vdobrovolskii / wl-coref

This repository contains the code for EMNLP-2021 paper "Word-Level Coreference Resolution"
MIT License
104 stars 37 forks source link

inference using predict.py #33

Closed ShahzebPurelogics closed 2 years ago

ShahzebPurelogics commented 2 years ago

i save a input.jsonlines file with this { "document_id": "tc_mydoc_001", "cased_words": ["Hi", "!", "Bye", "."], "sentid": [0, 0, 1, 1] } Now when i run command python predict.py roberta "input.jsonlines" "output.jsonlines" --weights "data/roberta(e20_2021.05.02_01.16)_release.pt" using the weights downloaded from dropbox, it loads default config which gives errors maybe because it by default start training and when it gets no training file it gives error. Can you please give me details how to run inference using custom input using pretrained weights file downloaded from dropbox?

vdobrovolskii commented 2 years ago

Hi! Try this solution: https://github.com/vdobrovolskii/wl-coref/issues/13#issuecomment-1044694236