--- base_model: jialinyyzz/humanizer base_model_relation: finetune license: apache-2.0 language: - en - zh tags: - draftfit - speculative-decoding - dflash - humanizer - transformers - safetensors - text-rewriting - rewriting - paraphrase - style-transfer - gemma4 --- # Draftfit · Humanizer 12B DFlash A validation checkpoint for **[Draftfit](https://github.com/JayYun98/Draftfit)** — train DFlash-family drafts jointly, convert compatible families, and continue training existing drafts. This **DFlash draft** accelerates [Humanizer](https://huggingface.co/jialinyyzz/humanizer); it requires the target model. Target / teacher: [Humanizer](https://huggingface.co/jialinyyzz/humanizer). Initial draft weights: [Gemma4 DFlash](https://huggingface.co/z-lab/gemma4-12B-it-DFlash). This is a target-specific speculative draft, not a standalone Humanizer fine-tune. ## Performance | Model | tok/s | Acceptance | vs. original | |---|---:|---:|---:| | Target only | 21.71 | — | — | | Original DFlash | 23.43 | 30.1% | reference | | Draftfit DSpark | 33.49 | 59.9% | +42.9% | | Draftfit DFlash | 32.62 | 59.9% | +39.2% | | Draftfit DFlash2 | 31.23 | 55.6% | +33.3% | ### By domain · tok/s | Model | General prose | Blog | Email | Report | Essay | |---|---:|---:|---:|---:|---:| | Target only | 21.52 | 21.51 | 21.44 | 22.00 | 21.79 | | Original DFlash | 25.22 | 24.38 | 23.75 | 22.01 | 22.98 | | Draftfit DSpark | 33.59 | 39.29 | 33.54 | 31.77 | 30.93 | | Draftfit DFlash | 32.92 | 38.81 | 32.80 | 29.75 | 29.13 | | Draftfit DFlash2 | 33.16 | 35.81 | 30.46 | 31.14 | 29.48 | Apple M5 Pro · MLX · mixed 4/8-bit target with FP32 floating computation · 4-bit drafts. Median throughput includes prefill and verification; acceptance is accepted/proposed draft tokens. Same 10 held-out inputs ×2 repeats per model, greedy, up to 64 output tokens, draft cap 3, lookup off. **All 160 requests matched target-only tokens**, including previous-checkpoint comparisons. Continued drafts improved median throughput by 2.6–4.9% over their previous checkpoints in this run; trial variability is substantial, so this is not a robust or long-form speedup claim. [Domain acceptance, previous checkpoints and details](evaluation.md). ### Acceptance by draft position | Model | Position 1 | Position 2 | Position 3 | |---|---:|---:|---:| | Original DFlash | 52.8% | 26.6% | 10.5% | | Draftfit DSpark | 80.6% | 59.4% | 39.3% | | Draftfit DFlash | 80.6% | 59.4% | 39.3% | | Draftfit DFlash2 | 74.0% | 57.1% | 35.5% | Same MLX runs above. Position k = accepted prefixes reaching k / rounds proposing position k. Earlier rejection also rejects later positions; these are not conditional token accuracies. [Counts and domain breakdown](evaluation.md#acceptance-by-draft-position). ## How to use **MLX — tested.** After [runtime setup](runtime-setup.md), pass prose to rewrite: ```bash hf download jayyun98/Humanizer-12B-Draftfit-DFlash run_mlx.py --local-dir ./humanizer-draft python ./humanizer-draft/run_mlx.py --text "The meeting has been moved to Friday." --check ``` **vLLM / SGLang — unverified for this checkpoint.** No CUDA serving or speed claim; use the tested MLX path. Basic precision showed occasional token differences; the supplied runner uses the tested FP32 setting. [Provenance](training.md) · [Apache-2.0](LICENSE) · [DFlash MIT](LICENSE-DFlash) · [Attribution](NOTICE).