flower
flower copied to clipboard
adap
Add W&B example
Describe the type of feature and its functionality.
In a few of the projects I've been involved I found it quite useful to integrate Weight&Biases in the FL training process. Would there be interest in adding an example using W&B to the examples/ directory?
Describe step by step what files and adjustments are you planning to include.
The idea would be to build upon the simulation_pytorch example and modify it (or create a new example) by adding W&B at the server-level and client-level. The elements that this example will include are:
- Basics about W&B biases (i.e. creating an account etc)
- W&B integration in the server so metrics such as global validation accuracy/loss at the end of each epoch are logged.
- Similarly, the server could log client-level specific metrics that has received in a given round.
- Finally, the example would show how all the data logged can be conveniently analysed in a W&B dashboard.
- I'd include demo code and link to a public W&B dashboard (so people can check what to expect from running the example).
- (potentially) I could add a more advance section in the example showing how to use W&B's sweep functionality.
Is there something else you want to add?
No response
@jafermarq hi , have you done implementation of basic pytorch example? if so .can you provide link for that?
