Can't get the results in the paper or the results of the pre-trained model provided

[fanqinshan0114]

I configured two different environments according to environment.yml and requirements.txt. Even if the seed is fixed, the training results in these two environments are inconsistent. why is that? And I still can't get the results in the paper or the results of the pre-trained model provided. Is this normal?

[Murrol]

First, the environments installed from the environment.yml and the requirements.txt are different. It's to help everyone run our model on different machines. For example, the pytorch versions are different. Version differences would affect the processed dataset and inference performance slightly.

Second, I need to see your results to check if it's acceptable.