function not found

[omaclean]

Hi, I'm trying to run through the listed steps, but I can't get foldtree2/learn_lightning.py to run because of this error

I ran
python learn_lightning.py --dataset <path_to_my_dataset.h5> --model-name my_model --output-dir <my_output_dir>

then I got this error almost immediately:

)
Total epochs: 50, Burn-in epochs: 0, After burn-in epochs: 50
Epoch 1/50: 0%| | 0/32 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/oscar/foldtree2/foldtree2/learn_lightning.py", line 258, in
edgeloss, _ = ft2.recon_loss_diag(data, edge_index, decoder, plddt=True, offdiag=True, key='edge_probs')
AttributeError: module 'foldtree2.src.foldtree2_ecddcd' has no attribute 'recon_loss_diag'

I checked the file foldtree2/src/foldtree2_ecddcd.py and confirmed that I couldn't find a function named recon_loss_diag, which seems to be the source of the error.

Thanks for your work on this tool!

[cactuskid]

hey. thanks for trying out foldtree2. im going to revise the readme a bit since the training isnt strictly necesary to apply the models that are already trained ( they are trained wtih 10k random structural cluster representatives from the AFDB clusters database to give a good representative sample of lots of different folds). This would be useful if you are doing something with sets of structures that are a bit exotic and not included in the AFDB ( think viruses and metagenomic dark matter like stuff in the ESM atlas ). Im refactoring a lot of the componenets of FT2 and it's still under heavy development so things may break every so often. I'll goover the training scripts again to make sure everything works if you need to train models on custom sets of structures.