Dataset split in YAGO

[JuliaGast]

needs to be modified for the automated downloaded dataset

dataset.py

        if ("yago" in self.name):

            _train_mask, _val_mask, _test_mask = self.generate_splits(full_data, val_ratio=0.1, test_ratio=0.10) 

[JuliaGast]

The problem is that I changed the dataset split for yago in dataset.py -> when the dataset is downloaded, the train/valid/test set is downloaded/split with the new split however, the negative samples are still downloaded with the old split; one needs to manually delete the negative sample pkl files and run tkgl_yago_ns_gen.py

Can we automatically provide the correct new split for these negative samples?

[JuliaGast]

Note: this might no longer be needed as soon as we implement the new method for time-aware filtered mrr computation without negative samples.

[shenyangHuang]

issue resolved