Open yinxusen opened 6 years ago
https://github.com/cs230-stanford/cs230-code-examples/blob/159df10a6187c7e1d6ec949c8e06d7f67f8f1cd2/tensorflow/nlp/model/model_fn.py#L68
The computation of accuracy and the metrics of accuracy below are lack of masking. In this way the accuracy could be wrong because of wrong predictions of padded tokens.
https://github.com/cs230-stanford/cs230-code-examples/blob/159df10a6187c7e1d6ec949c8e06d7f67f8f1cd2/tensorflow/nlp/model/model_fn.py#L68
The computation of accuracy and the metrics of accuracy below are lack of masking. In this way the accuracy could be wrong because of wrong predictions of padded tokens.