The output of the model is inconsistent

#2
by amirakhlaghiqqq - opened

Hey,
Thanks for your great model. I am trying to use PickScore to evaluate the performance of a prompt enhancement model. The problem is that, If I rerun the model on the same set of inputs the results are different. Do you have any ideas what could be the reason? I experiment with defining seed and enforcing the cuda to be deterministic, but the problem persist!

Best regards

Sign up or log in to comment