Unlike MTP, ngram-mod needs no draft context: common/speculative.cpp adds the
ngram types unconditionally while the model-backed drafters are gated on
ctx_dft != nullptr, so 'draft-dflash,ngram-mod' is structurally allowed (and the
config vector's fixed order -- ngram first, then the drafters -- makes CLI order
irrelevant).
The harness grows the two things this question needs: --configs, and an ECHO
shape. ngram-mod can only ever match text that is present in the prompt, so
measuring it on prose calls it useless and measuring DFlash on echo calls it
wasteful; both shapes now run, and the greedy text of each is compared across
configurations because speculation may not change answers.
Four configurations of the same binary, identical prompts at temperature 0, the
unit stopped between runs, on the deployed payload:
none 17.7 tok/s card1 26.07 GB
draft-mtp n=3 32.4 tok/s 58 % accepted card1 29.35 GB
draft-mtp n=7 26.7 tok/s 33 % accepted card1 29.65 GB
draft-dflash 7 43.5 tok/s 41 % accepted card1 29.42 GB
DFlash2 is worth 2.46x over autoregressive and 34 % over the best MTP, and all
four produced IDENTICAL greedy text, so the 'lossless verification' claim in the
runbook is now a check rather than an assumption. MTP is not the cheaper option
either: its head loads ~3.3 GB against the drafter's 2.06 GB file, and deeper
drafting loses (n=7 accepts a third of its proposals, n=3 more than half).
The two CANNOT be chained, and the failure is structural rather than tunable:
'spec-type draft-mtp,draft-dflash' dies at load with 'context type MTP requested
but model doesn't contain MTP layers', because a chain has ONE draft context and
MTP builds against whatever model that context holds -- here the DFlash2 file.
The serve script therefore passes --spec-draft-model in the dflash branch ONLY,
and the seam now asserts exactly that (the wiring whose violation is this
error), plus that the deployment stays on the measured winner.
scripts/bench-spec-ab.py makes the comparison reproducible: the benchmarking
policy wants the runtime's own harness for this, and the numbers above were
otherwise unreproducible one-offs.
Context: 262144 was measured before it was deployed -- 24.72 GB on GPU 0 and
29.14 GB on GPU 1 of 34.21 GB each, ~5 GB of headroom on the busier card, and a
140,069-token prompt prefilled through the single slot in 134 s to prove the
window is real. The 114688 wall was the vLLM predecessor's arithmetic, not this
runtime's.
Throughput, measured through the service: 1.0-1.4k tok/s prefill from 4k to
140k, 43.8 tok/s decode on a 512-token generation, 41-71 % DFlash2 acceptance,
and the prefix cache re-prefilling 4 of 998 tokens on a repeated prompt.
The signals move with the runtime, so the fleet's rules and dashboard are
re-cut rather than renamed: vllm: series do not exist any more, and two of the
old rules watched things llama.cpp does not export at all (KV-pool fill, p95
latency). Those are retired with the reason written down, and LlmSpecDecodeOff
takes one of the slots -- a lost drafter is the failure no client can see. The
observability seam now fails if any rule names a series the endpoint does not
export, which is what would have caught the stale vllm: expressions.
omp models.yml (outside the repo) is updated in the same pass: contextWindow
262144, and the full reasoning-effort map, because Qwen3.8's template answers
500 on 'high' -- verified against the live endpoint, low/medium/xhigh answer
200. Subagents run at high.
The vllm-radiance replica pair could not use both cards for one model: TP=2
needs a cross-card collective and the second R9700 cannot dispatch one. The
same wall blocks vLLM's pipeline parallelism, whose stage handoff is an RCCL
send/recv of the hidden state (distributed/parallel_state.py) -- a fact worth
writing down, because the obvious reading of the ask ('use PP to dodge the
atomics problem') is exactly backwards.
llama.cpp's layer split needs no collective at all: layers are distributed and
the boundary activation is a per-backend copy, once per forward pass. One
process, both cards, one model. The runtime is mainline llama.cpp at fb27a525
(the ref is pinned by DFlash2 AND qwen35 MTP; the ROCmFPX fork this host had
before has neither), built for gfx1201 against the private ROCm 7.14 toolchain
the earlier build already used, and bundled next to the binary.
Server: ajgazin Swift-Qwen3.8-27B-Uncensored UD-Q8_K_XL with the model's MTP
head in the file, plus the z-lab DFlash2 Q8_0 block-diffusion drafter at depth
7 (its block is 8, so 7 is the ceiling llama.cpp allows). Text only: vision
stays on frank-small.
