统一常见机器学习模型的保存与加载

统一常见机器学习模型的保存与加载

包括如下：
thundersvm
lightgbm（gpu）
deepforest
xgboost（gpu）
sklearn

代码：

import deepforest
import joblib
import lightgbm
import sklearn
from sklearn.datasets import make_classification
import thundersvm
import xgboost
import numpy as np


def model_save(estimator, save_name, save_path):
    # from deepforest import lgbcadeForestClassifier
    # from lightgbm.sklearn import LGBMClassifier
    # from sklearn.ensemble import RandomForestClassifier
    # from thundersvm import SVC
    # from xgboost import XGBClassifier
    # from sklearn.linear_model import LogisticRegression
    # from sklearn.naive_bayes import BernoulliNB
    # from sklearn.neural_network import MLPClassifier
    save_path = save_path + "/" + save_name
    if isinstance(estimator, deepforest.CascadeForestClassifier):
        print("save deepforest")
        estimator.save(save_path)
    elif isinstance(estimator, lightgbm.sklearn.LGBMClassifier):
        print("save lightgbm")
        joblib.dump(estimator, save_path + ".pkl")
    elif isinstance(estimator, sklearn.ensemble.RandomForestClassifier):
        print("save randomforest")
        joblib.dump(estimator, save_path + ".pkl")
    elif isinstance(estimator, thundersvm.SVC):
        print("save svm")
        estimator.save_to_file(save_path + ".pkl")
    elif isinstance(estimator, xgboost.XGBClassifier):
        print("save xgboost")
        estimator.save_model(save_path + ".json")
    elif isinstance(estimator, sklearn.linear_model.LogisticRegression):
        print("save logistic")
        joblib.dump(estimator, save_path + ".pkl")
    elif isinstance(estimator, sklearn.neural_network.MLPClassifier):
        print("save mlp")
        joblib.dump(estimator, save_path + ".pkl")
    else:
        raise Exception("estimator model saving error!")


train_x, train_y = make_classification(100, 5)
test_x = np.array([[1, 3, 5, 10, 5], [7, 52, 14, 61, 2]])

lgb = lightgbm.sklearn.LGBMClassifier()
lgb.fit(train_x, train_y)
model_save(lgb, "lgb", "./modeltest")
test_y = lgb.predict_proba(test_x)
print(test_y)
# lgb.load("./modeltest/lgb")
load_model = joblib.load("./modeltest/lgb.pkl")
test_y = load_model.predict_proba(test_x)
print(test_y)
print("=================================")

svm = thundersvm.SVC(probability=True)
svm.fit(train_x, train_y)
model_save(svm, "svm", "./modeltest")
test_y = svm.predict_proba(test_x)
print(test_y)

load_model = thundersvm.SVC(probability=True)
load_model.load_from_file("./modeltest/svm.pkl")
test_y = load_model.predict_proba(test_x)
print(test_y)
print("=================================")

xgb = xgboost.XGBClassifier(use_label_encoder=False)
xgb.fit(train_x, train_y)
model_save(xgb, "xgb", "./modeltest")
test_y = xgb.predict_proba(test_x)
print(test_y)
# lgb.load("./modeltest/lgb")
# load_model = joblib.load("./modeltest/svm.pkl")
load_model = xgboost.XGBClassifier()
load_model.load_model("./modeltest/xgb.json")
test_y = load_model.predict_proba(test_x)
print(test_y)
print("=================================")

log = sklearn.linear_model.LogisticRegression()
log.fit(train_x, train_y)
model_save(log, "log", "./modeltest")
test_y = log.predict_proba(test_x)
print(test_y)

load_model = joblib.load("./modeltest/log.pkl")
test_y = load_model.predict_proba(test_x)
print(test_y)