The high level API was introduced in https://github.com/umami-hep/puma/pull/128 and there were still some open points which need to be addressed but can be done in another pull request. This issue will keep track of these changes:
[ ] extract_tagger_scores function in Tagger class: rather than a single source which cannot be type hinted and a source_type, I would suggest to have df : DataFrame = None and path: Path = None arguments and the logic can be based on which of the two is defined. (see comment)
[x] making self.scores in Tagger class a pandas data frame (see comment)
[ ] extract_tagger_scores function should dynamically set self.is_light etc based on the passed/loaded jets (see comment)
Also we should extend the support beyond b/c signal jets, e.g. to the Xbb use case where we have VHbb/cc signal jets and then VHcc/bb, top, qcd backgrounds
The high level API was introduced in https://github.com/umami-hep/puma/pull/128 and there were still some open points which need to be addressed but can be done in another pull request. This issue will keep track of these changes:
Tagger
class can be interfaced with https://jsonargparse.readthedocs.io/extract_tagger_scores
function inTagger
class: rather than a single source which cannot be type hinted and a source_type, I would suggest to have df : DataFrame = None and path: Path = None arguments and the logic can be based on which of the two is defined. (see comment)self.scores
inTagger
class a pandas data frame (see comment)extract_tagger_scores
function should dynamically set self.is_light etc based on the passed/loaded jets (see comment)