pytorchltr
pytorchltr copied to clipboard
A better explanation about the n tensor
The input to an LTR loss function comprises three tensors:
scores
: a tensor of size (batch_size, number of scores)
relevance
: a tensor of size (batch_size, number of relevance scores)
n
: a tensor of size (batch_size) ( the number of docs per learning instance)
Could you provide a better explanation of n
?