Closed chrislouis0106 closed 3 years ago
Dear @THUCSTHanxu13 I have some questions to consult you. 1 As shown in this picture, , from the paper "Multi-Hop Knowledge Graph Reasoning with Reward Shaping", your test result in fb15k237 and wn18rr dataset is very different from the data in the paper. I'm sure you know the paper, so I want to consult you what is the reason. 2 About the custom dataset, I found the issue, but to my disappointment, I don't found that any notes or hints to complete the train and test process with a new custom dataset. So, can you give me some advice, thank you anyway.
Dear @THUCSTHanxu13 I have some questions to consult you. 1 As shown in this picture, , from the paper "Multi-Hop Knowledge Graph Reasoning with Reward Shaping", your test result in fb15k237 and wn18rr dataset is very different from the data in the paper. I'm sure you know the paper, so I want to consult you what is the reason. 2 About the custom dataset, I found the issue, but to my disappointment, I don't found that any notes or hints to complete the train and test process with a new custom dataset. So, can you give me some advice, thank you anyway. 3 About the print result, just like this. I don't sure where 0.418133 is from? And I try to modify the test.h file, like this, , and , but, not any changes. So, could you explain the print result and how to get more accurate parameters such as MRR and hit@1?
Thank you a lot.