[CUDA] Add binary objective for cuda_exp

[shiyu1994]

Add binary objective for cuda_exp. (#5163)

[guolinke]

it seems the cuda self-hosted machine in queue? BTW, do we have a test for the cuda binary objective?

[shiyu1994]

BTW, do we have a test for the cuda binary objective?

I think this can be covered by existing tests for binary objective on CPUs, for example,

LightGBM/tests/python_package_test/test_engine.py

Lines 61 to 83 in 9489f87

	def test_binary():
	X, y = load_breast_cancer(return_X_y=True)
	X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)
	params = {
	'objective': 'binary',
	'metric': 'binary_logloss',
	'verbose': -1,
	'num_iteration': 50 # test num_iteration in dict here
	}
	lgb_train = lgb.Dataset(X_train, y_train)
	lgb_eval = lgb.Dataset(X_test, y_test, reference=lgb_train)
	evals_result = {}
	gbm = lgb.train(
	params,
	lgb_train,
	num_boost_round=20,
	valid_sets=lgb_eval,
	callbacks=[lgb.record_evaluation(evals_result)]
	)
	ret = log_loss(y_test, gbm.predict(X_test))
	assert ret < 0.14
	assert len(evals_result['valid_0']['binary_logloss']) == 50
	assert evals_result['valid_0']['binary_logloss'][-1] == pytest.approx(ret)

because we are replacing the default device with cuda_exp in

LightGBM/.ci/test.sh

Line 204 in 9489f87

grep -q 'std::string device_type = "cuda_exp"' $BUILD_DIRECTORY/include/LightGBM/config.h || exit -1 # make sure that changes were really done

when testing for cuda_exp version.

[shiyu1994]

It is weird that this PR should now passes all the tests. But my local tests are failing with cuda_exp enabled. I'm taking a deeper check into this.

[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.