This makes sense. In that case why is my custom BERT Classification Model’s accuracy lower than the official BertForSequnceClassification?
This makes sense. In that case why is my custom BERT Classification Model’s accuracy lower than the official BertForSequnceClassification?