GPU support for lightgbm

[Jianyang Zhao (uzhao)]

I'm trying to use FLAML for lightgbm with GPU, but it failed due to max_bin too large. I don't see a place to setup max_bin, how can I limit this params? Thanks!

class GPULGBM(LGBMEstimator):
    def __init__(self, **config):
        super().__init__(device="gpu", **config)


automl = AutoML()
automl.add_learner(learner_name='gpulgbm', learner_class=GPULGBM)


automl_settings = {
    "time_budget": 60*3,
    "metric": "accuracy",
    "task": "classification",
    "log_file_name": "test.log",
    "estimator_list": ["gpulgbm"],
    "eval_method": "cv",
    "ensemble": True,
    "custom_hp": {
        "gpulgbm": {
            "max_bin": {
                "domain": tune.choice([8, 16, 32, 64])
            }
        }
    }
}

automl.fit(
    X_train=train_X,
    y_train=train_high_y,
    **automl_settings
)

[Jianyang Zhao (uzhao)]

Found log_max_bin in source

    "custom_hp": {
        "gpulgbm": {
            "log_max_bin": {
                "domain": tune.lograndint(lower=3, upper=7),
                "init_value": 5,
            },
        }
    },