Retired into the trash rather than deleted: the three vLLM quadlets, the
radiance entrypoint/healthcheck, the routing proxy and its tests, and the two
bench scripts that drove 'vllm bench serve' inside the retired image. The
deploy handler now stops them, waits for 7770 to be released, installs the
ROOT unit and refuses to start anything when the runtime or payload is
missing. The planner maps the scripts/ trio to frank-medium only, and the
scoping seam asserts the retired service dir is gone rather than merely
unmapped.
Rootless Quadlet on dankserver serving
just1moremodel/Qwen3.8-27B-Uncensored-MXFP4-awq (the radiance stack's
designated alternate target; ships its fp8 MTP head, no rewrite step) with
the DFlash2-FP8 drafter at SPEC=7 on the 2x R9700s (TP=2, host network,
:7770, alias frank-medium unchanged).
- dankserver/frank-medium/: quadlet unit, container entrypoint (fork of
serve-mxfp4.sh's patch block at pinned clone commit 97353293), healthcheck
- dispatcher retires the root llama.cpp unit, re-tags the registry image to
the localhost alias, installs the quadlet, waits on /health
- planner maps dankserver/frank-medium/* to frank-medium; scripts-era
mappings removed
- ROCmFPX-era env/serve/unit/download scripts staged in scripts/.AGENTTRASH
with .rocmfpx suffix; seams updated (frank_medium_quadlet.bats new)
The RTX 3090 left Dankserver; two AMD Radeon AI PRO R9700 cards (Navi 48,
gfx1201, 32,624 MiB each) replaced it. FreeToken/CUDA cannot run there at all,
so the `frank-small` service is retired and `frank-medium` takes the endpoint.
New service: charlie12345/ROCmFPX llama.cpp (built for gfx1201 against the
fork's self-contained ROCm 7.14 toolchain, so the host's stale /opt/rocm-6.3.3
never participates) serving kingjones777/Qwen3.8-27B-ROCmFPX-Q8_0-AGENT-GGUF --
a Q8_0_ROCMFPX_AGENT (ftype 115) checkpoint that stock llama.cpp cannot load.
Placeholder config: --split-mode tensor across both cards, 131072 context,
1 slot, f16 KV, MTP draft head on ROCm0.
deploy-remote.sh grows retire_root_llm_small(): the CUDA-era unit is stopped,
disabled and moved to ~/.AGENTTRASH/20260915-frank-small-retire/host-units
before frank-medium starts, because `systemctl disable` has been observed to
delete the unit file itself on this host (2026-08-26 incident).
The retired frank-small repo files (env, unit, serve script) move to
scripts/.AGENTTRASH/, and the planner, dispatcher, service allow-list and the
deployment test seams now name frank-medium.
Follow-up (same session, measured): llama-bench on the host with the correct
multi-device syntax (-dev ROCm0/ROCm1) overturned the tensor choice:
single R9700 pp4096 1160.9 / tg128 19.42 (d32768: 512 / 18.4)
layer 2x pp4096 1991.2 / tg128 18.76 (d32768: 859 / 17.6)
tensor 2x pp4096 271.8 / tg128 26.26 (d32768: 229 / 25.1)
This build has no RCCL, so tensor mode falls back to a "meta-backend
butterfly" allreduce and prefill collapses 7x. Served path with MTP:
layer 35.7-46.2 tok/s vs tensor 28.7. SPLIT_MODE=layer, draft pinned to
ROCm1 (the card holding the last target layers). Served prefix cache with
the draft head verified live: turn 2 reprocessed 22 tokens instead of 949.
The RTX 3090 left Dankserver; two AMD Radeon AI PRO R9700 cards (Navi 48,
gfx1201, 32,624 MiB each) replaced it. FreeToken/CUDA cannot run there at all,
so the `frank-small` service is retired and `frank-medium` takes the endpoint.
New service: charlie12345/ROCmFPX llama.cpp (built for gfx1201 against the
fork's self-contained ROCm 7.14 toolchain, so the host's stale /opt/rocm-6.3.3
never participates) serving kingjones777/Qwen3.8-27B-ROCmFPX-Q8_0-AGENT-GGUF --
a Q8_0_ROCMFPX_AGENT (ftype 115) checkpoint that stock llama.cpp cannot load.
Placeholder config: --split-mode tensor across both cards, 131072 context,
1 slot, f16 KV, MTP draft head on ROCm0.
deploy-remote.sh grows retire_root_llm_small(): the CUDA-era unit is stopped,
disabled and moved to ~/.AGENTTRASH/20260915-frank-small-retire/host-units
before frank-medium starts, because `systemctl disable` has been observed to
delete the unit file itself on this host (2026-08-26 incident).
The retired frank-small repo files (env, unit, serve script) move to
scripts/.AGENTTRASH/, and the planner, dispatcher, service allow-list and the
deployment test seams now name frank-medium.
Frank