AmeerHamza547/political-speech-classification-clf-llama-3.2-3b-GGUF

Political Speech Classification (Llama 3.2 3B) - GGUF

This repository contains the GGUF quantized version of a fine-tuned Llama 3.2 3B model specifically optimized for political speech transcript analysis. The model was fine-tuned using QLoRA via Unsloth and quantized into the highly efficient Q4_K_M format, making it perfect for fast, local inference on low-specification hardware (CPUs or low-VRAM laptops).

🎯 Model Capabilities

The model is optimized to perform two core downstream classification tasks using structured system prompts:

1. 1.Political Sentiment Classification: Analyzes a speech transcript and categorizes the underlying tone as positive, negative, or neutral.
2. 2.Agenda & Topic Detection: Maps the transcript across 6 key political categories and outputs a structured JSON array from: economy, healthcare, education, social_welfare, security, and infrastructure.

💾 Quantization Details

• —Format: GGUF (.gguf)
• —Method: Q4_K_M (4-bit medium quantization using mixed precision)
• —File Size: ~2.02 GB
• —Hardware Requirement: Extremely lightweight. Runs comfortably on systems with 4GB–8GB of system RAM/VRAM.

🚀 Local Deployment

1. Using Ollama

To run this model locally with Ollama, follow these quick steps:

1. 1.Download the .gguf file from the Files tab.
2. 2.Create a file named Modelfile in the same directory and add the following content:

FROM ./political-speech-classification-clf-llama-3.2-3b-Q4_K_M.gguf

# Set parameters
PARAMETER temperature 0.0
PARAMETER top_p 0.9

# You can adjust the system prompt depending on your task
SYSTEM "You are a political speech sentiment classifier. Analyze the speech and respond with ONLY one word: positive, negative, or neutral."