PSEUDOTENSOR / Jonathan McKinney

The optuna example is chosen so that the score is clearly bad and very different at the end. I'm not able to easily generate such a case. openml was included...

@jameslamb As I said, I already spent hours reducing it to this point. Saying I'm not willing is quite unfair of you and disappointing. All I said was I'm not...

I'll try to dump a csv and params just before the final training. Is there a problem with sharing a csv and a script?

I spent more time @jameslamb and reduce script as much as possible. 1) For my problem, the seed seems to make a considerable impact on performance. 2) When one passes...

Thanks! Given the instability in the model building to seed, I will take the approach https://github.com/microsoft/LightGBM/issues/5268#issuecomment-1146904708 for now, hacking in the booster into the sklearn object, since I really want...

@StrikerRUS Hi, noticed it was closed. One issue is reproducibility, but the other is the general instability and poor score. By instability, I mean the result is very bad or...

Here is more minimal MRE: ``` import pickle X, y = pickle.load(open("lgb257b.pkl", "rb")) params = dict(categorical_feature='', device_type='gpu', gpu_device_id=0, gpu_platform_id=0, min_data_in_bin=1, max_bin=255) model = lgb.LGBMClassifier(**params) model.fit(X, y, categorical_feature='') ``` FYI gpu_use_dp=True...

Hi, any thoughts? Seems like a clear MRE, but it's been 5 days and no response. Thanks.

@guolinke ?

``` File "/opt/h2oai/dai/python/lib/python3.8/site-packages/lightgbm_gpu/sklearn.py", line 712, in fit self._Booster = train(params, train_set, File "/opt/h2oai/dai/python/lib/python3.8/site-packages/lightgbm_gpu/engine.py", line 235, in train booster = Booster(params=params, train_set=train_set) File "/opt/h2oai/dai/python/lib/python3.8/site-packages/lightgbm_gpu/basic.py", line 2528, in __init__ _safe_call(_LIB.LGBM_BoosterCreate( File "/opt/h2oai/dai/python/lib/python3.8/site-packages/lightgbm_gpu/basic.py",...