Evaluate selecting test configs separately for each confidence level

Aperta
  • Qualcosa non va?
#7,386 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 2 giorni

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
5/5
Tempo stimato
Più di una settimana
Idoneità per principianti
25/100
Tipo di issue
Funzionalità
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
python
Ambito
machine-learning

Direzione di ricerca

Start by locating the HTTP service that returns config_groups, the config selection model evaluation, and the Firefox taskgraph strategies named bugbug-reduced-manifests-config-selection-*. Measure selection cost and regressions caught for each level against a selection using only groups at or above that level. The issue is resolved when the measurements show whether separate per-level selections are worthwhile and, if so, the service replies and taskgraph strategies use them.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

The HTTP service selects the configs of each confidence level (very_high, high, medium, low, very_low) after the ones of the levels above, reusing them, and returns a single config_groups for all levels. The Firefox taskgraph strategies that select configs (bugbug-reduced-manifests-config-selection-*) run the groups at or above their level on these configs.

This keeps the configs of a level independent from the groups below it, which don't run with it. However, a strategy at a lower level runs the groups of the levels above too, and their configs were selected without knowing which configs the lower groups would use: e.g. for the low strategy, a medium group can't reuse a config selected for a low group, so the low strategy may run more configs than a selection made for its groups alone.

We should measure how much this costs, e.g. in the config selection model's evaluation, comparing the cost and the regressions caught at each level with a selection made for the groups at or above it only.

If the difference is significant, the HTTP service could return a separate config selection for each level (a selection over all the groups at or above it), and the taskgraph strategies could read the one for their level. This would increase the size of the replies and the number of config selection model calls, and require changes in the taskgraph.

Lingua principale
Python
Stelle
578
Fork
366
Merge medio
2g 19h
PR unite (30g)
136

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di mozilla/bugbug

Tutte le issue di mozilla/bugbug

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.