About training Cons parsing with PTB dataset

yzhangcs / parser

:rocket: State-of-the-art parsers for natural language.

https://parser.yzhang.site/

MIT License

825 stars 138 forks source link

About training Cons parsing with PTB dataset #117

Closed Quang-elec44 closed 1 year ago

Quang-elec44 commented 1 year ago

Hi, I ran this command: python -m supar.cmds.crf_con train -b -d 0 -c crf-con-en -p model -f char, and it showed train: Dataset(n_sentences=1, n_batches=1, n_buckets=1). Then the loss value was really high. Can you tell me what is the problem? Is it necessary to download the dataset or your code will download it?

yzhangcs commented 1 year ago

@Quang-elec44 Yes, please check that the PTB files exist in your specified paths.

Quang-elec44 commented 1 year ago

@yzhangcs But where can i get this file: data/ptb/train.pid