Running/ training on own dataset about titan HOT 4 CLOSED

Hi Cedric,
thanks for the interest in our work. It depends on what you want to achieve so I cant formulate an interesting research question for you from the far.
It depends a bit on the statistics and size of the dataset but probably I would first predict the test data with the semifrozen model (zero-shot setting) and then finetune the pretrained semifrozen model on the train data and re-evaluate it on the test.
Hope this helps!

from titan.

HI thanks for replying.
So I did as the readme suggested and start directly with the finetuning.
But I get a lot of errors, which seem to be connected to smiles embedding

from titan.

You're feeding amino acid sequences where there should be SMILES sequences. Please read the docs and compare your input files to the example inputs

from titan.

Ok, I thought that you can also provide .csv files with the amino acids, because it is also like that in ./scripts/flexible_training.py
I'll transform my data into smile and try to run it again.
Thanks for your help and patience.

from titan.