James Thorne

No - I never needed to go through and retrain the entire system. For me, i got best value out of just putting it into a docker image and calling...

The retrieval step isn't the fastest: It's using Facebook's DrQA implementation that isn't really 'production ready'. I had to rewrite part of it so that it is data-parallel. More CPUs...

Hi. Did this PR ever go into master? I'm having a similar issue and it looks like I might have to fork and run an ad-hoc local build with @Linus12's...

The jack folder sits as a sibling to this folder. I've made a working Dockerfile if you'd like to use it: ``` FROM continuumio/miniconda3 ENTRYPOINT ["/bin/bash"] ENV NVIDIA_VISIBLE_DEVICES all ENV...

Hi, the intermediate output from the maskers (with IR selected evidence) are added to the Google drive folder. With IR evidence, they don't actually need the FEVER database, but the...

if `pipeline_evidence` is set (a list of 2-tuples (page name, + text)), the `evidence` field isn't used by the dataset loader

Yes, i meant pipeline_text. I think any extra values are just passed through to the `metadata` field and are ignored by the model

There's a few caveats to this. For the distant supervision objective, it's assume that the model doesn't have access to the reference correction, instead, it's trying to recover the input...

There's a supervised version as well which doesn't use any masking (see finetune_supervised.sh and finetune_supervised_pipeline.sh) if you want to mix supervision and masks, you could either train a supervised model...

I merged your PR. Can you make a wrapper to this function too? https://msdn.microsoft.com/en-gb/library/7fy67z6d(v=vs.110).aspx