cm00cm/qwen3-8b-dflash-draft-onpolicy-perfectblend

Qwen3-8B DFlash speculative-decoding draft (on-policy)

A DFlash speculative-decoding draft head for Qwen/Qwen3-8B, trained with SpecForge.

What it is

• Draft architecture: DFlashDraftModel (5 draft layers, hidden 4096), block size 16.
• Consumes 5 aux hidden states from the frozen target at layers [1, 9, 17, 25, 33] (fc input dim = 5 x 4096 = 20480); frozen shared embedding + LM head.

Training

• Data: on-policy — perfectblend prompts with responses regenerated by Qwen/Qwen3-8B itself (greedy, temperature 0, no-think), ~50k conversations.
• 6 epochs, num_anchors 512, block_size 16, lr 6e-4, cosine (0.04 warmup), loss-decay gamma 7, max_length 3072, train_dflash.py (hf backend).
• Final training block-accuracy ~0.6.
• Why on-policy matters: the DFlash training accuracy is scored against the target's own greedy continuations; off-policy (human) text caps the achievable accuracy at ~0.35, while on-policy data lifts it past 0.6.

Evaluation — serving accept length (tau)

Measured via sglang avg_spec_accept_length on gsm8k (100 questions, greedy, no-thinking, block size 16):

draft	gsm8k tau
this checkpoint (on-policy)	5.75
off-policy baseline	2.30
DFlash paper (Qwen3-8B)	6.54
EAGLE-3 (paper baseline)	2.96

Usage (sglang)

python -m sglang.launch_server \
  --model-path Qwen/Qwen3-8B --trust-remote-code \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path cm00cm/qwen3-8b-dflash-draft-onpolicy-perfectblend \
  --speculative-dflash-block-size 16