Sklearn-linear model_locator for a Cross Site Model Evaluation

[kandalto]

Python version (python3 -V)

3.10.13

NVFlare version (python3 -m pip list | grep "nvflare")

2.4

NVFlare branch (if running examples, please use the branch that corresponds to the NVFlare version, git branch)

2.4

Operating system

Ubuntu 20.04

Have you successfully run any of the following examples?

• hello-numpy-sag with simulator
• hello-pt with simulator
• hello-numpy-sag with POC
• hello-pt with POC

Please describe your question

Hello!

I want to add Cross Site Model Evaluation (CSE) (CrossSiteModelEval worflow) to the sklearn-linear example here:
https://github.com/NVIDIA/NVFlare/blob/main/examples/advanced/sklearn-linear/sklearn_linear_higgs.ipynb

However, I cannot find a readily available model_locator/formatter for sklearn-linear in the codebase. Going through the other examples of CSE I found CSE model_locators for pytorch and numpy models at
https://github.com/NVIDIA/NVFlare/blob/main/nvflare/app_opt/pt/file_model_locator.py and https://github.com/NVIDIA/NVFlare/blob/main/nvflare/app_common/np/np_model_locator.py

I suspect these files can be modified for sklearn-linear models and put in the custom folder of the config_fed_server file. However, it is not intuitive for me which changes I should make. Can you please confirm that these are the files than needs to be modified? What modifications should I make? Do I need just model_locator or formatter as well?

Maybe there are other already implemented solutions that I am missing?

Many thanks.

[chesterxgchen]

Thanks for your question, sorry for the delayed response. @ZiyueXu77 can you help answer this question

[ZiyueXu77]

Hi @kandalto, thanks for the question, and yes I think you can start with the numpy one, which is more general-purpose
https://github.com/NVIDIA/NVFlare/blob/main/nvflare/app_common/np/np_model_locator.py
You will need to modify it so as to satisfy sklearn's model format for linear model, please refer to these lines
https://github.com/NVIDIA/NVFlare/blob/dev/examples/advanced/sklearn-linear/jobs/sklearn_linear_base/app/custom/linear_learner.py#L67-L70