首頁 探索 說明
登入
sm121
/
recipes
2
0
複刻 0
檔案 問題管理 0 合併請求 0 Wiki
目錄樹: a3c43d9bb1
分支列表 標籤列表
recipes
/
recipes
Alexander Korolev a3c43d9bb1 beam lane: spec-k7 quality/perf reports + real-work operational contract 2 天之前
..
beam-1-0-nvfp4-4x a3c43d9bb1 beam lane: spec-k7 quality/perf reports + real-work operational contract 2 天之前
deepseek-v4-flash 73716e2912 recipes: deepseek-v4-flash 1 周之前
deepseek-v4-flash-dspark38 9e7c0341e7 recipes: deepseek-v4-flash-dspark38 1 周之前
deepseek-v4-flash-eugr 036c2f64d1 recipes: deepseek-v4-flash-eugr 1 周之前
deepseek-v4-flash-eugr-agent 6947c8d492 recipes: deepseek-v4-flash-eugr-agent 1 周之前
deepseek-v4-flash-src b6d3a7cba4 recipes: deepseek-v4-flash-src 1 周之前
dsv4-flash-vision 2f313d7481 recipes: dsv4-flash-vision 1 周之前
dsv4-flash-vision-eugr f680fb78e2 recipes: dsv4-flash-vision-eugr 1 周之前
dsv4-vision-aiden-2x 232c5512d1 recipes: dsv4-vision-aiden-2x 1 周之前
dsv4-vision-aiden-373 9ca52b3416 recipes: dsv4-vision-aiden-373 1 周之前
dsv4-vision-aiden-4x b8c6a8d594 recipes: dsv4-vision-aiden-4x 1 周之前
dsv4-vision-exp-vavrusa 8ab7464c7a recipes: dsv4-vision-exp-vavrusa 1 周之前
dsv41-flash-1017-2x-tensorfold ec17b00796 dsv41-flash-1017-2x-tensorfold: ship it as an sm121 recipe 4 天之前
dsv41-flash-1017-4x-sglang fe606b76bb dsv41-flash-1017-4x-sglang: land the 2026-10-18 A/B reports and the 24/24 concurrency bump 2 天之前
dsv41-flash-4x 1bdb4ee1d1 recipes: dsv41-flash-4x 1 周之前
dsv41-flash-4x-aiden11 db6954377d recipes: dsv41-flash-4x-aiden11 1 周之前
dsv41-flash-4x-b12x ab0f145dad recipes: dsv41-flash-4x-b12x 1 周之前
dsv41-flash-4x-dcp2 1b96f6def9 recipes: dsv41-flash-4x-dcp2 1 周之前
dsv41-flash-4x-mia 71fcd431a8 recipes: dsv41-flash-4x-mia 1 周之前
dsv41-flash-4x-sglang-aiden16 856115ecd1 recipes: dsv41-flash-4x-sglang-aiden16 1 周之前
dsv41-flash-8x 5a84abdb51 recipes: dsv41-flash-8x 1 周之前
dsv41-flash-eugr-0913-4x 67604ecc99 recipes: dsv41-flash-eugr-0913-4x 1 周之前
dsv41-flash-exl3-2x 9eca5e8c43 recipes: dsv41-flash-exl3-2x 1 周之前
dsv41-flash-jovian-r36-4x 19068d9ced recipes: dsv41-flash-jovian-r36-4x 1 周之前
dsv41-flash-jovian-r37-4x fe009aa890 recipes: dsv41-flash-jovian-r37-4x 1 周之前
gemma4-12b d5fa022173 recipes: gemma4-12b 1 周之前
glm51 eb306d3af4 recipes: glm51 1 周之前
glm52 7a3b8ccfbb recipes: glm52 1 周之前
glm52-hybrid-fp8-mxfp4 5ce31e739f recipes: glm52-hybrid-fp8-mxfp4 1 周之前
glm52-int4-int8 5d49756564 recipes: glm52-int4-int8 1 周之前
glm52-mxfp4-experts-gptq 314f2450f4 recipes: glm52-mxfp4-experts-gptq 1 周之前
