UMT
UMT copied to clipboard
if use bert-base-uncased model, will lead to significant decrease in performance
Hi, Can only the bert-base-cased model be used? And why does using the bert-base-uncased model result in a significant decrease in performance, with only 40% of f1?