Executing the command with "--lastwords 1" will report an error, and "lastwords" cannot be recognized. Also, when I was training the "reddit" data set, the correct rate was still 17% after 10,000 epoch training. When executing the baseline backdoor, the highest backdoor accuracy rate It's only 2.6%. What's the matter?
Executing the command with "--lastwords 1" will report an error, and "lastwords" cannot be recognized. Also, when I was training the "reddit" data set, the correct rate was still 17% after 10,000 epoch training. When executing the baseline backdoor, the highest backdoor accuracy rate It's only 2.6%. What's the matter?