vh/esh-pfi-infrastructure
1
0
Fork 0
Code Issues 7 Pull Requests Actions Packages Projects Releases Wiki Activity

feat(h2o-eval): round 2 eval seat on GPU 3 with floor-on/floor-off fronts for the judge investigation

Browse Source 
Prime (via albok-dev): understand why h2o failed and re-fit the judge. One
vLLM on GPU 3 behind two fronts: :8034 with the shim's yes/no floor at 0.801
exactly as the cutover ran (same judge stamp), :8035 with SHIM_NOUL_FLOOR=0 and
a distinct stamp (h2o-lightning-4b+noul-floor-0). Both refuse image URLs; the
vLLM --allowed-media-domains backstop is now verified (400, no fetch).
This commit is contained in:
vh
2026-10-12 14:29:15 -07:00
parent 501bda078d
commit 2b40805b84
4 changed files with 58 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
stacks/h2o-eval/README.md
+18 -2
View File
@@ -1,7 +1,23 @@
# h2o-eval (TEMPORARY) — fv-ml1 GPU 3, :8034
> **DOWN since 2026-10-12 1319 PT** (`docker compose down`), once the GPU 1 cutover's post-checks passed. The host keeps
> `/opt/docker/compose/h2o-eval` and the weights, so it can come back on GPU 3 for a transient test with `docker compose up -d`.
> **UP again since 2026-10-12 1428 PT, round 2** (Prime 1420, via albok-dev): "I'd like to understand why h2o failed, so
> stand it up on gpu3 and refit it and let's get an understanding." albok-dev runs the investigation and the judge
> re-fit (thread `01M4S026GWS354WZGXZ5E34F31`). **No load tests or other traffic while their runs are in.** Round 1
> (1300–1319) was the pre-cutover validation. Take it down with `docker compose down` when albok-dev is done.
>
> | port | front | floor | judge stamp (`/health` model) |
> |---|---|---|---|
> | `:8034` | `h2o-eval-front` | ON, 0.801, as the live cutover ran | `h2o-lightning-4b@b43f759b…`, identical to the cutover's |
> | `:8035` | `h2o-eval-front-nofloor` | OFF (`SHIM_NOUL_FLOOR=0`) | `h2o-lightning-4b+noul-floor-0@b43f759b…` |
>
> Both fronts sit over ONE vLLM (14,420 MiB, KV 135,070 tokens, util-sized because GPU 3 is otherwise empty). Each
> `/health` also reports `backend.noul_floor {value, source}`.
>
> Checked at 1428:
> - `/decide/shared` 200 on both, with the cutover's smoke values reproduced (adult 0.052, personal 0.876).
> - An https image URL is refused on both: 422 "not accepted here".
> - **The vLLM backstop answered for the first time.** An https image URL posted straight to vLLM got 400 "The URL must
> be from one of the allowed domains: ['media.invalid']", with no fetch.
h2o-lightning-4b @ b43f759b behind `services/h2o-decision-front` (copied into `conf/front/`), for albok-dev's
re-validation of Nemi's sensitivity judge before the GPU 1 cutover (Prime 2026-10-12 via brokkr). **Remove right after
