Callback to save local model weights at client when the FL round ends

[paloma-jim]

What is your question?

Hi,
I am a newbie in Flower. I have been implementing a training pipeline with two servers and around 20 clients with a custom dataset for 100 FL rounds in flower==1.9.0 with pytorch. My code is based on the example provided here:
After doing the code, i want to implement a callback function in each client that saves the last local model (after performing the last training round) and calculates a hash function based on the weights of this last model. Could you provide me some insights on how to proceed?
Thank you in advance

[danieljanes]

Hi @paloma-jim, before answering your question, I want to recommend updating to Flower 1.10 (a big upgrade over 1.9).

Regarding your question: you could easily do this in the Client's fit method itself (client.py, between lines 80 and 83).