 Alexander Korolev
|
0bbd167626
night 2026-10-23: engram localization closed — 91/100, prefill parity with production
|
hace 22 minutos |
|
Alexander Korolev
|
cb4d7d148f
docs: AGENTS + spec — local-view measured (prefill 750 -> 3,957 t/s on the 1017 lane)
|
hace 1 hora |
|
Alexander Korolev
|
01011f077d
staging local-view: fix the shared-cache symlink root, nest one view per repo, adopt a pre-staged tree
|
hace 1 hora |
|
Alexander Korolev
|
cd5e76112a
staging: local-view kind — keep a hot side table on local NVMe, gated by a pre-start free-space check
|
hace 2 horas |
|
Alexander Korolev
|
72358e130c
spec: stock vLLM/SGLang vs ours for DeepSeek-V4.1-Flash-1017
|
hace 6 horas |
|
Alexander Korolev
|
12e79273ef
scripts: flare-upstream-watch — FLARE moves daily, check it before every experiment
|
hace 6 horas |
|
Alexander Korolev
|
f9606ab210
night report 2026-10-22/23: full GLM-5.3 at 90/100 with MTP, DCP impossible, prefill ceiling measured, FLARE brought up, add-on 1.3.4-1.3.9
|
hace 6 horas |
|
Alexander Korolev
|
60ffddaa38
build: bake SM121_VERSION through a generated source, not option_env alone
|
hace 12 horas |
|
Alexander Korolev
|
89553b26ac
build: rebuild when SM121_VERSION changes (option_env! alone does not invalidate cargo's cache)
|
hace 12 horas |
|
Alexander Korolev
|
a9cffba5fa
web: build-version badge under the dashboard title
|
hace 13 horas |
|
Alexander Korolev
|
b984f59269
night plan 2026-10-22/23: reproduction of the long-prefill host hang, the three envelopes to measure (KV-per-token, prefill scratch, spec depth), and the guardrails
|
hace 20 horas |
|
Alexander Korolev
|
4c4e608ec1
report/web: model-column copy + Hugging Face buttons, and a favicon
|
hace 21 horas |
|
Alexander Korolev
|
73b27de372
spec: FLARE bring-up — DeepSeek-V4.1-Flash-1017 served by NVIDIA's flare on 4 Sparks
|
hace 1 día |
|
Alexander Korolev
|
87e885d967
night report 2026-10-21/22: dashboard fixes 1.3.0-1.3.3, recipe-relative hooks, TF 1.0.12 blocked (MissingEntry) with full diagnosis, GLM-5.3 full MXFP4xFP8 blocked at 4-node worker init, GLM-5.3-Flash 2x repro 95/100
|
hace 1 día |
|
Alexander Korolev
|
ff00fa4d1a
night report 2026-10-21/22 (in progress): dashboard fixes 1.3.0-1.3.3, recipe-relative hooks, TF 1.0.12 bring-up blocked with diagnosis, GLM-5.3 full MXFP4xFP8 lane staged
|
hace 1 día |
|
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)
|
hace 1 día |
|
Alexander Korolev
|
73b8ee0800
night plan 2026-10-21: TF 1.0.12 TP4 (upstream moved from our pinned 0.6.0), RedHat GLM-5.3-MXFP4xFP8 (415GB/4 nodes), local-inference-lab GitHub, daily HF deltas
|
hace 1 día |
|
Alexander Korolev
|
f38b60111d
recipes panel: running state follows the run ledger (stale rows after stop)
|
hace 1 día |
|
Alexander Korolev
|
ebf461d422
recipe-relative hook scripts (host scripts belong to recipes), bench --name, 500-line splits
|
hace 1 día |
|
Alexander Korolev
|
a24a83696e
pre-serve hooks: serve-local NODE_RANK + TensorFold staging fixes; AGENTS.md monitoring rule
|
hace 2 días |
|
Alexander Korolev
|
882625898f
web: window drag/resize now survives an <iframe> (pointer capture)
|
hace 2 días |
|
Alexander Korolev
|
fd30bf160a
web/report: settings button — per-agent client config for a recipe's served model
|
hace 2 días |
|
Alexander Korolev
|
06722fefd9
night shift 2026-10-21: long-context results (131K/327K needle, both correct), depth sweep, 384K context
|
hace 2 días |
|
Alexander Korolev
|
a334763f95
night shift 2026-10-21: Beam DFlash2 speculation unlocked via FLASHINFER (+40% decode, turn 2.6s); real agentic task on Beam via pi (11/11 tests); DeepInfer assessed as SM100/offline-only
|
hace 2 días |
|
Alexander Korolev
|
5e0d7c2306
specs: DeepInfer assessment — offline-only + SM100-only, not runnable on GB10; the transferable ideas (SWA bounded replay, confidence-scheduled verification)
|
hace 2 días |
|
Alexander Korolev
|
7a8c147767
night shift 2026-10-20: FLARE 1017 tested and rejected (2x slower decode, crashes on agent prompts); Beam-1.0 NVFP4 lane added at 90/100 (fleet best); four-lane comparison recorded
|
hace 2 días |
|
Alexander Korolev
|
5a4eec608c
spec: 1017 TensorFold lane is now an sm121 recipe; record the TF+vLLM blocker
|
hace 4 días |
|
Alexander Korolev
|
198e6137dd
engine: tensorfold driver + tensorfold_1017_prepare built-in (NODE_RANK exported to hooks)
|
hace 4 días |
|
Alexander Korolev
|
8f734c7926
spec: TensorFold 2x 1017 lane measured — 89/100 with half the nodes, decode 50.4 vs 39-43, but turn latency 6.7s vs 1.4s
|
hace 4 días |
|
Alexander Korolev
|
b17dc421d2
spec: 1017 reasoning-effort A/B — 75 buys nothing over 100 on the new model (86 vs 87, same 1.4s turn); depth sweep closed
|
hace 4 días |
