Merge pull request #296 from gAldeia/bug_fixes

Bug fixes in verbosity and logfile
cavalab · Nov 18, 2023 · 459e021 · 459e021
2 parents 51816f4 + b5bc524
commit 459e021
Show file tree

Hide file tree

Showing 8 changed files with 70 additions and 40 deletions.
diff --git a/configure b/configure
@@ -38,7 +38,7 @@ elif [ "$1" == "lpc_cuda" ] ; then
     BUILD_DIR="buildGPU"
     OMP=OFF
     EXTRA_FLAGS="-DCORE_USE_CUDA=ON"
-elif [ "$1" == "tests" ||  "$1" == "test"  ] ; then 
+elif [[ "$1" == "tests" ||  "$1" == "test"  ]] ; then 
     EXTRA_FLAGS="-DGTEST=ON"
 elif [ "$1" == "gpu" ] ; then
     BUILD_DIR="buildGPU"

diff --git a/feat/feat.py b/feat/feat.py
@@ -52,9 +52,9 @@ class Feat(BaseEstimator):
         'a': all
         'b': boolean only
         'f': floating point only
-    functions: string, optional (default: "")
+    functions: list[string], optional (default: [])
         A comma-separated string of operators to use to build features. 
-        If functions="", all the available functions are used. 
+        If functions=[], all the available functions are used. 
         Options: +, -, *, /, ^2, ^3, sqrt, sin, cos, exp, log, ^, logit, tanh, 
         gauss, relu, split, split_c, b2f, c2f, and, or, not, xor, =, <, <=, >,
         >=, if, ite
@@ -398,6 +398,9 @@ def _check_shape(self, X):
     def get_representation(self): return self.cfeat_.get_representation()
     def get_model(self, sort=True): return self.cfeat_.get_model(sort)
     def get_coefs(self): return self.cfeat_.get_coefs()
+    def get_n_params(self): return self.cfeat_.get_n_params()
+    def get_dim(self): return self.cfeat_.get_dim()
+    def get_n_nodes(self): return self.cfeat_.get_n_nodes()
 
 class FeatRegressor(Feat):
     """Convenience method that enforces regression options."""

diff --git a/src/feat.cc b/src/feat.cc
@@ -716,6 +716,13 @@ void Feat::run_generation(unsigned int g,
     if (params.max_stall > 0)
         update_stall_count(stall_count, updated_best);
 
+    if ( (use_arch || params.verbosity>1) || !logfile.empty()) {
+        // set objectives to make sure they are reported in log/verbose/arch
+        #pragma omp parallel for
+        for (unsigned int i=0; i<pop.size(); ++i)
+            pop.individuals.at(i).set_obj(params.objectives);
+    }
+
     logger.log("update archive...",2);
     if (use_arch) 
         archive.update(pop,params);
@@ -725,6 +732,9 @@ void Feat::run_generation(unsigned int g,
     else if(params.verbosity == 1)
         printProgress(fraction);
 
+    if (!logfile.empty())
+        log_stats(log);
+
     if (save_pop > 1)
         pop.save(this->logfile+".pop.gen" + 
                     to_string(params.current_gen) + ".json");
@@ -1357,7 +1367,8 @@ void Feat::calculate_stats(const DataRef& d)
     ArrayXf Complexities(this->pop.size()); 
     i = 0; 
     for (auto& p : this->pop.individuals)
-    { 
+    {
+        // Calculate to assure it gets reported in stats (even if's not used as an obj)
         Complexities(i) = p.get_complexity(); 
         ++i;
     }
@@ -1520,37 +1531,37 @@ void Feat::print_stats(std::ofstream& log, float fraction)
     }
 
     std::cout <<"\n\n";
-
-    if (!logfile.empty())
-    {
-        // print stats in tabular format
-        string sep = ",";
-        if (params.current_gen == 0) // print header
-        {
-            log << "generation"     << sep
-                << "time"           << sep
-                << "min_loss"       << sep 
-                << "min_loss_val"   << sep 
-                << "med_loss"       << sep 
-                << "med_loss_val"   << sep 
-                << "med_size"       << sep 
-                << "med_complexity" << sep 
-                << "med_num_params" << sep
-                << "med_dim\n";
-        }
+}
 
