Apoorv Saxena

Hi, in case you are trying on a custom dataset I would recommend you use the code from KGQA/RoBERTa. There you don't need bn0 etc. (which is needed for code...

No, it is not needed. Please refer Table 5 in the paper for the ablation study that shows that performance decreases when doing relation matching for half KG.

@lihuiliullh did you re-train the KG embeddings or use the pretrained checkpoint?

Hi, thanks for your interest! > Could give me some general params to assessment the kg embedding model? Like 0.4 or more for hits_at_1/hits_at_10 is enough to learn something from...

Can you pls give the exact command that you used again? A screenshot of output would also be helpful.

> I got a same problem. I used the training command on the 3-hop MetaQA data, but the model early stopped when it reached 14 rounds with the accuarcy rate...

can you please elaborate how you got that file?

Can you elaborate more on the exact changes you made in code?

which model did you use for the results? eg. EmbedKGQA_complex, model1 etc. It would be helpful if you could give the exact line numbers you changed + the command you...

Can you please try with an older version of pytorch? This post suggests pytorch 1.4 https://stackoverflow.com/questions/66590981/transformer-error-importing-packages-importerror-cannot-import-name-save-st