def test_gbm_regressor_backupsklearn(backend='auto'):
df = pd.read_csv("./open_data/simple.txt", delim_whitespace=True)
X = np.array(df.iloc[:, :df.shape[1] - 1], dtype='float32', order='C')
y = np.array(df.iloc[:, df.shape[1] - 1], dtype='float32', order='C')
import h2o4gpu
Solver = h2o4gpu.GradientBoostingRegressor
# Run h2o4gpu version of RandomForest Regression
gbm = Solver(backend=backend,
random_state=1234,
n_gpus=n_gpus(),
n_jobs=-1)
print("h2o4gpu fit()")
gbm.fit(X, y)
# Run Sklearn version of RandomForest Regression
from h2o4gpu.ensemble import GradientBoostingRegressorSklearn
gbm_sk = GradientBoostingRegressorSklearn(random_state=1234, max_depth=3)
print("Scikit fit()")
gbm_sk.fit(X, y)
if backend == "sklearn":
assert (gbm.predict(X) == gbm_sk.predict(X)).all() == True
print(
(a == b
for a, b in zip(gbm.staged_predict(X), gbm_sk.staged_predict(X))))
assert np.allclose(list(gbm.staged_predict(X)),
list(gbm_sk.staged_predict(X)))
assert (gbm.score(X, y) == gbm_sk.score(X, y)).all() == True
assert (gbm.apply(X) == gbm_sk.apply(X)).all() == True
print("Estimators")
print(gbm.estimators_)
print(gbm_sk.estimators_)
print("loss")
print(gbm.loss_)
print(gbm_sk.loss_)
assert gbm.loss_.__dict__ == gbm_sk.loss_.__dict__
print("init_")
print(gbm.init)
print(gbm_sk.init)
print("Feature importance")
print(gbm.feature_importances_)
print(gbm_sk.feature_importances_)
assert (gbm.feature_importances_ == gbm_sk.feature_importances_
).all() == True
print("train_score_")
print(gbm.train_score_)
print(gbm_sk.train_score_)
assert (gbm.train_score_ == gbm_sk.train_score_).all() == True
class GradientBoostingRegressor(object):
"""H2O GradientBoostingRegressor Solver
Selects between h2o4gpu.solvers.xgboost.GradientBoostingRegressor
and h2o4gpu.ensemble.gradient_boosting.GradientBoostingRegressorSklearn
Documentation:
import h2o4gpu.solvers ; help(h2o4gpu.xgboost.GradientBoostingRegressorO)
help(h2o4gpu.ensemble.gradient_boosting.GradientBoostingRegressorSklearn)
:param: backend : Which backend to use. Options are 'auto', 'sklearn',
'h2o4gpu'. Default is 'auto'.
Saves as attribute for actual backend used.
"""
def __init__(
self,
loss='ls',
learning_rate=0.1, # h2o4gpu
n_estimators=100, # h2o4gpu
subsample=1.0, # h2o4gpu
criterion='friedman_mse',
min_samples_split=2,
min_samples_leaf=1,
min_weight_fraction_leaf=0.0,
max_depth=3, # h2o4gpu
min_impurity_decrease=0.0,
min_impurity_split=None,
init=None,
random_state=None, # h2o4gpu
max_features='auto',
alpha=0.9,
verbose=0, # h2o4gpu
max_leaf_nodes=None,
warm_start=False,
presort='auto',
# XGBoost specific params
colsample_bytree=1.0, # h2o4gpu
num_parallel_tree=100, # h2o4gpu
tree_method='gpu_hist', # h2o4gpu
n_gpus=-1, # h2o4gpu
predictor='gpu_predictor', # h2o4gpu
backend='auto'): # h2o4gpu
import os
_backend = os.environ.get('H2O4GPU_BACKEND', None)
if _backend is not None:
backend = _backend
from ..typecheck.typechecks import assert_is_type
assert_is_type(backend, str)
# Fall back to Sklearn
# Can remove if fully implement sklearn functionality
self.do_sklearn = False
if backend == 'auto':
params_string = [
'loss', 'criterion', 'min_samples_split', 'min_samples_leaf',
'min_weight_fraction_leaf', 'min_impurity_decrease',
'min_impurity_split', 'init', 'max_features', 'alpha',
'max_leaf_nodes', 'presort'
]
params = [
loss, criterion, min_samples_split, min_samples_leaf,
min_weight_fraction_leaf, min_impurity_decrease,
min_impurity_split, init, max_features, alpha, max_leaf_nodes,
presort
]
params_default = [
'ls', 'friedman-mse', 2, 1, 0.0, 0.0, None, None, 'auto', 0.9,
None, 'auto'
]
i = 0
for param in params:
if param != params_default[i]:
self.do_sklearn = True
if verbose > 0:
print("WARNING: The sklearn parameter " +
params_string[i] +
" has been changed from default to " +
str(param) +
". Will run Sklearn GradientBoostingRegressor.")