glm52-nf3-hybrid-context 00cf3eae27 recipes: glm52-nf3-hybrid-context 1 周之前
glm52-nf3-hybrid-fast 9ae00b3be3 recipes: glm52-nf3-hybrid-fast 1 周之前
glm52-nf4-lloyd-experts-gptq e3e284829c recipes: glm52-nf4-lloyd-experts-gptq 1 周之前
glm53-flash-aiden-2x 43c5aa8df7 recipes: glm53-flash-aiden-2x 1 周之前
glm53-flash-aiden-4x dd9a1e5923 recipes: glm53-flash-aiden-4x 1 周之前
glm53-flash-aiden-4x-labdrafter 620bd21915 recipes: glm53-flash-aiden-4x-labdrafter 1 周之前
glm53-flash-dflash2-mxfp8 58294341e6 recipes: glm53-flash-dflash2-mxfp8 1 周之前
glm53-flash-eugr-2x 2b11fb66b0 recipes: glm53-flash-eugr-2x 1 周之前
glm53-flash-exl3 e2c6e275df recipes: glm53-flash-exl3 1 周之前
glm53-flash-fp4-local-2x ef684e8ae7 glm53-fp4local-2x: served-model-name glm53-fp4-local → glm53 (the fleet-wide alias convention) 5 天之前
glm53-flash-fp4local-2x-dflash 1b35a14ef6 dflash A/B outcome: eugr build rejects all draft attention backends under fp8 KV (FLASHINFER/TRITON_ATTN/B12X/FLASH_ATTN) — DFlash2-for-GLM lives in the kindling overlay stack; the card stack is MTP-only 5 天之前
glm53-flash-intel-w4a16 ddff9a6414 recipes: glm53-flash-intel-w4a16 1 周之前
glm53-flash-nvfp4 028efc1d8f recipes: glm53-flash-nvfp4 1 周之前
glm53-flash-nvfp4-4p67 6de1687d9b recipes: glm53-flash-nvfp4-4p67 1 周之前
glm53-flash-nvfp4-eugr b98b0a13a0 recipes: glm53-flash-nvfp4-eugr 1 周之前
glm53-flash-nvfp4-eugr-4x 8683ce6896 recipes: glm53-flash-nvfp4-eugr-4x 1 周之前
glm53-flash-nvfp4-fp8-dgpp-2x 201b853c79 dgpp recipes: pin [dgpp].repo/rev (3ad63d6b) for sm121 on-demand DGPP provisioning 1 周之前
glm53-flash-nvfp4-lab 254cfdce19 recipes: glm53-flash-nvfp4-lab 1 周之前
glm53-flash-nvfp4-lab-0rand 7118377bdb recipes: glm53-flash-nvfp4-lab-0rand 1 周之前
glm53-flash-nvfp4-lab-4x 965c93872c recipes: glm53-flash-nvfp4-lab-4x 1 周之前
glm53-flash-nvfp4-lab-4x-r38 864fb0fedb recipes: glm53-flash-nvfp4-lab-4x-r38 1 周之前
glm53-flash-nvfp4-lab-4x-r39 649bf787c2 recipes: glm53-flash-nvfp4-lab-4x-r39 1 周之前
glm53-flash-nvfp4-lab-dflash2 bf10a63a96 recipes: glm53-flash-nvfp4-lab-dflash2 1 周之前
glm53-flash-nvfp4-lab-r39 9d22ac7f6d recipes: glm53-flash-nvfp4-lab-r39 1 周之前
glm53-flash-nvfp4-r17abi c6c03fb3f2 recipes: glm53-flash-nvfp4-r17abi 1 周之前
glm53-flash-nvfp4-r18 721a39f5dd recipes: glm53-flash-nvfp4-r18 1 周之前
glm53-flash-nvfp4-r18-2x-nvfp4kv 645cea0a1e recipes: glm53-flash-nvfp4-r18-2x-nvfp4kv 1 周之前
glm53-flash-nvfp4-r18-2x-nvfp4kv-spark d3113b73be recipes: glm53-flash-nvfp4-r18-2x-nvfp4kv-spark 1 周之前
