Retrain a classifier head from verified identifications, and start the run from the job form by mohamedelabbas1996 · Pull Request #1494 · RolnickLab/antenna

feat(ml): record on an algorithm whether and how it can be retrained

mohamedelabbas1996

committed

2c47148

feat(ml): mirror whether a processing service can retrain an algorithm

mohamedelabbas1996

committed

2553720

feat(ml): filter algorithms by whether they can be retrained

mohamedelabbas1996

committed

4211495

feat(jobs): let a job type declare what it needs and how long it may go quiet

mohamedelabbas1996

committed

e858b46

feat(jobs): take a job type's params when a job is created

mohamedelabbas1996

committed

365fb61

fix(jobs): judge a stale job by its own type's deadline

mohamedelabbas1996

committed

d6a3cc3

Merge remote-tracking branch 'upstream/feat/occurrence-sets' into feat/retrain-training-set

mohamedelabbas1996

committed

a6e02c4

feat(main): add a project permission for retraining a classifier

mohamedelabbas1996

committed

14d2773

feat(main): let a project name the taxa list its classifier covers

mohamedelabbas1996

committed

3fef6c9

feat(ml): build a classifier head training set from verified occurrences

mohamedelabbas1996

committed

d42e6b6

feat(ml): summarise what a retrain would learn from, for the job form

mohamedelabbas1996

committed

ae5da96

feat(jobs): retrain a classifier head and take the result back

mohamedelabbas1996

committed

0640da9

fix(jobs): point the training job at the merged training package

mohamedelabbas1996

committed

4f253bf

feat(jobs): make the training occurrence set optional

mohamedelabbas1996

committed

6793f23

fix(jobs): retrain the newest version of the algorithm named

mohamedelabbas1996

committed

476cf88

feat(jobs): follow a training run's epochs while it fits

mohamedelabbas1996

committed

73a49db

feat(ml): write down the contract between a training job and a service

mohamedelabbas1996

committed

b01bc18

fix(jobs): take a training result one way only, so there is one shape to read

mohamedelabbas1996

committed

7faf0c1

feat(jobs): say that a training job dispatches to an external service

mohamedelabbas1996

committed

2f55e09

feat(ui): start a classifier retrain from the job form

mohamedelabbas1996

committed

a390c8d

fix(ui): do not send training settings a run could not use

mohamedelabbas1996

committed

6026f2f

fix(ui): name a training job in the jobs list

mohamedelabbas1996

committed

b8e9fe9

Navigation Menu

Repository navigation

Retrain a classifier head from verified identifications, and start the run from the job form - #1494

Retrain a classifier head from verified identifications, and start the run from the job form#1494
mohamedelabbas1996 wants to merge 27 commits into
feat/detection-embeddings-taskfrom
feat/retrain-classifier-head

Commits on Oct 8, 2026

feat(occurrences): keep a fixed list of occurrences as a set

feat(ui): filter occurrences by set

feat(ui): save the selected occurrences as a set

fix(ui): show the occurrence set filter in the filter panel

fix(ui): put creating a set in the toolbar rather than the selection bar

Commits on Oct 9, 2026

feat(ml): record on an algorithm whether and how it can be retrained

feat(ml): mirror whether a processing service can retrain an algorithm

feat(ml): filter algorithms by whether they can be retrained

feat(jobs): let a job type declare what it needs and how long it may go quiet

feat(jobs): take a job type's params when a job is created

fix(jobs): judge a stale job by its own type's deadline

Merge remote-tracking branch 'upstream/feat/occurrence-sets' into feat/retrain-training-set

feat(main): add a project permission for retraining a classifier

feat(main): let a project name the taxa list its classifier covers

feat(ml): build a classifier head training set from verified occurrences

feat(ml): summarise what a retrain would learn from, for the job form

feat(jobs): retrain a classifier head and take the result back

fix(jobs): point the training job at the merged training package

feat(jobs): make the training occurrence set optional

fix(jobs): retrain the newest version of the algorithm named

feat(jobs): follow a training run's epochs while it fits

feat(ml): write down the contract between a training job and a service

fix(jobs): take a training result one way only, so there is one shape to read

feat(jobs): say that a training job dispatches to an external service

feat(ui): start a classifier retrain from the job form

fix(ui): do not send training settings a run could not use

fix(ui): name a training job in the jobs list