yifan-gao / conversational-qg Goto Github PK

Hello, @evan-gao, I am very interesting about your work CQG, but I notice that your code in data is preprocessed, so I don't know how to split the original CoQA data. Could you explain more detail about the data format or give the original code about that? THX~

zengyj@id-d3:/conversational-QG/conversational-QG/scripts$ bash generate.sh
[2019-09-27 11:55:47,954 INFO] 8360 Example before filter
[2019-09-27 11:55:47,957 INFO] 8360 Example after filter
0%| | 0/4180 [00:00<?, ?it/s]/usr/local/anaconda3/lib/python3.6/site-packages/torchtext/data/field.py:359: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
var = torch.tensor(arr, dtype=self.dtype, device=device)
/home/zengyj/conversational-QG/conversational-QG/code/onmt/translate/translator.py:520: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
return torch.tensor(a, requires_grad=False)
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4180/4180 [11:56<00:00, 5.83it/s]
Traceback (most recent call last):
File "/home/zengyj/conversational-QG/conversational-QG/code/tools/eval/eval.py", line 3, in
from bleu.bleu import Bleu
File "/home/zengyj/conversational-QG/conversational-QG/code/tools/eval/bleu/bleu.py", line 14, in
import bleu_scorer
ModuleNotFoundError: No module named 'bleu_scorer'
zengyj@id-d3:/conversational-QG/conversational-QG/scripts$

----------------上面是报错的信息

import os,sys
sys.path.append("..")
sys.path.append(os.getcwd())

import bleu_scorer
from bleu.bleu_scorer import BleuScorer

我在 bleu.py里面已经了如上的代码 还是不能导入同级目录的代码呢

这个明明在一级目录呢

是什么问题，求解答

Hi,

Could you please share the pretrained models for this work?

Cheers!

Hi, @evan-gao . Thanks for the interesting work. When I trained a model use your code I got a result B1=36.86, B2=22.32, B3=15.77, R-L=46.37, I found that the result is below the pretrain model you given or the paper's report. So could you give some details about the train procedure？

Hi,

Thanks @Yifan-Gao sharing the code. It is a great job.

But, the question is how to generate the question from a custom text of the paragraph. Actually, you have directly provided the processed train and test data.

So, I am not able to use my custom test data in the model that needs to be processed.

Kindly request you to can you please provide function so that I can feed processed custom test data to model?

Thanks a lot,
Bhavika

Does the model with CorefGenerator in your code denote the CFNet model in your paper? And the model with CopyGenerator denote MSNet? I tried to disable the coref_vocab and coref_attn commands to run the base model, but a ZeroDivisionError occurs.