MiniCPM5-2B-DSpark
MiniCPM5-2B-DSpark is a DSpark
draft checkpoint trained for exact pairing with MiniCPM5-2B and its
tokenizer.
Model Specification
| Property | Value |
|---|---|
| Target model | MiniCPM5-2B |
| Draft layers | 5 |
| Draft parameters | 323,776,001 |
| Draft tokens per forward pass | 7 |
| Target hidden-state layers | [1, 10, 20, 30, 39] |
| Checkpoint precision | BF16 |
Training Recipe
Training responses were generated by MiniCPM5-2B from a mixture of general-domain, mathematics, and code prompts.
| Setting | Value |
|---|---|
| Training sequences | 1,959,525 |
| Training tokens | 7,054,154,509 |
| Epochs | 6 |
| Maximum sequence length | 12,288 |
| Optimizer | AdamW |
| Objective | CE + L1 + confidence loss |
Evaluation
Acceptance length is defined as the total number of completion tokens divided
by the total number of speculative verification steps. Evaluation uses natural
EOS termination and max_new_tokens=4096.
| Domain | T=0 | T=1.0 |
|---|---|---|
| Math | 6.0496 | 4.6050 |
| Code | 6.1106 | 4.4381 |
| General | 4.1585 | 3.0997 |
| Aggregate | 5.5174 | 4.0514 |
Usage with SGLang
python -m sglang.launch_server \
--model-path <target-model-path> \
--trust-remote-code \
--speculative-algorithm DSPARK \
--speculative-draft-model-path <draft-model-path> \
--speculative-dspark-block-size 7
Limitations and Disclaimer
This model has no autonomous intent or legal personhood; its outputs are text generated from statistical patterns and may be inaccurate, biased, or offensive, and may be manipulated by carefully crafted prompts ("jailbreaks") into producing unintended content. Its responses on sensitive topics such as politics, health, finance, and law are not reviewed by experts and should not be treated as professional advice.
This model is provided "AS IS", without warranty of any kind, express or implied, and the developers are not liable for any damages arising from its use. Users must employ the model only for lawful, compliant, and ethical purposes, configure their own safeguards, and label AI-generated content where required; deliberate jailbreaking, injection attacks, or inducing harmful output is prohibited, and any such testing is at the user's own risk.
License
This repository and MiniCPM model weights are released under the Apache-2.0 License.
Citation
Please cite our paper if you find our work valuable:
@article{minicpm4,
title={Minicpm4: Ultra-efficient llms on end devices},
author={MiniCPM, Team},
journal={arXiv preprint arXiv:2506.07900},
year={2025}
}