Description
I have 2 server, and I got completely differnt results when training using exact same data and parameters. The lgbm version are both v4.5.0. I found "Stopped training because there are no more leaves that meet the split requirements" in server 2. Which i believe is the reason for the huge difference, i set num_trees = 1000, and in server 2 dumped model we got n_trees is ~ 400(results in server 1 looks just fine).
Reproducible example
Environment info
LightGBM version or commit hash:
v4.5.0
Command(s) you used to install LightGBM
One is python 3.12 and another is 3.11
Additional Comments