Segmentation fault (core dumped) when using lncDC-train.py

[Jefferson-Chou]

I was using my own data to train models.
And I ran:
fold=1
python bin/lncDC-train.py -m fold${fold}_pc_train.fasta -c /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold${fold}_cds_train.fasta -l /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold${fold}_lnc_train.fasta -o /data1/yyzhou/LncADeep_mouse/identification/lncDC_perform/fold${fold} -t 1

Here are logs:

Process Start.
Checking if the training files exist ...
File /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_pc_train.fasta exist.
File /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_cds_train.fasta exist.
File /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_lnc_train.fasta exist.

Checking if the training files are in fasta format ...
/data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_pc_train.fasta format checking: PASS
/data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_cds_train.fasta format checking: PASS
/data1/yyzhou/LncADeep_mouse/identification/vs_src/fold1_lnc_train.fasta format checking: PASS

Initializing dataframe ...
Total Number of transcripts loaded: 172107
Calculating transcript lengths ...

Removing Non-valid transcripts (sequence that have non-ATGCatgc letters & sequence length less than 200 nt) ...
Number of valid transcripts for training: 171424

Extracting SIF and PF features ...
OpenBLAS warning: precompiled NUM_THREADS exceeded, adding auxiliary array for thread metadata.
To avoid this warning, please rebuild your copy of OpenBLAS with a larger NUM_THREADS setting
or set the environment variable OPENBLAS_NUM_THREADS to 64 or lower
train_lncDC.sh: line 2: 193916 Segmentation fault (core dumped) python bin/lncDC-train.py -m /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold${fold}_pc_train.fasta -c /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold${fold}_cds_train.fasta -l /data1/yyzhou/LncADeep_mouse/identification/vs_src/fold${fold}_lnc_train.fasta -o /data1/yyzhou/LncADeep_mouse/identification/lncDC_perform/fold${fold} -t 1

Whatever the thread parameter -t was set as, the error still occured.

[lim74]

Hi Jefferson,

Thank you for trying LncDC and reporting the issues. The error is likely due to OpenBLAS using lots of threads by default.

Could you try running the following code before you run lncDC-train.py in terminal?

export OPENBLAS_NUM_THREADS=1
export OMP_NUM_THREADS=1
export MKL_NUM_THREADS=1
export NUMEXPR_NUM_THREADS=1

Best,
Minghua