self.do_sklearn = True
i = i + 1
elif backend == 'sklearn':
self.do_sklearn = True
elif backend == 'h2o4gpu':
self.do_sklearn = False
self.backend = backend
from h2o4gpu.ensemble import GradientBoostingRegressorSklearn
self.model_sklearn = GradientBoostingRegressorSklearn(
loss=loss,
learning_rate=learning_rate, # h2o4gpu
n_estimators=n_estimators, # h2o4gpu
subsample=subsample, # h2o4gpu
criterion=criterion,
min_samples_split=min_samples_split,
min_samples_leaf=min_samples_leaf,
min_weight_fraction_leaf=min_weight_fraction_leaf,
max_depth=max_depth, # h2o4gpu
min_impurity_decrease=min_impurity_decrease,
min_impurity_split=min_impurity_split,
init=init,
random_state=random_state, # h2o4gpu
max_features=max_features,
alpha=alpha,
verbose=verbose, # h2o4gpu
max_leaf_nodes=max_leaf_nodes,
warm_start=warm_start,
presort=presort) # h2o4gpu)
# Parameters for gbm
silent = False
if verbose != 0:
silent = True
if random_state is None:
random_state = 0
import xgboost as xgb
self.model_h2o4gpu = xgb.XGBClassifier(
learning_rate=learning_rate, # h2o4gpu
n_estimators=n_estimators, # h2o4gpu
subsample=subsample, # h2o4gpu
max_depth=max_depth, # h2o4gpu
random_state=random_state, # h2o4gpu
silent=silent, # h2o4gpu
colsample_bytree=colsample_bytree, # h2o4gpu
num_parallel_tree=num_parallel_tree, # h2o4gpu
tree_method=tree_method, # h2o4gpu
n_gpus=n_gpus, # h2o4gpu
predictor=predictor, # h2o4gpu
backend=backend) # h2o4gpu
if self.do_sklearn:
print("Running sklearn GradientBoostingRegressor")
self.model = self.model_sklearn
else:
print("Running h2o4gpu GradientBoostingRegressor")
self.model = self.model_h2o4gpu
def apply(self, X):
print("WARNING: apply() is using sklearn")
return self.model_sklearn.apply(X)
def fit(self, X, y=None, sample_weight=None):
res = self.model.fit(X, y, sample_weight)
self.set_attributes()
return res
def get_params(self):
return self.model.get_params()
def predict(self, X):
if self.do_sklearn:
res = self.model.predict(X)
self.set_attributes()
return res
res = self.model.predict(X)
res[res < 0.5] = 0
res[res > 0.5] = 1
self.set_attributes()
return res.squeeze()
def score(self, X, y, sample_weight=None):
# TODO add for h2o4gpu
print("WARNING: score() is using sklearn")
if not self.do_sklearn:
self.model_sklearn.fit(X, y) # Need to re-fit
res = self.model_sklearn.score(X, y, sample_weight)
return res
def set_params(self, **params):
return self.model.set_params(**params)
def staged_predict(self, X):
print("WARNING: staged_predict() is using sklearn")
return self.model_sklearn.staged_predict(X)
def set_attributes(self):
""" Set attributes for class"""
from ..solvers.utils import _setter
s = _setter(oself=self, e1=NameError, e2=AttributeError)
s('oself.feature_importances_ = oself.model.feature_importances_')
s('oself.oob_improvement_ = oself.model.oob_improvement_')
s('oself.train_score_ = oself.model.train_score_')
s('oself.loss_ = oself.model.loss_')
s('oself.init = oself.model.init')
s('oself.estimators_ = oself.model.estimators_')
class GradientBoostingRegressor(object):
"""H2O GradientBoostingRegressor Solver
Selects between h2o4gpu.solvers.xgboost.GradientBoostingRegressor
and h2o4gpu.ensemble.gradient_boosting.GradientBoostingRegressorSklearn
Documentation:
import h2o4gpu.solvers ; help(h2o4gpu.xgboost.GradientBoostingRegressorO)
help(h2o4gpu.ensemble.gradient_boosting.GradientBoostingRegressorSklearn)
:param: backend : Which backend to use. Options are 'auto', 'sklearn',
'h2o4gpu'. Default is 'auto'.
Saves as attribute for actual backend used.
