Finetuning on DrugBankDDI

[JunhuaL]

I'm trying to adapt your work to finetune the model on the DrugBankDDI dataset but I haven't had much success. In multiple test runs, I couldn't get the testing f1 score to be above 0.5. Have you had similar issues like this in the past and do you have any possible resolutions?

Link to DrugBankDDI dataset:
https://github.com/AstraZeneca/chemicalx/tree/main/dataset/drugbankddi

[eda-ginger]

Hello, I would like to evaluate the performance of DrugBank and TWOSIDES using pretrained models as well, but I am facing difficulties in setting up the environment due to outdated Python libraries. If possible, could I get some advice on the environment you have implemented?