def saveParams(self, best_estimator):
#outdir = OutputMgr.checkCreateDSDir(self.estimator.dataset.name, self.estimator.nick)
outdirI = OutputMgr.getOutIncludeDir()
myFile = open(f"{outdirI}SVM_params.h", "w+")
myFile.write(f"#ifndef SVM_PARAMS_H\n")
myFile.write(f"#define SVM_PARAMS_H\n\n")
if self.estimator.is_regr == False:
myFile.write(f"#define N_CLASS {self.estimator.n_classes}\n\n")
myFile.write(
f"#define WEIGTH_DIM {best_estimator.coef_.shape[0]}\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(
f"#define N_FEATURE {best_estimator.coef_.shape[1]}\n")
myFile.write(f"#endif\n\n")
else:
myFile.write(f"#define N_CLASS 0\n\n")
myFile.write(f"#define WEIGTH_DIM 1\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(
f"#define N_FEATURE {best_estimator.coef_.shape[1]}\n")
myFile.write(f"#endif\n\n")
if self.estimator.is_regr == False:
myFile.write(
f"extern float support_vectors[WEIGTH_DIM][N_FEATURE];\n")
else: #Same as above as a declaration
myFile.write(
f"extern float support_vectors[WEIGTH_DIM][N_FEATURE];\n")
myFile.write(f"extern float bias[WEIGTH_DIM];\n")
myFile.write(f"\n#endif")
myFile.close()
outdirS = OutputMgr.getOutSourceDir()
myFile = open(f"{outdirS}SVM_params.c", "w+")
myFile.write(f"#include \"SVM_params.h\"\n")
import sys
sys.path.insert(1, 'utils')
import create_matrices
if self.estimator.is_regr == False:
stri = create_matrices.createMatrix('float', 'support_vectors',
best_estimator.coef_,
'WEIGTH_DIM', 'N_FEATURE')
else:
stri = create_matrices.createMatrix(
"float", "support_vectors",
np.reshape(best_estimator.coef_,
(1, best_estimator.coef_.shape[1])), 'WEIGTH_DIM',
'N_FEATURE')
myFile.write(stri)
stri = create_matrices.createArray("float", "bias",
best_estimator.intercept_,
'WEIGTH_DIM')
myFile.write(stri)
myFile.close()
def saveTestingSet(X_test, y_test, estimator):
import pandas as pd
if isinstance(X_test, pd.core.series.Series):
X_test = X_test.to_frame()
#outdir = OutputMgr.checkCreateDSDir(estimator.dataset.name, estimator.nick)
outdirI = OutputMgr.getOutIncludeDir()
myFile = open(f"{outdirI}testing_set.h","w+")
myFile.write(f"#ifndef TESTINGSET_H\n")
myFile.write(f"#define TESTINGSET_H\n\n")
myFile.write(f"#define N_TEST {y_test.shape[0]}\n\n")
'''myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(f"#define N_FEATURE {X_test.shape[1]}\n")
myFile.write(f"#endif\n\n")'''
myFile.write(f"#ifndef N_ORIG_FEATURE\n")
myFile.write(f"#define N_ORIG_FEATURE {X_test.shape[1]}\n") # Is this value correct??? xxx
myFile.write(f"#endif\n\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
myFile.write(f"extern {type_s} y_test[N_TEST];\n")
myFile.write(f"extern float X_test[N_TEST][N_ORIG_FEATURE];\n")
#
#if cfg.normalization!=None and cfg.regr and cfg.algo.lower() != 'dt':
# saveTestNormalization(myFile)
#
myFile.write(f"#endif")
myFile.close()
#outdirI = OutputMgr.checkCreateGeneralIncludeDir()
#from shutil import copyfile
#copyfile(f"{outdir}testing_set.h", f"{outdirI}testing_set.h")
outdirS = OutputMgr.getOutSourceDir()
myFile = open(f"{outdirS}testing_set.c","w+")