"""
def __init__(
self,
loss='ls',
learning_rate=0.1, # h2o4gpu
n_estimators=100, # h2o4gpu
subsample=1.0, # h2o4gpu
criterion='friedman_mse',
min_samples_split=2,
min_samples_leaf=1,
min_weight_fraction_leaf=0.0,
max_depth=3, # h2o4gpu
min_impurity_decrease=0.0,
min_impurity_split=None,
init=None,
random_state=None, # h2o4gpu
max_features='auto',
alpha=0.9,
verbose=0, # h2o4gpu
max_leaf_nodes=None,
warm_start=False,
presort='auto',
# XGBoost specific params
colsample_bytree=1.0, # h2o4gpu
num_parallel_tree=100, # h2o4gpu
tree_method='gpu_hist', # h2o4gpu
n_gpus=-1, # h2o4gpu
predictor='gpu_predictor', # h2o4gpu
backend='auto'): # h2o4gpu
import os
_backend = os.environ.get('H2O4GPU_BACKEND', None)
if _backend is not None:
backend = _backend
from ..typecheck.typechecks import assert_is_type
assert_is_type(backend, str)
# Fall back to Sklearn
# Can remove if fully implement sklearn functionality
self.do_sklearn = False
if backend == 'auto':
params_string = [
'loss', 'criterion', 'min_samples_split', 'min_samples_leaf',
'min_weight_fraction_leaf', 'min_impurity_decrease',
'min_impurity_split', 'init', 'max_features', 'alpha',
'max_leaf_nodes', 'presort'
]
params = [
loss, criterion, min_samples_split, min_samples_leaf,
min_weight_fraction_leaf, min_impurity_decrease,
min_impurity_split, init, max_features, alpha, max_leaf_nodes,
presort
]
params_default = [
'ls', 'friedman-mse', 2, 1, 0.0, 0.0, None, None, 'auto', 0.9,
None, 'auto'
]
i = 0
for param in params:
if param != params_default[i]:
self.do_sklearn = True
if verbose > 0:
print("WARNING: The sklearn parameter " +
params_string[i] +
" has been changed from default to " + str(param)
+ ". Will run Sklearn GradientBoostingRegressor.")
self.do_sklearn = True
i = i + 1
elif backend == 'sklearn':
self.do_sklearn = True
elif backend == 'h2o4gpu':
self.do_sklearn = False
self.backend = backend
from h2o4gpu.ensemble import GradientBoostingRegressorSklearn
self.model_sklearn = GradientBoostingRegressorSklearn(
loss=loss,
learning_rate=learning_rate, # h2o4gpu
n_estimators=n_estimators, # h2o4gpu
subsample=subsample, # h2o4gpu
criterion=criterion,
min_samples_split=min_samples_split,
min_samples_leaf=min_samples_leaf,
min_weight_fraction_leaf=min_weight_fraction_leaf,
max_depth=max_depth, # h2o4gpu
min_impurity_decrease=min_impurity_decrease,
min_impurity_split=min_impurity_split,
init=init,
random_state=random_state, # h2o4gpu
max_features=max_features,
alpha=alpha,
verbose=verbose, # h2o4gpu
max_leaf_nodes=max_leaf_nodes,
warm_start=warm_start,
presort=presort) # h2o4gpu)
# Parameters for gbm
silent = False
if verbose != 0:
silent = True
if random_state is None:
random_state = 0
import xgboost as xgb
self.model_h2o4gpu = xgb.XGBClassifier(
learning_rate=learning_rate, # h2o4gpu
n_estimators=n_estimators, # h2o4gpu
subsample=subsample, # h2o4gpu
max_depth=max_depth, # h2o4gpu
random_state=random_state, # h2o4gpu
silent=silent, # h2o4gpu
colsample_bytree=colsample_bytree, # h2o4gpu
num_parallel_tree=num_parallel_tree, # h2o4gpu
tree_method=tree_method, # h2o4gpu
n_gpus=n_gpus, # h2o4gpu
predictor=predictor, # h2o4gpu
backend=backend) # h2o4gpu
if self.do_sklearn:
print("Running sklearn GradientBoostingRegressor")
self.model = self.model_sklearn
else:
print("Running h2o4gpu GradientBoostingRegressor")
self.model = self.model_h2o4gpu
def apply(self, X):
print("WARNING: apply() is using sklearn")
return self.model_sklearn.apply(X)
def fit(self, X, y=None, sample_weight=None):
res = self.model.fit(X, y, sample_weight)
self.set_attributes()
return res
def get_params(self):
return self.model.get_params()
def predict(self, X):
if self.do_sklearn:
res = self.model.predict(X)
self.set_attributes()
return res
res = self.model.predict(X)
res[res < 0.5] = 0
res[res > 0.5] = 1
self.set_attributes()
return res.squeeze()
def score(self, X, y, sample_weight=None):
# TODO add for h2o4gpu
print("WARNING: score() is using sklearn")
if not self.do_sklearn:
self.model_sklearn.fit(X, y) # Need to re-fit
res = self.model_sklearn.score(X, y, sample_weight)
return res
def set_params(self, **params):
return self.model.set_params(**params)
def staged_predict(self, X):
print("WARNING: staged_predict() is using sklearn")
return self.model_sklearn.staged_predict(X)
def set_attributes(self):
""" Set attributes for class"""
from ..solvers.utils import _setter
s = _setter(oself=self, e1=NameError, e2=AttributeError)
s('oself.feature_importances_ = oself.model.feature_importances_')
s('oself.oob_improvement_ = oself.model.oob_improvement_')
s('oself.train_score_ = oself.model.train_score_')
s('oself.loss_ = oself.model.loss_')
s('oself.init = oself.model.init')
s('oself.estimators_ = oself.model.estimators_')