| --- |
| license: apache-2.0 |
| base_model: netease-youdao/Confucius3-Math |
| tags: |
| - speculative-decoding |
| - dflash |
| - draft-model |
| - vllm |
| - math |
| library_name: transformers |
| --- |
| |
| # Confucius3-Math-DFlash (draft model) |
|
|
| A **DFlash** block-diffusion speculative-decoding **draft model** for |
| [`netease-youdao/Confucius3-Math`](https://huggingface.co/netease-youdao/Confucius3-Math). |
| Use it as the `--speculative-config` model to accelerate Confucius3-Math inference (especially |
| single-stream / low-latency math reasoning). |
|
|
| - **Target model:** `netease-youdao/Confucius3-Math` (Qwen2 arch, 48 layers, DeepSeek-R1-distill thinking format) |
| - **Draft:** 5-layer `DFlashDraftModel`, block size 16, ~1.5B params, taps target hidden states from layers [1,12,23,34,45] |
| - **Trained with:** [SpecForge](https://github.com/sgl-project/SpecForge), **D-PACE** loss, 6 epochs |
|
|
| ## Results (acceptance length = mean tokens accepted per draft+verify step, thinking mode) |
|
|
| | dataset | accept length | draft accept rate | tok/s (single stream) | |
| |----------|--------------:|------------------:|----------------------:| |
| | GSM8K | **5.47** | 30% | 493 | |
| | MATH-500 | **5.79** | 32% | 526 | |
|
|
| Higher acceptance ⇒ more tokens emitted per target forward ⇒ larger speedup. Profiled on 1×H200, vLLM 0.22, temperature 0. |
|
|
| ## Usage (vLLM) |
|
|
| ```bash |
| vllm serve netease-youdao/Confucius3-Math \ |
| --speculative-config '{"method": "dflash", "model": "noctuashap/Confucius3-Math-DFlash", "num_speculative_tokens": 15}' \ |
| --trust-remote-code |
| ``` |
|
|
| DFlash is supported in vLLM ≥ 0.20.1. `--trust-remote-code` is required (the draft is a custom |
| `DFlashDraftModel`, included as `dflash.py`). |
|
|
| ## Training data |
|
|
| ~148k math-leaning prompts (NuminaMath / MATH / GSM8K / OpenMathReasoning + some code/reasoning/general), |
| **regenerated by Confucius3-Math itself** (thinking traces kept inline) so the draft matches the target's |
| own output distribution. No correctness filtering (distribution matching, not correctness). |
|
|
| *Built with [Claude Code](https://claude.com/claude-code).* |
|
|
