Hyperparameters used for the results in the paper

[Martin36]

I am trying to reproduce the results of the sentence retriever model, but can't seem to get the accuracy to the same level as in the paper. I tried using the hyperparameters from the config.py file, but the accuracy is significantly lower than what is mentioned in the paper and the training stops after about 5 epochs due to no increase in the performence. In the paper no hyperparameters are mentioned. Do you still have information regarding the hyperparameters from the paper?