rapid-artist-48509
04/04/2025, 11:34 PMjolly-florist-66728
04/04/2025, 11:34 PMcluster_resources.yaml
or similar configuration file used in your Flyte deployment.
Here's a general approach:
1. Modify the Cluster Resource Configuration: Update the cluster_resources.yaml
file to set the GPU limit for the desired project and domain. For example:
cluster_resources:
customData:
- production:
- projectQuotaCpu:
value: "5"
- projectQuotaMemory:
value: "4000Mi"
- projectQuotaGpu:
value: "4" # Set GPU limit here
2. Apply the Configuration: Use flytectl
or your deployment tool to apply the updated configuration to your Flyte cluster.
Would you like more detailed instructions on any specific part of this process?
Sources:
- https://www.union.ai/docs/byoc/user-guide/core-concepts/tasks/task-hardware-environment/customizing-task-resources/
- https://www.union.ai/docs/serverless/user-guide/core-concepts/workspaces/ 👍 0 button 👎 1 button 💡 Hint: Mention @ancient-wolf-19325 in the thread for followups.rapid-artist-48509
04/04/2025, 11:36 PMjolly-florist-66728
04/04/2025, 11:36 PMcluster_resources.yaml
, and applying it to your Flyte deployment.
1. Define the GPU Limit: Create or update a configuration file (e.g., cluster_resources.yaml
) with the desired GPU limit for your project and domain. Here's an example configuration:
cluster_resources:
customData:
- development:
- projectQuotaCpu:
value: "2"
- projectQuotaMemory:
value: "3000Mi"
- projectQuotaGpu:
value: "4" # Set GPU limit here
2. Apply the Configuration: Use flytectl
to apply the updated configuration to your Flyte cluster. This ensures that the GPU limit is enforced for the specified project and domain.
Would you like more detailed instructions on applying this configuration?
Sources:
- Flyte Configuration Guide
- Customizing Task Resource Configuration 👍 0 button 👎 1 button 💡 Hint: Mention @ancient-wolf-19325 in the thread for followups.