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

Use `UpdateAllowUnknown' for non-model related parameter. (#4961)

Browse Source 
* Use `UpdateAllowUnknown' for non-model related parameter.

Model parameter can not pack an additional boolean value due to binary IO
format.  This commit deals only with non-model related parameter configuration.

* Add tidy command line arg for use-dmlc-gtest.
This commit is contained in:
Jiaming Yuan
2019-10-23 05:50:12 -04:00
committed by GitHub
parent f24be2efb4
commit ac457c56a2
44 changed files with 189 additions and 112 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/cli_main.cc
View File

@@ -12,6 +12,8 @@
#include <xgboost/learner.h>
#include <xgboost/data.h>
#include <xgboost/logging.h>
#include <xgboost/parameter.h>
#include <dmlc/timer.h>
#include <iomanip>
#include <ctime>
@@ -30,7 +32,7 @@ enum CLITask {
kPredict = 2
};
struct CLIParam : public dmlc::Parameter<CLIParam> {
struct CLIParam : public XGBoostParameter<CLIParam> {
/*! \brief the task name */
int task;
/*! \brief whether evaluate training statistics */
@@ -123,7 +125,7 @@ struct CLIParam : public dmlc::Parameter<CLIParam> {
// customized configure function of CLIParam
inline void Configure(const std::vector<std::pair<std::string, std::string> >& _cfg) {
this->cfg = _cfg;
this->InitAllowUnknown(_cfg);
this->UpdateAllowUnknown(_cfg);
for (const auto& kv : _cfg) {
if (!strncmp("eval[", kv.first.c_str(), 5)) {
char evname[256];

4
src/gbm/gblinear.cc
View File

@@ -25,7 +25,7 @@ namespace gbm {
DMLC_REGISTRY_FILE_TAG(gblinear);
// training parameters
struct GBLinearTrainParam : public dmlc::Parameter<GBLinearTrainParam> {
struct GBLinearTrainParam : public XGBoostParameter<GBLinearTrainParam> {
std::string updater;
float tolerance;
size_t max_row_perbatch;
@@ -64,7 +64,7 @@ class GBLinear : public GradientBooster {
if (model_.weight.size() == 0) {
model_.param.InitAllowUnknown(cfg);
}
param_.InitAllowUnknown(cfg);
param_.UpdateAllowUnknown(cfg);
updater_.reset(LinearUpdater::Create(param_.updater, learner_param_));
updater_->Configure(cfg);
monitor_.Init("GBLinear");

4
src/gbm/gbtree.cc
View File

@@ -34,7 +34,7 @@ DMLC_REGISTRY_FILE_TAG(gbtree);
void GBTree::Configure(const Args& cfg) {
this->cfg_ = cfg;
tparam_.InitAllowUnknown(cfg);
tparam_.UpdateAllowUnknown(cfg);
model_.Configure(cfg);
@@ -295,7 +295,7 @@ class Dart : public GBTree {
void Configure(const Args& cfg) override {
GBTree::Configure(cfg);
if (model_.trees.size() == 0) {
dparam_.InitAllowUnknown(cfg);
dparam_.UpdateAllowUnknown(cfg);
}
}

4
src/gbm/gbtree.h
View File

@@ -48,7 +48,7 @@ namespace xgboost {
namespace gbm {
/*! \brief training parameters */
struct GBTreeTrainParam : public dmlc::Parameter<GBTreeTrainParam> {
struct GBTreeTrainParam : public XGBoostParameter<GBTreeTrainParam> {
/*!
* \brief number of parallel trees constructed each iteration
* use this option to support boosted random forest
@@ -95,7 +95,7 @@ struct GBTreeTrainParam : public dmlc::Parameter<GBTreeTrainParam> {
};
/*! \brief training parameters */
struct DartTrainParam : public dmlc::Parameter<DartTrainParam> {
struct DartTrainParam : public XGBoostParameter<DartTrainParam> {
/*! \brief type of sampling algorithm */
int sample_type;
/*! \brief type of normalization algorithm */

25
src/learner.cc
View File

@@ -5,13 +5,8 @@
* \author Tianqi Chen
*/
#include <dmlc/io.h>
#include <dmlc/timer.h>
#include <dmlc/any.h>
#include <xgboost/feature_map.h>
#include <xgboost/learner.h>
#include <xgboost/base.h>
#include <xgboost/logging.h>
#include <xgboost/generic_parameters.h>
#include <dmlc/parameter.h>
#include <algorithm>
#include <iomanip>
#include <limits>
@@ -21,6 +16,12 @@
#include <utility>
#include <vector>
#include "xgboost/feature_map.h"
#include "xgboost/learner.h"
#include "xgboost/base.h"
#include "xgboost/parameter.h"
#include "xgboost/logging.h"
#include "xgboost/generic_parameters.h"
#include "xgboost/host_device_vector.h"
#include "common/common.h"
#include "common/io.h"
@@ -103,7 +104,7 @@ struct LearnerModelParam : public dmlc::Parameter<LearnerModelParam> {
}
};
struct LearnerTrainParam : public dmlc::Parameter<LearnerTrainParam> {
struct LearnerTrainParam : public XGBoostParameter<LearnerTrainParam> {
// data split mode, can be row, col, or none.
DataSplitMode dsplit;
// flag to disable default metric
@@ -155,9 +156,9 @@ class LearnerImpl : public Learner {
auto old_tparam = tparam_;
Args args = {cfg_.cbegin(), cfg_.cend()};
tparam_.InitAllowUnknown(args);
tparam_.UpdateAllowUnknown(args);
generic_param_.InitAllowUnknown(args);
generic_param_.UpdateAllowUnknown(args);
generic_param_.CheckDeprecated();
ConsoleLogger::Configure(args);
@@ -208,7 +209,7 @@ class LearnerImpl : public Learner {
}
void Load(dmlc::Stream* fi) override {
generic_param_.InitAllowUnknown(Args{});
generic_param_.UpdateAllowUnknown(Args{});
tparam_.Init(std::vector<std::pair<std::string, std::string>>{});
// TODO(tqchen) mark deprecation of old format.
common::PeekableInStream fp(fi);
@@ -314,7 +315,7 @@ class LearnerImpl : public Learner {
cfg_.insert(n.cbegin(), n.cend());
Args args = {cfg_.cbegin(), cfg_.cend()};
generic_param_.InitAllowUnknown(args);
generic_param_.UpdateAllowUnknown(args);
gbm_->Configure(args);
obj_->Configure({cfg_.begin(), cfg_.end()});

3
src/linear/coordinate_common.h
View File

@@ -10,6 +10,7 @@
#include <limits>
#include "xgboost/data.h"
#include "xgboost/parameter.h"
#include "./param.h"
#include "../gbm/gblinear_model.h"
#include "../common/random.h"
@@ -17,7 +18,7 @@
namespace xgboost {
namespace linear {
struct CoordinateParam : public dmlc::Parameter<CoordinateParam> {
struct CoordinateParam : public XGBoostParameter<CoordinateParam> {
int top_k;
DMLC_DECLARE_PARAMETER(CoordinateParam) {
DMLC_DECLARE_FIELD(top_k)

4
src/linear/param.h
View File

@@ -5,7 +5,7 @@
*/
#ifndef XGBOOST_LINEAR_PARAM_H_
#define XGBOOST_LINEAR_PARAM_H_
#include <dmlc/parameter.h>
#include "xgboost/parameter.h"
namespace xgboost {
namespace linear {
@@ -20,7 +20,7 @@ enum FeatureSelectorEnum {
kRandom
};
struct LinearTrainParam : public dmlc::Parameter<LinearTrainParam> {
struct LinearTrainParam : public XGBoostParameter<LinearTrainParam> {
/*! \brief learning_rate */
float learning_rate;
/*! \brief regularization weight for L2 norm */

4
src/linear/updater_coordinate.cc
View File

@@ -26,9 +26,9 @@ class CoordinateUpdater : public LinearUpdater {
// set training parameter
void Configure(Args const& args) override {
const std::vector<std::pair<std::string, std::string> > rest {
tparam_.InitAllowUnknown(args)
tparam_.UpdateAllowUnknown(args)
};
cparam_.InitAllowUnknown(rest);
cparam_.UpdateAllowUnknown(rest);
selector_.reset(FeatureSelector::Create(tparam_.feature_selector));
monitor_.Init("CoordinateUpdater");
}

