If you are referring to using local mode via the SM PySDK, then pipeline deployment is not supported.
As an alternative, given your three inference containers, you could manually run the services locally and then implement a kind of facade function that invokes the three services in pipeline and manages input/output accordingly.
Is it possible to create Parallel Pipelines in Sagemakerasked 6 months ago
How to define concurrency in SageMaker real-time inferenceasked 8 months ago
how to inference parameters to a huggingface model hosted in sagemaker?asked a month ago
How can I use serverless inference with models generated by SageMaker Autopilot?asked a month ago
What is the difference between SageMaker Pipelines and SageMaker Step Function SDK?Accepted Answer
is it possible to create steps within the sagemaker pipeline via cloudformation?asked 18 days ago
Is it possible to test locally SageMaker Inference Pipelines?Accepted Answer
Using R model in SageMaker ML pipelinesAccepted Answerasked 8 months ago
SageMaker Neo with SageMaker Serverless Inference possible?asked a month ago
[Feature Request] Serverless Inference with VPC Configasked 8 months ago