Seldon Overview

Deploy machine learning locally with Docker or to a Kubernetes cluster. Scale to 1000s of models. Run data pipelines for models containing outlier and drift detection along with model explainers. Comes with tracing, metrics and a CLI. Supports tensorflow, pytorch, sklearn, ONNX, custom python models and many more ML artifacts. Run experiments to test new models and combine with any Kubernetes service mesh. Create synchronous and asynchronous inference pipelines backed by Kafka.

Instalação

instala o heml
Seldon

-> cria um cluster no GKE com o command line
gcloud container clusters create-auto ml-dev --region us-east1 --project=home-lab

-> Criei um namespace pro seldon
kubectl create namespace seldon-system

-> Instalei o seldon no namespace novo

helm install seldon-core seldon-core-operator \
    --repo https://storage.googleapis.com/seldon-charts \
    --set usageMetrics.enabled=true \
    --namespace seldon-system \
    --set istio.enabled=true

-> criei um namespace pra deploy em dev
kubectl create namespace ml-dev

-> peguei minhas credenciais no cluster pra ajustar meu kubectl
gcloud container clusters get-credentials ml-dep loyments --zone us-east1-b

-> me dei permissão de admin no cluster

kubectl create clusterrolebinding cluster-admin-binding \
    --clusterrole=cluster-admin \
    --user=$(gcloud config get-value core/account)

baixa e instala o istio
curl -L https://istio.io/downloadIstio | sh -

sudo mv istioctl /usr/local/bin

instala o istio no cluster:
istioctl install

pegar o endpoint:
kubectl -n istio-system get service istio-ingressgateway -o jsonpath='{.status.loadBalancer.ingress[0].ip}'

http://35.231.98.48:15021/seldon/ml-dev/iris/api/v1.0/doc/

MLOps with Seldon