---
base_model:
- mistralai/Mistral-7B-v0.1
---

# Mistral-7B-v0.1-4bit-mlx

A 4-bit LoRA‑fine‑tuned Mistral-7B model in Apple MLX format, created via the MLX Examples LoRA tutorial.

---

## 🚀 Overview

This model was created by following the MLX LoRA tutorial:

- **LoRA fine‑tuning** applied on Mistral‑7B using low‑rank adapters  
- **Model fusion** performed with `convert.py` to combine base weights and LoRA adapters  
- **Quantized to 4-bit** for memory efficiency on Apple Silicon devices  
- Packaged in MLX format, ready for downstream use via `mlx-lm`

The pipeline is fully automated based on `mlx-examples/lora/convert.py`, with no manual weight editing.

---

## 📦 Model Packaging

| File                | Description                                  |
|---------------------|----------------------------------------------|
| `weights.npz`       | Fused weights: base + LoRA adapters         |
| `config.json`       | Model configuration & quantization metadata |
| `tokenizer.model`   | SentencePiece tokenizer for Mistral-7B     |

---

## 💡 Usage

### Inference with `mlx-lm`

```bash
pip install mlx-lm