vllm-speculative-decoding SKILL.md skillA
vllm-speculative-decoding from air-gapped/skills is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add air-gapped/skills/vllm-speculative-decoding@git:20261018.4183bb7Pins git:20261018.4183bb7 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- air-gapped/skills
- Path
.claude/skills/vllm-speculative-decoding/SKILL.md- Versions recorded
- 11
- Latest version
- git:20261018.4183bb7, committed Oct 18, 2026
- First version recorded
- Apr 30, 2026
- Size
- 17,508 bytes
- Repository stars
- 6
- License
- MIT
Pick, configure, tune, monitor vLLM speculative decoding in production. Thirteen SpeculativeMethod options (ngram, ngram_gpu, medusa, mlp_speculator, draft_model, suffix, custom_class, eagle, eagle3, dflash, dspark, mtp, extract_hidden_states), `--speculative-config` JSON schema, which methods pair with which target model family, Prometheus acceptance metric surface, version gates (v0.11.1 EAGLE-3 preamble fix, v0.16 parallel drafting, v0.18 ngram_gpu, v0.19 dflash and zero-bubble)…
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# vLLM speculative decoding — operator skill For production vLLM operators deciding which speculative method fits a given model + workload, configuring it correctly, wiring the acceptance metrics into their dashboards, and diagnosing why a deployment isn't seeing the expected speedup. ## When spec-dec wins, when it loses Spec-dec amortises memory-bandwidth-bound decode by letting a cheap proposer guess k tokens that a single target-model forward can verify in parallel. - **Wins at low concurrency (BS 1–8)**: decode is bandwidth-bound, k verified tokens per target step → 1.5–3× throughput on a well-matched target+drafter. EAGLE-3 on Llama-3.1-8B: +32% TPOT over EAGLE-1 at BS=4 (vLLM v0.11.1+, PR #25916). DFlash on Qwen3-8B: 3.5× at BS=1, 1.6× at BS=32 (PR #36847). - **Hurts at high concurrency (BS ≥ 32)**: target becomes compute-bound, draft latency is no longer hidden, rejections waste GPU time. Red Hat, Snowflake and the P-EAGLE author all report this. Gate spec-dec to the low-concurrency tier of a disagg or multi-tenant deployment, or disable above a threshold. - **Domain mismatch sinks acceptance**: stock EAGLE-3 checkpoints are chat-tuned. …
Read the whole file at its exact version.
How to install
mdr add air-gapped/skills/vllm-speculative-decoding@sha256:df7e356f7bbc9c75The line at the top of this page pins the label git:20261018.4183bb7, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_7whuo5h5e3gjocra)
2 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 11 versions of this file, between Apr 30, 2026 and Oct 18, 2026, about one every 17 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20261018.4183bb7 latest | 2026-10-18 | 4183bb7 | 17,508 B | A | view · diff |
| git:20261018.25c40a4 | 2026-10-18 | 25c40a4 | 17,511 B | A | view · diff |
| git:20261018.a9c7736 | 2026-10-18 | a9c7736 | 17,615 B | A | view · diff |
| git:20261018.3079892 | 2026-10-18 | 3079892 | 18,173 B | A | view · diff |
| git:20261018.e9c2342 | 2026-10-18 | e9c2342 | 18,671 B | A | view · diff |
| git:20260923.466430c | 2026-09-23 | 466430c | 23,008 B | A | view · diff |
| git:20260915.69fc432 | 2026-09-15 | 69fc432 | 21,005 B | A | view · diff |
| git:20260811.3f40ad0 | 2026-08-11 | 3f40ad0 | 20,252 B | A | view · diff |
| git:20260721.b755a49 | 2026-07-21 | b755a49 | 18,070 B | A | view · diff |
| git:20260529.9f668b9 | 2026-05-29 | 9f668b9 | 15,535 B | A | view · diff |
| git:20260430.9141074 | 2026-04-30 | 9141074 | 15,571 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (17508 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
air-gapped/skills · 6 stars · license MIT · pushed 2026-11-04 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_7whuo5h5e3gjocra GET https://markdownregistry.com/api/v1/resolve?ref=air-gapped/skills/vllm-speculative-decoding GET https://markdownregistry.com/api/v1/blob/df7e356f7bbc9c7544b23130517cac3ca6d52aaa1a6b4974f695ff50ac00535c
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.