4
src/linear/updater_gpu_coordinate.cu
View File

@@ -1,5 +1,5 @@
/*!
* Copyright 2018 by Contributors
* Copyright 2018-2019 by Contributors
* \author Rory Mitchell
*/
@@ -36,7 +36,7 @@ class GPUCoordinateUpdater : public LinearUpdater { // NOLINT
// set training parameter
void Configure(Args const& args) override {
tparam_.InitAllowUnknown(args);
tparam_.UpdateAllowUnknown(args);
selector_.reset(FeatureSelector::Create(tparam_.feature_selector));
monitor_.Init("GPUCoordinateUpdater");
}

2
src/linear/updater_shotgun.cc
View File

@@ -15,7 +15,7 @@ class ShotgunUpdater : public LinearUpdater {
public:
// set training parameter
void Configure(Args const& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
if (param_.feature_selector != kCyclic &&
param_.feature_selector != kShuffle) {
LOG(FATAL) << "Unsupported feature selector for shotgun updater.\n"

7
src/logging.cc
View File

@@ -5,12 +5,13 @@
* \author Tianqi Chen
*/
#include <rabit/rabit.h>
#include <dmlc/parameter.h>
#include <xgboost/logging.h>
#include <iostream>
#include <map>
#include "xgboost/parameter.h"
#include "xgboost/logging.h"
#if !defined(XGBOOST_STRICT_R_MODE) || XGBOOST_STRICT_R_MODE == 0
// Override logging mechanism for non-R interfaces
void dmlc::CustomLogMessage::Log(const std::string& msg) {
@@ -51,7 +52,7 @@ bool ConsoleLogger::ShouldLog(LogVerbosity verbosity) {
}
void ConsoleLogger::Configure(Args const& args) {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
// Deprecated, but when trying to display deprecation message some R
// tests trying to catch stdout will fail.
if (param_.silent) {

1
src/metric/elementwise_metric.cu
View File

@@ -327,7 +327,6 @@ struct EvalEWiseBase : public Metric {
CHECK_EQ(preds.Size(), info.labels_.Size())
<< "label and prediction size not match, "
<< "hint: use merror or mlogloss for multi-class classification";
const auto ndata = static_cast<omp_ulong>(info.labels_.Size());
int device = tparam_->gpu_id;
auto result =

1
src/metric/metric_common.h
View File

@@ -5,7 +5,6 @@
#ifndef XGBOOST_METRIC_METRIC_COMMON_H_
#define XGBOOST_METRIC_METRIC_COMMON_H_
#include <dmlc/parameter.h>
#include "../common/common.h"
namespace xgboost {

1
src/metric/multiclass_metric.cu
View File

@@ -172,7 +172,6 @@ struct EvalMClassBase : public Metric {
CHECK_GE(nclass, 1U)
<< "mlogloss and merror are only used for multi-class classification,"
<< " use logloss for binary classification";
const auto ndata = static_cast<bst_omp_uint>(info.labels_.Size());
int device = tparam_->gpu_id;
auto result = reducer_.Reduce(*tparam_, device, nclass, info.weights_, info.labels_, preds);

10
src/objective/multiclass_obj.cu
View File

@@ -5,16 +5,18 @@
* \author Tianqi Chen
*/
#include <dmlc/omp.h>
#include <dmlc/parameter.h>
#include <xgboost/data.h>
#include <xgboost/logging.h>
#include <xgboost/objective.h>
#include <vector>
#include <algorithm>
#include <limits>
#include <utility>
#include "xgboost/parameter.h"
#include "xgboost/data.h"
#include "xgboost/logging.h"
#include "xgboost/objective.h"
#include "xgboost/json.h"
#include "../common/common.h"
#include "../common/math.h"
#include "../common/transform.h"

