在熊猫DataFrame上进行多重处理

时间:2020-09-14 00:40:45

标签: pandas multithreading dataframe parallel-processing multiprocessing

我在Dataframe列上应用了一个函数,但是我想使其变得更快,因为该函数在串行完成时会占用大量处理时间。

df[df['codes']=='None']['q'][:1].apply(lambda x: clf(x,candidate_labels))

通常,单行只需要2.52 secs来运行,但是使用多处理来运行下面的代码时,它需要花费更长的时间51.61 secs,而我大约需要2500 rows来进行处理,因此需要大量时间才能运行该功能。我希望至少加快20%的速度。

import multiprocessing
import pandas as pd
import numpy as np

def clf(x):
    ...
    return list
def _apply_df(args):
    df, func, kwargs = args
    return df.apply(func, **kwargs)

def apply_by_multiprocessing(df, func, **kwargs):
    workers = kwargs.pop('workers')
    pool = multiprocessing.Pool(processes=workers)
    result = pool.map(_apply_df, [(d, func, kwargs)
            for d in np.array_split(df, workers)])
    pool.close()
    return pd.concat(list(result))
    
if __name__ == '__main__':
    tart_time = time.time()
    res=apply_by_multiprocessing(df[df['codes']=='None']['q'][:1],clf, workers=4)  
    print(res)
    print("--- %s seconds ---" % (time.time() - start_time))
    ## run by 4 processors

我也尝试过不同的迭代进行多处理,但是似乎没有一个可以加快流程,因为它们会使我的代码变慢。

from pandarallel import pandarallel
import time
pandarallel.initialize(progress_bar=True)

start_time = time.time()
categories = df[df['codes']=='None']['q'][:10].parallel_apply(lambda x: clf(x,candidate_labels))
print("--- %s seconds ---" % (time.time() - start_time))

另一个实验:

import multiprocessing as mp

def clf:
    ...
    return list

if __name__ == '__main__':
    p = mp.Pool(processes=8)
    pool_results = p.map(clf, df[df['codes']=='None']['q'][:1])
    p.close()
    p.join()

1 个答案:

答案 0 :(得分:0)

也许您可以使用此功能: https://github.com/xieqihui/pandas-multiprocess

pip install pandas-multiprocess
from pandas_multiprocess import multi_process


args = {'workers': 4}
result = multi_process(func=clf, data=df, num_process=8, **args)