---
title: FAST_DS_1_5b
canonical_url: "https://www.modelscope.cn/models/ruiruiL/FAST_DS_1_5b"
md_url: "https://www.modelscope.cn/models/ruiruiL/FAST_DS_1_5b.md"
repository: ruiruiL/FAST_DS_1_5b
last_updated: 2025-12-23
model_type:
  - qwen2
architectures:
  - Qwen2ForCausalLM
library_name:
  - safetensors
  - pytorch
frameworks:
  - Pytorch
inference_backends:
  - "deploy_task text/emb"
  - "lmdeploy 0.9.1"
  - "lmdeploy_turbomind 0.9.1"
  - "sglang 0.5.2"
  - "vllm 0.9.2"
downloads: 51
stars: 1
---

# FAST_DS_1_5b

> FAST_DS_1_5b - ruiruiL 在 ModelScope 开源的模型。FAST-GRPO文本推理版

ruiruiL/FAST_DS_1_5b 是 ModelScope 魔搭社区上的机器学习模型，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。

- **Repository**: ruiruiL/FAST_DS_1_5b
- **Inference backends**: deploy_task text/emb, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1, sglang 0.5.2, vllm 0.9.2
- **Downloads**: 51
- **Stars**: 1
- **Last updated**: 2025-12-23

Source: https://www.modelscope.cn/models/ruiruiL/FAST_DS_1_5b

---

# FAST-1_5B Model Documentation

## Overview
This repository provides access to the **FAST-1_5B** model, which is built on the **deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B** base model.

## System Prompt
```
""" Please reason step by step, and put your final answer within \boxed{}"""
```
## Decoding Paramater

We recommend setting `temperature=0.6` to reproduce the reported performance. Note that performance may vary depending on the version of vLLM being used.

## Inference Guide

### Installation

Install the required dependencies:

```bash
pip install vllm==0.11.0
```

### Starting the Server

Start the vLLM server with the following command:

```bash
CUDA_VISIBLE_DEVICES=0 vllm serve /PATH/TO/FAST \
    --max-model-len 16384 \
    --dtype auto \
    --gpu_memory_utilization 0.85 \
    --trust-remote-code \
    --max-num-seqs 8 
```

Replace `/PATH/TO/FAST` with the actual path to your model.

### Simple Demo

```python
import base64

# Define the system prompt
from openai import OpenAI

# Define the system prompt
SYSTEM_PROMPT = """Please reason step by step, and put your final answer within \boxed{}"""

def simple_inference(query, model_path, max_tokens=16384, temperature=0.6):
    """
    Perform a simple text-only inference.
    Args:
        query (str): Text query for the model.
        model_path (str): The name or path of the model served by vLLM.
        max_tokens (int): Maximum number of tokens in the response.
        temperature (float): Sampling temperature for the model.
    Returns:
        str: The model's response.
    """
    
    # Prepare the chat request
    request = {
        "model": model_path, 
        "messages": [
            {"role": "system", "content": SYSTEM_PROMPT},
            {"role": "user", "content": query},
        ],
        "temperature": temperature,
        "max_tokens": max_tokens,
    }

    # Call the chat API
    try:
        # Initialize client pointing to local vLLM server
        client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
        
        chat_response = client.chat.completions.create(**request)
        return chat_response.choices[0].message.content
    except Exception as e:
        print(f"Error during inference: {e}")
        return None

# Example Usage
if __name__ == "__main__":
    model_name = "/PATH/TO/FAST" # Ensure this matches the served model name
    user_query = "Prove that the square root of 2 is irrational."
    
    response = simple_inference(user_query, model_name)
    
    print("Model Response:")
    print(response)
```