myFile.write(f"#include \"testing_set.h\"\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
import sys
sys.path.insert(1, 'utils')
import create_matrices
import numpy as np
stri = create_matrices.createArray(type_s, "y_test", np.reshape(y_test, (y_test.shape[0], )), 'N_TEST')
myFile.write(stri)
stri = create_matrices.createMatrix('float', 'X_test', X_test.values, 'N_TEST', 'N_ORIG_FEATURE') # changed by FB
myFile.write(stri)
myFile.close()
def saveTrainingSet(X_train, y_train, estimator):
#outdir = OutputMgr.checkCreateDSDir(estimator.dataset.name, estimator.nick)
outdirI = OutputMgr.getOutIncludeDir()
myFile = open(f"{outdirI}training_set.h","w+")
myFile.write(f"#define N_TRAIN {y_train.shape[0]}\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(f"#define N_ORIG_FEATURE {X_train.shape[1]}\n")
myFile.write(f"#endif\n\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
myFile.write(f"extern {type_s} y_train[N_TRAIN];\n")
myFile.write(f"extern float X_train[N_TRAIN][N_FEATURE];\n")
myFile.close()
#outdirI = OutputMgr.checkCreateGeneralIncludeDir()
#from shutil import copyfile
#copyfile(f"{outdir}training_set.h", f"{outdirI}training_set.h")
outdirS = OutputMgr.getOutSourceDir()
myFile = open(f"{outdirS}training_set_params.c","w+")
#myFile.write(f"#include \"AI_main.h\"\n")
myFile.write(f"#include \"training_set.h\"\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
import sys
sys.path.insert(1, 'utils')
import create_matrices
import numpy as np
stri = create_matrices.createArray(type_s, "y_train", np.reshape(y_train, (y_train.shape[0], )), 'N_TRAIN')
myFile.write(stri)
stri = create_matrices.createMatrix('float', 'X_train', X_train.values, 'N_TRAIN', 'N_FEATURE') # changed by FB
myFile.write(stri)
myFile.close()
def saveTestingSet(X_test, y_test, estimator, full=True):
outdir = OutputMgr.checkCreateDSDir(estimator.dataset.name,
estimator.nick)
if full:
myFile = open(f"{outdir}testing_set.h", "w+")
myFile.write(f"#ifndef TESTINGSET_H\n")
myFile.write(f"#define TESTINGSET_H\n\n")
else:
myFile = open(f"{outdir}minimal_testing_set.h", "w+")
myFile.write(f"#ifndef MINIMAL_TESTINGSET_H\n")
myFile.write(f"#define MINIMAL_TESTINGSET_H\n\n")
myFile.write(f"#define N_TEST {y_test.shape[0]}\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(f"#define N_FEATURE {X_test.shape[1]}\n")
myFile.write(f"#endif\n\n")
myFile.write(f"#ifndef N_ORIG_FEATURE\n")
myFile.write(f"#define N_ORIG_FEATURE {X_test.shape[1]}\n")
myFile.write(f"#endif\n\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
myFile.write(f"extern {type_s} y_test[N_TEST];\n")
myFile.write(f"extern float X_test[N_TEST][N_FEATURE];\n")
#
#if cfg.normalization!=None and cfg.regr and cfg.algo.lower() != 'dt':
# saveTestNormalization(myFile)
#
myFile.write(f"#endif")
myFile.close()
outdirI = OutputMgr.checkCreateGeneralIncludeDir()
from shutil import copyfile
if full:
copyfile(f"{outdir}testing_set.h", f"{outdirI}testing_set.h")
else:
copyfile(f"{outdir}minimal_testing_set.h",
f"{outdirI}minimal_testing_set.h")
