你好，请教一下知识蒸馏方面的问题 about light-reid HOT 2 CLOSED

from light-reid.

font{ line-height: 1.6; } ul,ol{ padding-left: 20px; list-style-position: inside; } please refer https://arxiv.org/abs/1503.02531 font{ line-height: 1.6; } font{ line-height: 1.6; } Guan'an WangInstitute of Automation, Chinese Academy of Sciences,No. 95 Zhongguancun East Rd., Beijing, China. 在2021年01月25日 22:06，xiexu666[email protected] 写道： 在代码中看到蒸馏部分的loss主要是 class KLLoss: '''KL Divergence''' def init(self, t=4): self.t = t def call(self, logits_s, logits_t): p_s = F.log_softmax(logits_s / self.t, dim=1) p_t = F.softmax(logits_t / self.t, dim=1) loss = F.kl_div(p_s, p_t, reduction='sum') * (self.t ** 2) / logits_s.shape[0] return loss 请问关于这里的原理相应论文可以用来学习吗，感谢~ —You are receiving this because you are subscribed to this thread.Reply to this email directly, view it on GitHub, or unsubscribe.

收到，非常感谢您的回复。

from light-reid.