Shashank Rajput

icon indicating a website link https://shashankrajput.github.io/

icon company @mosaicml @databricks

Results 1 comments of Shashank Rajput
trafficstars

WIP: Preventing the loss from being computed when the input token is EOS Token

> Then, if sequences are _not_ joined randomly, as in that TSP NN method, we definitely want to compute loss. Thanks for your comment! Yes, what you said makes sense....