stacks/h2o-eval/compose.yaml
+36 -4
View File
@@ -1,10 +1,17 @@
# TEMPORARY h2o-lightning-4b decision seat on fv-ml1 GPU 3 (Prime 2026-10-12 via brokkr: "stand up the h2o copy on
# gpu 3 if you need to test, it should replace the live seat as soon as possible and keep gpu3 only for transient
# tasks"). It exists so albok-dev can re-validate Nemi's sensitivity judge on h2o before the GPU 1 cutover; REMOVE it
# right after the cutover. INC (stacks/intern-decision) keeps serving production on GPU 1 meanwhile.
# TRANSIENT h2o-lightning-4b eval seat on fv-ml1 GPU 3 (GPU 3 is for transient work only).
# Round 1, 2026-10-12 1300-1319: albok-dev validated Nemi's judge on h2o before the GPU 1 cutover (Prime via brokkr).
# Round 2, from 2026-10-12 ~1430 (Prime 1420, via albok-dev): "I'd like to understand why h2o failed, so stand it up
# on gpu3 and refit it and let's get an understanding." albok-dev investigates the 154 texts h2o alone called adult
# and re-fits the judge. Two fronts over ONE vLLM:
# :8034 front the shim's yes/no floor ON (0.801), exactly as the live cutover ran; /health model stamp
# h2o-lightning-4b@b43f759b..., identical to the cutover's
# :8035 front-nofloor SHIM_NOUL_FLOOR=0 (floor OFF, brokkr's lever); stamp h2o-lightning-4b+noul-floor-0@b43f759b...
# Each /health also reports backend.noul_floor {value, source}. NO load tests or other traffic while albok-dev's
# runs are in (they name start/end on thread 01M4S026GWS354WZGXZ5E34F31). INC keeps :8033 on GPU 1, untouched.
# Serving exactly as benched (docs/pfi/decision-sweep-followon-bench-2026-10-12.md, Amendment 1): stock vLLM 0.30.0
# + the vendor's unmodified shim + the admission front (services/h2o-decision-front, copied into conf/front).
# Weights: /tank/aimodels/h2o-lightning-4b-b43f759b (hashes match HF at b43f759b; see the stack README).
# Inline images only (Prime 2026-10-12): LIGHTNING_IMAGE_URLS=off on both fronts + vLLM --allowed-media-domains.
name: h2o-eval
services:
vllm:
@@ -54,5 +61,30 @@ services:
- /tank/aimodels/h2o-lightning-4b-b43f759b:/model:ro
- /opt/docker/conf/h2o-eval/front:/front:ro
command: ["bash", "/front/run.sh"]
front-nofloor:
image: python:3.12-slim
container_name: h2o-eval-front-nofloor
restart: unless-stopped
network_mode: host
depends_on: [vllm]
env_file: .env
environment:
SHIM_VLLM: http://127.0.0.1:18199
SHIM_PORT: "18192"
SHIM_NOUL_FLOOR: "0" # the floor OFF (the shim: 0 turns it off)
LIGHTNING_IMAGE_URLS: "off" # Prime 2026-10-12: inline images only, the seat never fetches an image URL
FRONT_SHIM: http://127.0.0.1:18192
FRONT_VLLM: http://127.0.0.1:18199
FRONT_BIND: 0.0.0.0
FRONT_PORT: "8035"
FRONT_CAP: "32000"
FRONT_SHIM_FILE: /model/h2o_lightning_shim.py
FRONT_SERVE_CONFIG: /model/serve_config.json
FRONT_MODEL_NAME: h2o-lightning-4b+noul-floor-0 # a distinct judge stamp
FRONT_MODEL_REVISION: b43f759b6385d21182f87c3829931ec3d4dfa98a
volumes:
- /tank/aimodels/h2o-lightning-4b-b43f759b:/model:ro
- /opt/docker/conf/h2o-eval/front:/front:ro
command: ["bash", "/front/run.sh"]
volumes:
h2o-eval-vllm-cache:
stacks/intern-decision/README.md
+3 -3
View File
@@ -91,9 +91,9 @@ The shim switch was verified live at 1402, on the h2o seat with the front recrea
- inline red still answered red at 0.984;
- before the change, inline red/blue/green answered at 0.98+ and no-image came out near-uniform.
The vLLM backstop has **not** been exercised: the seat was rolled back before vLLM was recreated. Verify it on the next
switch, by posting an https image URL straight to vLLM's loopback port and expecting a refusal that names the allowed
domains. INC needs neither setting, because its served checkpoint has no vision tower.
The vLLM backstop was first exercised on the GPU 3 eval seat at 1428 (same image and flag): an https image URL
posted straight to vLLM got 400 "The URL must be from one of the allowed domains: ['media.invalid']", with no fetch.
Repeat that check on this seat's loopback port (`:18299`) on the next switch. INC needs neither setting, because its served checkpoint has no vision tower.
### Switching between the two