search

hakyimlab / PredictDBPipeline

16 stars 10 forks source link

command qsub not found #5

Open alisonrocco opened 4 years ago

alisonrocco commented 4 years ago

Hello, The computing cluster I am working with uses Slurm instead of SGE and requires the $SBATCH command to submit jobs to the queue. Because of this I am running into an error when attempting to run the train_models.py file. Is there a simple way to alter this file so I can run it on a slurm based system?

Heroico commented 4 years ago

Hi Allison,

I assume you are referring to the train_models.py file in the tutorial (i.e. joblogs/example/train_models.py). This file is a simple python script that you can modify in any text editor. I'm not familiar with Slurm so you would have to check the documentation for sbatch. sbatch is the actual slurm command for job submission, i assume $SBATCH is a convenience setup by your cluster's personel

You would need to modify lines 8-10:

CMD = 'qsub -v study={0},expr_RDS={1},geno={2},gene_annot={3},snp_annot={4},' + \
    'n_k_folds={5},alpha={6},out_dir={7},chrom={8},snpset={9},window={10} ' + \
    '-N {0}_model_chr{8} -d {11} train_model_by_chr.pbs'

I'm guessing from a quick glance at Slurm docs that you would have to replace:

You would also need to modify joblogs/example/train_model_by_chr.pbs and replace PBS directives for SLURM directives.

Also, bear in mind that this code is deprecated and exists only for documentation purposes. The current pipeline is here, and a brief description is here