glm53-flash-nvfp4-r18-4x e21150512c recipes: glm53-flash-nvfp4-r18-4x 1 周之前
glm53-flash-nvfp4-r18-4x-dcp4 e745a7a518 recipes: glm53-flash-nvfp4-r18-4x-dcp4 1 周之前
glm53-flash-nvfp4-r18-4x-nvfp4kv d09a402f76 recipes: glm53-flash-nvfp4-r18-4x-nvfp4kv 1 周之前
glm53-flash-nvfp4-r18-4x-nvfp4kv-spark 23b0ea817c recipes: glm53-flash-nvfp4-r18-4x-nvfp4kv-spark 1 周之前
glm53-flash-nvfp4-r18-dflash2 fd661c12f3 recipes: glm53-flash-nvfp4-r18-dflash2 1 周之前
glm53-flash-nvfp4-redhat a8c0fe30e8 recipes: glm53-flash-nvfp4-redhat 1 周之前
glm53-flash-nvfp4-spark b3c27a26d4 recipes: glm53-flash-nvfp4-spark 1 周之前
glm53-flash-r18-2x-nvfp4kv-spark-eager 8e2b58ba71 recipes: glm53-flash-r18-2x-nvfp4kv-spark-eager 1 周之前
glm53-flash-r27-4x-dcp4 7666ef6074 recipes: glm53-flash-r27-4x-dcp4 1 周之前
glm53-flash-r27-spark f6dcc71a9f recipes: glm53-flash-r27-spark 1 周之前
glm53-flash-r28-4x-dcp4 cd88bd5c0e recipes: glm53-flash-r28-4x-dcp4 1 周之前
glm53-flash-r28-4x-dcp4-b12xload 355bea9f6b recipes: glm53-flash-r28-4x-dcp4-b12xload 1 周之前
glm53-flash-r28-4x-dcp4-split2048 350c37b1ba recipes: glm53-flash-r28-4x-dcp4-split2048 1 周之前
glm53-flash-r28-dflash2-2x bc8a8789e6 recipes: glm53-flash-r28-dflash2-2x 1 周之前
glm53-flash-r28-spark 23369b3cd6 recipes: glm53-flash-r28-spark 1 周之前
glm53-flash-r31-2x 37d95669bc recipes: glm53-flash-r31-2x 1 周之前
glm53-flash-r31-4x-dcp4 7ea8df0127 recipes: glm53-flash-r31-4x-dcp4 1 周之前
glm53-flash-r35-2x-dflash2 798cead213 recipes: glm53-flash-r35-2x-dflash2 1 周之前
glm53-flash-r35-4x-dcp4 17270adcb0 recipes: glm53-flash-r35-4x-dcp4 1 周之前
glm53-flash-r35-4x-dflash2 f791e0b0e4 recipes: glm53-flash-r35-4x-dflash2 1 周之前
glm53-flash-spark-4x cd4b5ce28b recipes: glm53-flash-spark-4x 1 周之前
glm53-flash-sparkring-2x 76c360bfe0 recipes: glm53-flash-sparkring-2x 1 周之前
glm53-flash-sparkring-4x f754d25613 recipes: glm53-flash-sparkring-4x 1 周之前
glm53-flash-sparkrun 7e249192e0 recipes: glm53-flash-sparkrun 1 周之前
glm53-full-mxfp4-4x 832c488484 recipes: glm53-full-mxfp4-4x 1 周之前
glm53-full-mxfp4-4x-dspark8 becf0521e2 recipes: glm53-full-mxfp4-4x-dspark8 1 周之前
glm53-full-mxfp4-4x-probe 4284f95262 recipes: glm53-full-mxfp4-4x-probe 1 周之前
glm53-full-mxfp4-8x d524c77c75 recipes: glm53-full-mxfp4-8x 1 周之前
glm53-full-native-8x-dspark8 083b49ba41 recipes: glm53-full-native-8x-dspark8 1 周之前
glm53-full-tp4-gr8 541c5f6dc4 glm53-full-tp4-gr8: first-class gr8-engine recipe (engine=gr8, TP=4 switched, --tp/--peers generated, fixed API :8100) 1 周之前
