sklearn remove warning, there is no verbose_eval

[startakovsky]

Is there any way to remove warnings in the sklearn API? The fit function only takes verbose which seems to only toggle the display of the per iteration details.

[LightGBM] [Warning] min_data_in_leaf is set=74, min_child_samples=20 will be ignored. Current value: min_data_in_leaf=74.

[StrikerRUS]

You are trying to set one parameter twice (or you are using an alias of parameter which shouldn't be passed via **kwargs). Please pay attention to the aliases in docs: https://lightgbm.readthedocs.io/en/latest/Parameters.html#learning-control-parameters

[startakovsky]

Thanks @StrikerRUS , you seem to be responding specifically to this warning. Here is another warning: [LightGBM] [Warning] No further splits with positive gain, best gain: -inf.

I am seeing this line appear too many times and other lines during training and I am interested in completely suppressing this warning.

These warnings are generated from the C++ code, and doesn't seem to be suppressed by filtering warnings from Python's warnings library.

I have tried setting verbose=-1, but it is not suppressing the warning.

Any advice, hacks or ways I might be able to suppress all warnings?

[StrikerRUS]

@startakovsky Yeah, I was talking about that concrete warning.

Now ping @guolinke because you've already discussed this with him in #1157.

[startakovsky]

thanks @StrikerRUS

@guolinke how do I suppress the C++ generated warnings in the sklearn API?
#1157

[tnmichael309]

import warnings
warnings.filterwarnings('ignore')

[AdamsMi]

doesn't work (the filter warnings hack). This is really annoying, this warning. Can anybody finally tell us how to get rid of this message?

[StrikerRUS]

@AdamsMi filtering warnings cannot work in this case because warnings come directly from C++ code. Passing verbose=-1 or verbose_eval=-1 doesn't work?

[guolinke]

for sklearn interface, you can set verbose=-1 when defining the model (not in fit).
for lgb.train interface, you can set verbose=-1 in param dict.

[tnmichael309]

Sadly... I use warining filter along with @guolinke mentioned and it works.
In my case, my warning came from resetting categorical features.
I'm using jupyter notebook on windows.

[tnmichael309]

Maybe I used notebook and the warning was not shown on it but console only.

[marcioribp]

I am having the same problem @tnmichael309.

[LightGBM] [Warning] Met categorical feature which contains sparse values. Consider renumbering to consecutive integers started from zero

Tried everything suggested and nothing works, this warning is haunting me.

Using lgb interface on linux console.

[StrikerRUS]

@marcioribp Please refer to https://lightgbm.readthedocs.io/en/latest/Advanced-Topics.html#categorical-feature-support, #1636

Categorical features must be encoded as non-negative integers (int) less than Int32.MaxValue (2147483647). It is best to use a contiguous range of integers started from zero.

#1920 (comment)

If you don't have memory issues, you can ignore this.

[marcioribp]

@marcioribp Please refer to https://lightgbm.readthedocs.io/en/latest/Advanced-Topics.html#categorical-feature-support, #1636

Categorical features must be encoded as non-negative integers (int) less than Int32.MaxValue (2147483647). It is best to use a contiguous range of integers started from zero.

#1920 (comment)

If you don't have memory issues, you can ignore this.

@StrikerRUS
Thank you very much! As it turns out, I had to convert my category columns to int 32 as mentioned in the documentation page.