Closed zshy1205 closed 5 years ago
This might be helpful: https://github.com/allanj/pytorch_lstmcrf/issues/4#issuecomment-526368279
Thanks
Thanks @mxc19912008 . To briefly elaborate more about this: forward_labeled
calculates the score (in log-space) corresponds to the gold path. forward_unlabeled
calculate the sum of the scores (in log-space) that corresponds to all the possible paths.
Can you give some advice about read your code? why the loss use forward_labeld and foward_unlabeld?