 Alexander Korolev
|
559e8d4db9
glm53-mxfp4-fp8-4x-r39: DSpark speculator attempt recorded — dies in the draft KV builder on r39 (needs vLLM 0.32)
|
19 小時之前 |
|
Alexander Korolev
|
35de736d83
glm53-mxfp4-fp8-4x-r39: 196K window hung the host on a 150K prefill — back to 131072 / 6 GiB KV / 2048 chunks
|
20 小時之前 |
|
Alexander Korolev
|
baabab7674
glm53-mxfp4-fp8-4x-r39: DCP4 rejected by every MLA backend ('DCP not supported'); window now 196608 at a 10 GiB KV pin
|
22 小時之前 |
|
Alexander Korolev
|
866d05a06c
glm53-mxfp4-fp8-4x-r39: MTP k=5 measured — same throughput as k=3 (33-37 c1, 47.9 c2), acceptance 4.4/5
|
23 小時之前 |
|
Alexander Korolev
|
8efe2f3d82
glm53-mxfp4-fp8-4x-r39: MTP k=3 breaks the decode ceiling — 19.0 -> 33-35 t/s c1, 47.6 t/s c2
|
23 小時之前 |
|
Alexander Korolev
|
75276e8670
glm53-mxfp4-fp8-4x-karmic: A/B vs r39 — native b12x MoE experts buy +1% (19.2 vs 19.0 t/s)
|
1 天之前 |
|
Alexander Korolev
|
1c4d95db9d
glm53-mxfp4-fp8-4x-r39: measured on v2.8.0 — 88/100 (170/194), Hard Mode 46/56, 1 safety-critical, 8.8 s median turn
|
1 天之前 |
|
Alexander Korolev
|
76abd98c9c
glm53-mxfp4-fp8-4x-r39: GLM-5.3 (full) RedHat MXFP4xFP8 boots at TP4 on four Sparks
|
1 天之前 |
|
Alexander Korolev
|
fc210f50bd
glm53-kindling-2x: reproducibility run on v2.8.0 — 95/100 (185/194), Hard Mode 52/56 (93%), zero safety-critical, 3.6 s median turn
|
1 天之前 |
|
Alexander Korolev
|
817cb85d66
glm53-mxfp4-fp8-4x: GLM-5.3 (full, 78 layers) RedHatAI MXFP4xFP8 on 4 Sparks at TP4
|
1 天之前 |
|
Alexander Korolev
|
337f442e74
dsv41-flash-1017-2x-tensorfold: session cache A/B — same 89/100, median turn 6.7s -> 4.2s (responsiveness 23 -> 38)
|
1 天之前 |
|
Alexander Korolev
|
994a1ffa05
dsv41-flash-1017-2x-tensorfold: staging script moves into the recipe (scripts/prepare.sh) + TF_DSV41_SESSIONS=1 experiment
|
1 天之前 |
|
Alexander Korolev
|
7e9da804f3
glm53-kindling-2x: re-benchmarked on harness v2.8.0 — 94/100 (183/194), Hard Mode 50/56 (89%), zero safety-critical failures
|
2 天之前 |
|
Alexander Korolev
|
b1f054231b
beam lane: 384K context (13.6x concurrency) + needle-in-haystack at 131K and 327K tokens (both correct, 60s/175s TTFT)
|
2 天之前 |
|
Alexander Korolev
|
502080b104
beam lane: depth sweep (31-35 tok/s to 128K) + the second/third quality runs (92/100, Hard Mode 95%)
|
2 天之前 |
|
Alexander Korolev
|
3248479cdf
beam lane: real-workload TTFT measured (0.41/0.51 s with prefix cache)
|
2 天之前 |
|
Alexander Korolev
|
a3c43d9bb1
beam lane: spec-k7 quality/perf reports + real-work operational contract
|
2 天之前 |
|
Alexander Korolev
|
d135e829d3
beam lane: DFlash2 speculation works with FLASHINFER — +40% decode, responsiveness 34->56
|
2 天之前 |
|
Alexander Korolev
|
fe606b76bb
dsv41-flash-1017-4x-sglang: land the 2026-10-18 A/B reports and the 24/24 concurrency bump
|
2 天之前 |
|
Alexander Korolev
|
88b5f8ddc1
beam-1-0-nvfp4-4x: new lane — Reflection Beam-1.0-501B-A23B NVFP4 on 4 Sparks, 90/100 quality (fleet best)
|
2 天之前 |
|
Alexander Korolev
|
ec17b00796
dsv41-flash-1017-2x-tensorfold: ship it as an sm121 recipe
|
4 天之前 |
|
Alexander Korolev
|
dbe6e49aca
dsv41-flash-1017-2x-tensorfold: measure the upstream TensorFold dual-Spark lane — 89/100 quality (best 1017 lane) but responsiveness 23 (6.7s turn) vs the 4x SGLang lane's 87/75; decode +20-30%
|
4 天之前 |
|
Alexander Korolev
|
fe0d38641b
dsv41-flash-1017-4x-sglang: 1017 perf A/B — ep-size 2 (+9% decode), context 1M; document the engram MAX_IDS <-> chunked-prefill coupling
|
5 天之前 |
|
Alexander Korolev
|
fc1850811d
dsv41-sglang-1017: version the image tag v1 (not :local) — trackable and rollback-able
|
5 天之前 |
|
Alexander Korolev
|
aa4d33f1f2
dsv41-flash-1017-4x-sglang: bake the 1017 encoder patch into a thin wrapper image (docker-images/dsv41-sglang-1017, glm52 pattern: FROM aiden production-1.7.1 by digest + patch at build) — replaces the [mods]+manual-scp approach; recipe points at dsv41-sglang-1017:local
|
5 天之前 |
|
Alexander Korolev
|
fc1fdcaa50
dsv41-flash-1017-4x-sglang: [mods] patch aligning the 1.7.1 image's DSV4.1 encoder with the 1017 reference protocol — REASONING_EFFORT_MAPPINGS 25/50/75 → 50/75/100 (xhigh dropped); idempotent anchor-based patcher mounted per node; upstream has no fix yet (Aiden still 1.7.1, eugr repo unchanged, sglang upstream silent)
|
5 天之前 |
|
Alexander Korolev
|
ef684e8ae7
glm53-fp4local-2x: served-model-name glm53-fp4-local → glm53 (the fleet-wide alias convention)
|
5 天之前 |
|
Alexander Korolev
|
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 |
|
Alexander Korolev
|
a3004e8e0b
dflash A/B: attention_backend B12X→FLASH_ATTN (B12X rejects the draft's block_size; trying the standard backend)
|
5 天之前 |
|
Alexander Korolev
|
5939b55b07
dflash A/B: attention_backend TRITON_ATTN→B12X — on the eugr build only B12X accepts fp8 kv-cache (the main engine's backend); kindling's TRITON_ATTN works on our overlay image, not here
|
5 天之前 |
