recon_gguf

ReCon LLaMA 3.1 (GGUF) — Local Ollama Model

This repository contains a GGUF version of a fine-tuned LLaMA 3.1 model for generating contrastive explanations.

The model is optimized for local inference on laptops using Ollama.

---

What this model does

• Generates natural language explanations
• Specialized for contrastive OWL reasoning
• Built using:
  • LLaMA 3.1 (base)
  • LoRA fine-tuning
  • Merged (GGUF)

---

Requirements

• macOS / Linux / Windows
• At least 8 GB RAM recommended
• Ollama installed

Install Ollama

curl -fsSL https://ollama.com/install.sh | sh

---

Setup Instructions

1. Download the model from Hugging Face

Install the Hugging Face CLI if you do not already have it:

pip install -U "huggingface_hub[cli]"

Or download the .gguf file directly at https://huggingface.co/vanishingMonk/recon_gguf/tree/main.

Download the GGUF model:

huggingface-cli download vanishingMonk/recon_gguf model-recon.gguf --local-dir .

This will download the model into your current directory.

2. Create a Modelfile

Create a file named Modelfile:

nano Modelfile 

Paste:

FROM ./model-recon.gguf

PARAMETER temperature 0.5

SYSTEM """You explain contrastive explanations clearly and concisely."""

3. Build the model in Ollama

ollama create recon -f Modelfile

4. Run the model

ollama run recon

---

Example usage

User: Explain contrastive explanations in simple terms with an example

Model: Before I explain, do you know what a contrastive explanation is?

User: no

Model: No problem! Here's the simple version.

A contrastive explanation shows why one situation happened instead of another. It highlights the key difference between two possible outcomes and how that difference caused the result we see.

Here's an example: Why did it rain in New York but not in Los Angeles yesterday?

In this case, the contrast is the location (New York vs. Los Angeles). The weather forecast showed a high chance of rain only for New York. Because New York was in the path of a storm system, it rained there. Since Los Angeles was not affected by that same system, it remained dry.

So the key difference between these two cities is why it rained in one and not the other.