3.1 How to setup a Federated Learning project? (PI)

To setup a Federated Learning (FL) project, select the project you want to manage:

Then you could see the details of the project. You could then:

Upload AI model. (ref Chapter 3.2)
Upload initial AI model weights (optional). (ref Chapter 3.3)
Create a Federated Learning training plan. (ref Chapter 4.2)

It's PI's duty to manage a Federated Learning (FL) project. PI could create several Federated Learning (FL) plans under a FL project. For each FL plan, PI could select different settings for that training plan. For details about setting the FL plan, please refer to Chapter4.2 - How to setup a Federated Learning plan?

There are some things that PI needs to do on the FL project level: For example, to upload the initial AI model, and upload the initial AI model weights.

Note:

The AI model is required to park under the docker images, and using the docker-push commands to send the AI model from local site to the Aggregator. For detailed techniques of how to implement it, please refer to the Federated Validation Developers' Guidebook.

After the AI model docker images are implemented, then follow the commands on Chapter 3.2 to upload the AI model.

Chapter 3.3 will show you how to upload the initial AI model weights.

Note:

It's PI's responsibility to upload the AI initial model and AI initial model weights. It's operated on the Aggregator Dashboard under FL project setting. Please make sure you (as PI) have finished this step before you create a FL plan.