You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: examples/gke-tpu-v6/README.md
+11-11Lines changed: 11 additions & 11 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,6 @@
1
1
# GKE TPU V6 blueprint
2
2
3
-
This example shows how a TPU cluster with v6 machines and topology 4x4 can be created. The example also includes a `tpu-available-chips.yaml` that creates a kubernetes service and job. The job includes commands to install `jax` and run a simple command using jax, on the TPU.
3
+
This example shows how a TPU cluster with v6 machines and topology 4x4 can be created. The example also includes a `tpu-multislice.yaml` that creates a kubernetes service and job. The job includes commands to install `jax` and run a simple command using jax, on the TPU.
4
4
5
5
Key parameters when working with TPUs:
6
6
@@ -59,7 +59,7 @@ This section guides you through the cluster creation process, ensuring that your
59
59
*`BUCKET_NAME`: the name of the new Cloud Storage bucket.
60
60
*`COMPUTE_REGION`: the compute region where you want to store the state of the Terraform deployment.
61
61
62
-
1. In the [`community/examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml`](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/community/examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml) file, replace the following variables in the `terraform_backend_defaults` and `vars` sections to match the specific values for your deployment:
62
+
1. In the [`examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml`](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml) file, replace the following variables in the `terraform_backend_defaults` and `vars` sections to match the specific values for your deployment:
63
63
64
64
*`bucket`: the name of the Cloud Storage bucket you created in the previous step.
65
65
*`project_id`: your Google Cloud project ID.
@@ -72,10 +72,10 @@ This section guides you through the cluster creation process, ensuring that your
72
72
*`authorized_cidr`: The IP address range that you want to allow to connect with the cluster. This CIDR block must include the IP address of the machine to call Terraform.
73
73
*`reservation`: the name of the compute engine reservation of TPU v6 nodes.
74
74
75
-
To modify advanced settings, edit `community/examples/gke-tpu-v6/gke-tpu-v6.yaml`.
75
+
To modify advanced settings, edit `examples/gke-tpu-v6/gke-tpu-v6.yaml`.
76
76
77
77
1. To use on-demand capacity, you can remove the reservation usage by making the following changes.
78
-
1. Remove the `reservation` variable from the [`gke-tpu-v6-deployment.yaml`](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/community/examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml) file.
78
+
1. Remove the `reservation` variable from the [`gke-tpu-v6-deployment.yaml`](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/examples/gke-tpu-v6/gke-tpu-v6-deployment.yaml) file.
79
79
1. Remove the `reservation_affinity` block from the nodepool module.
80
80
81
81
1. Generate [Application Default Credentials (ADC)](https://cloud.google.com/docs/authentication/provide-credentials-adc#google-idp) to provide access to Terraform.
@@ -86,8 +86,8 @@ This section guides you through the cluster creation process, ensuring that your
1. After deployment, the blueprint will output instructions for running a fio benchmark job. This job serves as a validation test to confirm that the GCS mounts are working correctly for both reading and writing. Follow the printed instructions to run the test.
117
117
118
118
## Run the sample job
119
119
120
-
The [tpu-available-chips.yaml](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/community/examples/gke-tpu-v6/tpu-available-chips.yaml) file creates a service and a job resource in kubernetes. It is based on https://cloud.google.com/kubernetes-engine/docs/how-to/tpus#tpu-chips-node-pool. The workload returns the number of TPU chips across all of the nodes in a multi-host TPU slice.
120
+
The [tpu-multislice.yaml](https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/examples/gke-tpu-v6/tpu-multislice.yaml) file creates a service and a job resource in kubernetes. It is based on https://cloud.google.com/kubernetes-engine/docs/how-to/tpus#tpu-chips-node-pool. The workload returns the number of TPU chips across all of the nodes in a multi-host TPU slice.
121
121
122
122
1. Connect to your cluster:
123
123
@@ -127,7 +127,7 @@ The [tpu-available-chips.yaml](https://github.com/GoogleCloudPlatform/cluster-to
127
127
128
128
Replace the `REGION` and `PROJECT_ID` with the ones used in the blueprint.
129
129
130
-
1. Update the nodeSelector under the template spec of tpu-available-chips.yaml file. The values depend on the tpu accelerator and tpu topology used in the blueprint.
130
+
1. Update the nodeSelector under the template spec of tpu-multislice.yaml file. The values depend on the tpu accelerator and tpu topology used in the blueprint.
131
131
132
132
```yaml
133
133
nodeSelector:
@@ -138,7 +138,7 @@ The [tpu-available-chips.yaml](https://github.com/GoogleCloudPlatform/cluster-to
0 commit comments