Repository navigation

perf(eval): cut ranking_eval per-query overhead (~1% on trained models) - #731

Merged

qtuantruong merged 3 commits into

PreferredAI:masterfrom

dexhunter:perf/next-item-eval-ranking-loop

Nov 1, 2026

Merged

perf(eval): cut ranking_eval per-query overhead (~1% on trained models)#731
qtuantruong merged 3 commits into
PreferredAI:masterfrom
dexhunter:perf/next-item-eval-ranking-loop

Commits on Oct 31, 2026
- perf(eval): 22% faster NextItemEvaluation ranking loop (217->170 ms)
  
  dexhunter
  and
  aiden-weco
  committed
  6455c2c
- test(eval): cover both ground-truth paths in next-item ranking_eval
  
  dexhunter
  and
  aiden-weco
  committed
  c526ffb
- test(eval): check exact metric values in next-item ranking_eval test
  
  dexhunter
  and
  aiden-weco
  committed
  5b5ae2f