 Alexander Korolev
|
e9928e3ac7
night list: prod-3.73-vision = DSV4F-Vision aiden lane (replaces blocked eugr row)
|
пре 1 месец |
|
Alexander Korolev
|
03edc92027
night revisit: aiden lanes + prod-3.73-vision tag queued, statuses ledgered
|
пре 1 месец |
|
Alexander Korolev
|
d736662719
r27: explicit TORCH_CUDA_ARCH_LIST=12.1a in vllm-builder (stage-env loss -> sm120-only kernels)
|
пре 1 месец |
|
Alexander Korolev
|
afd95f742d
r27 lane: glm53-flash-r27-spark (best 2x config on the r27 engine)
|
пре 1 месец |
|
Alexander Korolev
|
f599efe741
skill: jovian-images (clone-at-pin builds, rev adoption, GB10 deltas, pitfall ledger)
|
пре 1 месец |
|
Alexander Korolev
|
35d8f090b2
r27 fetch-deps: recursive gitlink population (FlashMLA/qutlass/FlashKDA cutlass vendors; -d stub trap fixed)
|
пре 1 месец |
|
Alexander Korolev
|
3e63d19947
state doc: image build v2 (clone-at-pin) contract
|
пре 1 месец |
|
Alexander Korolev
|
9118a4648b
r27: WORKDIR in builder stages (stage split dropped inherited cwd)
|
пре 1 месец |
|
Alexander Korolev
|
97329c6b10
r27 build dir: clone-at-pin (no tarballs), gated vllm/b12x/lmcache stages, in-build deps fetch w/ pin gates
|
пре 1 месец |
|
Alexander Korolev
|
0a1a5dbcc1
typo
|
пре 1 месец |
|
Alexander Korolev
|
b564bd561d
ops layout: 2nodes-dac = production (operator agent), ds4 = dev lane
|
пре 1 месец |
|
Alexander Korolev
|
ca8bb6f448
campaign extended: 2x nvfp4 KV (154), 2x Spark (158, best 2x), 4x Spark blocked (mm_mxfp8 TP4); 4x 16G pins; JIT warm-start 9min
|
пре 1 месец |
|
Alexander Korolev
|
fe78fec205
add nvfp4kv spark variants (2x + 4x 16G, -Spark checkpoint on r18 img)
|
пре 1 месец |
|
Alexander Korolev
|
c296f735a3
4x cells: KV pin 10 -> 16 GiB (NVFP4 4x has the headroom; 2x stays 10)
|
пре 1 месец |
|
Alexander Korolev
|
79c40c9482
add r18 2x nvfp4_ds_mla KV lane (parity with the 4x cell)
|
пре 1 месец |
|
Alexander Korolev
|
cc19b5f581
persist JIT caches per node (flashinfer/torchinductor/vllm mounts in all glm53 lanes)
|
пре 1 месец |
|
Alexander Korolev
|
14b02ac5f5
history/night-shifts/2026-09-05.md: full combined score+perf table of the r18 campaign
|
пре 1 месец |
|
Alexander Korolev
|
24cfdc3a76
state doc: add --perf comparison table (all 7 lanes, llama-benchy pp/tg/TTFT)
|
пре 1 месец |
|
Alexander Korolev
|
5d587215dd
separation complete: r18 dir carries source tarballs; state doc records r12/r18 split + r17 rebuild caveat
|
пре 1 месец |
|
Alexander Korolev
|
b3d61b7af0
separate r18 from r12: dedicated docker-images/glm53-jovian-r18 dir + glm53-jovian-r18 tag; r12 dir restored to r17 pins; recipes renamed to r18 lanes
|
пре 1 месец |
|
Alexander Korolev
|
6327352f09
untrack regenerated data/benchmarks.sqlite (gitignore)
|
пре 1 месец |
|
Alexander Korolev
|
fe070bebdd
night shift: r18 built + full lane matrix (DCP4 161/176 best; eugr/DFlash2/dsv4fv-eugr documented)
|
пре 1 месец |
|
Alexander Korolev
|
4e6b20c673
dflash2 lane: drafter kv_cache_dtype auto (fp8 KV invalid on SM121 outside B12X)
|
пре 1 месец |
|
Alexander Korolev
|
a8ed91b8b5
dflash2 lane: pin drafter attention_backend FLASH_ATTN (auto-select poisoned by global B12X)
|
пре 1 месец |
|
Alexander Korolev
|
3976ebe72f
dflash2 lane: drop drafter B12X (kv_cache_dtype constraint; main keeps B12X)
|
пре 1 месец |
|
Alexander Korolev
|
bddf8d552e
add DFlash2 k=7 drafter lane (r18 img, A/B vs MTP)
|
пре 1 месец |
|
Alexander Korolev
|
fb4692fdab
dcp4 probe: add --cp-kv-cache-interleave-size 4 (GLM C4 DCP requirement)
|
пре 1 месец |
|
Alexander Korolev
|
2191c7018b
fix 4x recipes: tensor_parallel_size 4 (world size must cover nnodes)
|
пре 1 месец |
|
Alexander Korolev
|
b2135bab66
add r18 4x probe variants: DCP4 (TP4+decode-CP4) and nvfp4_ds_mla KV
|
пре 1 месец |
|
Alexander Korolev
|
eb8646ed43
add r17abi-4x probe recipe (4-node r18 cells: DCP4/RoCE/#646/nvfp4 KV)
|
пре 1 месец |
