CODEWITHSUNDEEP
pythonpython-3.xmachine-learninglightgbm
Amin Kiany
Amin Kiany

Reputation: 800

lightgbm<=3.3.1: early_stopping() got an unexpected keyword argument 'min_delta'

I am dealing with a Regression problem for which I used LGBMRegressor. I also utilize early_stopping as callback in .fit(...) as follows:

reg = LGBMRegressor(**lgbr_params)
reg.fit(train_valid[features], 
        train_valid[target], 
        verbose=100,
        eval_set=[(test[features], test[target])], 
                   eval_metric=['rmse', 'mae'],
                   callbacks=[early_stopping(stopping_rounds=10, first_metric_only=False)], )

which works quite fine. However, I would like to consider some sort of "tolerance" in my early_stopping callback function. According to lightgbm documentation, this is apparently possible using min_delta argument in early stopping callback function. When I add this to my code:

reg = LGBMRegressor(**lgbr_params)
reg.fit(train_valid[features], 
        train_valid[target], 
        verbose=100,
        eval_set=[(test[features], test[target])], 
                   eval_metric=['rmse', 'mae'],
                   callbacks=[early_stopping(stopping_rounds=10, first_metric_only=False), min_delta=[0.1, 0.1]])

I face with the following error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-39-834c9eaac9b2> in <module>
      1 reg = LGBMRegressor(**lgbr_params)
      2 reg.fit(train_valid[features], train_valid[target], eval_set=[(test[features], test[target])], eval_metric=['rmse'],
----> 3         callbacks=[early_stopping(stopping_rounds=10, first_metric_only=False, min_delta=[1.0])], verbose=100)
TypeError: early_stopping() got an unexpected keyword argument 'min_delta'

I am using windows, and tried to upgrade the lightbm to solve the issue, but failed. Any idea?

Upvotes: 4

Views: 7443

Answers (1)

Miguel Trejo
Miguel Trejo

Reputation: 6667

I was facing the same issue, as latest lightgbm release on pipy is 3.3.1, I had to install it from github.

  1. First install cmake

  2. Install lightgbm from github

git clone --recursive https://github.com/microsoft/LightGBM
cd LightGBM/python-package
python setup.py install

it shoud work now

from lightgbm import early_stopping

early_stopping(stopping_rounds=10, first_metric_only=False, min_delta=[0.1, 0.1])
>>> <function lightgbm.callback.early_stopping.<locals>._callback(env: lightgbm.callback.CallbackEnv) -> None>

Upvotes: 4

Related Questions