search

mljar / mljar-supervised

Python package for AutoML on Tabular Data with Feature Engineering, Hyper-Parameters Tuning, Explanations and Automatic Documentation
https://mljar.com
MIT License
3k stars 401 forks source link

A little 🐛 in the printing of XGBoost results #534

Open munozariasjm opened 2 years ago

munozariasjm commented 2 years ago

Hi!,

Just to report a little (and unimportant) bug. When the results of the XGBoost models are printed, I heavily suspect (😂) the rsme and the fitting time are being interchanged: 6_Xgboost rmse 11.868675 trained in 1264.67 seconds Clearly it should be: 6_Xgboost rmse 1264.67 trained in 11.868675 seconds If necessary, tell me, I can open a PR with the fix. Regards and keep the excellent work on!!!

pplonski commented 2 years ago

Hi @munozariasjm,

It should be good. Please check the code: https://github.com/mljar/mljar-supervised/blob/04a90ffbff33b2c93a7c212825b987e73b7f62fe/supervised/base_automl.py#L290-L294

pplonski commented 2 years ago

Why do you think that they are interchanged?

unreal79 commented 2 years ago

There is no bug, I'm sure. munozariasjm doesn't know what he's talking about, it seems.

I'd closed the "issue".

munozariasjm commented 2 years ago

Ok, @unreal79. Let's suppose there is absolutely no bug... Then please explain to me how come we can print the RMSE for this toy data obtained by the linear model is: 3_Linear rmse 0.0 trained in 13.17 seconds Here is the notebook.

I still don't know where these printing bugs are coming from, but it is not normal.

pplonski commented 2 years ago

Thank you @munozariasjm for example code to reproduce the issue. It might be a bug with get_final_loss() method.

Is anyone interested in tracking the bug?

Vamp1899 commented 2 years ago

@pplonski Not able to understand the error at first place ,can you please elaborate so i can start working on it?