mirror of
https://github.com/hanzo-ml/kserve.git
synced 2026-10-30 20:53:49 +00:00
fc20909f00
Signed-off-by: Mariah Holder <[email protected]> Signed-off-by: tomcli <[email protected]> Co-authored-by: Mark Winter <[email protected]> Co-authored-by: Tommy Li <[email protected]> Co-authored-by: Alex Lembiyeuski <[email protected]> Co-authored-by: Magdalena Kuhn <[email protected]>
53 lines
4.7 KiB
YAML
53 lines
4.7 KiB
YAML
name: Serve a model with KServe
|
|
description: Serve Models using KServe
|
|
inputs:
|
|
- {name: Action, type: String, default: 'create', description: 'Action to execute on KServe'}
|
|
- {name: Model Name, type: String, default: '', description: 'Name to give to the deployed model'}
|
|
- {name: Model URI, type: String, default: '', description: 'Path of the S3 or GCS compatible directory containing the model.'}
|
|
- {name: Canary Traffic Percent, type: String, default: '100', description: 'The traffic split percentage between the candidate model and the last ready model'}
|
|
- {name: Namespace, type: String, default: '', description: 'Kubernetes namespace where the KServe service is deployed.'}
|
|
- {name: Framework, type: String, default: '', description: 'Machine Learning Framework for Model Serving.'}
|
|
- {name: Runtime Version, type: String, default: 'latest', description: 'Runtime Version of Machine Learning Framework'}
|
|
- {name: Resource Requests, type: String, default: '{"cpu": "0.5", "memory": "512Mi"}', description: 'CPU and Memory requests for Model Serving'}
|
|
- {name: Resource Limits, type: String, default: '{"cpu": "1", "memory": "1Gi"}', description: 'CPU and Memory limits for Model Serving'}
|
|
- {name: Custom Model Spec, type: String, default: '{}', description: 'Custom model runtime container spec in JSON'}
|
|
- {name: Autoscaling Target, type: String, default: '0', description: 'Autoscaling Target Number'}
|
|
- {name: Service Account, type: String, default: '', description: 'ServiceAccount to use to run the InferenceService pod'}
|
|
- {name: Enable Istio Sidecar, type: Bool, default: 'True', description: 'Whether to enable istio sidecar injection'}
|
|
- {name: InferenceService YAML, type: String, default: '{}', description: 'Raw InferenceService serialized YAML for deployment'}
|
|
- {name: Watch Timeout, type: String, default: '300', description: "Timeout seconds for watching until InferenceService becomes ready."}
|
|
- {name: Min Replicas, type: String, default: '-1', description: 'Minimum number of InferenceService replicas'}
|
|
- {name: Max Replicas, type: String, default: '-1', description: 'Maximum number of InferenceService replicas'}
|
|
- {name: Request Timeout, type: String, default: '60', description: "Specifies the number of seconds to wait before timing out a request to the component."}
|
|
- {name: Enable ISVC Status, type: Bool, default: 'True', description: "Specifies whether to store the inference service status as the output parameter"}
|
|
|
|
outputs:
|
|
- {name: InferenceService Status, type: String, description: 'Status JSON output of InferenceService'}
|
|
implementation:
|
|
container:
|
|
image: quay.io/aipipeline/kserve-component:v0.12.0
|
|
command: ['python']
|
|
args: [
|
|
-u, kservedeployer.py,
|
|
--action, {inputValue: Action},
|
|
--model-name, {inputValue: Model Name},
|
|
--model-uri, {inputValue: Model URI},
|
|
--canary-traffic-percent, {inputValue: Canary Traffic Percent},
|
|
--namespace, {inputValue: Namespace},
|
|
--framework, {inputValue: Framework},
|
|
--runtime-version, {inputValue: Runtime Version},
|
|
--resource-requests, {inputValue: Resource Requests},
|
|
--resource-limits, {inputValue: Resource Limits},
|
|
--custom-model-spec, {inputValue: Custom Model Spec},
|
|
--autoscaling-target, {inputValue: Autoscaling Target},
|
|
--service-account, {inputValue: Service Account},
|
|
--enable-istio-sidecar, {inputValue: Enable Istio Sidecar},
|
|
--output-path, {outputPath: InferenceService Status},
|
|
--inferenceservice-yaml, {inputValue: InferenceService YAML},
|
|
--watch-timeout, {inputValue: Watch Timeout},
|
|
--min-replicas, {inputValue: Min Replicas},
|
|
--max-replicas, {inputValue: Max Replicas},
|
|
--request-timeout, {inputValue: Request Timeout},
|
|
--enable-isvc-status, {inputValue: Enable ISVC Status}
|
|
]
|