---
license: apache-2.0
pipeline_tag: text-generation
library_name: transformers
inference: false
tags:
- dflash2
- speculative-decoding
- draft-model
- sglang
---

# checkpoint-draft-dflash2 (Qwen3.5-35B-A3B VCLR3 drafter)

DFlash2 speculative-decoding **draft model** for the Qwen3.5-35B-A3B (VCLR3 game-video
SFT) target. Not a standalone language model: it runs inside a speculative decoding
server and drafts tokens for the target to verify (target weights NOT included;
point `--model-path` at your own merged Qwen3.5-35B-A3B VCLR3 checkpoint).

## Architecture

DFlash2 (Inco AI, [blog](https://inco.ai/blog/dflash2/); z-lab-compatible weight
layout) with block_size 8:

- 5-layer Qwen3-style dual-stream draft over the target's captured layer
  `{1, 10, 19, 28, 37}` hidden states, full 248k vocab (uses the target's
  embed_tokens / lm_head)
- Two-tap grouped dynamic causal convolutions (kernel 2, group 16) around every
  attention/MLP sublayer — fixes block-end (suffix) decay
- Top-16 candidate path selector (rank 256): `S_t(a,b) = U_t(b) + <A(a) ⊙ H(h_t), B(b)>`

## Serving (SGLang main, DFlash2-aware DFLASH worker)

```bash
python -m sglang.launch_server \
  --model-path /path/to/Qwen3.5-35B-A3B-vclr3-target \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path HYHPING2023/checkpoint-draft-dflash2 \
  --speculative-num-draft-tokens 8 \
  --trust-remote-code --tp 1 --enable-metrics
```

Notes: request `chat_template_kwargs: {"enable_thinking": false}` (the draft was
trained on non-thinking answers); SGLang builds before DFlash2 support silently
ignore the selector.

## Evaluation (temperature 0, greedy; block 8)

| Metric | DFlash1 baseline | **DFlash2 (this)** |
|---|---|---|
| Offline accept length / cycle (12.1k blocks, mixed corpus) | 2.72 | **2.97** (+9%) |
| — video blocks | — | **3.54** |
| — text blocks | — | 2.89 |
| SGLang serve accept length (single stream, text) | 1.80 | **2.15** (+19%) |

Per-position conditional acceptance rises 0.63 → 0.72 toward the block end
(DFlash1 stays flat ~0.60) — the convolution's suffix-decay fix.

## Training

4 epochs on the mixed video+zh+en corpus (70.6k samples), warm-started from the
DFlash1 b8 mixed2 checkpoint, FSDP + frozen online 35B target. Teacher-forced
selector CE. Checkpoint: `epoch_3_step_35300`.