بنقرة واحدة
serving-llms-dspark
DSpark: DeepSeek speculative decoding — 60-85% faster inference, identical outputs.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
القائمة
DSpark: DeepSeek speculative decoding — 60-85% faster inference, identical outputs.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Route a game brief to a gated, engine-agnostic pipeline.
Route orchestrated work across the installed seats.
Talk like Breadstick Ricky from Breadstick Ricky & The Boss — the default "Ricky" register for Muse. Load when Jeremiah wants Muse's normal conversational personality (excitable, confident, colorful, high-energy) — while staying honest and competent.
Run Karpathy's autoresearch loop — autonomous 5-minute pretraining experiments on the vendored nanochat-style engine — inside a disposable workspace, governed by muse's cost ceiling, VRAM feasibility gate, and benchmark gate, with every experiment on the flywheel. The winning train.py surfaces as an owner-gated proposal; val_bpb LOWER is better; nothing is ever applied without the owner.
Turn a prompt packet into a gated, ledgered Unreal Engine 5 offscreen render.
Poll RSS, JSON APIs, and GitHub with watermark dedup.
| name | serving-llms-dspark |
| description | DSpark: DeepSeek speculative decoding — 60-85% faster inference, identical outputs. |
| version | 1.0.0 |
| author | Nous Research |
| license | MIT |
| dependencies | [] |
| platforms | ["linux","macos"] |
| metadata | {"hermes":{"tags":["DSpark","DeepSpec","Speculative Decoding","Draft Model","Inference Acceleration","llama.cpp","vLLM","DeepSeek","Efficiency"]}} |
Use when local or self-hosted inference feels slow and you want 60-85% faster generation without changing outputs, models, or hardware. DSpark (DeepSeek + Peking University, MIT, arXiv:2606.19348) attaches a small draft module that proposes several tokens per step; the target model verifies them in one forward pass. Output quality is identical — the target model still decides every token.
Muse's plan builder lives at hermes_cli/local_models/dspark.py (runtime: dspark in the server-adapter layer): it picks llama.cpp or vLLM, resolves the right draft checkpoint, and emits the exact launch command.
Zero-install path — DeepSeek API. api.deepseek.com serves V4-Flash/V4-Pro with DSpark server-side since 2026-06-27. Routing to the deepseek provider already gets the speedup.
llama.cpp (CPU / Metal / consumer GPU):
# 1. Grab a DSpark draft GGUF (community conversion of the official draft)
huggingface-cli download ankk98/dspark-qwen3-8b-block7-Q4_K_M-GGUF
# 2. Serve target + draft
llama-server -m qwen3-8b-instruct-Q4_K_M.gguf --port 8080 \
--spec-draft-model dspark-qwen3-8b-block7-Q4_K_M.gguf \
--spec-draft-n-max 16 --spec-draft-n-min 1 --spec-draft-p-min 0.75
vLLM (GPU hosts):
vllm serve Qwen/Qwen3-8B \
--speculative-config '{"model": "mgoin/Qwen3-8B-speculator.dspark", "num_speculative_tokens": 5}'
Self-contained V4 checkpoints (draft module already attached; server-class GPUs):
vllm serve deepseek-ai/DeepSeek-V4-Flash-DSpark
Point muse at the endpoint as usual (provider: custom, base_url: http://127.0.0.1:8080/v1). Speculative decoding is invisible to clients.
| Target | Official draft (safetensors) | GGUF (llama.cpp) | speculators (vLLM) |
|---|---|---|---|
| Qwen3-4B | deepseek-ai/dspark_qwen3_4b_block7 | — | — |
| Qwen3-8B | deepseek-ai/dspark_qwen3_8b_block7 | ankk98/dspark-qwen3-8b-block7-Q4_K_M-GGUF | mgoin/Qwen3-8B-speculator.dspark |
| Qwen3-14B | deepseek-ai/dspark_qwen3_14b_block7 | — | — |
| Gemma4-12B-it | deepseek-ai/dspark_gemma4_12b_block7 | ankk98/dspark-gemma4-12b-block7-Q4_0-GGUF | — |
| DeepSeek-V4-Flash | self-contained: deepseek-ai/DeepSeek-V4-Flash-DSpark | — | (serves directly) |
| DeepSeek-V4-Pro | self-contained: deepseek-ai/DeepSeek-V4-Pro-DSpark | — | (serves directly) |
For any other target, train a custom draft with DeepSpec — data prep → multi-GPU draft training → 9-benchmark eval (defaults assume 1 node × 8 GPUs).
num_speculative_tokens / draft window: DeepSeek's deployed config drafts 5 ("DSpark-5"). llama.cpp's --spec-draft-n-max 16 lets acceptance run longer on easy text.Use DSpark when: the model you already run has an official or community draft (see catalog), or you can afford a DeepSpec training run for a custom one.
Use alternatives instead:
hermes_cli/local_models/dspark.py, tests at tests/hermes_cli/test_dspark_runtime.py