It then uses TensorFlow.js to run the DistilBERT-cased model fine-tuned for ... F1 score on SQuAD v1.1 dev set, compared to 88.7 for BERT-base …
確定! 回上一頁