When I ran a machine learning task in a container, it created a large number of threads, which led to a particularly high load on the cpu, making the data load slow.
but when I run the job in the node(which the container in), It behaves normally and loads data much faster。
my dataload config as:
