Biowulf High Performance Computing at the NIH
bonito on Biowulf

A PyTorch Basecaller for Oxford Nanopore Reads. According to ONT this is a research release

provided as technology demonstrators to provide early access to features or stimulate Community development of tools. Support for this software will be minimal and is only provided directly by the developers.
Important Notes

Models available

Version Models
0.3.6 dna_r9.4.1

Note: fast: fast model hac: high accuracy sup: super high accuracy

Interactive job
Interactive jobs should be used for debugging, graphics, or applications that cannot be run as batch jobs.

Allocate an interactive session and run the program. Note that 0.3.6 runs on p100 GPUs but >=0.5.0 requires v100 or newer. Sample session:

[user@biowulf]$ sinteractive --gres=lscratch:50,gpu:v100x:1 --mem=12g --cpus-per-task=6
salloc.exe: Pending job allocation 46116226
salloc.exe: job 46116226 queued and waiting for resources
salloc.exe: job 46116226 has been allocated resources
salloc.exe: Granted job allocation 46116226
salloc.exe: Waiting for resource configuration
salloc.exe: Nodes cn3144 are ready for job

[user@cn3144]$ cd /lscratch/$SLURM_JOB_ID
[user@cn3144]$ module load bonito/0.5.1
[user@cn3144]$ cp -rL ${BONITO_TEST_DATA:-none}/* .
[user@cn3144]$ ls -lh
total 4.0K
drwxr-xr-x 3 user group 4.0K Feb  8 11:07 Zymo-GridION-EVEN-BB-SN
[user@cn3144]$ find Zymo-GridION-EVEN-BB-SN -name '*.fast5' -printf '.' | wc -c
[user@cn3144]$ ### basecalling command for bonito 0.3.6
[user@cn3144]$ bonito basecaller --fastq --recursive \
                    --device cuda dna_r9.4.1 Zymo-GridION-EVEN-BB-SN > reads.fastq
> loading model
> calling: 20829 reads [49:08,  7.08 reads/s]
[user@cn3144]$ ### basecalling command for bonito >=0.5.0
[user@cn3144]$ bonito basecaller --recursive --device cuda \
    dna_r9.4.1_e8.1_hac@v3.3 Zymo-GridION-EVEN-BB-SN > reads.fastq
> loading model dna_r9.4.1_e8.1_hac@v3.3
> completed reads: 160000
> duration: 0:24:34
> samples per second 4.7E+06
> done

[user@cn3144]$ exit
salloc.exe: Relinquishing job allocation 46116226

bonito achieved ~21 reads/s on a V100 or V100X GPU and the basecaller apparently does not scale to more than 1 GPU

Batch job
Most jobs should be run as batch jobs.

Create a batch input file (e.g., which uses the input file ''. For example:


module load bonito/0.5.0 || exit 1
cd /lscratch/$SLURM_JOB_ID || exit 1
cp -rL ${BONITO_TEST_DATA:-none}/* .
bonito basecaller --recursive --device cuda \
    dna_r9.4.1_e8.1_hac@v3.3 Zymo-GridION-EVEN-BB-SN > reads.fastq
mv reads.fastq $wd

Submit this job using the Slurm sbatch command.

sbatch --cpus-per-task=6 --mem=12g --gres=gpu:v100x:1,lscratch:50
Swarm of Jobs
A swarm of jobs is an easy way to submit a set of independent commands requiring identical resources.

Create a swarmfile (e.g. bonito.swarm). For example:

bonito basecaller --fastq --recursive --device cuda dna_r9.4.1 run1 > reads1.fastq
bonito basecaller --fastq --recursive --device cuda dna_r9.4.1 run2 > reads2.fastq

Submit this job using the swarm command.

swarm -f bonito.swarm -g 12 -t 6 --gres=gpu:v100x:1 --module bonito/0.3.6
-g # Number of Gigabytes of memory required for each process (1 line in the swarm command file)
-t # Number of threads/CPUs required for each process (1 line in the swarm command file).
--module bonito Loads the bonito module for each subjob in the swarm