Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Unable to use TPU on GKE using on-demand quota #621

Open
samos123 opened this issue Aug 2, 2024 · 0 comments · May be fixed by #622
Open

Unable to use TPU on GKE using on-demand quota #621

samos123 opened this issue Aug 2, 2024 · 0 comments · May be fixed by #622

Comments

@samos123
Copy link
Contributor

samos123 commented Aug 2, 2024

Currently axlearn either adds a nodeSelector for spot=true or it adds a nodeSelector for reservation:

        if tier == "0" and cfg.reservation is not None:
            logging.info("Found tier=%s in env. Using reservation=%s", tier, cfg.reservation)
            selector.update({"cloud.google.com/reservation-name": cfg.reservation})
        else:
            logging.info("Found tier=%s in env. Using spot quota", tier)
            selector.update({"cloud.google.com/gke-spot": "true"})
            tolerations.append(
                {
                    "key": "cloud.google.com/gke-spot",
                    "operator": "Equal",
                    "value": "true",
                    "effect": "NoSchedule",
                }
            )

It should be possible to launch a job using on-demand TPU, however today that's not possible unless you remove this line:

selector.update({"cloud.google.com/gke-spot": "true"})
@samos123 samos123 changed the title Unable to use TPU using on-demand Unable to use TPU on GKE using on-demand quota Aug 2, 2024
samos123 added a commit to samos123/axlearn that referenced this issue Aug 3, 2024
@samos123 samos123 linked a pull request Aug 3, 2024 that will close this issue
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant