You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello!
As a data scientist, I'm exploring the possibilities offered by Substra to train models.
I would like to launch an example of federated learning orchestrated by a compute plan but I didn't find so much information in the readme and docs. So I don't totally understand how the compute plan works and how it can generate a federated learning experiment.
Describe the ideal feature
The generic explanation about the compute plan is clear and examples help to understand how a part of it can be implemented. Nevertheless, it would be worth to detail all options possible with a compute plan (transfer learning by stipulating with model to use for initialization, aggregation after each epoch, after each batch, etc.).
Is there any already existing similar feature?
The example in the folder Examples of the repo Substra is useful to understand some subtilities but it only uses Transfer Learning without aggregation step.
The text was updated successfully, but these errors were encountered:
Introduction
Hello!
As a data scientist, I'm exploring the possibilities offered by Substra to train models.
I would like to launch an example of federated learning orchestrated by a compute plan but I didn't find so much information in the readme and docs. So I don't totally understand how the compute plan works and how it can generate a federated learning experiment.
Describe the ideal feature
The generic explanation about the compute plan is clear and examples help to understand how a part of it can be implemented. Nevertheless, it would be worth to detail all options possible with a compute plan (transfer learning by stipulating with model to use for initialization, aggregation after each epoch, after each batch, etc.).
Is there any already existing similar feature?
The example in the folder Examples of the repo Substra is useful to understand some subtilities but it only uses Transfer Learning without aggregation step.
The text was updated successfully, but these errors were encountered: