Expose relevance_score for ranker models through `pipeline.eval()`

[shalinshah1993]

Is your feature request related to a problem? Please describe. There has been prior discussions about exposing model scores for re-ranker and i believe it was added. However, that only works with pipeline.run() API since it returns documents with score metadata. In practice, often, we run large scale evaluation and in that cases, it has a lot of wrapping around pipeline.run() (see here)

Describe the solution you'd like A very nice solution would be -> in the EvalutionResults df or csv file, there is additional column stored called score which comes from document.score

Describe alternatives you've considered Hacky way to do it right now, is to take Retriever.csv in eval run add a column with score (re-run ranker inference) then sort based on score and add it as df to EvaluationResults so we save Ranker.csv

This is using Haystack 1.x

[anakin87]

Hello. Haystack 1.x is in maintenance mode and the idea is to fix bugs, but not to introduce new features. Have you considered migrating to 2.x? https://docs.haystack.deepset.ai/docs/migration

[shalinshah1993]

is the above possible in 2.x?