ProSST icon indicating copy to clipboard operation
ProSST copied to clipboard

PdbQuantizer is too slow

Open yourh opened this issue 11 months ago • 2 comments

It's too slow and even cannot process multiple PDBs at once. I cannot understand how the authors use this code to process the AlphaFoldDB including millions proteins for PLM training. I guess the authors may have a much faster version but they don't release it.

yourh avatar Jan 19 '25 03:01 yourh

Thank you for raising this important concern. We're excited to share that we've just merged a significant optimization contributed by @mdanzi. Which took a batch of 100 proteins from running in about 7 hours to running in about 80 seconds. thanks again to mdanzi for the excellent contribution!

Tpan1039-ui avatar Feb 26 '25 07:02 Tpan1039-ui

We have updated the structure serialization module, and it is now very fast. Please try it out, thank you!

tyang816 avatar May 29 '25 13:05 tyang816