Skip to content

Commit e830d7d

Browse files
authored
Minor fix for the SkyPilot tutorial (#1001)
* test * minor fix for the SkyPilot tutorial * minor fix for the SkyPilot tutorial * minor fix for the SkyPilot tutorial * minor fix for the SkyPilot tutorial * minor fix for the SkyPilot tutorial
1 parent 889ec98 commit e830d7d

File tree

4 files changed

+11
-11
lines changed

4 files changed

+11
-11
lines changed

tutorials-and-examples/skypilot/README.md

+1-1
Original file line numberDiff line numberDiff line change
@@ -106,7 +106,7 @@ If you encounter an error, please consult the [offical documentation](https://do
106106

107107
You can find the available GPUs in a GKE cluster.
108108
```bash
109-
sky show-gpus
109+
sky show-gpus --cloud kubernetes
110110
```
111111

112112
3. Find the context names

tutorials-and-examples/skypilot/dws-and-kueue/README.md

+7-7
Original file line numberDiff line numberDiff line change
@@ -81,10 +81,10 @@ Server Version: v1.30.6-gke.1596000
8181
```
8282
If not, you can change the version in Terraform with the `kubectl_version` variable
8383
## Install and configure Kueue
84-
1. Install Kueue from the official manifest.\
85-
Note that `--server-side` switch . Without it the client cannot render the CRDs because of annotation size limitations.
84+
1. Install [Kueue](https://kueue.sigs.k8s.io/) from the official manifest.\
85+
Note that `--server-side` switch . Without it the client cannot render the CRDs because of annotation size limitations. For more configuration options visit [Kueue's installation guide](https://kueue.sigs.k8s.io/docs/installation/).
8686
```bash
87-
VERSION=v0.7.0
87+
VERSION=v0.10.2
8888
kubectl apply --server-side -f https://github.com/kubernetes-sigs/kueue/releases/download/$VERSION/manifests.yaml
8989
```
9090
2. Configure Kueue for pod provisioning by patching the Kueue configmap.
@@ -140,7 +140,7 @@ allowed_clouds:
140140
kubernetes:
141141
# Use the context's name
142142
allowed_contexts:
143-
- gke_${PROJECT_NAME}_europe-${LOCATION}_${CLUSTER_NAME}
143+
- gke_${PROJECT_NAME}_${LOCATION}_${CLUSTER_NAME}
144144
autoscaler: gke
145145
```
146146
And verify again:
@@ -251,7 +251,7 @@ This section details how to fine-tune Gemma 2B for SQL generation on GKE Autopil
251251
The [finetune.py](https://github.com/GoogleCloudPlatform/ai-on-gke/blob/skypilot_dws_kueue/tutorials-and-examples/skypilot/dws-and-kueue/code/finetune.py) script uses QLoRA with 4-bit quantization to fine-tune Gemma 2B on SQL generation tasks.
252252

253253
### Configure GCS Storage Access
254-
The infrastructure Terraform configuration in [main.tf](https://github.com/GoogleCloudPlatform/ai-on-gke/blob/skypilot_dws_kueue/tutorials-and-examples/skypilot/dws-and-kueue/main.tf) includes Workload Identity and GCS bucket setup:
254+
The infrastructure Terraform configuration in [main.tf](https://github.com/GoogleCloudPlatform/ai-on-gke/blob/main/tutorials-and-examples/skypilot/dws-and-kueue/main.tf) includes Workload Identity and GCS bucket setup:
255255
```
256256
module "skypilot-workload-identity" {
257257
source = "terraform-google-modules/kubernetes-engine/google//modules/workload-identity"
@@ -409,8 +409,8 @@ No resource satisfying Kubernetes({'L4': 1}) on Kubernetes.
409409
sky.exceptions.ResourcesUnavailableError: Kubernetes cluster does not contain any instances satisfying the request: 1x Kubernetes({'L4': 1}).
410410
To fix: relax or change the resource requirements.
411411
412-
Hint: sky show-gpus to list available accelerators.
413-
sky check to check the enabled clouds.
412+
Hint: `sky show-gpus --cloud kubernetes` to list available accelerators.
413+
`sky check` to check the enabled clouds.
414414
```
415415
Make sure you added `autoscaling: gke` to the sky config in step [Install SkyPilot](#install-skypilot)
416416

tutorials-and-examples/skypilot/dws-and-kueue/example_environment.tfvars

+2-2
Original file line numberDiff line numberDiff line change
@@ -26,8 +26,8 @@ gcs_bucket = "skypilot-model-bucket"
2626
# For Autopilot clusters
2727
autopilot_cluster = true
2828

29-
# For Standard clusters, configure GPU node pools:
30-
#autopilot_cluster = false
29+
# For Standard clusters, configure GPU node pools:
30+
# autopilot_cluster = false
3131

3232
# If using Standard cluster please uncomment the
3333
# following gpu_pools block to enable queued_provisioning

tutorials-and-examples/skypilot/dws-and-kueue/main.tf

+1-1
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
# Copyright 2023 Google LLC
1+
# Copyright 2025 Google LLC
22
#
33
# Licensed under the Apache License, Version 2.0 (the "License");
44
# you may not use this file except in compliance with the License.

0 commit comments

Comments
 (0)