unclear how to prepare data for lambdarank_position_bias_regularization option

[greyowl]

documentation

I want to train a learning to rank model with position bias regularization option(lambdarank_position_bias_regularization) turned on. But the documentation is unclear on how the position data needs to be prepared for training.

Can I get a pointer on how to prepare training data with position information? I am planning to use python

[jameslamb]

Please see https://lightgbm.readthedocs.io/en/latest/Advanced-Topics.html#support-for-position-bias-treatment

[jameslamb]

And there are some Python examples in tests:

LightGBM/tests/python_package_test/test_engine.py

Line 727 in e9a6c79

def test_ranking_with_position_information_with_file(tmp_path):

[shiyu1994]

You may use a .position file
https://lightgbm.readthedocs.io/en/latest/Advanced-Topics.html#support-for-position-bias-treatment
or you may specify the position argument in Python API of Dataset
https://lightgbm.readthedocs.io/en/latest/pythonapi/lightgbm.Dataset.html#lightgbm.Dataset