Skip to content

[Bug] CPU Mem utilization grows with training, when Dataloader num_workers>0 #415

Description

@BradZhone

Describe the bug

CPU memory utilization grows with training and finally cause OOM when num_workers of Dataloader greater than 0.
Especially when more datasets are used, this mem growth phenomenon becomes more obvious.

Environment

torch 2.3.0+cu121

Other information

No response

Metadata

Metadata

Assignees

Labels

bugSomething isn't working

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions