| .. |
|
DESIGN_PLAN_LATEST.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
IMPACT_LATEST.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
REFACTOR_LATEST.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
SECURITY_PLAN_LATEST.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
TEST_PLAN_LATEST.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
dgx-spark-agent-serving-findings.md
|
3aa9d380b2
feat(recipes): agent-serving profile + GMU/env fixes from benchmark findings
|
2 luni în urmă |
|
dsv41-1017-aiden-image-audit.md
|
d390fd2e87
audit: #15 resolved — the effort-mapping patch baked into the versioned dsv41-sglang-1017:v1 image (glm52 pattern), verified live in the running engine; TensorFold 1017 dual-Spark thread 386121 noted for the watch list
|
6 zile în urmă |
|
dsv41-1017-tuning.md
|
5e0d7c2306
specs: DeepInfer assessment — offline-only + SM100-only, not runnable on GB10; the transferable ideas (SWA bounded replay, confidence-scheduled verification)
|
2 zile în urmă |
|
engine-drivers.md
|
28f909758a
engine-drivers spec: correct DGPP provisioning model (rank0-only install, idempotent, space-nodes, script-file tee) + DGPP deploy prereqs from live w2 run
|
1 săptămână în urmă |
|
glm53-recipe-state.md
|
f944aea35c
kindling 4x: first full campaign — quality 166/176 (94/100, best on cluster), perf table, boot bugs ledger (spec-config form, UMA gate coin flip, control-file dirs); spec section + report artifacts (json+log snapshot, depth suite in flight)
|
3 săptămâni în urmă |
|
ha-addon-recipe-repos.md
|
18eb6e8254
cluster config is an HA option (sm121_config_extra), not a second repo
|
1 săptămână în urmă |
|
qwen38-recipe-state.md
|
65f366427a
Add Qwen3.8-27B recipes + measured state doc
|
2 luni în urmă |
|
sparkring-mimo-lanes.md
|
8fc0188614
stacks: adopt upstream updates (kindling re-vendor, SparkRing 10-04, karmic tip 636afde9)
|
2 săptămâni în urmă |
|
staging-feature.md
|
05d095ac96
staging: sparse-copy format (aiden16 DSV41_ENGRAM_DIR contract) + packed-Mia bench notes
|
3 săptămâni în urmă |
|
tech-stack.md
|
3379f83f32
chore(bigpowers): adopt bigpowers lifecycle and conventions
|
2 luni în urmă |
|
tensorfold-1.0.12-bringup.md
|
cd1fc19c1e
specs: TensorFold 1.0.12 4-node bring-up — engine+kit+weights+rank-cache all verified, blocked on a loader MissingEntry (documented with the exact inputs and our deviations)
|
1 zi în urmă |
|
upstream-issues-2026-09-09.md
|
0d7c9f334e
upstream issue descriptions: b12x eh_proj loader + DFlash2 dtype (retired in sparkring, native MTP is the path); 386B capacity note
|
1 lună în urmă |
|
upstream-watch-2026-10-14.md
|
6b56ffa3d7
upstream watch 2026-10-14: b12x merged into FlashInfer 0.7.x (r40 path), gr8/DGPP/veloGB10 engines, MiMo 10-09 re-cuts + sparkring rebase, Qwen3.8-Flash-Next 2x EP lane, Ling-3.1-flash (MIT) as GLM-5.3-Flash competitor
|
1 săptămână în urmă |
|
upstream-watch-2026-10-15.md
|
28af1b0d4d
upstream-watch 10-15: MiMo 4x rebase done (90/100, +19% decode)
|
1 săptămână în urmă |
|
upstream-watch-2026-10-16.md
|
f4b4f62683
upstream-watch 10-16: Ling fp4 research — ling3 parser, fp8-as-source (Ling is mixed bf16+fp8 ckpt), fp4 format spec (mxfp4 experts+fp8 dense), single-Spark serving via cookbook, fp8->mxfp4 converter path + reference files, Qwen4-27B warm-up plan
|
1 săptămână în urmă |
|
upstream-watch-2026-10-17.md
|
8e3aff4749
upstream-watch 10-17: eugr/spark-vllm-b12x nightly (vLLM 0.1.dev21830, 10-15, 391 archs) checked — NO MuseSpark, NO Step5 (only MuseGlimmer + Step3 family). Confirms SGLang #44530 series is earliest path for Muse-816B. Bonus: b12x has DSpark speculators (Glm53DSpark, Qwen3OmniDSpark, K3DSpark, Gemma4DSpark)
|
6 zile în urmă |
|
web-dashboard.md
|
f011388d0a
web: SSE streams can no longer die silently or stay dead — 15s keepalive comments on both endpoints, live-chip watchdog rebuilds a stale stream / a gave-up EventSource, bounce on visibility/online; supervisor isolates ticks and survives poisoned locks
|
1 lună în urmă |
Alexander Korolev
cd1fc19c1e
specs: TensorFold 1.0.12 4-node bring-up — engine+kit+weights+rank-cache all verified, blocked on a loader MissingEntry (documented with the exact inputs and our deviations)