0

我试图并行化以下代码：

def extract_features(dict_w, key, d_f):
    lowess = sm.nonparametric.lowess
    f_v = []
    for c in np.setdiff1d(dict_w[key].columns, ["t", "ts"]):
        dict_w[key][c + "_lowess"] = lowess(
            dict_w[key][c],
            dict_w[key].iloc[:, ~dict_w[key].columns.duplicated()]["ts"],
            is_sorted=True,
            frac=0.05,
        )[:, 1]
        f_v.extend(np.zeros(1,5))

    d_f[key] = np.array(f_v)

我尝试了以下方法：

def extract_features(key, dict_w, shared_dic):
    
    lowess = sm.nonparametric.lowess
    f_v = []
    
    for c in np.setdiff1d(dict_w[key].columns, ["t", "ts"]):
        lw =  lowess(
            dict_w[key][c],
            dict_w[key].iloc[:, ~dict_w[key].columns.duplicated()]["ts"],
            is_sorted=True,
            frac=0.05,)[:, 1]
        #print(lw)
        dict_w[key][c + "_lowess"] = lw
        #print("test ") 
        
        print(lw)
        

    shared_dic[key] = np.array(f_v)
    print(len(shared_dic[key]))


def parallelize(dict_w, max_workers, shared_dic):
    with concurrent.futures.ProcessPoolExecutor(max_workers=max_workers) as pool:
        start = time.perf_counter()
        for key in dict_w.keys():

            pool.submit(extract_features, key, dict_w,  shared_dic)

    logger.info("Execution time (parallel) = {}".format(time.perf_counter() - start))

问题是，如果我输出第一个 print(lw)，我可以看到 lowess 函数的结果，但是如果我dict_w[key][c + "_lowess"] = lw在终端中没有任何输出之后放置 print(lw)。两者dict_w和shared_dic定义如下：

manager = Manager()
dict_w = manager.dict()
shared_dic = manager.dict()
4

0 回答 0