search

ray-project / ray

Ray is a unified framework for scaling AI and Python applications. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
https://ray.io
Apache License 2.0
33.01k stars 5.59k forks source link

Add AcceleratorType:None resources for nodes without GPUs #18523

Open ericl opened 3 years ago

ericl commented 3 years ago

(1) The current accelerator auto-detection should add "None" if no GPU is detected. This allows tasks to force placement on CPU-only nodes if needed. (2) We should document this as a way to resolve the issue of CPU tasks being scheduled onto GPU nodes (along with the workaround of setting num_cpus=0 on those nodes).

More generally, we should consider adding anti-affinity support to the scheduler.

scv119 commented 2 years ago

@ericl is this still relevant?

ericl commented 2 years ago

We should do it, but it's not particularly pressing.

On Tue, Apr 19, 2022, 5:26 PM Chen Shen @.***> wrote:

@ericl https://github.com/ericl is this still relevant?

— Reply to this email directly, view it on GitHub https://github.com/ray-project/ray/issues/18523#issuecomment-1103293622, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAADUSWDORNUX5NPEL7HJWTVF5FLLANCNFSM5D3P3HOQ . You are receiving this because you were mentioned.Message ID: @.***>