search

chaidiscovery / chai-lab

Chai-1, SOTA model for biomolecular structure prediction
https://www.chaidiscovery.com
Other
1.09k stars 138 forks source link

Straightforward way to generate more than 5 models? #64

Open dieglm opened 3 weeks ago

dieglm commented 3 weeks ago

Hi! Thank you for this amazing tool, also for the easy installation! I was wondering if there is any easy way of generating more than 5 models per case, besides looping over the execution of run_inference, is there any option to specify a custom number of generated models? Thank you!

arogozhnikov commented 2 weeks ago

we will have this in example script, for now you can look at this PR:

https://github.com/chaidiscovery/chai-lab/pull/2/files

Nicholas-Freitas commented 2 days ago

Hi, I wanted to add to this question. From the PR you linked, it appears that the new script wraps run_inference in a for loop. Am I correct that if I wanted to generate 100 predictions for a protein, I could just run the run_inference function 20 times? (5 predictions per run, x 20 runs).

Additionally, would these results be equivalent to setting num_diffn_samples to 100? Or would there be correlation between the 5 generated structures in each run? Not sure how the starting seeds are generated.

Thanks!