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

fix, to support hip

Browse Source
This commit is contained in:
amdsc21 2023-05-02 17:43:11 +02:00
parent 5446c501af
commit e4538cb13c
2 changed files with 26 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/data/iterative_dmatrix.cu
View File

@ -66,7 +66,7 @@ void IterativeDMatrix::InitFromCUDA(Context const* ctx, BatchParam const& p,
do { do {
// We use do while here as the first batch is fetched in ctor // We use do while here as the first batch is fetched in ctor
// ctx_.gpu_id = proxy->DeviceIdx(); // ctx_.gpu_id = proxy->DeviceIdx();
CHECK_LT(ctx_.gpu_id, common::AllVisibleGPUs()); CHECK_LT(ctx->gpu_id, common::AllVisibleGPUs());
#if defined(XGBOOST_USE_CUDA) #if defined(XGBOOST_USE_CUDA)
dh::safe_cuda(cudaSetDevice(get_device())); dh::safe_cuda(cudaSetDevice(get_device()));

30
src/objective/lambdarank_obj.cu
View File

@ -33,6 +33,12 @@
#include "xgboost/logging.h" #include "xgboost/logging.h"
#include "xgboost/span.h" // for Span #include "xgboost/span.h" // for Span
#if defined(XGBOOST_USE_HIP)
#include <hipcub/hipcub.hpp>
namespace cub = hipcub;
#endif
namespace xgboost::obj { namespace xgboost::obj {
DMLC_REGISTRY_FILE_TAG(lambdarank_obj_cu); DMLC_REGISTRY_FILE_TAG(lambdarank_obj_cu);
@ -291,7 +297,11 @@ void Launch(Context const* ctx, std::int32_t iter, HostDeviceVector<float> const
HostDeviceVector<GradientPair>* out_gpair) { HostDeviceVector<GradientPair>* out_gpair) {
// boilerplate // boilerplate
std::int32_t device_id = ctx->gpu_id; std::int32_t device_id = ctx->gpu_id;
#if defined(XGBOOST_USE_CUDA)
dh::safe_cuda(cudaSetDevice(device_id)); dh::safe_cuda(cudaSetDevice(device_id));
#elif defined(XGBOOST_USE_HIP)
dh::safe_cuda(hipSetDevice(device_id));
#endif
auto n_groups = p_cache->Groups(); auto n_groups = p_cache->Groups();
info.labels.SetDevice(device_id); info.labels.SetDevice(device_id);
@ -374,7 +384,11 @@ void LambdaRankGetGradientNDCG(Context const* ctx, std::int32_t iter,
HostDeviceVector<GradientPair>* out_gpair) { HostDeviceVector<GradientPair>* out_gpair) {
// boilerplate // boilerplate
std::int32_t device_id = ctx->gpu_id; std::int32_t device_id = ctx->gpu_id;
#if defined(XGBOOST_USE_CUDA)
dh::safe_cuda(cudaSetDevice(device_id)); dh::safe_cuda(cudaSetDevice(device_id));
#elif defined(XGBOOST_USE_HIP)
dh::safe_cuda(hipSetDevice(device_id));
#endif
auto const d_inv_IDCG = p_cache->InvIDCG(ctx); auto const d_inv_IDCG = p_cache->InvIDCG(ctx);
auto const discount = p_cache->Discount(ctx); auto const discount = p_cache->Discount(ctx);
@ -442,7 +456,11 @@ void LambdaRankGetGradientMAP(Context const* ctx, std::int32_t iter,
linalg::VectorView<double> li, linalg::VectorView<double> lj, linalg::VectorView<double> li, linalg::VectorView<double> lj,
HostDeviceVector<GradientPair>* out_gpair) { HostDeviceVector<GradientPair>* out_gpair) {
std::int32_t device_id = ctx->gpu_id; std::int32_t device_id = ctx->gpu_id;
#if defined(XGBOOST_USE_CUDA)
dh::safe_cuda(cudaSetDevice(device_id)); dh::safe_cuda(cudaSetDevice(device_id));
#elif defined(XGBOOST_USE_HIP)
dh::safe_cuda(hipSetDevice(device_id));
#endif
info.labels.SetDevice(device_id); info.labels.SetDevice(device_id);
predt.SetDevice(device_id); predt.SetDevice(device_id);
@ -481,7 +499,11 @@ void LambdaRankGetGradientPairwise(Context const* ctx, std::int32_t iter,
linalg::VectorView<double> li, linalg::VectorView<double> lj, linalg::VectorView<double> li, linalg::VectorView<double> lj,
HostDeviceVector<GradientPair>* out_gpair) { HostDeviceVector<GradientPair>* out_gpair) {
std::int32_t device_id = ctx->gpu_id; std::int32_t device_id = ctx->gpu_id;
#if defined(XGBOOST_USE_CUDA)
dh::safe_cuda(cudaSetDevice(device_id)); dh::safe_cuda(cudaSetDevice(device_id));
#elif defined(XGBOOST_USE_HIP)
dh::safe_cuda(hipSetDevice(device_id));
#endif
info.labels.SetDevice(device_id); info.labels.SetDevice(device_id);
predt.SetDevice(device_id); predt.SetDevice(device_id);
@ -496,15 +518,13 @@ void LambdaRankGetGradientPairwise(Context const* ctx, std::int32_t iter,
Launch(ctx, iter, predt, info, p_cache, delta, ti_plus, tj_minus, li, lj, out_gpair); Launch(ctx, iter, predt, info, p_cache, delta, ti_plus, tj_minus, li, lj, out_gpair);
} }
namespace { struct ReduceOp : thrust::binary_function<thrust::tuple<double, double> const&, thrust::tuple<double, double>
struct ReduceOp { const&, thrust::tuple<double, double>> {
template <typename Tup> thrust::tuple<double, double> __host__ XGBOOST_DEVICE operator()(thrust::tuple<double, double> const& l, thrust::tuple<double, double> const& r) {
Tup XGBOOST_DEVICE operator()(Tup const& l, Tup const& r) {
return thrust::make_tuple(thrust::get<0>(l) + thrust::get<0>(r), return thrust::make_tuple(thrust::get<0>(l) + thrust::get<0>(r),
thrust::get<1>(l) + thrust::get<1>(r)); thrust::get<1>(l) + thrust::get<1>(r));
} }
}; };
} // namespace
void LambdaRankUpdatePositionBias(Context const* ctx, linalg::VectorView<double const> li_full, void LambdaRankUpdatePositionBias(Context const* ctx, linalg::VectorView<double const> li_full,
linalg::VectorView<double const> lj_full, linalg::VectorView<double const> lj_full,