outdirS = OutputMgr.checkCreateGeneralSourceDir()
if full:
myFile = open(f"{outdirS}testing_set.c", "w+")
else:
myFile = open(f"{outdirS}minimal_testing_set.c", "w+")
#myFile.write(f"#include \"AI_main.h\"\n")
if full:
myFile.write(f"#include \"testing_set.h\"\n")
else:
myFile.write(f"#include \"minimal_testing_set.h\"\n")
if estimator.is_regr:
type_s = 'float'
else:
type_s = 'int'
import sys
sys.path.insert(1, 'utils')
import create_matrices
import numpy as np
stri = create_matrices.createArray(
type_s, "y_test", np.reshape(y_test, (y_test.shape[0], )),
'N_TEST')
myFile.write(stri)
stri = create_matrices.createMatrix('float', 'X_test', X_test.values,
'N_TEST',
'N_FEATURE') # changed by FB
myFile.write(stri)
myFile.close()
def savePPParams(scaler, reduce_dims, estimator):
outdir = OutputMgr.checkCreateDSDir(estimator.dataset.name, estimator.nick)
if scaler == None:
sx = np.ones(estimator.dataset.X.shape[1])
else:
sx = scaler.scale_
if isinstance(scaler, preprocessing.StandardScaler):
ux = scaler.mean_
if reduce_dims == None:
pca_components = np.identity(estimator.dataset.X.shape[1])
else:
pca_components = reduce_dims.components_
myFile = open(f"{outdir}PPParams.h","w+")
myFile.write(f"#ifndef PPPARAMS_H\n")
myFile.write(f"#define PPPARAMS_H\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(f"#define N_FEATURE {pca_components.shape[0]}\n")
myFile.write(f"#endif\n\n")
myFile.write(f"#ifndef N_ORIG_FEATURE\n")
myFile.write(f"#define N_ORIG_FEATURE {pca_components.shape[1]}\n")
myFile.write(f"#endif\n\n")
myFile.write(f"extern float pca_components[N_FEATURE][N_ORIG_FEATURE];\n")
myFile.write(f"\n")
if scaler!=None:
if isinstance(scaler, preprocessing.StandardScaler):
myFile.write(f"#define STANDARD_NORMALIZATION\n\n")
myFile.write(f"extern float s_x[N_ORIG_FEATURE];\n")
myFile.write(f"extern float u_x[N_ORIG_FEATURE];\n")
elif isinstance(scaler, preprocessing.MinMaxScaler):
myFile.write(f"#define MINMAX_NORMALIZATION\n\n")
myFile.write(f"extern float s_x[N_ORIG_FEATURE];\n")
'''
if cfg.normalization!=None and cfg.regr and cfg.algo.lower() != 'dt':
saveTestNormalization(myFile)
'''
myFile.write(f"#endif")
myFile.close()
outdirI = OutputMgr.checkCreateGeneralIncludeDir()
from shutil import copyfile
copyfile(f"{outdir}PPParams.h", f"{outdirI}PPParams.h")
outdirS = OutputMgr.checkCreateGeneralSourceDir()
myFile = open(f"{outdirS}preprocess_params.c","w+")
#myFile.write(f"#include \"AI_main.h\"\n")
myFile.write(f"#include \"PPParams.h\"\n")
import sys
sys.path.insert(1, 'utils')
import create_matrices
stri = create_matrices.createMatrix('float', 'pca_components', pca_components, 'N_FEATURE', 'N_ORIG_FEATURE')
myFile.write(stri)
myFile.write(f"\n")
if scaler!=None:
if isinstance(scaler, preprocessing.StandardScaler):
myFile.write(f"#define STANDARD_NORMALIZATION\n\n")
stri = create_matrices.createArray('float', "s_x", np.reshape(s_x, (s_x.size, )), 'N_ORIG_FEATURE')
myFile.write(stri)
stri = create_matrices.createArray('float', "u_x", np.reshape(u_x, (u_x.size, )), 'N_ORIG_FEATURE')
myFile.write(stri)
