Stack Exchange network consists of 183 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.
Visit Stack Exchange
Stack Internal
Knowledge at work
Bring the best of human thought and AI automation together at your work.
Explore Stack Internal The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE_loss because you can always make $-(x-c)^2$ smaller by choosing $x$ farther from $c$. If $c$ is your target, this means your model is getting further from your goal.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood, whence MSE is a rescaling.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE_loss because you can always make $-(x-c)^2$ smaller by choosing $x$ farther from $c$. If $c$ is your target, this means your model is getting further from your goal.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE_loss because you can always make $-(x-c)^2$ smaller by choosing $x$ farther from $c$. If $c$ is your target, this means your model is getting further from your goal.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood, whence MSE is a rescaling.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSEMSE_loss because you can always make $-x^2$$-(x-c)^2$ smaller by choosing $x$ farther from 0$c$. If $c$ is your target, this means your model is getting further from your goal.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE because you can always make $-x^2$ smaller by choosing $x$ farther from 0.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE_loss because you can always make $-(x-c)^2$ smaller by choosing $x$ farther from $c$. If $c$ is your target, this means your model is getting further from your goal.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood.
The problem is return -MSE_loss + KLDiv_Loss. You don't want to minimize -MSE because you can always make $-x^2$ smaller by choosing $x$ farther from 0.
Use return MSE_loss + KLDiv_Loss instead. You can show that this is correct by starting from a Gaussian likelihood for your target tgts and manipulating the algebra to obtain the negative log-likelihood.
