Run on GPU Error

[erdaqorri]

Hi,
Thank you for developing Parafold.

I successfully ran the CPU part using the script you provided, however when I try to run the GPU part I get the error attached to the file here. This is the script that I am using to submit the job to the cluster.

#!/bin/bash

#SBATCH --job-name=parafold_gpu # Remove the file extension from the input filename
#SBATCH --output=parafold_gpu.out # Remove the file extension from the input filename
#SBATCH --nodes=1
#SBATCH --cpus-per-task=8
#SBATCH --mem=80GB
#SBATCH --partition=gpu
#SBATCH --gres=gpu:1

./run_alphafold.sh
-d /home/p_af2qe/monomer_af2_db
-o output
-m model_1,model_2,model_3,model_4,model_5
-p monomer_ptm
-i /home/p_af2qe/ParallelFold/input/mono_set1/GA98.fasta
-t 1800-01-01 \

Thank you for your help!

Alexa

parafold_gpu_part_err.txt

[hermannschwaerzlerUIBK]

Hi Alexa,

I had the very same problem. I was able to solve it like this:

• if not already done: conda activate parafold
• conda install -c nvidia cuda-nvcc=11.3

This installs some missing utilities, one of which is ptxas.

Regards,
Hermann

[erdaqorri]

Hi Hermann,

Thank you for your reply!
I have just installed the version of cuda you mentioned. I am waiting for the job to run. I will most likely provide a follow up comment in case it works (for future users).

Cheers,
Alexa

[erdaqorri]

@hermannschwaerzlerUIBK

I also struggled to make the singularity image work for about 2 months. However, only two weeks ago I managed to successfully run in our HPC using the nondocker version (https://github.com/kalininalab/alphafold_non_docker). It is very easy to set up, it seems stable, and it works both on CPU and gpu partitions for me.

P.S it is not the 2.3.2 version but the 2.3.1

I hope this helps, feel free to ask if you need some suggestions.

Cheers,
Alexa