xgboost/old_src/learner/learner-inl.hpp
2016-01-16 10:24:00 -08:00

548 lines
19 KiB
C++

/*!
* Copyright 2014 by Contributors
* \file learner-inl.hpp
* \brief learning algorithm
* \author Tianqi Chen
*/
#ifndef XGBOOST_LEARNER_LEARNER_INL_HPP_
#define XGBOOST_LEARNER_LEARNER_INL_HPP_
#include <algorithm>
#include <vector>
#include <utility>
#include <string>
#include <limits>
#include "../sync/sync.h"
#include "../utils/io.h"
#include "./objective.h"
#include "./evaluation.h"
#include "../gbm/gbm.h"
namespace xgboost {
/*! \brief namespace for learning algorithm */
namespace learner {
/*!
* \brief learner that performs gradient boosting for a specific objective function.
* It does training and prediction.
*/
class BoostLearner : public rabit::Serializable {
public:
BoostLearner(void) {
obj_ = NULL;
gbm_ = NULL;
name_obj_ = "reg:linear";
name_gbm_ = "gbtree";
silent = 0;
prob_buffer_row = 1.0f;
distributed_mode = 0;
updater_mode = 0;
pred_buffer_size = 0;
seed_per_iteration = 0;
seed = 0;
save_base64 = 0;
}
virtual ~BoostLearner(void) {
if (obj_ != NULL) delete obj_;
if (gbm_ != NULL) delete gbm_;
}
/*!
* \brief add internal cache space for mat, this can speedup prediction for matrix,
* please cache prediction for training and eval data
* warning: if the model is loaded from file from some previous training history
* set cache data must be called with exactly SAME
* data matrices to continue training otherwise it will cause error
* \param mats array of pointers to matrix whose prediction result need to be cached
*/
inline void SetCacheData(const std::vector<DMatrix*>& mats) {
utils::Assert(cache_.size() == 0, "can only call cache data once");
// assign buffer index
size_t buffer_size = 0;
for (size_t i = 0; i < mats.size(); ++i) {
bool dupilicate = false;
for (size_t j = 0; j < i; ++j) {
if (mats[i] == mats[j]) dupilicate = true;
}
if (dupilicate) continue;
// set mats[i]'s cache learner pointer to this
mats[i]->cache_learner_ptr_ = this;
cache_.push_back(CacheEntry(mats[i], buffer_size, mats[i]->info.num_row()));
buffer_size += mats[i]->info.num_row();
}
char str_temp[25];
utils::SPrintf(str_temp, sizeof(str_temp), "%lu",
static_cast<unsigned long>(buffer_size)); // NOLINT(*)
this->SetParam("num_pbuffer", str_temp);
this->pred_buffer_size = buffer_size;
}
/*!
* \brief set parameters from outside
* \param name name of the parameter
* \param val value of the parameter
*/
inline void SetParam(const char *name, const char *val) {
using namespace std;
// in this version, bst: prefix is no longer required
if (strncmp(name, "bst:", 4) != 0) {
std::string n = "bst:"; n += name;
this->SetParam(n.c_str(), val);
}
if (!strcmp(name, "silent")) silent = atoi(val);
if (!strcmp(name, "dsplit")) {
if (!strcmp(val, "col")) {
this->SetParam("updater", "distcol");
distributed_mode = 1;
} else if (!strcmp(val, "row")) {
this->SetParam("updater", "grow_histmaker,prune");
distributed_mode = 2;
} else {
utils::Error("%s is invalid value for dsplit, should be row or col", val);
}
}
if (!strcmp(name, "updater_mode")) updater_mode = atoi(val);
if (!strcmp(name, "prob_buffer_row")) {
prob_buffer_row = static_cast<float>(atof(val));
utils::Check(distributed_mode == 0,
"prob_buffer_row can only be used in single node mode so far");
this->SetParam("updater", "grow_colmaker,refresh,prune");
}
if (!strcmp(name, "eval_metric")) evaluator_.AddEval(val);
if (!strcmp("seed", name)) {
seed = atoi(val); random::Seed(seed);
}
if (!strcmp("seed_per_iter", name)) seed_per_iteration = atoi(val);
if (!strcmp("save_base64", name)) save_base64 = atoi(val);
if (!strcmp(name, "num_class")) {
this->SetParam("num_output_group", val);
}
if (!strcmp(name, "nthread")) {
omp_set_num_threads(atoi(val));
}
if (gbm_ == NULL) {
if (!strcmp(name, "objective")) name_obj_ = val;
if (!strcmp(name, "booster")) name_gbm_ = val;
mparam.SetParam(name, val);
}
if (gbm_ != NULL) gbm_->SetParam(name, val);
if (obj_ != NULL) obj_->SetParam(name, val);
if (gbm_ == NULL || obj_ == NULL) {
cfg_.push_back(std::make_pair(std::string(name), std::string(val)));
}
}
// this is an internal function
// initialize the trainer, called at InitModel and LoadModel
inline void InitTrainer(bool calc_num_feature = true) {
if (calc_num_feature) {
// estimate feature bound
unsigned num_feature = 0;
for (size_t i = 0; i < cache_.size(); ++i) {
num_feature = std::max(num_feature,
static_cast<unsigned>(cache_[i].mat_->info.num_col()));
}
// run allreduce on num_feature to find the maximum value
rabit::Allreduce<rabit::op::Max>(&num_feature, 1);
if (num_feature > mparam.num_feature) mparam.num_feature = num_feature;
}
char str_temp[25];
utils::SPrintf(str_temp, sizeof(str_temp), "%d", mparam.num_feature);
this->SetParam("bst:num_feature", str_temp);
}
/*!
* \brief initialize the model
*/
inline void InitModel(void) {
this->InitTrainer();
// initialize model
this->InitObjGBM();
// reset the base score
mparam.base_score = obj_->ProbToMargin(mparam.base_score);
// initialize GBM model
gbm_->InitModel();
}
/*!
* \brief load model from stream
* \param fi input stream
* \param calc_num_feature whether call InitTrainer with calc_num_feature
*/
inline void LoadModel(utils::IStream &fi, // NOLINT(*)
bool calc_num_feature = true) {
utils::Check(fi.Read(&mparam, sizeof(ModelParam)) != 0,
"BoostLearner: wrong model format");
{
// backward compatibility code for compatible with old model type
// for new model, Read(&name_obj_) is suffice
uint64_t len;
utils::Check(fi.Read(&len, sizeof(len)) != 0, "BoostLearner: wrong model format");
if (len >= std::numeric_limits<unsigned>::max()) {
int gap;
utils::Check(fi.Read(&gap, sizeof(gap)) != 0, "BoostLearner: wrong model format");
len = len >> static_cast<uint64_t>(32UL);
}
if (len != 0) {
name_obj_.resize(len);
utils::Check(fi.Read(&name_obj_[0], len) != 0, "BoostLearner: wrong model format");
}
}
utils::Check(fi.Read(&name_gbm_), "BoostLearner: wrong model format");
// delete existing gbm if any
if (obj_ != NULL) delete obj_;
if (gbm_ != NULL) delete gbm_;
this->InitTrainer(calc_num_feature);
this->InitObjGBM();
char tmp[32];
utils::SPrintf(tmp, sizeof(tmp), "%u", mparam.num_class);
obj_->SetParam("num_class", tmp);
gbm_->LoadModel(fi, mparam.saved_with_pbuffer != 0);
if (mparam.saved_with_pbuffer == 0) {
gbm_->ResetPredBuffer(pred_buffer_size);
}
}
// rabit load model from rabit checkpoint
virtual void Load(rabit::Stream *fi) {
// for row split, we should not keep pbuffer
this->LoadModel(*fi, false);
}
// rabit save model to rabit checkpoint
virtual void Save(rabit::Stream *fo) const {
// for row split, we should not keep pbuffer
this->SaveModel(*fo, distributed_mode != 2);
}
/*!
* \brief load model from file
* \param fname file name
*/
inline void LoadModel(const char *fname) {
utils::IStream *fi = utils::IStream::Create(fname, "r");
std::string header; header.resize(4);
// check header for different binary encode
// can be base64 or binary
utils::Check(fi->Read(&header[0], 4) != 0, "invalid model");
// base64 format
if (header == "bs64") {
utils::Base64InStream bsin(fi);
bsin.InitPosition();
this->LoadModel(bsin, true);
} else if (header == "binf") {
this->LoadModel(*fi, true);
} else {
delete fi;
fi = utils::IStream::Create(fname, "r");
this->LoadModel(*fi, true);
}
delete fi;
}
inline void SaveModel(utils::IStream &fo, bool with_pbuffer) const { // NOLINT(*)
ModelParam p = mparam;
p.saved_with_pbuffer = static_cast<int>(with_pbuffer);
fo.Write(&p, sizeof(ModelParam));
fo.Write(name_obj_);
fo.Write(name_gbm_);
gbm_->SaveModel(fo, with_pbuffer);
}
/*!
* \brief save model into file
* \param fname file name
* \param with_pbuffer whether save pbuffer together
*/
inline void SaveModel(const char *fname, bool with_pbuffer) const {
utils::IStream *fo = utils::IStream::Create(fname, "w");
if (save_base64 != 0 || !strcmp(fname, "stdout")) {
fo->Write("bs64\t", 5);
utils::Base64OutStream bout(fo);
this->SaveModel(bout, with_pbuffer);
bout.Finish('\n');
} else {
fo->Write("binf", 4);
this->SaveModel(*fo, with_pbuffer);
}
delete fo;
}
/*!
* \brief check if data matrix is ready to be used by training,
* if not initialize it
* \param p_train pointer to the matrix used by training
*/
inline void CheckInit(DMatrix *p_train) {
int ncol = static_cast<int>(p_train->info.info.num_col);
std::vector<bool> enabled(ncol, true);
// set max row per batch to limited value
// in distributed mode, use safe choice otherwise
size_t max_row_perbatch = std::numeric_limits<size_t>::max();
if (updater_mode != 0 || distributed_mode == 2) {
max_row_perbatch = 32UL << 10UL;
}
// initialize column access
p_train->fmat()->InitColAccess(enabled,
prob_buffer_row,
max_row_perbatch);
const int kMagicPage = 0xffffab02;
// check, if it is DMatrixPage, then use hist maker
if (p_train->magic == kMagicPage) {
this->SetParam("updater", "grow_histmaker,prune");
}
}
/*!
* \brief update the model for one iteration
* \param iter current iteration number
* \param train reference to the data matrix
*/
inline void UpdateOneIter(int iter, const DMatrix &train) {
if (seed_per_iteration != 0 || rabit::IsDistributed()) {
random::Seed(this->seed * kRandSeedMagic + iter);
}
this->PredictRaw(train, &preds_);
obj_->GetGradient(preds_, train.info, iter, &gpair_);
gbm_->DoBoost(train.fmat(), this->FindBufferOffset(train), train.info.info, &gpair_);
}
/*!
* \brief whether model allow lazy checkpoint
*/
inline bool AllowLazyCheckPoint(void) const {
return gbm_->AllowLazyCheckPoint();
}
/*!
* \brief evaluate the model for specific iteration
* \param iter iteration number
* \param evals datas i want to evaluate
* \param evname name of each dataset
* \return a string corresponding to the evaluation result
*/
inline std::string EvalOneIter(int iter,
const std::vector<const DMatrix*> &evals,
const std::vector<std::string> &evname) {
std::string res;
char tmp[256];
utils::SPrintf(tmp, sizeof(tmp), "[%d]", iter);
res = tmp;
for (size_t i = 0; i < evals.size(); ++i) {
this->PredictRaw(*evals[i], &preds_);
obj_->EvalTransform(&preds_);
res += evaluator_.Eval(evname[i].c_str(), preds_, evals[i]->info, distributed_mode == 2);
}
return res;
}
/*!
