cripac-dig / hgls Goto Github PK

Thank your code for this paper. But there are some bugs/mistakes existing in the code.
Obviously, in the running of the normal code, the overheads of running datasets ICEWS18, ICEWS05-15, and GDELT are too huge to run. So you use two different ways of data loading for the above three datasets and dataset ICESW14s (It also refers to the “In order to speed up training and testing, for ICEWS18, ICEWS05-15, and GDELT datasets, data in the required format can be constructed in advance before training and testing: python save_data.py --data=DATA_NAME” in your readme). That also can be seen in your code (main,py)

For dataset icews14s, you use the “co.collate_rel” function while for other datasets you use the “collate_new” function. The point is that there exists problems in this part, resulting in the failure of running code. So the followers can not run your code to reproduce your work.

Also, there exist other bugs in the code which hinder the reproductivity of this work.
Such as this part in main.py (which can not correspond to the naming rule in save_data.py )

Thank your code for this paper, but I failed to reproduce .Can someone tell me where the results in the paper came from? I ran 500 epochs on ICEWS14s and the results are shown in the following figure, but did not achieve the expected results. I slightly modified the code so that ICEWS14s can also be loaded from folder

When will it be available?🤭