-        log << params.current_gen  << sep
-            << timer.Elapsed().count() << sep
-            << stats.min_loss.back()          << sep
-            << this->min_loss_v        << sep
-            << stats.med_loss.back()  << sep
-            << stats.med_loss_v.back() << sep
-            << stats.med_size.back()   << sep
-            << stats.med_complexity.back() << sep
-            << stats.med_num_params.back() << sep
-            << stats.med_dim.back()        << "\n"; 
-    } 
+void Feat::log_stats(std::ofstream& log)
+{
+    // print stats in tabular format
+    string sep = ",";
+    if (params.current_gen == 0) // print header
+    {
+        log << "generation"     << sep
+            << "time"           << sep
+            << "min_loss"       << sep 
+            << "min_loss_val"   << sep 
+            << "med_loss"       << sep 
+            << "med_loss_val"   << sep 
+            << "med_size"       << sep 
+            << "med_complexity" << sep 
+            << "med_num_params" << sep
+            << "med_dim"        << "\n";
+    }
+    log << params.current_gen          << sep
+        << timer.Elapsed().count()     << sep
+        << stats.min_loss.back()       << sep
+        << this->min_loss_v            << sep
+        << stats.med_loss.back()       << sep
+        << stats.med_loss_v.back()     << sep
+        << stats.med_size.back()       << sep
+        << stats.med_complexity.back() << sep
+        << stats.med_num_params.back() << sep
+        << stats.med_dim.back()        << "\n"; 
 }
+
 //TODO: replace these with json
 json Feat::get_stats()
 { 

diff --git a/src/feat.h b/src/feat.h
@@ -433,7 +433,8 @@ class Feat
         void calculate_stats(const DataRef& d);
         void print_stats(std::ofstream& log,
                          float fraction);      
-
+        void log_stats(std::ofstream& log);
+
         // gets weights via univariate initial models
         vector<float> univariate_initial_model(DataRef &d, int n_feats);
         /// method to fit inital ml model            

diff --git a/src/pop/archive.cc b/src/pop/archive.cc
@@ -78,10 +78,6 @@ namespace FT{
 
             vector<Individual> tmp = pop.individuals;
 
-            #pragma omp parallel for
-            for (unsigned int i=0; i<tmp.size(); ++i)
-                tmp.at(i).set_obj(params.objectives);
-
             for (const auto& p : individuals)
                 tmp.push_back(p);
 

diff --git a/src/pybind.cc b/src/pybind.cc
@@ -149,6 +149,9 @@ PYBIND11_MODULE(_feat, m)
         .def("save", &Feat::save)
         .def("load", &Feat::load)
         .def("get_representation", &Feat::get_representation)
+        .def("get_n_params", &Feat::get_n_params)
+        .def("get_dim", &Feat::get_dim)
+        .def("get_n_nodes", &Feat::get_n_nodes)
         .def("get_model", &Feat::get_model, py::arg("sort") = true)
         .def("get_eqn", &Feat::get_eqn, py::arg("sort") = true)
         ;

diff --git a/src/sel/tournament.cc b/src/sel/tournament.cc
@@ -13,7 +13,7 @@ namespace FT{
          * @class Tournament
          */
 
-        Tournament::Tournament(bool surv){ name = "nsga2"; survival = surv; };
+        Tournament::Tournament(bool surv){ name = "tournament"; survival = surv; };
 
         Tournament::~Tournament(){}
 

diff --git a/tests/selectionTests.cc b/tests/selectionTests.cc
@@ -1,10 +1,22 @@
 #include "testsHeader.h"
 
-TEST(Selection, SelectionOperator)
+class SelectionTest : public testing::TestWithParam<std::string> {
+protected:
+    void SetUp() override {
+        selectionType = GetParam();
+    }
+
+    std::string selectionType;
+};
+
+TEST_P(SelectionTest, SelectionOperator)
 {
     Feat feat = make_estimator(100, 100, "LinearRidgeRegression", false, 1, 666);
     feat.set_scorer("mae");
 
+    feat.set_selection(selectionType);
+    ASSERT_STREQ(selectionType.c_str(), feat.selector.get_type().c_str());
+
     MatrixXf X(7,2); 
     X << 0,1,  
          0.47942554,0.87758256,  
@@ -58,3 +70,7 @@ TEST(Selection, SelectionOperator)
 
     ASSERT_EQ(parents.size(), feat.get_pop_size());
 }
+
+INSTANTIATE_TEST_SUITE_P(AllSelectionTypes, SelectionTest,
+    testing::Values("lexicase", "fair_lexicase", "simanneal", "tournament",
+                    "offspring", "random", "nsga2"));