Mix training and inference infra and manifests #1487
Conversation
… gemma deployment manifest
Done |
|
|
||
|
|
||
| gcloud artifacts repositories add-iam-policy-binding fine-tuning \ | ||
| --role=roles/artifactregistry.reader \ |
There was a problem hiding this comment.
we don't need this anymore, correct?
There was a problem hiding this comment.
yes, not needed anymore. Corrected
| #internalCertManagement: | ||
| # enable: false | ||
| # webhookServiceName: "" | ||
| # webhookSecretName: "" |
There was a problem hiding this comment.
Question: There are some lines commented out through out this file. Is this intentional?
There was a problem hiding this comment.
Yes, the team felt it was important to keep the context of the comments.
| containers: | ||
| - name: gpu-job | ||
| imagePullPolicy: Always | ||
| image: us-docker.pkg.dev/google-samples/containers/gke/gemma-fine-tuning:v1.0.0 |
There was a problem hiding this comment.
Just a note for future developers (any passerby): This source code for this image exists at https://github.com/GoogleCloudPlatform/accelerated-platforms/tree/974f2eff748d00d2566024d6ec4dd7f309f641c5/use-cases/model-fine-tuning-pipeline/fine-tuning/pytorch/src.
|
|
||
|
|
||
| cd gke-platform | ||
| sed -ie 's/"deletion_protection": true/"deletion_protection": false/g' terraform.tfstate |
There was a problem hiding this comment.
Nitpick (feel free to ignore): Consider setting deletion_protection to false from the very start (in the google_container_cluster resources), since all Terraform in this repo are to be cleaned up at the end of the tutorial.
There was a problem hiding this comment.
Just a thought (no immediate action needed). This is out-of-scope for this pull-request, but we should consider modularizing the Terraform in this git repo. These gke_standard and gke_autopilot folders look similar to existing gke_standard and existing gke_autopilot. I'll add a comment in #861.
There was a problem hiding this comment.
Awesome work on these samples! 👏
Looks good to me.
I trust you've tested these samples appropriately for functionality.
Left a few comments, but nothing major.
Judging from internal discussions, I'm guessing this is ready for merge (even though this PR is still in draft mode). Merging...
NimJay
requested review from
alizaidis,
yoshi-approver and
a team
as code owners
Description
This PR adds samples for a tutorial related to mixed training and inference in a single cluster.
Tasks