* \brief simple evaluation function, with a specified metric
* \param data input data
* \param metric name of metric
* \return a pair of <evaluation name, result>
*/
std::pair<std::string, float> Evaluate(const DMatrix &data, std::string metric) {
if (metric == "auto") metric = obj_->DefaultEvalMetric();
IEvaluator *ev = CreateEvaluator(metric.c_str());
this->PredictRaw(data, &preds_);
obj_->EvalTransform(&preds_);
float res = ev->Eval(preds_, data.info);
delete ev;
return std::make_pair(metric, res);
}
/*!
* \brief get prediction
* \param data input data
* \param output_margin whether to only predict margin value instead of transformed prediction
* \param out_preds output vector that stores the prediction
* \param ntree_limit limit number of trees used for boosted tree
* predictor, when it equals 0, this means we are using all the trees
* \param pred_leaf whether to only predict the leaf index of each tree in a boosted tree predictor
*/
inline void Predict(const DMatrix &data,
bool output_margin,
std::vector<float> *out_preds,
unsigned ntree_limit = 0,
bool pred_leaf = false) const {
if (pred_leaf) {
gbm_->PredictLeaf(data.fmat(), data.info.info, out_preds, ntree_limit);
} else {
this->PredictRaw(data, out_preds, ntree_limit);
if (!output_margin) {
obj_->PredTransform(out_preds);
}
}
}
/*!
* \brief online prediction function, predict score for one instance at a time
* NOTE: use the batch prediction interface if possible, batch prediction is usually
* more efficient than online prediction
* This function is NOT threadsafe, make sure you only call from one thread
*
* \param inst the instance you want to predict
* \param output_margin whether to only predict margin value instead of transformed prediction
* \param out_preds output vector to hold the predictions
* \param ntree_limit limit the number of trees used in prediction
* \sa Predict
*/
inline void Predict(const SparseBatch::Inst &inst,
bool output_margin,
std::vector<float> *out_preds,
unsigned ntree_limit = 0) const {
gbm_->Predict(inst, out_preds, ntree_limit);
if (out_preds->size() == 1) {
(*out_preds)[0] += mparam.base_score;
}
if (!output_margin) {
obj_->PredTransform(out_preds);
}
}
/*! \brief dump model out */
inline std::vector<std::string> DumpModel(const utils::FeatMap& fmap, int option) {
return gbm_->DumpModel(fmap, option);
}
protected:
/*!
* \brief initialize the objective function and GBM,
* if not yet done
*/
inline void InitObjGBM(void) {
if (obj_ != NULL) return;
utils::Assert(gbm_ == NULL, "GBM and obj should be NULL");
obj_ = CreateObjFunction(name_obj_.c_str());
gbm_ = gbm::CreateGradBooster(name_gbm_.c_str());
this->InitAdditionDefaultParam();
// set parameters
for (size_t i = 0; i < cfg_.size(); ++i) {
obj_->SetParam(cfg_[i].first.c_str(), cfg_[i].second.c_str());
gbm_->SetParam(cfg_[i].first.c_str(), cfg_[i].second.c_str());
}
if (evaluator_.Size() == 0) {
evaluator_.AddEval(obj_->DefaultEvalMetric());
}
}
/*!
* \brief additional default value for specific objs
*/
inline void InitAdditionDefaultParam(void) {
if (name_obj_ == "count:poisson") {
obj_->SetParam("max_delta_step", "0.7");
gbm_->SetParam("max_delta_step", "0.7");
}
}
/*!
