m6anet-run_inference error

[rezarahman12]

Dear m6Anet Thank you and congratulations on getting m6Anet published in Nat. methods. I tried m6anet-run_inference on single replicate and it worked well. But, when I have given three replicates of a sample (wildtype), a error came, similar to #38 . The below is my script-

module load m6anet

m6anet-run_inference --input_dir /scratch/project_mnt/S0077/xPore/rawdata/xpore_data/m6anet/dataprep/wt-rep1 \ /scratch/project_mnt/S0077/xPore/rawdata/xpore_data/m6anet/dataprep/wt-rep2 \ /scratch/project_mnt/S0077/xPore/rawdata/xpore_data/m6anet/dataprep/wt-rep3 \

                 --out_dir /scratch/project_mnt/S0077/xPore/rawdata/xpore_data/m6anet/m6a_infer \
                 --infer_mod_rate \
                 --n_processes 1 

The screenshot of the error is given below-

I ran the job in HPC cluster server using a PBS script, and the m6anet was loaded through module load m6anet. I do not know how to add your suggestion in the issue #38 "Inserting the torch.multiprocessing.set_sharing_strategy('file_system') line in the /home/stakatis/miniconda3/envs/m6anet/lib/python3.9/site-packages/m6anet/scripts/run_inference.py ".

Could you please advise any solution to this problem? Kind regards Reza

[chrishendra93]

hi @rezarahman12 , thanks for your kind words!

Recently I found that setting --n_processes 0 seems to solve this problem for me.https://pytorch.org/docs/stable/data.html suggests that setting num_workers to 0, in our case this is n_processes, will disable multiprocesses which might help with our problem over here. Let me know if this works for you

[rezarahman12]

Hi Christopher, Thank you so much for your quick response and time. It worked for me. Thank you.

I have another question regarding converting the transcript ID and its position obtained through m6Anet into genes and genomic position. So, I'm creating another new issue on that.

Thanks a lot. Best regards Reza