DeepSpec

DeepSpec is a full-stack codebase for training and evaluating draft models for speculative decoding. It contains data preparation utilities, draft model implementations, training code, and evaluation scripts.

Environment

Install the Python dependencies:

python -m pip install -r requirements.txt

Data preparation additionally requires an inference engine to serve the target model when regenerating answers; see scripts/data/README.md for details.

Workflow

Run the stages in order — each stage's output feeds the next:

1. Data Preparation — download prompts and regenerate target answers with LightLLM or another compatible server.
2. Training — train a draft model while a frozen target computes hidden features online; precomputed caches remain optional.
3. Evaluation — measure speculative-decoding acceptance on benchmark tasks.

Data Preparation

See scripts/data/README.md for the step-by-step data pipeline:

1. download and split training data,
2. regenerate answers,
3. choose online target features (default) or prepare an optional target cache.

Training

online=1 train_data_path=train_datasets/qwen3_4b/perfectblend_train_regen.jsonl \
  bash scripts/train/train.sh

Online mode keeps the frozen target on each training GPU and computes its hidden states for each input batch. Only the draft is trained. This avoids dumping a full target-feature cache; it does not generate new answers during training. LightLLM serves the target during the earlier answer-regeneration stage. To use precomputed features, set online=0 and target_cache_dir instead.

train.sh launches train.py, which spawns one worker per visible GPU. Select the algorithm and target model by pointing config_path at one of the configs under config/ (e.g. config/dspark/dspark_qwen3_4b.py); see the script header for the full list of configs, how to override config_path / target_cache_dir, and how to use --opts to override individual config fields. Checkpoints are written to ~/checkpoints/<project_name>/<exp_name>/step_*.

Hardware: the default configs and scripts assume a single node with 8 GPUs. For fewer GPUs, reduce CUDA_VISIBLE_DEVICES.

Evaluation

bash scripts/eval/eval.sh

eval.sh runs eval.py against a trained draft checkpoint over the speculative-decoding benchmarks in eval_datasets/ (gsm8k, math500, aime25, humaneval, mbpp, livecodebench, mt-bench, alpaca, arena-hard-v2). Set:

• target_name_or_path — the target model the draft was trained against (e.g. Qwen/Qwen3-4B),
• draft_name_or_path — the draft checkpoint, e.g. ~/checkpoints/deepspec/dspark_block7_qwen3_4b/step_latest, or one of the Hugging Face repo IDs listed in Released Checkpoints.

Released Checkpoints

The checkpoints below are the ones used for Table 1 in the paper. Each checkpoint was trained on open-perfectblend data generated by its corresponding target model in non-thinking mode, and is the direct output of the corresponding training configuration under config/.

Algorithm	Qwen/Qwen3-4B	Qwen/Qwen3-8B	Qwen/Qwen3-14B	google/gemma-4-12B-it
Eagle3	deepseek-ai/eagle3_qwen3_4b_ttt7	deepseek-ai/eagle3_qwen3_8b_ttt7	deepseek-ai/eagle3_qwen3_14b_ttt7	deepseek-ai/eagle3_gemma4_12b_ttt7
DFlash	deepseek-ai/dflash_qwen3_4b_block7	deepseek-ai/dflash_qwen3_8b_block7	deepseek-ai/dflash_qwen3_14b_block7	deepseek-ai/dflash_gemma4_12b_block7
DSpark	deepseek-ai/dspark_qwen3_4b_block7	deepseek-ai/dspark_qwen3_8b_block7	deepseek-ai/dspark_qwen3_14b_block7	deepseek-ai/dspark_gemma4_12b_block7

Important

If you cite these results in a new paper, align your setup with the training settings in this repository; otherwise, the comparison is not meaningful. For domain-specific use, fine-tune the draft model again for better results, especially if the target model is expected to run in thinking mode.

Supported Algorithms

Currently, DeepSpec includes three draft models: DSpark, DFlash and Eagle3.

License

DeepSpec is released under the MIT License. It includes code adapted from third-party projects under their own licenses; see NOTICE for the full attribution.

Acknowledgements

DeepSpec builds on the ideas and code of several excellent open-source projects:

• SpecForge (Apache-2.0) — the overall training framework and Eagle3 implementation; portions of the Eagle3 modeling, loss, optimizer, attention, and evaluation code are adapted from it. Adapted files carry an in-file attribution comment, and the full notice is recorded in NOTICE.
• DFlash (MIT) — the DFlash draft-model design and training recipe.
• Qwen3 and Gemma — the target model families supported in this repo.

We thank the authors and maintainers of these projects. Contributions of new algorithms are welcome.

Citation

@misc{cheng2026dsparkconfidencescheduledspeculativedecoding,
      title={DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation}, 
      author={Xin Cheng and Xingkai Yu and Chenze Shao and Jiashi Li and Yunfan Xiong and Yi Qian and Jiaqi Zhu and Shirong Ma and Xiaokang Zhang and Jiasheng Ye and Qinyu Chen and Chengqi Deng and Jiping Yu and Damai Dai and Zhengyan Zhang and Yixuan Wei and Yixuan Tan and Wenkai Yang and Runxin Xu and Yu Wu and Zhean Xu and Xuanyu Wang and Muyang Chen and Rui Tian and Xiao Bi and Zhewen Hao and Shaoyuan Chen and Huanqi Cao and Wentao Zhang and Anyi Xu and Huishuai Zhang and Dongyan Zhao and Wenfeng Liang},
      year={2026},
      eprint={2607.05147},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2607.05147}, 
}