glm53-kindling-2x 6c784467fe glm53-kindling-2x: second perf pass (post-campaign cooldown) — d0 c1 51.4 vs 39.6 hot: run-variance baseline 1 周之前
glm53-kindling-4x 0acfa6c680 new-engine quality gates on v2.8.0 (97 scen, hardmode, engine-default effort): gr8 full GLM-5.3 TP=4 = 89/100 (decode 49.9 t/s code c1, prefill 1,294 cold/50.9K cached, agg c4 64.6); DGPP Qwen3.8-Flash-Next NVFP4 w1 = 88/100 (def-effort; the low-effort run 86 on 2.7.1.dev7 is a separate datapoint) 1 周之前
glm53-kindling-4x-karmic a941d39ad7 recipes: glm53-kindling-4x-karmic 1 周之前
glm53-kindling-6x d096210e09 recipes: glm53-kindling-6x 1 周之前
glm53-kindling-8x c87238bfd6 8x lane reports: kindling8x-conc-48g (2026-10-12, 95/100 anchor) run package — md/json/log/manifest/recipe.toml 1 周之前
laguna-m1-nvfp4 c5cba4e650 recipes: laguna-m1-nvfp4 1 周之前
laguna-s2.1-nvfp4 08d6c8814f recipes: laguna-s2.1-nvfp4 1 周之前
ling-31-flash-fp4 4837e6b6f3 ling-31-flash-fp4: first valid bench 82/100 (159/194, v2.8.0, TP4 dgx58-switch) 6 天之前
ling-31-flash-nvfp4 81eb20a763 ling recipe: pointer to the fp4 smaller-cut path (~110GB -> one Spark) + ling3 parser 1 周之前
mimo-v2.5-nvfp4 deccffc1f8 recipes: mimo-v2.5-nvfp4 1 周之前
mimo26-flash-mopd-2x b40b36db49 recipes: mimo26-flash-mopd-2x 1 周之前
mimo26-flash-mopd-2x-sparkring b26288e726 recipes: mimo26-flash-mopd-2x-sparkring 1 周之前
mimo26-flash-mopd-4x d1cf94a909 recipes: mimo26-flash-mopd-4x 1 周之前
mimo26-flash-mopd-4x-sparkring 040e9f26b6 mimo26-flash-mopd-4x-sparkring: rebase bench (2026.10.4 image, v2.8.0) — 90/100, tg 66.2/80.7 c1/c2 (+19% vs dev-20261004) 1 周之前
mimo26-flash-mopd-4x-vllm b4dee4e77d recipes: mimo26-flash-mopd-4x-vllm 1 周之前
mimo26-flash-rl-2x 7c691cbe00 recipes: mimo26-flash-rl-2x 1 周之前
mimo26-pro-mopd-8x-sparkring fa3714aceb recipes: mimo26-pro-mopd-8x-sparkring 1 周之前
mimo26-pro-rl-8x 823f22af9c recipes: mimo26-pro-rl-8x 1 周之前
minimax-m3 6fc8740f3e recipes: minimax-m3 1 周之前
minimax-m3-mxfp8 925eb96aad recipes: minimax-m3-mxfp8 1 周之前
minimax-m3-nvfp4 b1227da6c7 recipes: minimax-m3-nvfp4 1 周之前
nemotron-35-bolt-74b-24gb 5e97656e8d bolt-1x bench: 84/100 (163/194), v2.8.0, pluto → dgx3 (bench --node fix; first valid run after two wrong-head attempts) — card image + LiLiCorr dflash 7 spec-decode, GMU 0.90; fails TC-60 (sleeper injection BCC), TC-74 (premature batching), TC-85 (premature notification); vs card anchor 88/100 (-4, card eval conditions unknown) 5 天之前
