def test_lightgbm_kserve():
service_name = "isvc-lightgbm"
predictor = V1beta1PredictorSpec(
min_replicas=1,
lightgbm=V1beta1LightGBMSpec(
storage_uri="gs://kfserving-examples/models/lightgbm/iris",
resources=V1ResourceRequirements(
requests={"cpu": "100m", "memory": "256Mi"},
limits={"cpu": "100m", "memory": "256Mi"},
),
),
)
isvc = V1beta1InferenceService(
api_version=constants.KSERVE_V1BETA1,
kind=constants.KSERVE_KIND,
metadata=client.V1ObjectMeta(
name=service_name, namespace=KSERVE_TEST_NAMESPACE
),
spec=V1beta1InferenceServiceSpec(predictor=predictor),
)
kserve_client = KServeClient(config_file=os.environ.get("KUBECONFIG", "~/.kube/config"))
kserve_client.create(isvc)
kserve_client.wait_isvc_ready(service_name, namespace=KSERVE_TEST_NAMESPACE)
res = predict(service_name, "./data/iris_input_v3.json")
assert res["predictions"][0][0] > 0.5
kserve_client.delete(service_name, KSERVE_TEST_NAMESPACE)
def _get_resource_requirements(self):
"""Get the resource requirements that are configured.
This will construct and return the V1ResourceRequirements object to be used as the value
of resource property of a container. The resource values are those provided when the S3
data mover was initialized.
"""
resources = V1ResourceRequirements()
requests = {}
if self.cpu_request is not None:
requests['cpu'] = self.cpu_request
if self.memory_request is not None:
requests['memory'] = self.memory_request
if requests:
resources.requests = requests
limits = {}
if self.cpu_limit is not None:
limits['cpu'] = self.cpu_limit
if self.memory_limit is not None:
limits['memory'] = self.memory_limit
if limits:
resources.limits = limits
return resources
