project-nomad/admin/app/services
chriscrosstalk 5afb4ba7d0
fix(benchmark): warm the AI model before timed runs for reproducible scores (#1140)
The AI benchmark evicted resident models (forcing the benchmark model cold)
and then went straight into the timed median-of-N loop with no warm-up. On a
cold box the model-load + GPU spin-up cost landed inside the timed runs
(observed: 173s TTFT / 5.83 tok/s vs a ~80 tok/s warm steady-state), and
consecutive runs weren't isolated (a prior run left the model warm). Because
the AI channel is uncapped and ~30% of the composite, the same machine could
post a ~2x-different NOMAD Score depending on warm/cold state (888 vs 1958
observed back-to-back).

Add one discarded warm-up inference after eviction and before the timed loop
so every timed run measures warm, steady-state throughput. Cold and warm
invocations now converge on the same score. Best-effort: a warm-up hiccup
never fails the run.

Closes #1139
2026-07-23 20:07:16 -07:00
..
app_auto_update_service.ts feat(supply-depot): opt-in automatic updates for installed apps 2026-06-23 04:46:52 +00:00
auto_update_service.ts feat(supply-depot): opt-in automatic updates for installed apps 2026-06-23 04:46:52 +00:00
benchmark_service.ts fix(benchmark): warm the AI model before timed runs for reproducible scores (#1140) 2026-07-23 20:07:16 -07:00
benchmark_telemetry.ts feat(benchmark): end-of-run score reveal + NVIDIA GPU-util overlay (#1087) 2026-07-20 12:07:45 -07:00
chat_service.ts fix(chat): prefer selected model for suggestions, fall back to smallest 2026-06-23 04:47:00 +00:00
collection_manifest_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
collection_update_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
condition_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
container_registry_service.ts fix(updates): resolve relative registry pagination URL so tag listing doesn't crash (#945) 2026-06-23 04:46:59 +00:00
content_auto_update_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
countries_service.ts feat(Maps): regional map downloads via go-pmtiles extract (#780) 2026-05-20 10:16:00 -07:00
creator_pack_service.ts feat(creator-packs): gated per-creator video packs, offline via Kiwix (#1106) 2026-07-20 16:21:18 -07:00
custom_app_guard.ts feat: supply depot 2026-06-23 04:46:45 +00:00
docker_service.ts fix(amd): coerce gfx1103 (780M) to HSA_OVERRIDE 11.0.0 so it stays on GPU (#1134) 2026-07-22 11:07:42 -07:00
docs_service.ts docs: update in-app docs for v1.33 Supply Depot + auto-updates 2026-06-23 04:47:04 +00:00
download_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
drug_reference_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00
kiwix_catalog_service.ts feat(content): opt-in automatic updates for installed ZIM & map content 2026-06-23 04:47:01 +00:00
kiwix_library_service.ts fix(kiwix): self-heal a missing or corrupt library XML on startup 2026-06-23 04:46:56 +00:00
map_service.ts fix(maps): warn when world basemap missing instead of silent grey map (#1104) 2026-07-20 18:25:39 -07:00
nomad_md_service.ts feat(AI): nomad.md for custom instructions (#1127) 2026-07-21 11:15:58 -07:00
ollama_service.ts feat(AI): per-model thinking toggle with global default (off) (#1079) 2026-07-20 16:33:09 -07:00
queue_service.ts fix(queue): share one ioredis connection across BullMQ queues and workers (#1009) 2026-06-23 04:47:08 +00:00
rag_service.ts fix(rag): stop re-creating payload indexes on every embedded document (#1135) 2026-07-22 12:54:30 -07:00
system_service.ts feat(debug-info): add storage, docker, GPU health, and auto-update diagnostics (#1102) 2026-07-19 22:25:53 -07:00
system_update_service.ts feat(system): add opt-in automatic updates for the core NOMAD app 2026-06-23 04:46:51 +00:00
zim_extraction_service.ts chore(KB): filter non-content sections + render tables in ZIM extraction (#1044) 2026-07-18 09:03:53 -07:00
zim_service.ts Add offline FDA drug reference (labels, interaction view, conditions, remedies) (#1040) 2026-07-21 11:06:18 -07:00