search

sokrypton / ColabFold

Making Protein folding accessible to all!
MIT License
2.01k stars 504 forks source link

Crashing of complex with 600 residues #179

Open gezmi opened 2 years ago

gezmi commented 2 years ago

We are running the following protein sequence with AlphaFold2-ptm, using the AlphaFold2_mmseqs2 notebook. It is approximately 629 AA long. I have a Colab Pro account, so RAM should not be an issue for a lot longer one. Something similar MSTSLAVVRLDRELPMPTRAHDGDAGVDLYSAENVELAPGQRALVSTGIAVAIPHGMVGLVHPRSGLAARVGLSIVNSPGTIDAGYRGEIKVSLINLDPQTPVVISRGDRIAQLLVQRVELPELVEVTSFDEAGLADTTRGDGGHGSSGGHASL : MSTSLAVVRLDRELPMPTRAHDGDAGVDLYSAENVELAPGQRALVSTGIAVAIPHGMVGLVHPRSGLAARVGLSIVNSPGTIDAGYRGEIKVSLINLDPQTPVVISRGDRIAQLLVQRVE LPELVEVTSFDEAGLADTTRGDGGHGSSGGHASL : MSTSLAVVRLDRELPMPTRAHDGDAGVDLYSAENVELAPGQRALVSTGIAVAIPHGMVGLVHPRSGLAARVGLSIVNSPGTIDAGYRGEIKVSLINLDPQTPVVISRGDRIAQLLVQRVELPELVEVTSFDEAGLADTTRGDGGHGSSGGHASL: MTQVASPVVQARYSPVELSAALGLFPPTDEQAAVIAAPPGPLVVIAGAGAGKTETMAARVVWLVANGFATPSQVLGLTFTRKAAGQLLRRVRTRLARLAGAGLAPGSGASDESATVSTYHAFAGTLLREHGLLLPVEPDTRLLSETELWQLAYDV

Expected Behavior

Normal run of ColabFold

Current Behavior

When running model 2, the crashing of the runtime, with out of memory error.

ajasja commented 2 years ago

The GPU memory is usually the problem. You probably got a card with too little GPU ram. You can run it on pure CPU it will just take a lot longer (10-20x).

martin-steinegger commented 2 years ago

It worked for me. Attached is the zip. Maybe you could try to factory reset? test_544a3.result.zip