2
src/predictor/gpu_predictor.cu
View File

@@ -1,13 +1,13 @@
/*!
* Copyright 2017-2018 by Contributors
*/
#include <dmlc/parameter.h>
#include <thrust/copy.h>
#include <thrust/device_ptr.h>
#include <thrust/device_vector.h>
#include <thrust/fill.h>
#include <memory>
#include "xgboost/parameter.h"
#include "xgboost/data.h"
#include "xgboost/predictor.h"
#include "xgboost/tree_model.h"

1
src/tree/gpu_hist/row_partitioner.cu
View File

@@ -1,4 +1,3 @@
/*!
* Copyright 2017-2019 XGBoost contributors
*/

6
src/tree/param.h
View File

@@ -7,20 +7,20 @@
#ifndef XGBOOST_TREE_PARAM_H_
#define XGBOOST_TREE_PARAM_H_
#include <dmlc/parameter.h>
#include <xgboost/data.h>
#include <cmath>
#include <cstring>
#include <limits>
#include <string>
#include <vector>
#include "xgboost/parameter.h"
#include "xgboost/data.h"
namespace xgboost {
namespace tree {
/*! \brief training parameters for regression tree */
struct TrainParam : public dmlc::Parameter<TrainParam> {
struct TrainParam : public XGBoostParameter<TrainParam> {
// learning step size for a time
float learning_rate;
// minimum loss change required for a split

16
src/tree/split_evaluator.cc
View File

@@ -5,6 +5,7 @@
*/
#include <dmlc/json.h>
#include <dmlc/registry.h>
#include <algorithm>
#include <unordered_set>
#include <vector>
@@ -15,7 +16,8 @@
#include <utility>
#include "xgboost/logging.h"
#include "xgboost/host_device_vector.h"
#include "xgboost/parameter.h"
#include "param.h"
#include "split_evaluator.h"
#include "../common/common.h"
@@ -67,7 +69,7 @@ bool SplitEvaluator::CheckFeatureConstraint(bst_uint nodeid, bst_uint featureid)
}
//! \brief Encapsulates the parameters for ElasticNet
struct ElasticNetParams : public dmlc::Parameter<ElasticNetParams> {
struct ElasticNetParams : public XGBoostParameter<ElasticNetParams> {
bst_float reg_lambda;
bst_float reg_alpha;
// maximum delta update we can add in weight estimation
@@ -105,7 +107,7 @@ class ElasticNet final : public SplitEvaluator {
}
}
void Init(const Args& args) override {
params_.InitAllowUnknown(args);
params_.UpdateAllowUnknown(args);
}
SplitEvaluator* GetHostClone() const override {
@@ -185,7 +187,7 @@ XGBOOST_REGISTER_SPLIT_EVALUATOR(ElasticNet, "elastic_net")
split evaluator
*/
struct MonotonicConstraintParams
: public dmlc::Parameter<MonotonicConstraintParams> {
: public XGBoostParameter<MonotonicConstraintParams> {
std::vector<bst_int> monotone_constraints;
DMLC_DECLARE_PARAMETER(MonotonicConstraintParams) {
@@ -212,7 +214,7 @@ class MonotonicConstraint final : public SplitEvaluator {
void Init(const Args& args)
override {
inner_->Init(args);
params_.InitAllowUnknown(args);
params_.UpdateAllowUnknown(args);
Reset();
}
@@ -337,7 +339,7 @@ XGBOOST_REGISTER_SPLIT_EVALUATOR(MonotonicConstraint, "monotonic")
split evaluator
*/
struct InteractionConstraintParams
: public dmlc::Parameter<InteractionConstraintParams> {
: public XGBoostParameter<InteractionConstraintParams> {
std::string interaction_constraints;
bst_uint num_feature;
@@ -371,7 +373,7 @@ class InteractionConstraint final : public SplitEvaluator {
void Init(const Args& args)
override {
inner_->Init(args);
params_.InitAllowUnknown(args);
params_.UpdateAllowUnknown(args);
Reset();
}

