💗 MiryoLove-7B

A Fine-Tuned Language Model by MiryoBun

MiryoLove-7B is a fine-tuned language model developed as part of the MIRYOLOVE AI project by MiryoBun. It is based on Qwen2.5-7B-Instruct, an original language model developed and published by Qwen, a model family from Alibaba Cloud.

MiryoLove-7B builds upon the original Qwen model through additional fine-tuning using a small dataset focused on programming, technical explanations, learning, and technical translation.

Important attribution: The original Qwen2.5-7B-Instruct model and its underlying pretrained weights are developed by Qwen / Alibaba Cloud. MiryoBun is responsible for the MiryoLove-7B project and its fine-tuning and merged model weights. MiryoLove-7B is not an original foundation model created from scratch.

---

✨ Features

• 💻 Programming Assistance — explanations and examples involving Python, C, C++, Java, and Assembly.
• 📚 Learning Support — explanations of programming concepts and suggestions for practice tasks.
• 🌐 Technical Translation — technical text translation between English, Russian, and Ukrainian.
• 🧠 Instruction Following — fine-tuning intended to improve responses to programming and technical prompts.
• 💗 MIRYOLOVE AI — an independent project developed by MiryoBun using the Qwen model as its foundation.

🛠️ Model Information

Property	Details
Model name	MiryoLove-7B
Project developer	MiryoBun
Original model developer	Qwen / Alibaba Cloud
Base model	Qwen2.5-7B-Instruct
Model family	Qwen2.5
Parameter count	Approximately 7 billion
Fine-tuning method	LoRA
Final model format	Merged FP16 weights
Languages	English, Russian, Ukrainian
Main tasks	Programming, technical explanations, translation
License	Apache-2.0; see the original model license and applicable terms

🧬 Relationship to Qwen

MiryoLove-7B is based on Qwen2.5-7B-Instruct, developed by Qwen / Alibaba Cloud.

The project uses the original model as its foundation and adds task-focused fine-tuning. The LoRA fine-tuning was merged into the base model to produce the merged FP16 model weights published in this repository.

Please distinguish the two projects:

• Qwen / Alibaba Cloud: Developers and publishers of the original Qwen2.5-7B-Instruct model.
• MiryoBun: Developer of the MiryoLove-7B project and its additional fine-tuning.
• MiryoLove-7B: A derivative fine-tuned model based on Qwen2.5-7B-Instruct.

MiryoLove-7B is an independent project and is not presented as an official Alibaba Cloud or Qwen release.

🎯 Intended Use

MiryoLove-7B is intended for:

• Learning programming concepts and reading code examples.
• Getting short explanations of Python, C, C++, Java, and Assembly.
• Practicing programming with example tasks.
• Translating technical sentences between English, Russian, and Ukrainian.
• Exploring technical questions and basic educational topics.

💬 Example Prompts

The following are example prompts you can try with the model. They are not claimed to be verified outputs from MiryoLove-7B.

Programming

Explain what a Python function is and show a simple example.

Learning

Explain the difference between a variable and a constant, then give me a practice task.

Technical Translation

Translate this technical sentence from English into Russian: "The program reads data from a file."

⚠️ Limitations

MiryoLove-7B was fine-tuned on a relatively small dataset of approximately 130 examples. Its performance can vary depending on the task.

• Generated code may contain errors and should be tested.
• Technical explanations may be incomplete or inaccurate.
• Uncommon technical terms may be translated incorrectly.
• The model may misunderstand complex instructions or generate incorrect information.
• Comprehensive benchmark results are not currently provided in this model card.

The model should be evaluated on the intended task before being used in applications where accuracy is important.

📦 Model Files and Usage

The model files are available in this repository:

Repository: https://huggingface.co/MiryoBun/MiryoLove-7B

The published model uses merged FP16 weights. Loading and running these weights requires sufficient system memory or GPU memory, depending on the inference setup and context length.

For usage instructions, review the model repository files and the documentation for the inference framework you choose.

📜 License and Attribution

MiryoLove-7B is based on Qwen2.5-7B-Instruct, developed and published by Qwen / Alibaba Cloud.

• Original model: Qwen2.5-7B-Instruct
• Original model publisher: Qwen / Alibaba Cloud
• Fine-tuning project: MiryoLove-7B by MiryoBun
• Fine-tuning method: LoRA
• Published format: Merged FP16 weights
• License: Apache-2.0, subject to the applicable license terms

Users should review the original Qwen model license and the license information included in the repository before redistributing or using the model.

---

💗 MiryoLove-7B
Built by MiryoBun, based on Qwen.
_{Part of the MIRYOLOVE AI project.}