Instructions to use Kamran-56/Qwen2.5-3B-PromptRefiner with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kamran-56/Qwen2.5-3B-PromptRefiner with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Kamran-56/Qwen2.5-3B-PromptRefiner")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Kamran-56/Qwen2.5-3B-PromptRefiner", dtype="auto")

PEFT
How to use Kamran-56/Qwen2.5-3B-PromptRefiner with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Kamran-56/Qwen2.5-3B-PromptRefiner with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Kamran-56/Qwen2.5-3B-PromptRefiner"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kamran-56/Qwen2.5-3B-PromptRefiner",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Kamran-56/Qwen2.5-3B-PromptRefiner

SGLang

How to use Kamran-56/Qwen2.5-3B-PromptRefiner with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Kamran-56/Qwen2.5-3B-PromptRefiner" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kamran-56/Qwen2.5-3B-PromptRefiner",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Kamran-56/Qwen2.5-3B-PromptRefiner" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kamran-56/Qwen2.5-3B-PromptRefiner",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Kamran-56/Qwen2.5-3B-PromptRefiner with Docker Model Runner:
```
docker model run hf.co/Kamran-56/Qwen2.5-3B-PromptRefiner
```

Qwen2.5-3B-PromptRefiner

A fine-tuned version of Qwen2.5-1.5B-Instruct specifically trained to transform basic, vague user prompts into high-quality, structured, and effective prompts that get significantly better responses from AI systems.

Model Details

Model Description

Developed by: Kamran (Kamran-56)
Model type: Causal Language Model (fine-tuned with LoRA/PEFT)
Language(s): English
License: MIT
Finetuned from: Qwen/Qwen2.5-1.5B-Instruct
Dataset used: Kamran-56/prompt-refinement-dataset

Uses

Direct Use

This model takes a basic user-written prompt as input and returns an enhanced, detailed, and well-structured version of the same prompt. It is designed to be used as a middleware layer between the user and any AI chatbot.

Downstream Use

Chrome extensions that auto-enhance prompts on AI platforms like ChatGPT, Claude, and Gemini
API middleware that improves prompts before forwarding to LLMs
Productivity tools that help non-technical users write better prompts

Out-of-Scope Use

This model is not designed to answer questions or generate content
Not suitable for tasks outside of prompt enhancement
Not trained on non-English prompts

How to Get Started with the Model

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

BASE_MODEL = "Qwen/Qwen2.5-1.5B-Instruct"
LORA_MODEL = "Kamran-56/Qwen2.5-3B-PromptRefiner"

tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)

base_model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)

model = PeftModel.from_pretrained(base_model, LORA_MODEL)
model.eval()

def enhance_prompt(bad_prompt):
    input_text = f"""<|im_start|>system
You are a world-class prompt engineer with deep expertise across all domains 
including coding, writing, business, creativity, and research. Transform the 
given basic prompt into a highly specific, structured, and effective prompt.
Every enhanced prompt MUST include a clear Role, Context, Task, Format, and 
Constraints. Return ONLY the enhanced prompt, nothing else.<|im_end|>
<|im_start|>user
{bad_prompt}<|im_end|>
<|im_start|>assistant
"""
    inputs = tokenizer(input_text, return_tensors="pt").to("cuda")
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=200,
            do_sample=True,
            temperature=0.7,
            top_p=0.9,
            repetition_penalty=1.2,
            pad_token_id=tokenizer.eos_token_id
        )
    full_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return full_output.split("assistant")[-1].strip()

# Example
print(enhance_prompt("write a poem"))

Training Details

Training Data

Fine-tuned on Kamran-56/prompt-refinement-dataset which contains 1,561 input→output pairs of basic prompts mapped to high-quality enhanced prompts generated using Llama 3.1 via Groq API.

Training Procedure

Training Hyperparameters

Training regime: bf16 mixed precision
Fine-tuning method: LoRA (PEFT)
LoRA rank (r): 8
LoRA alpha: 16
Target modules: q_proj, k_proj, v_proj, o_proj
Learning rate: 5e-5
Epochs: 3
Batch size: 4
Gradient accumulation steps: 4
Effective batch size: 16
LR scheduler: Cosine
Warmup steps: 50
Max sequence length: 512

Speeds, Sizes, Times

Hardware: Kaggle P100 GPU (16GB VRAM)
Training time: ~45 minutes
Final training loss: 1.24

Evaluation

Testing Data

Evaluated manually on 12 diverse prompts spanning coding, writing, professional, creative, and general categories.

Results

Input Prompt	Quality Before	Quality After
"write a poem"	Vague, no structure	Role + Context + Format + Constraints added
"fix my code"	No context	Step-by-step debugging structure with format
"write an email"	Generic	Professional tone, structure, constraints defined

Average quality rating: 8.5/10 based on manual evaluation measuring specificity, structure, role clarity, and format instructions.

Citation

@model{qwen25_promptrefiner,
  author    = {Kamran},
  title     = {Qwen2.5-3B-PromptRefiner},
  year      = {2025},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/Kamran-56/Qwen2.5-3B-PromptRefiner}
}

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for Kamran-56/Qwen2.5-3B-PromptRefiner

Base model

Qwen/Qwen2.5-1.5B

Finetuned

Qwen/Qwen2.5-1.5B-Instruct

Adapter

(1024)

this model

Kamran-56
/

Qwen2.5-3B-PromptRefiner