 Alexander Korolev
|
b5905ac702
docs(agents): clarify path-ref resolves to container-side HF cache path
|
il y a 3 mois |
|
Alexander Korolev
|
dcb28faf1e
fix(docker): rewrite resolved HF cache path to container mount target
|
il y a 3 mois |
|
Alexander Korolev
|
8c68124cce
feat(recipe): GLM-5.2-MXPF4-Experts-GPTQ recipe
|
il y a 3 mois |
|
Alexander Korolev
|
a7cd759af1
feat(recipe): model-subfolder path refs via {{name}} substitution
|
il y a 3 mois |
|
Alexander Korolev
|
6cd245f9a8
feat(recipe): model-subfolder mounts resolved per-node from HF cache
|
il y a 3 mois |
|
Alexander Korolev
|
ac19fa45b8
fix(serve): dry-run skips SSH probing so it works with cluster off
|
il y a 3 mois |
|
Alexander Korolev
|
2b0deeaffa
feat(recipe): extra_mounts with config alias resolution + tilde + env injection
|
il y a 3 mois |
|
Alexander Korolev
|
7519fcef2a
chore(deepseek-v4-flash): disable temperature/top_p chat-template kwargs
|
il y a 3 mois |
|
Alexander Korolev
|
75a7055af5
feat(glm52): build on production-hybrid-1.3 base with pinned deps
|
il y a 3 mois |
|
Alexander Korolev
|
d102308e93
fix(docker): correct build alive-detection and harden build flow
|
il y a 3 mois |
|
Alexander Korolev
|
d14021070c
feat(recipe): enable DFlash speculative decoding for Laguna S 2.1
|
il y a 3 mois |
|
Alexander Korolev
|
e692ec1813
feat(recipe): update ghcr.io/timothystewart6/vllm-gb10 to v0.25.1-gb10.1
|
il y a 3 mois |
|
Alexander Korolev
|
e844cb481a
feat(recipe): add Laguna S 2.1 NVFP4 (single-node DGX Spark)
|
il y a 3 mois |
|
Alexander Korolev
|
b1a2b5f937
fix(recipe): set XDG_CACHE_HOME to /root/.cache for JIT persistence
|
il y a 3 mois |
|
Alexander Korolev
|
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe
|
il y a 3 mois |
|
Alexander Korolev
|
b85d797be1
feat(recipe): add GLM-5.2 Int4-Int8Mix recipe with overlay image
|
il y a 3 mois |
|
Alexander Korolev
|
3f08ff868e
refactor(core): extract Docker command helpers, rewrite host_keys with toml_edit, extract per-command handlers
|
il y a 3 mois |
|
Alexander Korolev
|
12f1fa44f9
feat!: rename binary from sm121-launcher to sm121
|
il y a 3 mois |
|
Alexander Korolev
|
280be2c119
refactor(cli): rename --power-off to --off on stop subcommand
|
il y a 3 mois |
|
Alexander Korolev
|
38d4e5568f
refactor(cli): rename --auto-power-on to --on (-O) on serve
|
il y a 3 mois |
|
Alexander Korolev
|
097f22082f
feat(cli): add build subcommand for standalone image build
|
il y a 3 mois |
|
Alexander Korolev
|
7fe2423a26
feat(build): checksum-based rebuild trigger with sm121.checksum label
|
il y a 3 mois |
|
Alexander Korolev
|
6e984caabd
fix(deepseek-v4): add build.sh for simple docker build
|
il y a 3 mois |
|
Alexander Korolev
|
57d686b17b
fix(deepseek-v4): add NCCL ABI fix + instanttensor overlay image
|
il y a 3 mois |
|
Alexander Korolev
|
43f00731a5
fix(glm52-nf3): restore b12x editable install to include integration/ and attention/ subpackages
|
il y a 3 mois |
|
Alexander Korolev
|
4e24cafb19
fix: revert kernel.py overlay (uses private b12x symbols)
|
il y a 3 mois |
|
Alexander Korolev
|
ff8340a20a
fix: add real CosmicRaisins DCP patches + sm120-return-lse.patch
|
il y a 3 mois |
|
Alexander Korolev
|
d8048a2b1d
fix: sync overlay with madeby561/vllm-glm52-nvfp4-nf3-hybrid:v3
|
il y a 3 mois |
|
Alexander Korolev
|
126fe81115
fix: drop_caches after image sync (clear docker load page cache)
|
il y a 3 mois |
|
Alexander Korolev
|
810d952c72
fix: b12x non-editable install (files land in site-packages)
|
il y a 3 mois |
