I have read the paper and want to if I can reproduce the results written on the paper. However, the accuracy that I got was below what the paper says.
The precision@1 that I got by running python3.6 -m webaskb_run.py splitqa
SimpQA: 20.48877522023302
SplitQA 27.62148337595908
SplitQAOracle 32.96391020176186
I suspected the data I used is an old version, so I ran python3.6 -m webaskb_run.py run_ptrnet
to try producing the intermediate documents. However, I got another error.
IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number
Is the code I downloaded not the latest version? I want to confirm this point before I spend unpredictable amount of time debugging.