jeandut

Yes it would be equivalent mathematically in this case, but would induce lag because of the docker spawning at each traintuple processing and be a bit more complicated in terms...

Hello, thanks for the great package ! You should really consider supporting newer versions of Python as it is breaking lots of workflows. Your survival utils are really good and...

When launching very long computations (1 "run" is expensive) it would greatly help to understand how much time is required to finish ( to go up to `--max-runs`).

I only found out about this part of the doc from your comment and actually had to "debug" this behavior by diving deep into the code. Maybe adding a simple...

Note that 1. would only work for `objective_value` and not for other metrics potentially.

Yes outlier detection is a research domain in itself I agree and it depends heavily on what the user wants. It would be much simpler from an implementation perspective to...

Hi ! Thanks for the answer maybe the term "hard to debug" was a bit of an overstatement I apologize. It's just that coming from a "mini-batch SGD DL" background,...

Also the syntax for the yaml config follows the CLI, aka MySolver[learning_rate=0.1, l1_ratio=3.], which does not follow [mlcollections](https://github.com/google/ml_collections) standard of using nested dicts (also it would allow nested params aka...

Hello @charlesbvll nice work ! I would also be glad to see another flower example in FLamby alongside substra, fedbiomed and fedml: https://github.com/owkin/FLamby/tree/main/integration

Okay will copy checkpoint in the function however striving for functional purity is a bit of an unattainable ideal.