search

SiyuanWangw / InductiveLR

8 stars 0 forks source link

The format of the datasets in KGReasoning does not seem to match the code in DataProcessing #1

Closed Vin1115 closed 1 year ago

Vin1115 commented 1 year ago

The format of the datasets in KGReasoning does not seem to match the code in DataProcessing, despite the claim in the README that the transductive datasets are available from KGReasoning. For example, KGReasoning's datasets contain only id2ent.pkl, not ind2ent.pkl from DataProcessing. It also does not contain traintriples{task}.pkl as indicated in the DataProcessing's code. with open(f'{data_path}/ind2rel.pkl', 'rb') as rel_handle: with open(f'{data_path}/train_triples_{task}.pkl', 'rb') as handle:

Could you please answer me if I misunderstood something or provide the dataset in the correct format?

yanween commented 11 months ago

The format of the datasets in KGReasoning does not seem to match the code in DataProcessing, despite the claim in the README that the transductive datasets are available from KGReasoning. For example, KGReasoning's datasets contain only id2ent.pkl, not ind2ent.pkl from DataProcessing. It also does not contain traintriples{task}.pkl as indicated in the DataProcessing's code. with open(f'{data_path}/ind2rel.pkl', 'rb') as rel_handle: with open(f'{data_path}/train_triples_{task}.pkl', 'rb') as handle:

Could you please answer me if I misunderstood something or provide the dataset in the correct format?

Hello, I have the same problem, how did you solve it?

Vin1115 commented 11 months ago

The format of the datasets in KGReasoning does not seem to match the code in DataProcessing, despite the claim in the README that the transductive datasets are available from KGReasoning. For example, KGReasoning's datasets contain only id2ent.pkl, not ind2ent.pkl from DataProcessing. It also does not contain traintriples{task}.pkl as indicated in the DataProcessing's code. with open(f'{data_path}/ind2rel.pkl', 'rb') as rel_handle: with open(f'{data_path}/train_triples_{task}.pkl', 'rb') as handle: Could you please answer me if I misunderstood something or provide the dataset in the correct format?

Hello, I have the same problem, how did you solve it?

The datasets used by this method are not the betaE datasets, but the datasets provided in the original query2box code. Please refer to https://github.com/hyren/query2box.

yanween commented 11 months ago

The format of the datasets in KGReasoning does not seem to match the code in DataProcessing, despite the claim in the README that the transductive datasets are available from KGReasoning. For example, KGReasoning's datasets contain only id2ent.pkl, not ind2ent.pkl from DataProcessing. It also does not contain traintriples{task}.pkl as indicated in the DataProcessing's code. with open(f'{data_path}/ind2rel.pkl', 'rb') as rel_handle: with open(f'{data_path}/train_triples_{task}.pkl', 'rb') as handle: Could you please answer me if I misunderstood something or provide the dataset in the correct format?

Hello, I have the same problem, how did you solve it?

The datasets used by this method are not the betaE datasets, but the datasets provided in the original query2box code. Please refer to https://github.com/hyren/query2box.

Thanks !