decstionback / aaai_2020_commonsenseqa Goto Github PK

The last line of this file is not in a complete json format. It seems something were dropped by mistake.

When I run run.py, it shows 'json.decoder.JSONDecodeError', like the picture:

Does someone have any methods to duel with it？

excuse me!when I run code,it have a error:

File "./AAAI_2020_CommonsenseQA/pytorch_transformers/tokenization_xlnet.py", line 85, in init
self.sp_model = spm.SentencePieceProcessor()
UnboundLocalError: local variable 'spm' referenced before assignment

Could you help me?Thank you very much!

run.py中convert_examples_to_features方法内部存在下面代码：
node_size=len(temp)-1 matrix=np.zeros((150,150)) for i,val in enumerate(adj_matrix): for j,v in enumerate(adj_matrix[i]): if v==1 and i<50 and j<50 and i<node_size and j<node_size: matrix[i,j]=1
此处将adj_matrix中值为1的位置赋值给了matrix，从所使用的if语句看，这个地方的node_size大小为问题+选项三跳以内路径的数量，此处对于matrix赋值的时候为什么还要使用问题+选项的三跳路径数量来限制matrix，for循环中adj_matrix是和wiki有关的，node_size是和conceptnet有关的，用conceptnet事实的数量限制wiki事实的数量有些许不理解，望得到回复。

Thank you for uploading the code.
Could you please show how you did your data preparation to generate (train.jsonl.concept, train.jsonl.wikigraphnew, train.jsonl_NL)?
@DecstionBack

您好，对wiki数据进行预处理的阶段我有点看不懂，请问：
1.使用 Elastic 搜索引擎 在检索查询和所有句子之间进行排序 ，这个排序方法是什么呀？
2.对wiki数据进行预处理的代码在开源里好像没找到哇，数据直接就是预处理过了的？