|
|
---
|
|
|
library_name: transformers
|
|
|
tags:
|
|
|
- transformers
|
|
|
- text-generation-inference
|
|
|
- text-generation
|
|
|
- reasoning
|
|
|
- r1-reasoning
|
|
|
- fine-tuned
|
|
|
license: mit
|
|
|
datasets:
|
|
|
- openai/gsm8k
|
|
|
language:
|
|
|
- zho
|
|
|
- eng
|
|
|
- fra
|
|
|
- spa
|
|
|
- por
|
|
|
- deu
|
|
|
- ita
|
|
|
- rus
|
|
|
- jpn
|
|
|
- kor
|
|
|
- vie
|
|
|
- tha
|
|
|
- ara
|
|
|
base_model:
|
|
|
- Qwen/Qwen2.5-7B-Instruct
|
|
|
pipeline_tag: text-generation
|
|
|
---
|
|
|
|
|
|
# **Qwen-2.5-7B-Reasoning (Fine-Tuned by HyperX-Sen)**
|
|
|
|
|
|
## 🚀 **Model Overview**
|
|
|
This model is a fine-tuned version of **Qwen/Qwen2.5-7B-Instruct**, specifically optimized for **advanced reasoning tasks**. Fine-tuned on the **OpenAI GSM8K dataset**, it significantly enhances multi-step reasoning and problem-solving capabilities.
|
|
|
|
|
|
## 🔧 **Fine-Tuning Details**
|
|
|
- **Base Model:** [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct)
|
|
|
- **Fine-tuned by:** HyperX-Sen
|
|
|
- **Dataset:** [GSM8K (Grade School Math 8K)](https://huggingface.co/datasets/openai/gsm8k)
|
|
|
- **Hardware:** 2× Tesla T4 GPUs
|
|
|
- **Objective:** Improve complex reasoning and logical deduction
|
|
|
|
|
|
## 📈 **Performance Improvements**
|
|
|
Through fine-tuning on **GSM8K**, the model has improved in:
|
|
|
- **Mathematical reasoning**
|
|
|
- **Step-by-step logical deduction**
|
|
|
- **Commonsense reasoning**
|
|
|
- **Word problem-solving**
|
|
|
|
|
|
This makes it ideal for applications requiring **high-level reasoning**, such as **AI tutoring, research assistance, and problem-solving AI agents**.
|
|
|
|
|
|
## 🛠 **How to Use**
|
|
|
```python
|
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
|
|
# Load the model and tokenizer
|
|
|
model_name = "HyperX-Sen/Qwen-2.5-7B-Reasoning"
|
|
|
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
|
|
SYSTEM_PROMPT = """
|
|
|
Respond in the following format:
|
|
|
<reasoning>
|
|
|
...
|
|
|
</reasoning>
|
|
|
<answer>
|
|
|
...
|
|
|
</answer>
|
|
|
"""
|
|
|
|
|
|
# Define the conversation
|
|
|
messages = [
|
|
|
{"role": "system", "content": f"{SYSTEM_PROMPT}"},
|
|
|
{"role": "user", "content": "What are the potential impacts of artificial intelligence on employment?"}
|
|
|
]
|
|
|
|
|
|
# Format the chat input
|
|
|
input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
|
|
|
|
|
# Tokenize the formatted input
|
|
|
inputs = tokenizer(input_text, return_tensors="pt").to(model.device)
|
|
|
|
|
|
# Generate the response
|
|
|
output = model.generate(**inputs, max_length=512, do_sample=True, temperature=0.7)
|
|
|
|
|
|
# Decode and display the response
|
|
|
response = tokenizer.decode(output[0], skip_special_tokens=True)
|
|
|
print(response)
|
|
|
```
|
|
|
|
|
|
## 🙌 **Acknowledgments**
|
|
|
A huge thanks to **Qwen** for providing the powerful **Qwen2.5-7B-Instruct** model, which served as the base for this fine-tuned version.
|
|
|
|
