[PWGDQ] extend the input features in MLResponse class#16228
[PWGDQ] extend the input features in MLResponse class#16228aferrero2707 wants to merge 4 commits into
Conversation
Use a single macro for retrieving features from tracks. Also removed alternative feature getter functions.
|
O2 linter results: ❌ 197 errors, |
|
Error while checking build/O2Physics/staging for 7a935eb at 2026-05-13 08:17: Full log here. |
|
Error while checking build/O2Physics/o2 for 7a935eb at 2026-05-13 08:17: Full log here. |
7a935eb to
bdb2ee0
Compare
The additional features are needed to match those currently used for training the ML models for the MFT-MCH matching
bdb2ee0 to
37b267d
Compare
|
Error while checking build/O2Physics/o2 for 37b267d at 2026-05-13 09:10: Full log here. |
|
Error while checking build/O2Physics/staging for 37b267d at 2026-05-13 09:17: Full log here. |
|
Error while checking build/O2Physics/o2 for b35c763 at 2026-05-13 09:47: Full log here. |
|
Error while checking build/O2Physics/staging for b35c763 at 2026-05-13 09:47: Full log here. |
|
You are supposed to test your changes before you open a PR. |
b35c763 to
aac67b9
Compare
|
Error while checking build/O2Physics/staging for aac67b9 at 2026-05-13 11:33: Full log here. |
aac67b9 to
a9954cf
Compare
|
Error while checking build/O2Physics/o2 for a9954cf at 2026-05-13 12:03: Full log here. |
The interface for the ML-based matching has changed and requires additional tables for the tracks and collision, which are not yet included in the tableMaker code. Since the ML-based matching is still experimental and not intended for physics analysis, the skimming code based on the ML output is disabled for the moment, waiting for a proper fix for the missing tables.
a9954cf to
ef97f86
Compare
|
@mcoquet642 @lucamicheletti93 the additional features included in the MLResponse class require some dynamic tables that are currently not included in the tableMaker tasks. Since the ML matching is highly experimental and not yet used in any analysis, I propose to disable the corresponding code for the time being (see last commit in this PR) and work on a proper fix in a separate PR. |
As I said, I would prefer to keep a dummy input feature method in the response class, and use it in tableMaker, instead of disabling the code |
The additional features are needed to match those currently used for training the ML models for the MFT-MCH matching.