Support IPA-based refiner #28
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What?
Instead of En-transformer refiner, IPA (https://github.com/lucidrains/invariant-point-attention) refiner has been integrated.
Why?
To see the effectiveness of different experiments. IPA has lied in the core of AF2 structure module. It is worthy to give a try.
How?
Simply follow existing RGN2_transformer implementation. The difference is output of coordinates instead of angles. However, the compatibility of angles (used by losses) has been done via discrete Frenet-Serret equations.
Testing?
dRMSD can reach around 5. Experimental hyper-para's include "structure_module_depth" and other IPA parameters. Running script has been included.
Screenshots
At the moment the xtension branch of mp_nerf repo has been integrated via source code. Some functions (e.g. mp_nerf.ml_utils.backbone_forcefield()) has not been implemented. Not a big deal. I leave it for further improvement.
Anything Else?
Integrate LSTM output of 4 angles has been discussed but not implemented. LSTM can give nice prediction of secondary structures. Worthy to give a trial.