This major version update consolidates the definition of the AGC v2 task and the accompanying reference implementation, which both have been stable for a while, with a corresponding tag. Compared to v1 of the task, the main addition is ML inference (a setup for training is provided as well). A few other minor things changed in the task definition, in particular cuts as described at https://agc.readthedocs.io/en/latest/versionsdescription.html.
The implementation of this AGC v2 task additionally adopts correctionlib for systematic uncertainty handling, with corresponding JSON file provided in this repository.
After the release of the AGC v1 task via tag 1.0.0, subsequent developments in the v1 series happened on the agc-v1 branch so the best comparison to previous tags is the comparison of 2.0.0 to 1.0.0 instead of the more recent 1.4.0.
Full list of changes:
- chore: Use conda-forge servicex for all platforms (#230)
- docs: update reference section and list proceedings (#229)
- feat: Add pixi project configuration (#227)
- CMS example with ServiceX 3 client (#225)
- docs: link additional AGC-related projects (#226)
- feat: improved axis labels for HEPData figures (#223)
- feat: create tables for HEPData (#215)
- feat: add a client for Purdue Analysis Facility (#219)
- fixed info for triton (#218)
- feat: update metadata for cores per worker (#216)
- refactor: save available histograms and avoid hardcoding names (#214)
- feat: improve handling of optional dependencies (#213)
- fix: changed duplication of model even (#209)
- do not delete models by default, download mc (#208)
- Add new client setup for CMS AF dev using Dask Gateway (#207)
- fix: load models only if not already loaded (#203)
- feat: add training for cms open data ttbar (#205)
- fix: avoid column overtouching in ML input feature calculation (#204)
- feat: track inference settings (#202)
- feat: extend metrics to support more file sources (#200)
- feat: save histograms without rebinning and rebin via cabinetry (#197)
- fix: make ServiceX dataset grouping compatible with root protocol (#167)
- fix: pick up correct single top sample in statistical model (#194)
- feat: updating client code for obtaining Dask client on EAF (#192)
- feat: move xcache prefix setting to config (#191)
- feat: add option for URI prefixes (#189)
- docs: clarify ServiceX setup instructions (#188)
- feat: add support input for input files on EOS (#187)
- docs: add columnflow implementation to readthedocs (#184)
- docs: Add columnflow implementation example of AGC (#183)
- fix: b-tagging threshold comparison (#182)
- docs: add v1.2 description and fix cut descriptions (#180)
- feat: do not require pseudodata hists in validate_histograms.py (#177)
- feat: remove unnecessary newline printing in validate_histograms.py (#176)
- docs: updated version documentation (#173)
- feat: updated validation script to check partial bin sums (#171)
- feat: dependency management with cloudpickle (#158)
- docs: fix binning for analysis task description (#169)
- feat: provide reference files for various numbers of files processed (#157)
- (enrico/main) feat: improve integration of ml task (#153)
- ci: add ruff linter check (#154)
- fix: ruff linter errors on plotEvents.py (#155)
- fix: ruff linter errors (#148)
- feat: add utility tool to validate histogram contents against a reference (#149)
- feat: teach construct_fileset to cache remote files in local directory (#145)
- docs: improve task description (#138)
- fix: changed wjets scale variation percentages (#152)
- feat: updating init code for obtaining Dask client on EAF (#150)
- fix: imports, remove unused variable (#146)
- docs: update links for RDF implementation (#147)
- added facility instructions (#135)
- fix: do not import tritonclient.grpc unless needed (#139)
- test: add example of streaming ServiceX output as ROOT file data (#137)
- feat: update object labels and add readme for ttbar analysis (#132)
- docs: add notebooks used for AGC-2023 workshop (#133)
- feat: add ML inference option to ttbar analysis pipeline (#122)
- feat: adopt correctionlib for applying systematic variations (#119)
Full Changelog: v1.0.0...v2.0.0