Skip to content

[ml-service] Remote launch/request #415

Description

@gichan-jang

Currently, using ml-service-remote, registering models and pipelines on remote devices has been implemented.
Next, let's implement the function of launching the pipeline from the sender device (RTOS) or requesting interference using registered model on remote devices.

  • Pipeline launch: launch and set/get state of the pipeline
  • Inference request
    • Singleshot (send input data with model config file (in/out dim, type, etc.) if necessary)
    • Pipeline (using appsrc, get caps or model config file)
  • Permission: Who will get permission and how to manage it?
  • Scheduling: How to deal the many request? thread pool or load balancer, etc.
  • pull mode: Get model or pipeline from connected devices.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions