enable max_memory_size in scikit predict

Xtra-Computing · Aug 4, 2018 · 158cdc5 · 158cdc5
1 parent efbf58b
commit 158cdc5
Show file tree

Hide file tree

Showing 3 changed files with 20 additions and 6 deletions.
diff --git a/python/thundersvmScikit.py b/python/thundersvmScikit.py
@@ -72,6 +72,8 @@ def __init__(self, kernel, degree,
         self.max_mem_size = max_mem_size
         thundersvm.model_new.restype = c_void_p
         self.model = thundersvm.model_new(SVM_TYPE.index(self._impl))
+        if self.max_mem_size != -1:
+            thundersvm.set_memory_size(c_void_p(self.model), self.max_mem_size)
 
     def label_validate(self, y):
 
@@ -238,7 +240,6 @@ def _validate_for_predict(self, X):
         return X
 
     def predict(self, X):
-
         X = self._validate_for_predict(X)
         predict = self._sparse_predict if self._sparse else self._dense_predict
         return predict(X)
@@ -294,7 +295,7 @@ def _dense_predict(self, X):
         thundersvm.dense_predict(
             samples, features, data,
             c_void_p(self.model),
-            self.predict_label_ptr)
+            self.predict_label_ptr, self.verbose)
 
         self.predict_label = np.array([self.predict_label_ptr[index] for index in range(0, X.shape[0])])
         return self.predict_label
@@ -310,7 +311,7 @@ def _sparse_predict(self, X):
         thundersvm.sparse_predict(
             X.shape[0], data, indptr, indices,
             c_void_p(self.model),
-            self.predict_label_ptr)
+            self.predict_label_ptr, self.verbose)
 
         predict_label = [self.predict_label_ptr[index] for index in range(0, X.shape[0])]
 

diff --git a/src/thundersvm/model/svmmodel.cpp b/src/thundersvm/model/svmmodel.cpp
@@ -334,5 +334,6 @@ int SvmModel::get_working_set_size(int n_instances, int n_features) {
 }
 
 void SvmModel::set_max_memory_size(size_t size) {
-    this->param.max_mem_size = size;
+    if(size > 0)
+		this->param.max_mem_size = static_cast<size_t>(size) << 20;
 }
diff --git a/src/thundersvm/thundersvm-scikit.cpp b/src/thundersvm/thundersvm-scikit.cpp
@@ -121,7 +121,11 @@ extern "C" {
         n_classes[0] = model->get_n_classes();
     }
 
-    int sparse_predict(int row_size, float* val, int* row_ptr, int* col_ptr, SvmModel *model, float* predict_label){
+    int sparse_predict(int row_size, float* val, int* row_ptr, int* col_ptr, SvmModel *model, float* predict_label, int verbose){
+        if(verbose)
+            el::Loggers::reconfigureAllLoggers(el::ConfigurationType::Enabled, "true");
+        else
+            el::Loggers::reconfigureAllLoggers(el::ConfigurationType::Enabled, "false");
         DataSet predict_dataset;
         predict_dataset.load_from_sparse(row_size, val, row_ptr, col_ptr, (float *)NULL);
         vector<float_type> predict_y;
@@ -219,7 +223,11 @@ extern "C" {
 
     }
 
-    int dense_predict(int row_size, int features, float* data, SvmModel *model, float* predict_label){
+    int dense_predict(int row_size, int features, float* data, SvmModel *model, float* predict_label, int verbose){
+        if(verbose)
+            el::Loggers::reconfigureAllLoggers(el::ConfigurationType::Enabled, "true");
+        else
+            el::Loggers::reconfigureAllLoggers(el::ConfigurationType::Enabled, "false");
         DataSet predict_dataset;
         predict_dataset.load_from_dense(row_size, features, data, (float*) NULL);
         vector<float_type> predict_y;
@@ -333,4 +341,8 @@ extern "C" {
     void get_n_classes(SvmModel *model, int *n_classes){
         n_classes[0] = model->get_n_classes();
     }
+
+    void set_memory_size(SvmModel *model, int m_size){
+        model->set_max_memory_size(m_size);
+    }
 }