 Alexander Korolev
|
b16bf62dee
bench: spark 2x @512K + 10 GiB KV (157/176, A/B vs re-cut 512K)
|
hai 1 mes |
|
Alexander Korolev
|
ed4bcc4369
2x re-cut: 512K context via max-num-seqs 4 + 10 GiB KV pin (bench 160/176)
|
hai 1 mes |
|
Alexander Korolev
|
ed49b2d94a
docker: image sync falls back 200G→10G with a warning (fabric ssh fix hint)
|
hai 1 mes |
|
Alexander Korolev
|
1b605f259b
docker: image sync inner hop over the fabric (ib_ip) when available
|
hai 1 mes |
|
Alexander Korolev
|
20a3c768c9
agents: re-cut measured results + node-to-node transfer / hf-cache / pkill gotchas
|
hai 1 mes |
|
Alexander Korolev
|
e9c009a7c3
bench: NVFP4-lab re-cut — 2x 164/176 (93), 4x 159/176 (90)
|
hai 1 mes |
|
Alexander Korolev
|
144077f8a4
recipes: nvfp4-lab 2x — 256K context (2.4 GiB KV budget at GMU 0.88)
|
hai 1 mes |
|
Alexander Korolev
|
4bbf75aa57
recipes: GLM-5.3-Flash NVFP4-lab re-cut lanes (2x + 4x, r35)
|
hai 1 mes |
|
Alexander Korolev
|
6a2f3c34b5
report: per-block copy buttons in rendered md (hover-only, full-text btn yields)
|
hai 1 mes |
|
Alexander Korolev
|
05ef2c1e43
web: fan spin hub-centered (fill-box origin wobbled the glyph)
|
hai 1 mes |
|
Alexander Korolev
|
0ebe775ceb
skill: web-topology cable mode (S-curve + bus/bridges)
|
hai 1 mes |
|
Alexander Korolev
|
498137d8b2
history: 09-18 end-state note (power off policy)
|
hai 1 mes |
|
Alexander Korolev
|
44b41a9520
recipes: spark 4x lane on r35 + mxfp8 kernel fallback (r38 pool-deadlock)
|
hai 1 mes |
|
Alexander Korolev
|
272842fcc9
web: topology row-1 S-curve + bus/bridges cable mode (legend toggle)
|
hai 1 mes |
|
Alexander Korolev
|
1db84cef5c
fix: pass base prefix in web report router call (compile break)
|
hai 1 mes |
|
Alexander Korolev
|
54f5c559e0
repo browser: absolute link hrefs + relative child URLs + clean crumbs
|
hai 1 mes |
|
Alexander Korolev
|
20bf94b024
web: window switcher + toml→repo navigation
|
hai 1 mes |
|
Alexander Korolev
|
f5eaa24a19
bench: r35 Spark 2x 162/176, 4x 161/176 (interim mode baselines)
|
hai 1 mes |
|
Alexander Korolev
|
6813ba085c
web: 8-port switch layout + toml viewer for no-run recipes
|
hai 1 mes |
|
Alexander Korolev
|
79727cca30
spark lanes: r35 + mxfp8 kernel fallback; r38 pool-deadlock bug report
|
hai 1 mes |
|
Alexander Korolev
|
70a9ec9c48
spark lanes: aiden production-1.2 image (r35 mm_mxfp8 n>=128 rejects the kda gemms)
|
hai 1 mes |
|
Alexander Korolev
|
f1209f78f2
glm-5.3-flash-spark lanes on the karmic-kraken-beta image (r12 gone, non-rebuildable); 4x variant; gmu 0.85
|
hai 1 mes |
|
Alexander Korolev
|
8ba943b935
full GLM-5.3 native lane on 8x: aiden prod-1.2 image, safetensors loader, dspark k8 @64k
|
hai 1 mes |
|
Alexander Korolev
|
c144bbb479
P2.1: 8x full GLM up @512K DCP1 (KV 972K tokens, 1.85x; no DCP bug), quality gate clean; bench running; NFS rides enP7s7 mgmt NIC (operator's Mikrotik question)
|
hai 1 mes |
|
Alexander Korolev
|
e625c2abd0
P2.3 result: 8x DS4.1 in-memory engram bench 161/176, 2,866 pp / 50.2 tg / 986 ms (prefill +75%, TTFT -43% vs 4x prod)
|
hai 1 mes |
|
Alexander Korolev
|
a5f50a4185
P2.3: 8x DS4.1 in-memory engram up (GMU 0.85 fix, dgx6-8 cache symlink, 122-min NFS load, KV 2.58M tokens @4.93x, quality gate clean); bench 8x running
|
hai 1 mes |
|
Alexander Korolev
|
5937043c71
plan: SGLang DS4.1 4x lane queued (aidenle recipe, production since 09-17 on their fleet); 8x DS4.1 in-memory engram boot in progress
|
hai 1 mes |
|
Alexander Korolev
|
e6e4b43c66
P1 verdict: lab-tree garbling trait (0913 A/B garbled like r37); bug report finalized; dsv41-flash-8x recipe (in-memory engram, GMU 0.91)
|
hai 1 mes |
|
Alexander Korolev
|
8b8f906a82
P1: eugr 0913 A/B recipe (measured r37 contract on lab's infernal-invocation image); boot D infeasible on 4x (engram-off OOM incident, power-cycled, recovered)
|
hai 1 mes |
|
Alexander Korolev
|
4b8e9815bc
P0 result: full GLM-5.3 + DSpark K8 bench 153/176, d0c1 892 pp / 33.2 tg (1.75x decode) / 2,193 ms vs no-spec anchor
|
hai 1 mes |
