Does the training data explain a particular error in text generation? We can use LIT to analyze a T5 model on the CNN-DM summarization task. LIT's scalar plot ...
確定! 回上一頁