4
src/tree/tree_model.cc
View File

@@ -423,7 +423,7 @@ XGBOOST_REGISTER_TREE_IO(JsonGenerator, "json")
return new JsonGenerator(fmap, attrs, with_stats);
});
struct GraphvizParam : public dmlc::Parameter<GraphvizParam> {
struct GraphvizParam : public XGBoostParameter<GraphvizParam> {
std::string yes_color;
std::string no_color;
std::string rankdir;
@@ -462,7 +462,7 @@ class GraphvizGenerator : public TreeGenerator {
public:
GraphvizGenerator(FeatureMap const& fmap, std::string const& attrs, bool with_stats) :
TreeGenerator(fmap, with_stats) {
param_.InitAllowUnknown(std::map<std::string, std::string>{});
param_.UpdateAllowUnknown(std::map<std::string, std::string>{});
using KwArg = std::map<std::string, std::map<std::string, std::string>>;
KwArg kwargs;
if (attrs.length() != 0) {

2
src/tree/updater_basemaker-inl.h
View File

@@ -31,7 +31,7 @@ namespace tree {
class BaseMaker: public TreeUpdater {
public:
void Configure(const Args& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
}
protected:

4
src/tree/updater_colmaker.cc
View File

@@ -26,7 +26,7 @@ DMLC_REGISTRY_FILE_TAG(updater_colmaker);
class ColMaker: public TreeUpdater {
public:
void Configure(const Args& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
spliteval_.reset(SplitEvaluator::Create(param_.split_evaluator));
spliteval_->Init(args);
}
@@ -773,7 +773,7 @@ class ColMaker: public TreeUpdater {
class DistColMaker : public ColMaker {
public:
void Configure(const Args& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
pruner_.reset(TreeUpdater::Create("prune", tparam_));
pruner_->Configure(args);
spliteval_.reset(SplitEvaluator::Create(param_.split_evaluator));

9
src/tree/updater_gpu_hist.cu
View File

@@ -16,6 +16,7 @@
#include <vector>
#include "xgboost/host_device_vector.h"
#include "xgboost/parameter.h"
#include "xgboost/span.h"
#include "../common/common.h"
@@ -38,7 +39,7 @@ DMLC_REGISTRY_FILE_TAG(updater_gpu_hist);
// training parameters specific to this algorithm
struct GPUHistMakerTrainParam
: public dmlc::Parameter<GPUHistMakerTrainParam> {
: public XGBoostParameter<GPUHistMakerTrainParam> {
bool single_precision_histogram;
// number of rows in a single GPU batch
int gpu_batch_nrows;
@@ -969,9 +970,9 @@ class GPUHistMakerSpecialised {
public:
GPUHistMakerSpecialised() : initialised_{false}, p_last_fmat_{nullptr} {}
void Configure(const Args& args, GenericParameter const* generic_param) {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
generic_param_ = generic_param;
hist_maker_param_.InitAllowUnknown(args);
hist_maker_param_.UpdateAllowUnknown(args);
device_ = generic_param_->gpu_id;
CHECK_GE(device_, 0) << "Must have at least one device";
@@ -1107,7 +1108,7 @@ class GPUHistMakerSpecialised {
class GPUHistMaker : public TreeUpdater {
public:
void Configure(const Args& args) override {
hist_maker_param_.InitAllowUnknown(args);
hist_maker_param_.UpdateAllowUnknown(args);
float_maker_.reset();
double_maker_.reset();
if (hist_maker_param_.single_precision_histogram) {

2
src/tree/updater_prune.cc
View File

@@ -30,7 +30,7 @@ class TreePruner: public TreeUpdater {
// set training parameter
void Configure(const Args& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
syncher_->Configure(args);
}
// update the tree, do pruning

2
src/tree/updater_quantile_hist.cc
View File

@@ -38,7 +38,7 @@ void QuantileHistMaker::Configure(const Args& args) {
pruner_.reset(TreeUpdater::Create("prune", tparam_));
}
pruner_->Configure(args);
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
is_gmat_initialized_ = false;
// initialize the split evaluator

2
src/tree/updater_refresh.cc
View File

@@ -22,7 +22,7 @@ DMLC_REGISTRY_FILE_TAG(updater_refresh);
class TreeRefresher: public TreeUpdater {
public:
void Configure(const Args& args) override {
param_.InitAllowUnknown(args);
param_.UpdateAllowUnknown(args);
}
char const* Name() const override {
return "refresh";