Deepseek-R1-8B-MikroTik-Distilled

MikroTik RouterOS Expert Assistant - DeepSeek-R1-Distill-Llama-8B LoRA

Model Description

This repository contains a fine-tuned version of the unsloth/DeepSeek-R1-Distill-Llama-8B large language model, specialized to act as an expert assistant for MikroTik RouterOS.

Fine-tuned for: Expert-level knowledge and assistance with MikroTik RouterOS networking, including configuration, troubleshooting, and understanding advanced features.

Intended Use: This model is designed for:

• Answering technical questions about MikroTik RouterOS.
• Providing step-by-step configuration guidance for MikroTik devices.
• Assisting network engineers with troubleshooting MikroTik network issues.
• Serving as a technical resource for users seeking expert-level MikroTik RouterOS information.

Model Architecture: LoRA (Low-Rank Adaptation) fine-tuned DeepSeek-R1-Distill-Llama-8B (8B parameters, Llama architecture). Converted to GGUF format for efficient inference with tools like llama.cpp.

Quantization: GGUF format, [Specify Quantization Type Here, e.g., q4_km] quantization.

Usage

The model is provided in GGUF format for efficient inference using llama.cpp and compatible tools.

Download GGUF Model:

Download the mikrotik-expert-model.gguf file from this repository.

Inference with llama.cpp/main (Example):

./main -m mikrotik-expert-model.gguf -p "Below is an instruction that describes a complex and critical task... ### Question: How do I configure a basic DHCP server on MikroTik? ### Response:" -n 500

Datasets Used for Fine-tuning:

The model was fine-tuned using a combination of these datasets:

• vivek-dodia/synthetic-data-gemini-2.0-ComplexTroubleshooting
• vivek-dodia/synthetic-data-gemini-2.0-ComplexConfigurations
• vivek-dodia/mikrotik-openAPI
• vivek-dodia/mikrotik-threads
• vivek-dodia/synthetic-data-deepseekR1-ComplexTroubleshooting
• vivek-dodia/mikrotik-github-repos
• vivek-dodia/mikrotik-gitlab-repos
• vivek-dodia/mikrotik-docs
• ramixpe/bgp_questions_v3

Acknowledgements:

• Fine-tuned using the Unsloth library.
• Based on the DeepSeek-R1-Distill-Llama-8B base model by Unsloth AI and DeepSeek AI.
• Built using the Transformers, Datasets, and TRL libraries from Hugging Face.

Contact

• [email protected]