Run on Singularity¶
For those who wish to avoid the administrator rights requirements often associated with Docker usage (or the security hazards when used on a shared computing system), Singularity is a good option. Singularity is a container solution just like Docker, except it is designed specifically to offer secure deployment on shared cluster environments.
You can pull a Singularity container much like how you would pull a Docker container:
singularity pull docker://fcpindi/c-pac:latest
This will produce a Singularity container image in your current directory, named something like c-pac_latest.sif
. You can instead specify the local filename you want before the docker://
URI.
Running a Singularity image is similar to running a Docker image, except -B
maps local directories to a location in the Singularity image instead of -v
:
singularity run \
-B /Users/You/local_bids_data:/bids_dataset \
-B /Users/You/some_folder:/outputs \
-B /tmp:/tmp \
FCP-INDI-C-PAC-master-latest.simg \
/bids_dataset \
/outputs \
participant
Again, you can also provide an AWS S3 link for the data:
singularity run \
-B /Users/You/some_folder:/outputs \
-B /tmp:/tmp \
FCP-INDI-C-PAC-master-latest.simg \
s3://fcp-indi/data/Projects/ADHD200/RawDataBIDS \
/outputs \
participant
As mentioned above, in addition to the default pipeline, C-PAC comes packaged with a growing library of pre-configured pipelines that are ready to use. Once again, you can use the --preconfig
flag with Singularity to run any of the pre-configured pipelines. See the full selection of pre-configured pipelines here.
singularity run \
-B /Users/You/some_folder:/outputs \
-B /tmp:/tmp \
FCP-INDI-C-PAC-master-latest.simg \
s3://fcp-indi/data/Projects/ADHD200/RawDataBIDS \
/outputs \
participant \
--preconfig anat-only