Hi All,
Working on training a classification model but instead of using standard linear+logSoftmax+NLLLoss I would like to use Euclidean+MulConstant+MultiMarginLoss. This was straight forward with torch7, but in pytorch all these modules don’t work with autograd. How can I go about it ? I searched the discuss forum but couldn’t find anything relevant. Apologies if I missed it.
Kind regards
Sagar M W