headacheboy

Hi, When postprocessing, I use the latest version of [RikVN/AMR](https://github.com/RikVN/AMR), and the api to get wiki label has been changed. I'm wondering whether these decrease the performance of model...

Hi, I get 81.0 now. But I am unable to get the best result 81.4... Could you provide your modification of post-processing and the dictionary on AMR2.0? Thank you!

Hello, We first build the vocabulary and extract word embedding from a pretrained GloVe. It can be downloaded [here](http://nlp.stanford.edu/data/glove.twitter.27B.zip). Details can be check on the [project page of Glove](https://nlp.stanford.edu/projects/glove/).

Hi, I didn't try my model without Bert. I used to face this issue, but it seems fixed when I add 1e-8 before the torch.log() funtion. Codes are in _compute_loss()_...

Hi, Attributes are in _extract_ folder

Hi, Original image data can be downloaded through **Image Link** in the readme.