geoffzhang opened this issue 4 years ago · 0 comments
你好,请问在训练的过程中heatmap的loss比其他三个的loss(加权后)要大一个数量级,这是正常的吗?