File size: 2,391 Bytes
31edbe3
 
 
c53d936
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
---

license: apache-2.0
---


# QwQ-Math-1.5B-Persona

## Introduction

QwQ-Math-1.5B-Persona is finetuned from Qwen2.5-Math-1.5B-Instruct on 1 million math persona data (see [this paper](https://arxiv.org/abs/2406.20094) for details about how to construct the data).

Currently QwQ-Math-1.5B-Persona is meant to serve as a draft model for losslessly accelerating the inference of QwQ-32B, but you may also use it as a standalone model.

## Quickstart

Here is a code snippet for using QwQ-Math-1.5B-Persona to accelerate the inference of QwQ 32B:

```python

from transformers import AutoModelForCausalLM, AutoTokenizer



model = AutoModelForCausalLM.from_pretrained(

    "Qwen/QwQ-32B-Preview",

    torch_dtype="auto",

    device_map={'': 0}

)



draft_model = AutoModelForCausalLM.from_pretrained(

    "Geralt-Targaryen/QwQ-Math-1.5B-Persona",

    torch_dtype="auto",

    device_map={'': 0}

)



tokenizer = AutoTokenizer.from_pretrained("Qwen/QwQ-32B-Preview")



prompt = "How many r in strawberry."

messages = [

    {"role": "system", "content": "You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step."},

    {"role": "user", "content": prompt}

]

text = tokenizer.apply_chat_template(

    messages,

    tokenize=False,

    add_generation_prompt=True

)

model_inputs = tokenizer([text], return_tensors="pt").to(model.device)



generated_ids = model.generate(

    **model_inputs,

    max_new_tokens=512,

    assistant_model=draft_model

)

generated_ids = [

    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)

]



response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

```

For the more advanced SVIP draft length policy, please refer to [this GitHub repo](https://github.com/Geralt-Targaryen/SVIP).

## Citation

If you find QwQ-Math-1.5B-Persona to be helpful, please cite the following paper.

```

@misc{zhang2024svip,

      title={Draft Model Knows When to Stop: A Self-Verification Length Policy for Speculative Decoding},

      author={Ziyin Zhang and Jiahao Xu and Tian Liang and Xingyu Chen and Zhiwei He and Rui Wang and Zhaopeng Tu},

      year={2024},

      eprint={2411.18462},

      archivePrefix={arXiv},

      primaryClass={cs.CL},

      url={https://arxiv.org/abs/2411.18462},

}

```