---
license: apache-2.0
tags:
- pruned
- writing
- optimized
- wanda
- activation-pruning
base_model: Qwen/Qwen2.5-3B-Instruct
pipeline_tag: text-generation
---

# Qwen2.5-3B-Instruct-writing-aggressive

> 🎯 **WRITING-optimized** | 📦 **Aggressive** pruning | ⚡ **35% weights pruned**

This model is a **aggressively pruned** version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), specialized for **WRITING** tasks using activation-aware weight pruning (Wanda-style).

## ✨ Key Features

- **Specialization**: Optimized for Writing tasks
- **Pruning Method**: Wanda-style (|W| × |activation|) importance scoring
- **Size Reduction**: 35% weights pruned
- **Use Case**: Maximum compression for edge deployment

## 📊 Performance Comparison

| Category | Original | Pruned | Change |
|----------|----------|--------|--------|
| Python | 100.0% | 0.0% | ↓ 100.0% |
| Html | 6.7% | 6.7% | → |
| Trivia | 66.7% | 0.0% | ↓ 66.7% |
| Math | 60.0% | 66.7% | ↑ 6.7% |
| Reasoning | 100.0% | 53.3% | ↓ 46.7% |
| Medical | 86.7% | 13.3% | ↓ 73.3% |
| Linux | 100.0% | 0.0% | ↓ 100.0% |
| **Writing** | 73.3% | 46.7% ⭐ | ↓ 26.7% |

**Average**: 74.2% → 23.3% (-50.8%)

**Writing Retention**: 63.6% of original performance

![Comparison Graph](comparison_graph.png)

## 🚀 Quick Start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-writing-aggressive")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-writing-aggressive")

# Example usage
inputs = tokenizer("Your prompt here", return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## 📋 Technical Details

| Property | Value |
|----------|-------|
| Base Model | [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) |
| Specialization | Writing |
| Prune Mode | Aggressive |
| Pruning Method | Activation-based weight pruning (Wanda) |
| Weight Reduction | 35% weights pruned |

## 🔗 Related Models

This model is part of the **Qwen2.5-3B-Instruct** pruned model collection. Variants:
- **Safe** - Conservative pruning (~10-20%), high accuracy retention
- **Aggressive** - Maximum compression (~40-50%), best for edge deployment

## 📜 License

This model inherits the license from the base model [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).

---
*Generated by ZANNPS [Zeto Automatic Neural Network Pruning System]*