baker-laboratory / RoseTTAFold-All-Atom

Other
652 stars 118 forks source link

RFAA can't run any RNA models? #110

Open jeramon opened 6 months ago

jeramon commented 6 months ago

Hi! TL;DR has anyone been able to run a protein/RNA complex and if so how?

Does this:

This repo currently does not support making RNA MSAs or pairing protein MSAs with RNA MSAs but this is functionality that we are keen to add. For now, please use RF-NA for modeling cases requiring paired protein-RNA MSAs.

mean that we can't run any protein-RNA complex? Asking because I can't run any model that contains RNA and I'm not asking it to make an RNA MSA (and indeed I shouldn't be able to because I'm only using protein databases). The first error I get is described in issues #57 and #109 and the problem there seems to by that 'U' gets indexed as '85' instead of '30.' But even after getting past this I run into a different issue that seems to be a recycling problem:

Error executing job with overrides: [] Traceback (most recent call last): File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/run_inference.py", line 206, in main runner.infer() File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/run_inference.py", line 155, in infer outputs = self.run_model_forward(input_feats) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/run_inference.py", line 121, in run_model_forward outputs = recycle_step_legacy(self.model, File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/training/recycling.py", line 30, in recycle_step_legacy output_i = ddp_model(input_i) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/RFAA_env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, *kwargs) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/model/RoseTTAFoldModel.py", line 368, in forward msa, pair, xyz, alpha_s, xyz_allatom, state, symmsub, quat = self.simulator( File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/RFAA_env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(args, kwargs) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/model/Track_module.py", line 1109, in forward msa_full, pair, xyz, state, alpha, symmsub, quat = self.extra_block[i_m](msa_full, pair, File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/RFAA_env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, kwargs) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/model/Track_module.py", line 963, in forward xyz, state, alpha, quat = self.str2str( File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/RFAA_env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, *kwargs) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/RFAA_env/lib/python3.10/site-packages/torch/amp/autocast_mode.py", line 14, in decorate_autocast return func(args, kwargs) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/model/Track_module.py", line 533, in forward xyz_frame = xyz_frame_from_rotation_mask(xyz, rotation_mask, atom_frames) File "/home/jamon/software/RFAA/RoseTTAFold-All-Atom/rf2aa/util.py", line 220, in xyz_frame_from_rotation_mask frames_reindex[:, i, :] = (i+atom_frames[..., i, :, 0])*natoms + atom_frames[..., i, :, 1] IndexError: index 0 is out of bounds for dimension 1 with size 0

Maybe this is all related to that first issue but I thought I'd put my entire rabbit-hole journey here in case it's helpful. Thanks!