创建Dask集群时默认n_workers？

Question

简单的问题。如果我使用以下代码创建 Dask 集群：

from dask.distributed import Client

client = Client()

它将创造多少工人？我在一台机器上运行这段代码，它创建了 4 个工作线程。我在服务器上运行相同的代码，它创建了 8 个工作线程。它是否只是根据可用资源创造尽可能多的东西？在源代码中，文档字符串中没有列出

n_workers

的默认值。我正在尝试了解如何自动创建集群，而不必提前知道我可用的资源。

class LocalCluster(SpecCluster):
"""Create local Scheduler and Workers

This creates a "cluster" of a scheduler and workers running on the local
machine.

Parameters
----------
n_workers: int
    Number of workers to start
memory_limit: str, float, int, or None, default "auto"
    Sets the memory limit *per worker*.

    Notes regarding argument data type:

    * If None or 0, no limit is applied.
    * If "auto", the total system memory is split evenly between the workers.
    * If a float, that fraction of the system memory is used *per worker*.
    * If a string giving a number of bytes (like ``"1GiB"``), that amount is used *per worker*.
    * If an int, that number of bytes is used *per worker*.

Answer 1

dask.distributed

使用

os.cpu_count()

获取机器上可用的 CPU 数量，然后将其与每个工作线程的线程数一起使用来计算工作线程的数量。