elif isinstance(scaler, preprocessing.MinMaxScaler):
myFile.write(f"#define MINMAX_NORMALIZATION\n\n")
stri = create_matrices.createArray('float', "s_x", np.reshape(s_x, (s_x.size, )), 'N_ORIG_FEATURE')
myFile.write(stri)
myFile.close()
def saveParams(self, best_estimator):
# outdir = OutputMgr.checkCreateDSDir(self.estimator.dataset.name, self.estimator.nick)
outdirI = OutputMgr.getOutIncludeDir()
n_nodes = getattr(best_estimator.tree_, 'node_count')
values = getattr(best_estimator.tree_, 'value')
children_left = getattr(best_estimator.tree_, 'children_left')
children_right = getattr(best_estimator.tree_, 'children_right')
feature = getattr(best_estimator.tree_, 'feature')
threshold = getattr(best_estimator.tree_, 'threshold')
target_classes = np.unique(self.estimator.dataset.y)
myFile = open(f"{outdirI}DT_params.h", "w+")
myFile.write(f"#ifndef DT_PARAMS_H\n")
myFile.write(f"#define DT_PARAMS_H\n\n")
if self.estimator.is_regr == False:
myFile.write(f"#define N_CLASS {self.estimator.n_classes}\n")
leaf_nodes = children_left == children_right
n_leaves = np.count_nonzero(leaf_nodes)
myFile.write(f"#define N_LEAVES {n_leaves}\n\n")
else:
myFile.write(f"#define N_CLASS 0\n\n")
myFile.write(f"#define VALUES_DIM {values.shape[2]}\n\n")
myFile.write(f"#define N_NODES {n_nodes}\n\n")
myFile.write(f"extern int children_left[N_NODES];\n")
myFile.write(f"extern int children_right[N_NODES];\n")
myFile.write(f"extern int feature[N_NODES];\n")
myFile.write(f"extern float threshold[N_NODES];\n")
if self.estimator.is_regr == True:
myFile.write(f"extern int values[N_NODES][VALUES_DIM];\n")
else:
myFile.write(f"extern int target_classes[N_CLASS];\n")
myFile.write(f"extern int leaf_nodes[N_LEAVES][2];\n")
myFile.write(f"\n#endif")
myFile.close()
#outdirI = OutputMgr.checkCreateGeneralIncludeDir()
#from shutil import copyfile
#copyfile(f"{outdir}DT_params.h", f"{outdirI}DT_params.h")
outdirS = OutputMgr.getOutSourceDir()
myFile = open(f"{outdirS}DT_params.c", "w+")
myFile.write(f"#include \"DT_params.h\"\n")
import sys
sys.path.insert(1, 'utils')
import create_matrices
stri = create_matrices.createArray("int", "children_left",
children_left, 'N_NODES')
myFile.write(stri)
stri = create_matrices.createArray("int", "children_right",
children_right, 'N_NODES')
myFile.write(stri)
stri = create_matrices.createArray("int", "feature", feature,
'N_NODES')
myFile.write(stri)
stri = create_matrices.createArray("float", "threshold", threshold,
'N_NODES')
myFile.write(stri)
if self.estimator.is_regr == True:
stri = create_matrices.createMatrix2('int', 'values', values,
'N_NODES', 'VALUES_DIM')
myFile.write(stri)
else:
stri = create_matrices.createArray("int", "target_classes",
target_classes, 'N_CLASS')
myFile.write(stri)
argmaxs = np.argmax(values[leaf_nodes][:, 0],
axis=1).reshape(-1, 1)
leaf_nodes_idx = np.asarray(np.nonzero(leaf_nodes)).T
leaf_nodes = np.concatenate((leaf_nodes_idx, argmaxs), axis=1)
stri = create_matrices.createMatrix('int', 'leaf_nodes',
leaf_nodes, 'N_LEAVES', '2')
myFile.write(stri)
myFile.close()
def savePPParams(scaler, reduce_dims, estimator):
