Historique des commits

Auteur SHA1 Message Date
  Alexander Korolev 8c292e8a5d vision 2x: point SPARK_HF_HOME at the shared NAS HF cache (DFlash2/DSpark drafts live there; the NVMe copy lacked the DSpark draft) il y a 1 mois
  Alexander Korolev b2ec9376cc vision 2x: drop the revision pin (offline refs/main resolves; pinned hash broke the draft lookup) il y a 1 mois
  Alexander Korolev dc13bd0afe vision 2x: aiden image via HIS env-script entrypoint — mods derive NODE_RANK/MASTER_ADDR from VLLM_HOST_IP + sed off the $@ tail (env-driven contract, launcher-tail dropped) il y a 1 mois
  Alexander Korolev f2ba69ee5d vision 2x: use his /opt/vllm-src/.venv python (dist-packages wheel lacks the multimodal wrap) + his env block il y a 1 mois
  Alexander Korolev 41f2808d7a vision 2x: cudagraph cap from sizes list (pydantic consistency); MAX env kept il y a 1 mois
  Alexander Korolev a0c57b9e2f aiden lanes: entrypoint=[] for built-in-entrypoint images (launcher doubles CMD otherwise); vision 2x uses proven overlay img + aiden DSpark-greedy additions (his env-script needs a dedicated pass) il y a 1 mois
  Alexander Korolev ce4727d2a3 aiden lanes: 2x+4x recipes from aidenle.com guides (glm prod-1.1 DFlash2-7 + dsv4 vision prod-3.72 DSpark-6); 2x to verify, 4x deferred (FP8 ckpt downloading) il y a 1 mois
  Alexander Korolev fd4c74dcc1 checkpoint A/B on ds4: NVFP4 lab ckpt = same 88 quality on r17 stack but +4-7 tok/s; lab day-0 stack retains 93; r17abi recipe added il y a 1 mois
  Alexander Korolev 9079638ee4 lab recipes: instanttensor baked into lab image + enabled (verified: 631s vs 700s, two-pass loader, all boot markers + smoke 3901) il y a 1 mois
  Alexander Korolev 04984a169d ledger: one-backend bisect solved (linear b12x = the toxin); ship config locked il y a 1 mois
  Alexander Korolev 2d379f9752 final bisect: moe-b12x kept (88, d8192+2.8), linear-b12x is the toxin (44), align perf-negative — recipe locked il y a 1 mois
  Alexander Korolev 582343c9c6 r17 composition ship anchor: backend-flags are the quality toxin (bisect 46/47/74 vs 87); knobs env clean; graphs+MTP fixed on public r17 refs; ledger il y a 1 mois
  kilork 08a930e477 r12 image: adopt eugr lab build patches (12x patch_vllm_*.py) + pre-staged build sub-deps (SRC_DIR escapes) for flaky node github; ledger: lil render reference + perf push state il y a 1 mois
  kilork cc41c48e03 glm53 spark recipe: full lil policy alignment (b12x linear/moe backends, k=5 MTP + moe_backend, instanttensor load, mamba-cache-mode align, INSTANTTENSOR env) + mounted pre-baked JIT caches (~/sm121-cache/glm53-jit-121f, dsv4 pattern); boot 35m -> 7m, exact-391 verified il y a 1 mois
  kilork 5283d11ce1 clean r12 image shipped: 90/100 c1 (was 88 day-0-base), entrypoint+MAX_JOBS boot fixes, ledger il y a 1 mois
  kilork 236ce5a7f3 spark recipe: entrypoint (clean image is ENTRYPOINT []) + MAX_JOBS=2 (JIT nvcc OOM at boot); clean r12 image serving, exact-391 verified il y a 1 mois
  kilork 3d084c54e4 cli: proper --no-follow flag (SetTrue opt-out; the SetFalse+default combo inverted semantics and lacked the negation) il y a 1 mois
  kilork 2d17cd0356 r12 clean build: SHA-pinned PR heads (550 closed), deepgemm pre-clone, lil launch-policy diff recorded il y a 1 mois
  kilork 9395d6d955 r12 clean build: flashinfer-python 0.6.17 from PyPI (whl-index path dead); recipe MTP:3 comment clarified (built-in MTP head, not dspark) il y a 1 mois
  kilork 8de6340763 clean r12 rebuild: raw-cuda base per solar pattern (r12-specific refs, torch 2.13, flashinfer 0.6.17, b12x PR-head composition); docker_image_dir fixes; bench 88/100 (69-suite) il y a 1 mois
  kilork ac5c6c9108 NVFP4-Spark serving on our r12 fork build: spec-config attention_backend=B12X fixes MTP draft selection; intel recipe retargeted; ledger il y a 1 mois
  kilork 191add7784 docker-images: glm53-jovian-r12 — in-docker build of lab fork (vllm base 54f6e9826c + PR heads 550-557 merges; b12x master + PR heads 260-270 merges; day-0 arm64-cu130 base; nvrtc/headers cmake fixes); teb c1 on vavrusa lane: 88/100 (default 69-scenario set) il y a 1 mois
  kilork 56c99b6a33 recipes: dsv4-vision-exp-vavrusa — native image_url lane (eugr nightly 20260810 + instanttensor-draft + vision-exp mods), DSpark-6, NVMe weights; verified text-exact + real-photo caption (325 img tokens) il y a 1 mois
  kilork 344479943a recipes: intel glm53 W4A16-AutoRound (tjstewart image, gptq_marlin, entrypoint) + lab NVFP4-Spark (mtp-5 embedded; blocked on whole-network W4A4 loader); ledger: rtx6kpro source contract for own fork build il y a 1 mois
  kilork de62327a2b ledger: dsv4vision tool-eval-bench c1 (81/100, tg 34.1/31.0/29.1, deployability 74, 3 safety-critical fails) il y a 1 mois
  kilork 8d49ca7159 dsv4vision: env-gated openai mm branch (VLLM_DSV4_MM), worker shape-follow patches, local-NVMe weights (dgx3+dgx4), cache-clear runbook; text lane verified exact on new boots il y a 1 mois
  kilork 4df980f57a publish: preserve nested overlay dirs (patches/ still flattened; --gist refused for nested); dsv4-vision: full OpenAI MM processor trio (info/processor/dummy+updates), model embed_multimodal + embed_input_ids merge, placeholder id 129264; verified sentinel/perm math in-image il y a 1 mois
  kilork d34e7758bc dsv4-vision lane: text serving verified on eugr fork (34.9 tok/s, accept 0.714); OpenAI-mm processor gap documented il y a 1 mois
  kilork 1617621e8c dsv4-flash-vision lane (sfxnz day-0 plugin port) + MNBT1024 assertion result il y a 1 mois
  kilork 65916245e2 lab-quant variant of 0rand winning config (forum 381350 #119): GMU 0.89, MNBT 2048, KV auto-size il y a 1 mois