pip install joblib

from joblib import dump, load

# 保存数据
data = {"name": "Joblib", "description": "高效计算库"}
dump(data, "data.joblib")

# 加载数据
loaded_data = load("data.joblib")
print(loaded_data)

{'name': 'Joblib', 'description': '高效计算库'}

from joblib import Parallel, delayed

# 定义一个计算任务
def compute_square(n):
    return n ** 2

# 使用 Parallel 和 delayed 实现并行计算
results = Parallel(n_jobs=4)(delayed(compute_square)(i) for i in range(10))
print(results)

[0, 1, 4, 9, 16, 25, 36, 49, 64, 81]

from joblib import Memory
import time

# 定义缓存存储路径
memory = Memory(location="./cachedir", verbose=0)

# 缓存的函数
@memory.cache
def slow_function(x):
    time.sleep(2)  # 模拟耗时操作
    return x ** 2

# 第一次运行（计算并缓存）
print(slow_function(10))  # 耗时 2 秒

# 第二次运行（直接从缓存中读取）
print(slow_function(10))  # 几乎瞬间完成

100  # 第一次调用耗时 2 秒
100  # 第二次调用从缓存读取，耗时几乎为 0

from joblib import Memory
import pandas as pd

memory = Memory(location="./cachedir", verbose=0)

@memory.cache
def preprocess_data(filepath):
    print("正在加载和处理数据...")
    df = pd.read_csv(filepath)
    # 假设这里有一些耗时的清洗和转换操作
    return df

data = preprocess_data("large_dataset.csv")

from sklearn.datasets import make_classification
from sklearn.ensemble import RandomForestClassifier
from joblib import Parallel, delayed

# 生成数据集
X, y = make_classification(n_samples=1000, n_features=20)

# 定义训练函数
def train_model(seed):
    model = RandomForestClassifier(random_state=seed)
    model.fit(X, y)
    return model

# 并行训练 5 个模型
models = Parallel(n_jobs=5)(delayed(train_model)(seed) for seed in range(5))

import numpy as np
from joblib import Parallel, delayed

# 定义积分任务
def integrate(f, a, b, n=1000):
    x = np.linspace(a, b, n)
    y = f(x)
    return np.sum(y) * (b - a) / n

# 定义被积函数
def func(x):
    return x ** 2

# 并行化多个积分任务
results = Parallel(n_jobs=4)(
    delayed(integrate)(func, i, i + 1) for i in range(4)
)
print(results)