You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
We trained a pairwise model using LightGBMRanker with below settings, the model performance is poor, it event didn't split the records correctly(about 98% records in a single leaf node). But when we train the model with python lib with the same settings, it works fine.
@jinmfeng001 could you provide some sample dataset (maybe some dummy ranking data that looks similar to your actual data and which can be run on synapseml and python versions of lightgbm ranker) that we could reproduce the issue on? This could be happening for a variety of reasons (maybe some different parameters? or possibly even some bug?), but without a way to reproduce the issue it's hard to diagnose. Do you notice any difference if you set the parameter setUseSingleDatasetMode(false)? That was the biggest change in the recent past, but I'm not sure if it's related to your issue.
Describe the bug
We trained a pairwise model using LightGBMRanker with below settings, the model performance is poor, it event didn't split the records correctly(about 98% records in a single leaf node). But when we train the model with python lib with the same settings, it works fine.
tree infos:
To Reproduce
The parameters from the model dump file are:
Expected behavior
A clear and concise description of what you expected to happen.
Info (please complete the following information):
** Stacktrace**
Additional context
Add any other context about the problem here.
AB#1767712
The text was updated successfully, but these errors were encountered: