Commit Verlauf

Autor SHA1 Nachricht Datum
  Alexander Korolev bbe9605380 Plan: sm121 web dashboard (design doc) vor 2 Monaten
  Alexander Korolev 4275e1cd7c config: global-config mgmt commands + ssh creds from cluster sources vor 2 Monaten
  Alexander Korolev 8446509baf Global config: recipe_sources + cluster_sources vor 2 Monaten
  Alexander Korolev 48bbec1f66 serve: --clear flag to wipe host-side extra_mount folders (sudo) vor 2 Monaten
  Alexander Korolev 65f366427a Add Qwen3.8-27B recipes + measured state doc vor 2 Monaten
  Alexander Korolev 50251da4ce Bump GLM recipes to aidendle94 production-hybrid-1.6 vor 2 Monaten
  Alexander Korolev 5a9795c612 Rebase deepseek-v4-flash :fixed overlay onto production-3.76.1 vor 2 Monaten
  Alexander Korolev 08871e3dfa Add --fit flag to serve: scale recipe to N nodes or whole cluster vor 2 Monaten
  Alexander Korolev 551ea7d8d7 feat(nccl): auto-detect NCCL_IB_GID_INDEX per node, opt-out via config pin vor 2 Monaten
  Alexander Korolev 42032fbe12 deepseek-v4-flash: Aiden base 3.75 + encoder fix baked in, 1M ctx, match 0rand compose tuning vor 2 Monaten
  Alexander Korolev 7f5a038adc Tune deepseek-v4-flash-eugr: prefill threshold, prefix cache, reasoning_effort=max vor 2 Monaten
  Alexander Korolev c6eda47a99 feat(deepseek-v4-flash-eugr): use high thinking by default vor 2 Monaten
  Alexander Korolev 43c9bd346f recipe(deepseek-v4-flash-eugr): second-variant test config vor 2 Monaten
  Alexander Korolev 5d90e025f0 revert: keep experimental eugr recipe changes out of main (user is still testing) vor 2 Monaten
  Alexander Korolev 08d1e9fe03 feat(serve): add --mirror — serve from a per-node local HF mirror vor 2 Monaten
  Alexander Korolev 4eff3e41b0 docs(skills): add nvidia-forum-reading skill vor 2 Monaten
  Alexander Korolev a432ba42be docs(skills): add sm121-launcher and vllm-mod-patches skills vor 2 Monaten
  Alexander Korolev 070bf6666a feat(deepseek-v4-eugr): optional reasoning_content key patch (default-off) vor 2 Monaten
  Alexander Korolev 0c7b29fbd6 feat(deepseek-v4-eugr): minimal reasoning_effort low/high/max patch (not full rewrite) vor 2 Monaten
  Alexander Korolev 25b18fcd56 fix(deepseek-v4): drafter b12x WO routing for real DSpark acceptance + GMU 0.78 vor 2 Monaten
  Alexander Korolev e296685e8c fix(deepseek-v4): let DSpark survive mixed-context batches (tool-eval-bench) vor 2 Monaten
  Alexander Korolev 5de97ee3aa fix(deepseek-v4): get DeepSeek-V4-Flash DSpark booting on production-3.8 (ds4) vor 2 Monaten
  Alexander Korolev d2026c2ddf feat(deepseek-v4): raise max-model-len to 1M + pin KV cache bytes (eugr) vor 2 Monaten
  Alexander Korolev 3aa9d380b2 feat(recipes): agent-serving profile + GMU/env fixes from benchmark findings vor 2 Monaten
  Alexander Korolev 0362145ef5 docs(specs): investigation — production-3.8 image update failures vor 2 Monaten
  Alexander Korolev a8e4932fc4 feat(deepseek-v4): add eugr b12x recipe variant with baked draft-loader mod vor 2 Monaten
  Alexander Korolev 5bc841f211 tmp: rename to simplify vor 2 Monaten
  Alexander Korolev 3a783ee58e Revert "feat(deepseek-v4): move :fixed overlay to production-3.8 base + enable DSPARK tiny decode" vor 2 Monaten
  Alexander Korolev 72f40b3da2 feat(deepseek-v4): move :fixed overlay to production-3.8 base + enable DSPARK tiny decode vor 2 Monaten
  Alexander Korolev e6f70ecef1 fix(deepseek-v4): forbid echoing injected content to close TC-34 vor 2 Monaten