uw-ipd / RoseTTAFold2NA

RoseTTAFold2 protein/nucleic acid complex prediction
MIT License
310 stars 69 forks source link

Unresolved references, network/loss.py not used #28

Open redsphinx opened 1 year ago

redsphinx commented 1 year ago

I want to finetune the model on my own data and some references are missing. Some examples:

Where is the loss.py code actually used?

fdimaio commented 1 year ago

Hello, loss.py is part of training. We have not yet included our training code, but can add it in the future.

redsphinx commented 1 year ago

Do you have an estimate for when the training code could be uploaded? Or any other resource I can look to for training/fine-tuning my own model?

luukhd2 commented 1 year ago

Hi all, I have the same issue, could the training code be uploaded, or sent privately if it is not fully ready? If not, is there an approach you recommend for looking at fine-tuning the pre-trained model?

redsphinx commented 1 year ago

Hey guys, I have another question related to the loss.py. In the paper you outline mutiple loss functions. I can't seem to find the seq loss (masked amino-acid recovery loss)