[python-package] fix mypy errors about early stopping rounds

[jameslamb]

Contributes to #3867 .

Fixes the following mypy errors.

engine.py:214: error: Argument "stopping_rounds" to "early_stopping" has incompatible type "Optional[Any]"; expected "int"  [arg-type]
engine.py:705: error: Argument "stopping_rounds" to "early_stopping" has incompatible type "Optional[Any]"; expected "int"  [arg-type]

mypy is rightly complaining that there's no guarantee that params["early_stopping_round"] will contain an integer here:

LightGBM/python-package/lightgbm/engine.py

Lines 211 to 214 in 2fe2bf0

	if "early_stopping_round" in params:
	callbacks_set.add(
	callback.early_stopping(
	stopping_rounds=params["early_stopping_round"],

However, we do protect against that on the C++ side.

from sklearn.datasets import make_regression
import lightgbm as lgb

X, y = make_regression()
dtrain = lgb.Dataset(X, label=y)
bst = lgb.train(
    params={"early_stopping_rounds": "too-many"},
    train_set=dtrain,
)

# lightgbm.basic.LightGBMError: Parameter early_stopping_round should be of type int, got "too-many"

So this PR proposes:

• silencing those errors with # type: ignore comments
• adding a unit test to enforce that lightgbm continues raising that informative error
  • found via git grep 'should be of type' that there wasn't currently one

We could, alternatively, address these errors by doing something like the following on the Python side:

def _integer_value_from_params(name: str, params: Dict[str, ny]) -> int:
    """
    Extract an integer value from ``params``, or raise an error f it can't be cast to integer.

    This method intentionally does not try to resolve parameter liases with e.g.
    ``_choose_param_value()``, to avoid unnecessary duplicate rocessing.
    """
    val = params[name]
    if isinstance(val, (int, float)):
        return int(val)
    else:
        raise ValueError(f"Could not cast value '{val}' for parameter '{params}' to integer.")

# ...
    stopping_rounds = _integer_value_from_params("early_stopping_round", params)

But I think it's preferable to do that validation only on the C++ side, so that all LightGBM wrappers benefit from it and to limit duplication of logic between the Python side and C++ side.

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed.
To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues
including a reference to this.