Skip to content

Does version 2.6.3 of WSL still not support multiple NUMA nodes? #14015

@luoxiaopiaoye

Description

@luoxiaopiaoye

Windows Version

Windows: 10.0.22621.4317

WSL Version

2.6.3.0

Are you using WSL 1 or WSL 2?

  • WSL 2
  • WSL 1

Kernel Version

6.6.87.2-1

Distro Version

Ubuntu 24.04.3 LTS

Other Software

lightgbm 4.6.0

Repro Steps

When I was training with LightGBM, I noticed that even though I had specified njobs=192, it seems that WSL can still only use one NUMA node and cannot utilize all system resources. Can this be configured?
Do I need to provide other information to identify the problem, or is the current version like this?

Information displayed in lscpu:
Virtualization features:
Virtualization: AMD-V
Hypervisor vendor: Microsoft
Virtualization type: full
Caches (sum of all):
L1d: 3 MiB (96 instances)
L1i: 3 MiB (96 instances)
L2: 48 MiB (96 instances)
L3: 16 MiB (1 instance)
NUMA:
NUMA node(s): 1
NUMA node0 CPU(s): 0-191

<img width="506" height="357" alt="Image" src="https://github.com/user-attachments/assets/d62b3d30-754a-4120-b95b-9481f9e72db1"

Expected Behavior

I'm on WSL, and I can't seem to find the relevant settings in the config file either.

Actual Behavior

When training a large dataset with 900,000 rows using lightgbm, it is unable to utilize all CPU cores.

Diagnostic Logs

No response

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions