Kernel dying mac os Mojave 10.14

[efago]

After grappling with installation of LightGBM, which I couldn't get an easy installation with anaconda, I couldn't run lightgbm on jupyter because it is crashing instantly. Please give me your feedback, I am stuck. :(

Environment info

Operating System: Mac Mojave 14

CPU/GPU model: 16 GB 2400 MHz DDR4, 2.2 GHz Intel Core i7

C++/Python/R version: python 3.7

LightGBM version or commit hash: 2.2

Error message

The kernel appears to have died. It will restart automatically.

Reproducible examples

Steps to reproduce

param = {'num_leaves': 111,
'min_data_in_leaf': 149,
'objective':'regression',
'max_depth': 9,
'learning_rate': 0.005,
"boosting": "gbdt",
"feature_fraction": 0.7522,
"bagging_freq": 1,
"bagging_fraction": 0.7083 ,
"bagging_seed": 11,
"metric": 'rmse',
"lambda_l1": 0.2634,
"random_state": 133,
"verbosity": -1}
folds = KFold(n_splits=5, shuffle=True, random_state=15)
oof = np.zeros(len(train))

for fold_, (trn_idx, val_idx) in enumerate(folds.split(train.values, target.values)):
train_mat = lgb.Dataset(train.iloc[trn_idx][features],
label=target.iloc[trn_idx],
)
val_mat = lgb.Dataset(train.iloc[val_idx][features],
label=target.iloc[val_idx],
categorical_feature=categorical_feats
)

num_round = 10000
clf = lgb.train(param,
                trn_data,
                num_round,
                valid_sets = [train_mat, val_mat],
                verbose_eval=100,
                early_stopping_rounds = 200)

oof[val_idx] = clf.predict(train.iloc[val_idx][features], num_iteration=clf.best_iteration)

feature_importance = clf.feature_importance()

[drabrucelee]

I am also stuck at that point using the predict, hopping for feedback!

[efago]

I hope someone with a better knowledge would help.

[StrikerRUS]

@efago @drabrucelee What about simple python command? Does it crashes too?

Please post a log from the Jupyter console. You haven't share a MCVE, so without any logs it's hard to say what's going wrong here.

[efago]

@StrikerRUS thanks for your reply here's the log:
[I 22:26:42.885 NotebookApp] Starting buffering for f1001099-4595-4b6d-89ec-e3347aa5fc23:6033709f02634fb0a31115fc5f8011d4
[I 22:26:43.321 NotebookApp] Adapting to protocol v5.1 for kernel f1001099-4595-4b6d-89ec-e3347aa5fc23
OMP: Error #15: Initializing libomp.dylib, but found libiomp5.dylib already initialized.
OMP: Hint This means that multiple copies of the OpenMP runtime have been linked into the program. That is dangerous, since it can degrade performance or cause incorrect results. The best thing to do is to ensure that only a single OpenMP runtime is linked into the process, e.g. by avoiding static linking of the OpenMP runtime in any library. As an unsafe, unsupported, undocumented workaround you can set the environment variable KMP_DUPLICATE_LIB_OK=TRUE to allow the program to continue to execute, but that may cause crashes or silently produce incorrect results. For more information, please see http://openmp.llvm.org/

[efago]

@StrikerRUS I have fixed the problem. During the installation of lightgbm, I was having problems and I installed and uninstalled lightgbm a couple of times without taking care of openmp. So I ended up with multiple openmp packages in my system and it was the cause of the crash. Thanks for your help.

[StrikerRUS]

@efago Glad that you've solved your problem!

We have a workaround for this OpenMP issue in case of conda + Homebrew in our FAQ: https://lightgbm.readthedocs.io/en/latest/FAQ.html#lightgbm - Question 10.

[drabrucelee]

Thanks a lot, the problem is solved.

[deepdalytics]

@efago I seem to have the same problem. How did you remove the additional openmp packages/clean up afterwards? Thank you