sent-conv-torch
sent-conv-torch copied to clipboard
the static performance is better than nonstatic performance
I want to ask if all datasets use the same hyperparameters? I didn't modify the hyperparameters and train model on the TREC, but the score I got is showing like: static: dev scores: 0.922, test score:0.936 while nonstatic: dev scores: 0.938, test score:0.928
Or could you explain the score in scores table?,It's dev score?