Segfault saving model.

[DebinUser42]

I am using Debin to train a CRF via py/train_crf.py and I get a segfault in Nice2Predict. Nice2Predict software library versions are from the Debin docker image.

Calling Nice2Predict/bazel-bin/n2p/training/train_json --input 128_kfold_10/1/crf/feature.json --log_dir 128_kfold_10/1/crf --valid_labels c_valid_labels --out_model 128_kfold_10/1/crf/x86_64/ --num_threads 8 --training_method pl --max_labels_z 8
*** Aborted at 1564576336 (unix time) try "date -d @1564576336" if you are using GNU date ***
PC: @     0x7f02e7c946f8 fwrite
*** SIGSEGV (@0x0) received by PID 206441 (TID 0x7f02e9115880) from PID 0; stack trace: ***
    @     0x7f02e8ad1390 (unknown)
    @     0x7f02e7c946f8 fwrite
    @           0x45b6cf GraphInference::SaveModel()
    @           0x410e25 LearningMain<>()
    @           0x40d97a main
    @     0x7f02e7c46830 __libc_start_main
    @           0x40d4a9 _start
    @                0x0 (unknown)
Segmentation fault (core dumped)

The same segfault happens using a single thread. The feature.json file I am using can be found here.