Activity in progress: Porting software to AMD to enable nf-core ProteinFold at Pawsey

Details

• Deep Learning models work natively with NVIDIA hardware.
• Setonix at Pawsey contains a large number of AMD GPUs.
• Models can be made to work with AMD hardware available at Pawsey.

Completed

• Build AMD compatible container for AlphaFold2.
• Build AMD compatible container for ColabFold.
• Build AMD compatible container for Boltz-1.
• Build AMD compatible container for ESMFold.
• Build AMD compatible container for Boltz-2.
• Adapt containers for compatibility with nf-core proteinfold.
• Test nfcore proteinfold workflow at Pawsey.

In Progress

• ProteinFold config optimized for Pawsey.

Future

• Port software to AMD that has been prioritised by the community.

Acknowledgements

This work was supported by Australian BioCommons and the Pawsey Supercomputing Research Centre (https://ror.org/04f2f0537). Australian BioCommons receives NCRIS funding through Bioplatforms Australia. Pawsey Supercomputing Research Centre receives funding from the Australian Government and the Government of Western Australia.