Repository navigation
error in mc_esm_gearnet #20
Description
Howdy, thank u for ur awesome work in Enhancing Protein Language Models with Structure-based Encoder and Pre-training. I am running the pretaining experiment now, and I am facing an issue of "Can't find atom_feature in features.atom". I paste the error statement below.
It seems like it cannot recognize the atom_feature: null or bond_feature: null. Do I need to change the source code for implementing these two arguments?
Any help will be grateful!
Activity
Oxer11 commentedon Apr 3, 2023
Hi! Thanks for your interest in our work.
The config
atom_feature: nullandbond_feature: nullmean that we want to setatom_feature=Noneandbond_feature=Noneso as to avoid the contruction of atom and bond features. This should be recognized by yaml library. It's quite weird that your code tried to findatom_featureinfeatures.atom. Have you ever changed the config file? If not, maybe you can first print the value ofcfg.dataset.atom_featurein thepretrain.py.hxu105 commentedon Apr 4, 2023
Hello, thanks for your response. I checked the atom_feature in the config, and it prints None
I also had one question in lazy mode, lazy mode is to load pdb file each time we call get_item function, however, the init will set up the self.data by loading the pickle file, so proteins have already existed in our dataset and taken some CPU memory. Why does lazy mode claim to save CPU cost?
Oxer11 commentedon Apr 5, 2023
This is weird. Have you faced the same issue when pre-training with GearNet-Edge instead of ESM-GearNet? Could you share the full log file with me?
I'm sorry that I haven't considered the lazy mode for CPU. Typically, we assume there is no
.pklfile when using lazy mode. So maybe you can try to remove the.pklfile and the lazy mode should work.