 Alexander Korolev
|
6e3b31bccd
task: Render the sm121 health cluster report readably in plain format from the unified HealthReport (TASK_0003)
|
1 ماه پیش |
|
Alexander Korolev
|
453ee960c8
ENFORCE GUIDELINES: Build shared reusable localized renderer module with plain/md/html renderers over a pure-data report model — all three renderers resolve fluent messages via t! with any new keys added to both locales/en and locales/ru; scope wiring to health now, do not retrofit doctor/tmux status | decisions (explicit user choices — these OVERRIDE the spec doc wherever they conflict; follow them exactly): all three renderers (plain/md/html) resolve fluent messages via t! from a pure-data report model, with any new message keys added to both locales/en and locales/ru; health only — build the md/html renderer as a shared reusable module and wire --format into sm121 health now, do not retrofit doctor/tmux status in this series (TASK_0002)
|
1 ماه پیش |
|
Alexander Korolev
|
fbb37658dd
task: Build shared reusable localized renderer module with plain/md/html renderers over a pure-data report model — all three renderers resolve fluent messages via t! with any new keys added to both locales/en and locales/ru; scope wiring to health now, do not retrofit doctor/tmux status | decisions (explicit user choices — these OVERRIDE the spec doc wherever they conflict; follow them exactly): all three renderers (plain/md/html) resolve fluent messages via t! from a pure-data report model, with any new message keys added to both locales/en and locales/ru; health only — build the md/html renderer as a shared reusable module and wire --format into sm121 health now, do not retrofit doctor/tmux status in this series (TASK_0002)
|
1 ماه پیش |
|
Alexander Korolev
|
0f0177f302
task: Extract printing-free unified HealthReport collection pass in docker::health — cluster verdict, per-node net/containers, tmux sessions, running-recipes summary as one pure-data model — build one unified per-cluster report model (cluster verdict + per-node table with net/containers + tmux sessions + running-recipes summary) rendered by the shared module, and keep the collection pass standalone under docker::health reusing only cluster_state::verdict, with no web::probe coupling | decisions (explicit user choices — these OVERRIDE the spec doc wherever they conflict; follow them exactly): build one unified per-cluster report model, collection free of printing; keep the collection pass standalone under docker::health reusing only cluster_state::verdict — do not couple to web::probe (TASK_0001)
|
1 ماه پیش |
|
Alexander Korolev
|
8936fde0cd
queue rocenante research (b12x #295): zero-port, b12x pin bump + env toggle
|
1 ماه پیش |
|
Alexander Korolev
|
00099bcfff
rust frontend evaluated and deferred (600s registration timeout vs our boot times); fork stays selective
|
1 ماه پیش |
|
Alexander Korolev
|
bc8bb0d90a
r27: b12x loader pinned_wc -> managed (discord patch, 1350->2000 tk/s prefill claim), hard-gated
|
1 ماه پیش |
|
Alexander Korolev
|
17a91a9b19
cleanup: remove legacy r12/r18 build dirs (non-rebuildable) + blocked r27-dflash2 recipe
|
1 ماه پیش |
|
Alexander Korolev
|
c957beb710
eager A/B done (159 vs 158, rumor not confirmed); aiden 4x 163 = campaign record; night closed
|
1 ماه پیش |
|
Alexander Korolev
|
6f39548ba3
aiden 4x benched: 163/176 campaign record; r18 2x eager A/B recipe
|
1 ماه پیش |
|
Alexander Korolev
|
d806bd2c2f
aiden 4x: use the extended image (production-1.2 + instanttensor), restore instanttensor load
|
1 ماه پیش |
|
Alexander Korolev
|
9e55382f40
aiden 4x: drop --load-format instanttensor (not in production-1.2 image)
|
1 ماه پیش |
|
Alexander Korolev
|
bbe82a903e
aiden 4x un-deferred: production-1.2 image, zai FP8 weights now cached
|
1 ماه پیش |
|
Alexander Korolev
|
9c648a2ee4
state doc: night-3 close — final campaign table, DFlash2 blocker, queue
|
1 ماه پیش |
|
Alexander Korolev
|
5b47e80fd6
r27 dflash2 BLOCKED: C4 parent-page stride assert (tree-level, both r18+r27), root cause documented
|
1 ماه پیش |
|
Alexander Korolev
|
b3babad4a2
r27 DFlash2 lane: r18-dflash2 clone retargeted (method dflash, k=7, tree-picked backends)
|
1 ماه پیش |
|
Alexander Korolev
|
191c035248
r27 4x DCP4 benched: 155 (-6 vs r18), KV 7.7M @ 16GiB; r27 still trailing r18 quality
|
1 ماه پیش |
|
Alexander Korolev
|
1e1df7eb06
r27 4x DCP4 recipe (r18-4x-dcp4 clone retargeted at the r27 engine)
|
1 ماه پیش |
|
Alexander Korolev
|
9cfed14f41
vision-373 lane benched: 161/176 new 2x record; launcher entrypoint-image support (model='', NODE_RANK/MASTER_ADDR env)
|
1 ماه پیش |
|
Alexander Korolev
|
9cb56de106
state doc: night 2 results (aiden 159 first canonical, r27 156), next queue
|
1 ماه پیش |
|
Alexander Korolev
|
2d74fdaca9
aiden 2x benched (159, first canonical aiden), r27 lane 156 (-2 vs r18), reports logged
|
1 ماه پیش |
|
Alexander Korolev
|
8ad35cba78
night part 2: aiden 2x benched (159, first canonical), r27 lane 156 (-2 vs r18), reports saved
|
1 ماه پیش |
|
Alexander Korolev
|
e9928e3ac7
night list: prod-3.73-vision = DSV4F-Vision aiden lane (replaces blocked eugr row)
|
1 ماه پیش |
|
Alexander Korolev
|
03edc92027
night revisit: aiden lanes + prod-3.73-vision tag queued, statuses ledgered
|
1 ماه پیش |
|
Alexander Korolev
|
d736662719
r27: explicit TORCH_CUDA_ARCH_LIST=12.1a in vllm-builder (stage-env loss -> sm120-only kernels)
|
1 ماه پیش |
|
Alexander Korolev
|
afd95f742d
r27 lane: glm53-flash-r27-spark (best 2x config on the r27 engine)
|
1 ماه پیش |
|
Alexander Korolev
|
f599efe741
skill: jovian-images (clone-at-pin builds, rev adoption, GB10 deltas, pitfall ledger)
|
1 ماه پیش |
|
Alexander Korolev
|
35d8f090b2
r27 fetch-deps: recursive gitlink population (FlashMLA/qutlass/FlashKDA cutlass vendors; -d stub trap fixed)
|
1 ماه پیش |
|
Alexander Korolev
|
3e63d19947
state doc: image build v2 (clone-at-pin) contract
|
1 ماه پیش |
|
Alexander Korolev
|
9118a4648b
r27: WORKDIR in builder stages (stage split dropped inherited cwd)
|
1 ماه پیش |
