Problems with pretraining task

[Jackson8787]

Hi, thanks for the dataset, I have some problems with the pretraining task.

1. In line 6 of the file "/src/data/jsons/MVSA_info.json" , "ANP_clss_dir"should be"ANP_class_dir"right?
2. In line 22 of the file "/src/data/tokenization_new.py" , './E2E-MABSA' can't work ,and I found it can be replace as './facebook/bart-base'
3. I met the problem when I running MVSA_pretrain.sh .How can I fix it?

[lyhuohuo]

Sorry, i forget to upload the full text data. I will upload to google drive soon. I fix some errors of my code and you can download the new code.

[lyhuohuo]

Upload is completed.

[Jackson8787]

mvsa.rar in google drive seems have no change since 3/29,and I have the same problem after I use the mvsa.rar

[lyhuohuo]

The text data is named data_new.rar, you can check in google drive.

[Jackson8787]

I can see IJCAI2019_data.tar.gz , pytorch_model.bin in https://drive.google.com/drive/folders/1rm0FtHOTMUfZfRjWIE9Ukn_1D5MDXQy3?usp=sharing and mvsa.rar in https://drive.google.com/file/d/14Dum0b4gX94OyEkbrz8q72-MzXGVH8OE/view?usp=share_link. I'm not sure if you set data_new.rar as public ?

[lyhuohuo]

I have updated the true url of google grive and you can download from the new url.

[Jackson8787]

By the way, I prepared txt files in data_new when I run MVSA_pretrain.shbefore ,so I think this error might due to other reason.

[lyhuohuo]

You can compare your files with mine, i can run my code without any error. I update my code yesterday, you may download the new code.