Dask：我如何将我的代码与 dask 延迟并行化？

Question

这是我第一次尝试并行处理，我一直在研究 Dask，但我在实际编码时遇到了麻烦。

我看过他们的示例和文档，我认为 dask.delayed 效果最好。我试图用 delayed(function_name) 包装我的函数，或者添加一个 @delayed 装饰器，但我似乎无法让它正常工作。与其他方法相比，我更喜欢 Dask，因为它是在 python 中制作的，而且它（假定的）简单。我知道 dask 在 for 循环中不起作用，但他们说它可以在循环内工作。

我的代码通过包含对其他函数的输入的函数传递文件，如下所示：

from dask import delayed
filenames = ['1.csv', '2.csv', '3.csv', etc. etc. ]
for count, name in enumerate(filenames)"
    name = name.split('.')[0]
    ....

然后做一些预处理例如：

    preprocess1, preprocess2 = delayed(read_files_and_do_some_stuff)(name)

然后我调用构造函数并将 pre_results 传递给函数调用：

    fc = FunctionCalls()
    Daily = delayed(fc.function_runs)(filename=name, stringinput='Daily',
                             input_data=pre_result1, model1=pre_result2)

我在这里所做的是将文件传递到 for 循环中，进行一些预处理，然后将文件传递到两个模型中。

关于如何并行化这个的想法或技巧？我开始遇到奇怪的错误，而且我不知道如何修复代码。该代码确实按原样工作。我使用了一堆 pandas 数据帧、系列和 numpy 数组，我不想返回并更改所有内容以使用 dask.dataframes 等

我评论中的代码可能难以阅读。这是一种更格式化的方式。

在下面的代码中，当我输入 print(mean_squared_error) 时，我只得到：Delayed('mean_squared_error-3009ec00-7ff5-4865-8338-1fec3f9ed138')

from dask import delayed
import pandas as pd
from sklearn.metrics import mean_squared_error as mse
filenames = ['file1.csv']

for count, name in enumerate(filenames):
    file1 = pd.read_csv(name)
    df = pd.DataFrame(file1)
    prediction = df['Close'][:-1]
    observed = df['Close'][1:]
    mean_squared_error = delayed(mse)(observed, prediction)

Answer 1

您需要调用 dask.compute 来最终计算结果。参见 dask.delayed documentation。

顺序码

import pandas as pd
from sklearn.metrics import mean_squared_error as mse
filenames = [...]

results = []
for count, name in enumerate(filenames):
    file1 = pd.read_csv(name)
    df = pd.DataFrame(file1)  # isn't this already a dataframe?
    prediction = df['Close'][:-1]
    observed = df['Close'][1:]
    mean_squared_error = mse(observed, prediction)  
    results.append(mean_squared_error)

并行代码

import dask
import pandas as pd
from sklearn.metrics import mean_squared_error as mse
filenames = [...]

delayed_results = []
for count, name in enumerate(filenames):
    df = dask.delayed(pd.read_csv)(name)
    prediction = df['Close'][:-1]
    observed = df['Close'][1:]
    mean_squared_error = dask.delayed(mse)(observed, prediction)
    delayed_results.append(mean_squared_error)

results = dask.compute(*delayed_results)

Answer 2

IMO，比公认的答案更清晰的解决方案是这个片段。

from dask import compute, delayed
import pandas as pd
from sklearn.metrics import mean_squared_error as mse
filenames = [...]

def compute_mse(file_name):
    df = pd.read_csv(file_name)
    prediction = df['Close'][:-1]
    observed = df['Close'][1:]
    return mse(observed, prediction)

delayed_results = [delayed(compute_mse)(file_name) for file_name in filenames]
mean_squared_errors = compute(*delayed_results, scheduler="processes")

Dask：我如何将我的代码与 dask 延迟并行化？

Dask: How would I parallelize my code with dask delayed?

parallel-processing

multithreading

python-3.x

python-multiprocessing

dask

顺序码

并行代码