---
license: mit
language:
- en
base_model:
- microsoft/Phi-3-mini-4k-instruct
tags:
- gguf
- llama.cpp
- phi-3
pipeline_tag: text-classification
---
# 📧 Email Classifier (Phi-3 Mini Fine-Tuned GGUF)

## 🧠 Model Overview

This model is a **fine-tuned version of Phi-3 Mini (4K Instruct)** optimized for **email classification tasks** such as spam detection and categorization.

* **Base Model:** Phi-3 Mini 4K Instruct
* **Fine-tuning:** Supervised fine-tuning on email dataset
* **Format:** GGUF (for efficient local inference)
* **Quantization:** Q5_K_M

---

## 📚 Dataset

The model was trained on:

* **Dataset:** `json22322/email-classifier`
* Contains labeled email samples for classification tasks (e.g., spam vs non-spam)

---

## 🎯 Task

* Email classification
* Spam detection
* Priority categorization

---

## ⚙️ Usage

### Using llama.cpp

```bash
./main -m Phi-3-mini-4k-instruct-Q5_K_M.gguf -p "Classify this email: Subject: ... Body: ..."
```

---

## 🧪 Example

**Input:**

```
Subject: Win a free iPhone now!!!
Body: Click here to claim your reward.
```

**Output:**

```
Spam
```

---

## ⚠️ Limitations

* Performance depends on dataset quality
* May not generalize to unseen domains
* Sensitive to prompt phrasing

---

## 📦 Files

| File   | Description                |
| ------ | -------------------------- |
| *.gguf | Quantized fine-tuned model |

---

## 📌 Notes

* Optimized for **local inference (CPU-friendly)**
* Compatible with **llama.cpp and GGUF runtimes**

---

## 📜 License

Same as base model (Phi-3).

---

## 🙌 Acknowledgements

* Microsoft for Phi-3
* Hugging Face
* llama.cpp

---