How do I execute each step separately in RF and XGBoost

Zama Concrete ML
1

Hi, I have a question with executing each steps separately in RF and XGBoost.
I applied the code in How do I encrypt, run and decrypt separately? then I got the error here y = model.quantized_module_.dequantize_output(q_y). It says there are no quantized_module_ in RF or XGBoost but I couldn’t find other solutions. So I wonder how can I run this step in RF and XGBoost. Thanks in advance.

2 Likes
2

Hello @minho, thanks for your interest in Concrete ML!

Most models, like XGBoost and RandomForest, have both a quantize_input and dequantize_output method that should do what you are looking for.

Linear models don’t have dequantize_output exposed directly in 0.5.x, and for those you should use what is explained in your link.

This will be fixed in the next release.

Don’t hesitate if you have further questions.

1 Like
3

Solved it. Thanks for your help.

1 Like