Open Mr-Tieguigui opened 2 years ago
This project loads from a local copy of the bart-base model. Either this model can be downloaded from huggingface into the directory '\watermarking-lexical-substitution\bart-base' or you can remove '\watermarking-lexical-substitution' to automatically download the model from huggingface.
Please know that this project is still in the research phase, so reproducibility may be limited. Additionally, the BART watermarker failed to produce desired results, and its inclusion was to show the difficulty of training models for watermarking.
I have downloaded the bart-base, but it still has the problem:
Can't load tokenizer for '\watermarking-lexical-substitution\bart-base'. Make sure that:
'\watermarking-lexical-substitution\bart-base' is a correct model identifier listed on 'https://huggingface.co/models'
or '\watermarking-lexical-substitution\bart-base' is the correct path to a directory containing relevant tokenizer files
File "\watermarking-lexical-substitution\BART_Watermarker\hparams.py", line 13, in
tokenizer = BartTokenizerFast.from_pretrained(os.path.join(os.getcwd(), 'bart-base'))
File "\watermarking-lexical-substitution\BART_Watermarker\model.py", line 11, in
import hparams
File "\watermarking-lexical-substitution\BART_Watermarker\playground.py", line 1, in
from model import MainModel
What's the problem?