When we use the last_hidden_state
of BertModel.forward
, do we need to normalize that results for such job like Sentiment Analysis? (Here is a distribution of last_hidden_state
over a sample text)
When we use the last_hidden_state
of BertModel.forward
, do we need to normalize that results for such job like Sentiment Analysis? (Here is a distribution of last_hidden_state
over a sample text)