* \brief get un-transformed prediction
* \param data training data matrix
* \param out_preds output vector that stores the prediction
* \param ntree_limit limit number of trees used for boosted tree
* predictor, when it equals 0, this means we are using all the trees
*/
inline void PredictRaw(const DMatrix &data,
std::vector<float> *out_preds,
unsigned ntree_limit = 0) const {
gbm_->Predict(data.fmat(), this->FindBufferOffset(data),
data.info.info, out_preds, ntree_limit);
// add base margin
std::vector<float> &preds = *out_preds;
const bst_omp_uint ndata = static_cast<bst_omp_uint>(preds.size());
if (data.info.base_margin.size() != 0) {
utils::Check(preds.size() == data.info.base_margin.size(),
"base_margin.size does not match with prediction size");
#pragma omp parallel for schedule(static)
for (bst_omp_uint j = 0; j < ndata; ++j) {
preds[j] += data.info.base_margin[j];
}
} else {
#pragma omp parallel for schedule(static)
for (bst_omp_uint j = 0; j < ndata; ++j) {
preds[j] += mparam.base_score;
}
}
}
/*! \brief training parameter for regression */
struct ModelParam{
/* \brief global bias */
float base_score;
/* \brief number of features */
unsigned num_feature;
/* \brief number of classes, if it is multi-class classification */
int num_class;
/*! \brief whether the model itself is saved with pbuffer */
int saved_with_pbuffer;
/*! \brief reserved field */
int reserved[30];
/*! \brief constructor */
ModelParam(void) {
std::memset(this, 0, sizeof(ModelParam));
base_score = 0.5f;
num_feature = 0;
num_class = 0;
saved_with_pbuffer = 0;
}
/*!
* \brief set parameters from outside
* \param name name of the parameter
* \param val value of the parameter
*/
inline void SetParam(const char *name, const char *val) {
using namespace std;
if (!strcmp("base_score", name)) base_score = static_cast<float>(atof(val));
if (!strcmp("num_class", name)) num_class = atoi(val);
if (!strcmp("bst:num_feature", name)) num_feature = atoi(val);
}
};
// data fields
// stored random seed
int seed;
// whether seed the PRNG each iteration
// this is important for restart from existing iterations
// default set to no, but will auto switch on in distributed mode
int seed_per_iteration;
// save model in base64 encoding
int save_base64;
// silent during training
int silent;
// distributed learning mode, if any, 0:none, 1:col, 2:row
int distributed_mode;
// updater mode, 0:normal, reserved for internal test
int updater_mode;
// cached size of predict buffer
size_t pred_buffer_size;
// maximum buffered row value
float prob_buffer_row;
// evaluation set
EvalSet evaluator_;
// model parameter
ModelParam mparam;
// gbm model that back everything
gbm::IGradBooster *gbm_;
// name of gbm model used for training
std::string name_gbm_;
// objective function
IObjFunction *obj_;
// name of objective function
std::string name_obj_;
// configurations
std::vector< std::pair<std::string, std::string> > cfg_;
// temporal storages for prediction
std::vector<float> preds_;
// gradient pairs
std::vector<bst_gpair> gpair_;
protected:
// magic number to transform random seed
static const int kRandSeedMagic = 127;
// cache entry object that helps handle feature caching
struct CacheEntry {
const DMatrix *mat_;
size_t buffer_offset_;
size_t num_row_;
CacheEntry(const DMatrix *mat, size_t buffer_offset, size_t num_row)
:mat_(mat), buffer_offset_(buffer_offset), num_row_(num_row) {}
};
// find internal buffer offset for certain matrix, if not exist, return -1
inline int64_t FindBufferOffset(const DMatrix &mat) const {
for (size_t i = 0; i < cache_.size(); ++i) {
if (cache_[i].mat_ == &mat && mat.cache_learner_ptr_ == this) {
if (cache_[i].num_row_ == mat.info.num_row()) {
return static_cast<int64_t>(cache_[i].buffer_offset_);
}
}
}
return -1;
}
// data structure field
/*! \brief the entries indicates that we have internal prediction cache */
std::vector<CacheEntry> cache_;
};
} // namespace learner
} // namespace xgboost
#endif // XGBOOST_LEARNER_LEARNER_INL_HPP_