Open keiran-rowell-unsw opened 1 month ago
https://github.com/Australian-Structural-Biology-Computing/proteinfold/blob/8ce05982ff57ba3722c4e87093307b68f9923d43/modules/local/run_alphafold2.nf#L67C23-L67C28
Should be truly random as each 5 models on 5 trained networks need to explore different starting areas of the NN parameter space.
We can keep 53343 as a magic number for debugging purposes and new hardware (tell M. Grogan)?
https://github.com/Australian-Structural-Biology-Computing/proteinfold/blob/8ce05982ff57ba3722c4e87093307b68f9923d43/modules/local/run_alphafold2_pred.nf#L49C1-L49C30
Occurs in _pred as well.
Fixed in df3b0cdd2b40984e9c0a79a57fecd8e91b90312d and 7e753ad05f39abea71923065c366ff81cf7695a6.
Will run tests and open pull request
https://github.com/Australian-Structural-Biology-Computing/proteinfold/blob/8ce05982ff57ba3722c4e87093307b68f9923d43/modules/local/run_alphafold2.nf#L67C23-L67C28
Should be truly random as each 5 models on 5 trained networks need to explore different starting areas of the NN parameter space.
We can keep 53343 as a magic number for debugging purposes and new hardware (tell M. Grogan)?