From: Exploring the potential of ChatGPT in medical dialogue summarization: a study on consistency with human preferences
Parameters
Values
encoder
(classifier/transformer/rnn)
batch size
(1000/2000/3000)
train steps
10,000
dropout
0.1
learning rate
\(2e^{-3} \cdot min\left( step^{-0.5}, step \cdot warmup^{-1.5} \right)\)
warmup
(1000/10,000)
optimizer
adam