Jessylg27/specialized-coding-logic-llm

Specialized Coding Logic LLM (32B)

This model is a specialized fine-tuned version of Qwen/Qwen2.5-Coder-32B-Instruct. It has been optimized to enhance logical reasoning and code generation capabilities.

🧠 Model Description

Specialized Coding Logic LLM builds upon the powerful Qwen 2.5 Coder architecture (32B parameters). It has been fine-tuned using the DeepThink-Code-Lite dataset to improve its ability to:

• —Solve complex algorithmic problems.
• —Follow multi-step logical instructions.
• —Generate cleaner and more optimized code.

📊 Dataset

This model was trained on the custom dataset: 👉 [Jessylg27/DeepThink-Code-Lite](https://huggingface.co/datasets/Jessylg27/DeepThink-Code-Lite)

🚀 Quick Start

You can use this model directly with the Hugging Face pipeline.

from transformers import pipeline

# Define the model ID
model_id = "Jessylg27/specialized-coding-logic-llm"

# Initialize the pipeline
generator = pipeline("text-generation", model=model_id, device_map="auto")

# Prompt the model
question = "Write a Python function to solve the Traveling Salesman Problem using dynamic programming."
output = generator([{"role": "user", "content": question}], max_new_tokens=512, return_full_text=False)[0]

print(output["generated_text"])

🛠️ Training procedure

This model was trained with SFT (Supervised Fine-Tuning) using the TRL library and Unsloth for efficient training.

Framework versions

• —PEFT: 0.18.1
• —TRL: 0.24.0
• —Transformers: 4.57.3
• —Pytorch: 2.8.0+cu128
• —Datasets: 4.3.0
• —Tokenizers: 0.22.2

📜 Citations

If you use this model or the TRL library, please cite:

@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{[https://github.com/huggingface/trl](https://github.com/huggingface/trl)}}
}