nustm / vlp-mabsa Goto Github PK

您好，论文很有新意，尤其是关于图像的意见信息生成。其中DeepSentiBank，请问您有其他版本吗，现有的版本运行时有很多问题，或许获取aspect-opinion对的细节您能提供吗？谢谢！

请问transformers是必须要用3.4.0版本么？ 实验环境是linux系统么

I got the error message below when running sh 15_pretrain_full.sh.

OSError: Model name './E2E-MABSA' was not found in tokenizers model name list (facebook/bart-base, facebook/bart-large, facebook/bart-large-mnli, facebook/bart-large-cnn, facebook/bart-large-xsum, yjernite/bart_eli5). We assumed './E2E-MABSA' was a path, a model identifier, or url to a directory containing vocabulary files named ['vocab.json', 'merges.txt'] but couldn't find such vocabulary files at this path or url.

How can I fix it?

Nice work and nice repository ! But I still have some doubts about the repository~

1. Could you please provide some more instructions about how to process the original MVSA dataset using the tools you mentioned?
   For example, what steps you have taken using the twitter_nlp to perform NER and how did you use sentiwordnet to matching the opinion words, finally what results we got after above processing.
   Same thing with Faster-RCNN and ANPs extractor.
2. Could you please provide some sample data of the processed MVSA?
   It will be great if you could provide some example data into BaiduNetdisk, because I still have no idea about the exact data format with only MVSA_descriptions.txt provided, thus could not reproduce the pretraining part of your code.

Thanks a lot!

line 1106, in BartModel
@add_start_docstrings_to_callable(BART_INPUTS_DOCSTRING)
NameError: name 'add_start_docstrings_to_callable' is not defined

作者您好，我目前在研究相关的方向，对您这篇文章比较感兴趣，想请问您可以开源其他子任务的代码吗，貌似目前的代码只涉及一个下游任务，想再复现一下这篇论文

你好，请想问一下，我下载数据集后，直接运行15_pretrain_full.sh 和 17_pretain_full.sh 得到的结果低于论文结果大概一个百分点，请想问一下，我是还需要修改其他地方吗？

Hi,sorry for bothering.
I think it would be better if you put this project on Google Colab,and add link on the README.
This will make it more accessible to others and easier to reproduce the experiment.
Thanks.

hello,
I would like to ask if the current code supports the downstream MASC task, because there is a difference between its input and the trunk task input, because the MASC task is known by the entity

你好，关于图像特征的来源，我有几个问题，希望可以解答一下
在百度网盘的微调数据集里，IJCAI2019_data/region_box/15/train文件夹下，有_att和_box两个文件夹：
1、_box文件夹下xxxx.npz是（36，4）的矩阵，是xxxx.jpg图片中36个矩形特征区域的左上角和右下角的坐标吗？
2、_att文件夹下xxxx.npz是（36，2048）的矩阵，是xxxx.jpg图片中36个矩形特征区域所对应的特征吗？2048维的特征是怎么得到的？

I would like to know how to use Faster-RCNN to extract the region feature(only retain 36 regions with highest Confidence) as the input feature and the dimension of each region feature is 2048，Can you give me a small demo if possible?

你好，中文不好请见谅。
我想要在实验中将论文中提到的5个pre training model
分别关闭一次，然后测试得到的成绩
我在training.py中的第60行开始将pre training model 註解化，如下
mlm_infos={ """2/22 disable MLM key: value.to(device) for key, value in batch['MLM'].items() """ } if 'MLM' in batch else None,
可是结果并没有变化，请问是哪裡做错了？

There are many empty pictures in IJCAI2019_data\twitter2017_images. Could you upload them again

我运行sh 15_pretrain_full.sh，但是出现以下错误
7e-5 Traceback (most recent call last): File "MAESC_training.py", line 19, in <module> from src.model.MAESC_model import MultiModalBartModel_AESC File "/home/jovyan/multimodal_aspect_sentiment/VLP-MABSA/src/model/MAESC_model.py", line 2, in <module> from fastNLP.modules import Seq2SeqEncoder, Seq2SeqDecoder, State ImportError: cannot import name 'Seq2SeqEncoder' from 'fastNLP.modules' (/home/jovyan/environment/anaconda3/body/envs/vlpmabsa/lib/python3.7/site-packages/fastNLP/modules/__init__.py)
该怎么做？你们用的是什么版本的fastNLP？

您好，
我下载预训练数据后发现里面有一些 &amp , &lt 这样被转义后的token，这些您有做 unescape 么？

感谢

Hello, I have seen some contents of KM-BART in this article. Have you ever reproduced KM-BART? I met some problems in the process of reproducing, such as the failure to find SBU data.

hi, the model in Baidu Netdist means trained in pre_train tasks, but not fine_tine in twitter2015 and twitter2017 ?