Avoid resetting seed for every configuration. (#6349)

Author: trivialfis

doc/parameter.rst

@@ -412,6 +412,10 @@ Specify the learning task and the corresponding learning objective. The objectiv
 
   - Random number seed.  This parameter is ignored in R package, use `set.seed()` instead.
 
+* ``seed_per_iteration`` [default=false]
+
+  - Seed PRNG determnisticly via iterator number, this option will be switched on automatically on distributed mode.
+
 ***********************
 Command Line Parameters
 ***********************

include/xgboost/generic_parameters.h

@@ -14,10 +14,11 @@ namespace xgboost {
 struct GenericParameter : public XGBoostParameter<GenericParameter> {
   // Constant representing the device ID of CPU.
   static int32_t constexpr kCpuId = -1;
+  static int64_t constexpr kDefaultSeed = 0;
 
  public:
   // stored random seed
-  int64_t seed;
+  int64_t seed { kDefaultSeed };
   // whether seed the PRNG each iteration
   bool seed_per_iteration;
   // number of threads to use if OpenMP is enabled
@@ -46,7 +47,7 @@ struct GenericParameter : public XGBoostParameter<GenericParameter> {
 
   // declare parameters
   DMLC_DECLARE_PARAMETER(GenericParameter) {
-    DMLC_DECLARE_FIELD(seed).set_default(0).describe(
+    DMLC_DECLARE_FIELD(seed).set_default(kDefaultSeed).describe(
         "Random number seed during training.");
     DMLC_DECLARE_ALIAS(seed, random_state);
     DMLC_DECLARE_FIELD(seed_per_iteration)

src/learner.cc

@@ -202,6 +202,7 @@ DMLC_REGISTER_PARAMETER(LearnerTrainParam);
 DMLC_REGISTER_PARAMETER(GenericParameter);
 
 int constexpr GenericParameter::kCpuId;
+int64_t constexpr GenericParameter::kDefaultSeed;
 
 void GenericParameter::ConfigureGpuId(bool require_gpu) {
 #if defined(XGBOOST_USE_CUDA)
@@ -239,6 +240,9 @@ using ThreadLocalPredictionCache =
     dmlc::ThreadLocalStore<std::map<Learner const *, PredictionContainer>>;
 
 class LearnerConfiguration : public Learner {
+ private:
+  std::mutex config_lock_;
+
  protected:
   static std::string const kEvalMetric;  // NOLINT
 
@@ -252,7 +256,6 @@ class LearnerConfiguration : public Learner {
   LearnerModelParam learner_model_param_;
   LearnerTrainParam tparam_;
   std::vector<std::string> metric_names_;
-  std::mutex config_lock_;
 
  public:
   explicit LearnerConfiguration(std::vector<std::shared_ptr<DMatrix> > cache)
@@ -283,7 +286,11 @@ class LearnerConfiguration : public Learner {
 
     tparam_.UpdateAllowUnknown(args);
     auto mparam_backup = mparam_;
+
     mparam_.UpdateAllowUnknown(args);
+
+    auto initialized = generic_parameters_.GetInitialised();
+    auto old_seed = generic_parameters_.seed;
     generic_parameters_.UpdateAllowUnknown(args);
     generic_parameters_.CheckDeprecated();
 
@@ -297,7 +304,9 @@ class LearnerConfiguration : public Learner {
     }
 
     // set seed only before the model is initialized
-    common::GlobalRandom().seed(generic_parameters_.seed);
+    if (!initialized || generic_parameters_.seed != old_seed) {
+      common::GlobalRandom().seed(generic_parameters_.seed);
+    }
 
     // must precede configure gbm since num_features is required for gbm
     this->ConfigureNumFeatures();

tests/cpp/test_learner.cc

@@ -11,6 +11,7 @@
 #include <xgboost/version_config.h>
 #include "xgboost/json.h"
 #include "../../src/common/io.h"
+#include "../../src/common/random.h"
 
 namespace xgboost {
 
@@ -333,4 +334,26 @@ TEST(Learner, Seed) {
   ASSERT_EQ(std::to_string(seed),
             get<String>(config["learner"]["generic_param"]["seed"]));
 }
+
+TEST(Learner, ConstantSeed) {
+  auto m = RandomDataGenerator{10, 10, 0}.GenerateDMatrix(true);
+  std::unique_ptr<Learner> learner{Learner::Create({m})};
+  learner->Configure();  // seed the global random
+
+  std::uniform_real_distribution<float> dist;
+  auto& rng = common::GlobalRandom();
+  float v_0 = dist(rng);
+
+  learner->SetParam("", "");
+  learner->Configure();  // check configure doesn't change the seed.
+  float v_1 = dist(rng);
+  CHECK_NE(v_0, v_1);
+
+  {
+    rng.seed(GenericParameter::kDefaultSeed);
+    std::uniform_real_distribution<float> dist;
+    float v_2 = dist(rng);
+    CHECK_EQ(v_0, v_2);
+  }
+}
 }  // namespace xgboost