def gaussian_kl_divergence_keepbatch(self, mean, ln_var):
var = F.exp(ln_var)
kld = F.sum(mean * mean + var - ln_var - 1, axis=1) * 0.5
return kld
微信
公众号
扫码关注公众号