#outdir = OutputMgr.checkCreateDSDir(estimator.dataset.name, estimator.nick)
outdirI = OutputMgr.getOutIncludeDir()
if scaler == None:
sx = np.ones(estimator.dataset.X.shape[1])
else:
sx = scaler.scale_
if isinstance(scaler, preprocessing.StandardScaler):
ux = scaler.mean_
elif isinstance(scaler, preprocessing.MinMaxScaler):
mx = scaler.min_
n_orig_feature = estimator.dataset.X.shape[1]
if reduce_dims != None:
pca_components = reduce_dims.components_
pca_means = reduce_dims.mean_
n_feature = pca_components.shape[0]
else:
n_feature = n_orig_feature
myFile = open(f"{outdirI}PPParams.h","w+")
myFile.write(f"#ifndef PPPARAMS_H\n")
myFile.write(f"#define PPPARAMS_H\n\n")
myFile.write(f"#ifndef N_FEATURE\n")
myFile.write(f"#define N_FEATURE {n_feature}\n")
myFile.write(f"#endif\n\n")
myFile.write(f"#ifndef N_ORIG_FEATURE\n")
myFile.write(f"#define N_ORIG_FEATURE {n_orig_feature}\n")
myFile.write(f"#endif\n\n")
if reduce_dims != None:
myFile.write(f"#define DO_PCA 1\n")
myFile.write(f"extern float pca_components[N_FEATURE][N_ORIG_FEATURE];\n")
myFile.write(f"extern float pca_means[N_ORIG_FEATURE];\n")
myFile.write(f"\n")
if scaler!=None:
if isinstance(scaler, preprocessing.StandardScaler):
myFile.write(f"#define STANDARD_SCALING\n\n")
myFile.write(f"extern float s_x[N_ORIG_FEATURE];\n")
myFile.write(f"extern float u_x[N_ORIG_FEATURE];\n")
elif isinstance(scaler, preprocessing.MinMaxScaler):
myFile.write(f"#define MINMAX_SCALING\n\n")
myFile.write(f"extern float s_x[N_ORIG_FEATURE];\n")
myFile.write(f"extern float m_x[N_ORIG_FEATURE];\n")
'''
if cfg.normalization!=None and cfg.regr and cfg.algo.lower() != 'dt':
saveTestNormalization(myFile)
'''
myFile.write(f"#endif")
myFile.close()
#outdirI = OutputMgr.checkCreateGeneralIncludeDir()
#from shutil import copyfile
#copyfile(f"{outdir}PPParams.h", f"{outdirI}PPParams.h")
outdirS = OutputMgr.getOutSourceDir()
myFile = open(f"{outdirS}preprocess_params.c","w+")
myFile.write(f"#include \"PPParams.h\"\n")
import sys
sys.path.insert(1, 'utils')
import create_matrices
if reduce_dims != None:
stri = create_matrices.createMatrix('float', 'pca_components', pca_components, 'N_FEATURE', 'N_ORIG_FEATURE')
myFile.write(stri)
stri = create_matrices.createArray('float', "pca_means", pca_means, 'N_ORIG_FEATURE')
myFile.write(stri)
myFile.write(f"\n")
if scaler!=None:
if isinstance(scaler, preprocessing.StandardScaler):
myFile.write(f"#define STANDARD_SCALING\n\n")
stri = create_matrices.createArray('float', "s_x", np.reshape(sx, (sx.shape[0], )), 'N_ORIG_FEATURE')
myFile.write(stri)
stri = create_matrices.createArray('float', "u_x", np.reshape(ux, (ux.shape[0], )), 'N_ORIG_FEATURE')
myFile.write(stri)
elif isinstance(scaler, preprocessing.MinMaxScaler):
myFile.write(f"#define MINMAX_SCALING\n\n")
stri = create_matrices.createArray('float', "s_x", np.reshape(sx, (sx.shape[0], )), 'N_ORIG_FEATURE')
myFile.write(stri)
stri = create_matrices.createArray('float', "m_x", np.reshape(mx, (mx.shape[0], )), 'N_ORIG_FEATURE')
myFile.write(stri)
myFile.close()