Closed Whitepinkwolf closed 3 months ago
As a trainable global model is used in splitVFL instead of a non-trainable global function, you can simply modify the config json file for aggVFL as the following to construct a splitVFL setting.
{
....
"model_list":{
....
"apply_trainable_layer": 1,
"global_model": "ClassificationModelHostTrainableHead"
}
}
You can further change the "global_model" to other trainable models that meets your need.
We have also added several example configureation files of splitVFL under folder ./src/configs/main_task/2_party/splitVFL/
. Please check our latest commit a9594d8badf7d250d77c5746e5a1de01b336fd47
on branch main
for finding them.
I can only find the settings for aggVFL in the main_task folder, but not splitVFL. Could you please provide more guidance on splitVFL?