nemotron-35-super-vl-120b-nvfp4 903ed35881 Add 3 single-Spark recipes from NVIDIA forum (10-16): Qwen3.8-Flash-Next-FP4-Local (b12x+MTP n4, 89/100), Nemotron-3.5-Super-VL-120B-NVFP4 (LiLiCorr dflash7, 1M ctx, 90/100), Nemotron-3.5-Bolt-74B-A9B-24GB (QAD 2-bit, LiLiCorr, 88/100). All solo/TP1, eugr/spark-vllm-b12x digest-pinned, port 30000. 6 天之前
qwen-agentworld-35b d9da512064 recipes: qwen-agentworld-35b 1 周之前
qwen38-27b 2ad2dca44a recipes: qwen38-27b 1 周之前
qwen38-27b-fp8 8eb51217f2 recipes: qwen38-27b-fp8 1 周之前
qwen38-27b-fp8-dspark 1f0b497e94 recipes: qwen38-27b-fp8-dspark 1 周之前
qwen38-27b-int4 7026edae8a recipes: qwen38-27b-int4 1 周之前
qwen38-27b-nvfp4 b7ab9951be recipes: qwen38-27b-nvfp4 1 周之前
qwen38-27b-nvfp4-dgpp 201b853c79 dgpp recipes: pin [dgpp].repo/rev (3ad63d6b) for sm121 on-demand DGPP provisioning 1 周之前
qwen38-27b-nvfp4-sglang-dflash2 7f68837c14 recipes: qwen38-27b-nvfp4-sglang-dflash2 1 周之前
qwen38-27b-nvfp4-sglang-dspark 46b3d2e981 recipes: qwen38-27b-nvfp4-sglang-dspark 1 周之前
qwen38-27b-nvfp4-unsloth 13dcb59e81 recipes: qwen38-27b-nvfp4-unsloth 1 周之前
qwen38-27b-nvfp4-unsloth-dspark cf340b90c0 recipes: qwen38-27b-nvfp4-unsloth-dspark 1 周之前
qwen38-flash-next-fp4-local 903ed35881 Add 3 single-Spark recipes from NVIDIA forum (10-16): Qwen3.8-Flash-Next-FP4-Local (b12x+MTP n4, 89/100), Nemotron-3.5-Super-VL-120B-NVFP4 (LiLiCorr dflash7, 1M ctx, 90/100), Nemotron-3.5-Bolt-74B-A9B-24GB (QAD 2-bit, LiLiCorr, 88/100). All solo/TP1, eugr/spark-vllm-b12x digest-pinned, port 30000. 6 天之前
qwen38-flash-next-fp8 e6920a2126 recipes: qwen38-flash-next-fp8 1 周之前
qwen38-flash-next-int4 fa5c0c9552 recipes: qwen38-flash-next-int4 1 周之前
qwen38-flash-next-int4-min d2ae2db64a recipes: qwen38-flash-next-int4-min 1 周之前
qwen38-flash-next-nvfp4 0acfa6c680 new-engine quality gates on v2.8.0 (97 scen, hardmode, engine-default effort): gr8 full GLM-5.3 TP=4 = 89/100 (decode 49.9 t/s code c1, prefill 1,294 cold/50.9K cached, agg c4 64.6); DGPP Qwen3.8-Flash-Next NVFP4 w1 = 88/100 (def-effort; the low-effort run 86 on 2.7.1.dev7 is a separate datapoint) 1 周之前
solar-open2-nvfp4 bedea1c6c4 recipes: solar-open2-nvfp4 1 周之前
step5-preview-nvfp4 b4bcb389ec live 8x probe results: Ling fp8 feasible on GB10 (serves+generates; bench needs tool-call config + thinking control); Step-5 ruled out (Step5 DSA needs SM90/SM103, GB10=SM121) 1 周之前
test-composed 6e465cdca4 recipes: test-composed 1 周之前
test-mods b506357455 recipes: test-mods 1 周之前
test-sleep 005ec37c6f recipes: test-sleep 1 周之前