docker build -t ml-train:latest -f Dockerfile . docker run --rm -v "$PWD/artifacts:/app/artifacts" ml-train:latest
docker build -t ml-serve:latest -f Dockerfile .
docker run --rm -p 8080:8080 -v "$PWD/artifacts:/app/artifacts" ml-serve:latest
docker run --rm -p 8080:8080
-v "$PWD/artifacts:/opt/ml/model"
inference:latest serve
curl -i http://localhost:8080/ping
curl -X POST http://localhost:8080/invocations
-H "Content-Type: application/json"
-d '{"features":[14.0,20.0,90.0,600.0,0.10,0.12,0.10,0.05,0.18,0.06,0.40,1.20,2.50,40.0,0.01,0.02,0.02,0.01,0.02,0.003,16.0,28.0,110.0,900.0,0.14,0.30,0.25,0.12,0.28,0.08]}'
curl -X POST http://localhost:8080/predict
-H "Content-Type: application/json"
-d '{
"features": [
14.0,20.0,90.0,600.0,0.10,0.12,0.10,0.05,0.18,0.06,
0.40,1.20,2.50,40.0,0.01,0.02,0.02,0.01,0.02,0.003,
16.0,28.0,110.0,900.0,0.14,0.30,0.25,0.12,0.28,0.08
]
}'
ECR private repo is required for sagemaker echo "dummy" > dummy.txt aws s3 cp dummy.txt s3://mlops-devops4solutions/training-input/dummy.txt aws sagemaker create-training-job --cli-input-json file://train-job.json
aws sagemaker create-model \
--cli-input-json file://create-model.json
--region us-east-1
aws sagemaker create-endpoint-config
--cli-input-json file://endpoint-config.json
--region us-east-1
aws sagemaker create-endpoint
--cli-input-json file://endpoint.json
--region us-east-1
aws sagemaker describe-endpoint
--endpoint-name mlops-demo-endpoint-002
--region us-east-1
An error occurred (ValidationException) when calling the CreateModel operation: Unsupported manifest media type application/vnd.oci.image.index.v1+json for image 936379345511.dkr.ecr.us-east-1.amazonaws.com/mlops-train:latest. Ensure that valid manifest media type is used for specified image. update the workflow
aws sagemaker-runtime invoke-endpoint
--endpoint-name mlops-demo-endpoint-002
--content-type application/json
--body fileb://payload.json
out.json
--region us-east-1
cat out.json
aws sagemaker delete-endpoint --endpoint-name mlops-demo-endpoint-002 --region us-east-1 aws sagemaker delete-endpoint-config --endpoint-config-name mlops-demo-epc-002 --region us-east-1 aws sagemaker delete-model --model-name mlops-demo-model-002 --region us-east-1
"EndpointStatus": "Failed",
"FailureReason": "CannotStartContainerError. Please ensure the model container for variant AllTraffic starts correctly when invoked with 'docker run serve'",
mkdir -p tmp_model
Local testing
aws sagemaker create-model-package-group
--model-package-group-name mlops-demo-group
--model-package-group-description "MLOps demo model registry"
--region us-east-1
aws sagemaker create-model-package
--cli-input-json file://register-model.json
--region us-east-1