpu.cu
View File

@ -43,7 +43,7 @@ struct GPURegLossParam : public dmlc::Parameter<GPURegLossParam> {
// GPU kernel for gradient computation // GPU kernel for gradient computation
template<typename Loss> template<typename Loss>
__global__ void get_gradient_k __global__ void get_gradient_k
(bst_gpair *__restrict__ out_gpair, uint *__restrict__ label_correct, (bst_gpair *__restrict__ out_gpair, unsigned int *__restrict__ label_correct,
const float * __restrict__ preds, const float * __restrict__ labels, const float * __restrict__ preds, const float * __restrict__ labels,
const float * __restrict__ weights, int n, float scale_pos_weight) { const float * __restrict__ weights, int n, float scale_pos_weight) {
int i = threadIdx.x + blockIdx.x * blockDim.x; int i = threadIdx.x + blockIdx.x * blockDim.x;
@ -76,7 +76,7 @@ class GPURegLossObj : public ObjFunction {
// manages device data // manages device data
struct DeviceData { struct DeviceData {
dvec<float> labels, weights; dvec<float> labels, weights;
dvec<uint> label_correct; dvec<unsigned int> label_correct;
// allocate everything on device // allocate everything on device
DeviceData(bulk_allocator<memory_type::DEVICE>* ba, int device_idx, size_t n) { DeviceData(bulk_allocator<memory_type::DEVICE>* ba, int device_idx, size_t n) {
@ -175,7 +175,7 @@ class GPURegLossObj : public ObjFunction {
safe_cuda(cudaGetLastError()); safe_cuda(cudaGetLastError());
// copy output data from the GPU // copy output data from the GPU
uint label_correct_h; unsigned int label_correct_h;
thrust::copy_n(d.label_correct.tbegin(), 1, &label_correct_h); thrust::copy_n(d.label_correct.tbegin(), 1, &label_correct_h);
bool label_correct = label_correct_h != 0; bool label_correct = label_correct_h != 0;

22
src/tree/updater_gpu_hist.cu
View File

@ -12,8 +12,8 @@
#include <vector> #include <vector>
#include "../common/compressed_iterator.h" #include "../common/compressed_iterator.h"
#include "../common/device_helpers.cuh" #include "../common/device_helpers.cuh"
#include "../common/host_device_vector.h"
#include "../common/hist_util.h" #include "../common/hist_util.h"
#include "../common/host_device_vector.h"
#include "../common/timer.h" #include "../common/timer.h"
#include "param.h" #include "param.h"
#include "updater_gpu_common.cuh" #include "updater_gpu_common.cuh"
@ -361,8 +361,7 @@ struct DeviceShard {
std::fill(ridx_segments.begin(), ridx_segments.end(), Segment(0, 0)); std::fill(ridx_segments.begin(), ridx_segments.end(), Segment(0, 0));
ridx_segments.front() = Segment(0, ridx.size()); ridx_segments.front() = Segment(0, ridx.size());
this->gpair.copy(begin + row_begin_idx, this->gpair.copy(begin + row_begin_idx, begin + row_end_idx);
begin + row_end_idx);
subsample_gpair(&gpair, param.subsample, row_begin_idx); subsample_gpair(&gpair, param.subsample, row_begin_idx);
hist.Reset(); hist.Reset();
} }
@ -527,7 +526,7 @@ class GPUHistMaker : public TreeUpdater {
private: private:
void UpdateHelper(HostDeviceVector<bst_gpair>* gpair, DMatrix* dmat, void UpdateHelper(HostDeviceVector<bst_gpair>* gpair, DMatrix* dmat,
const std::vector<RegTree*>& trees) { const std::vector<RegTree*>& trees) {
GradStats::CheckInfo(dmat->info()); GradStats::CheckInfo(dmat->info());
// rescale learning rate according to size of trees // rescale learning rate according to size of trees
float lr = param.learning_rate; float lr = param.learning_rate;
@ -607,7 +606,8 @@ class GPUHistMaker : public TreeUpdater {
monitor.Start("InitDataReset", dList); monitor.Start("InitDataReset", dList);
omp_set_num_threads(shards.size()); omp_set_num_threads(shards.size());
// TODO(canonizer): make it parallel again once HostDeviceVector is thread-safe // TODO(canonizer): make it parallel again once HostDeviceVector is
// thread-safe
for (int shard = 0; shard < shards.size(); ++shard) for (int shard = 0; shard < shards.size(); ++shard)
shards[shard]->Reset(gpair, param.gpu_id); shards[shard]->Reset(gpair, param.gpu_id);
monitor.Stop("InitDataReset", dList); monitor.Stop("InitDataReset", dList);
@ -722,7 +722,7 @@ class GPUHistMaker : public TreeUpdater {
shards[cpu_thread_id]->gpair.tend()); shards[cpu_thread_id]->gpair.tend());
} }
auto sum_gradient = auto sum_gradient =
std::accumulate(tmp_sums.begin(), tmp_sums.end(), bst_gpair()); std::accumulate(tmp_sums.begin(), tmp_sums.end(), bst_gpair_precise());
// Generate root histogram // Generate root histogram
for (auto& shard : shards) { for (auto& shard : shards) {
@ -733,7 +733,9 @@ class GPUHistMaker : public TreeUpdater {
// Remember root stats // Remember root stats
p_tree->stat(root_nidx).sum_hess = sum_gradient.GetHess(); p_tree->stat(root_nidx).sum_hess = sum_gradient.GetHess();
p_tree->stat(root_nidx).base_weight = CalcWeight(param, sum_gradient); auto weight = CalcWeight(param, sum_gradient);
p_tree->stat(root_nidx).base_weight = weight;
(*p_tree)[root_nidx].set_leaf(param.learning_rate * weight);
// Store sum gradients // Store sum gradients
for (auto& shard : shards) { for (auto& shard : shards) {
@ -879,8 +881,8 @@ class GPUHistMaker : public TreeUpdater {
return false; return false;
} }
bool UpdatePredictionCache(const DMatrix* data, bool UpdatePredictionCache(
HostDeviceVector<bst_float>* p_out_preds) override { const DMatrix* data, HostDeviceVector<bst_float>* p_out_preds) override {
return false; return false;
} }
@ -894,6 +896,8 @@ class GPUHistMaker : public TreeUpdater {
: nid(nid), depth(depth), split(split), timestamp(timestamp) {} : nid(nid), depth(depth), split(split), timestamp(timestamp) {}
bool IsValid(const TrainParam& param, int num_leaves) const { bool IsValid(const TrainParam& param, int num_leaves) const {
if (split.loss_chg <= rt_eps) return false; if (split.loss_chg <= rt_eps) return false;
if (split.left_sum.GetHess() == 0 || split.right_sum.GetHess() == 0)
return false;
if (param.max_depth > 0 && depth == param.max_depth) return false; if (param.max_depth > 0 && depth == param.max_depth) return false;
if (param.max_leaves > 0 && num_leaves == param.max_leaves) return false; if (param.max_leaves > 0 && num_leaves == param.max_leaves) return false;
return true; return true;