|
|
--- |
|
|
license: apache-2.0 |
|
|
tags: |
|
|
- pruned |
|
|
- math |
|
|
- optimized |
|
|
- wanda |
|
|
- activation-pruning |
|
|
base_model: Qwen/Qwen2.5-3B-Instruct |
|
|
pipeline_tag: text-generation |
|
|
--- |
|
|
|
|
|
# Qwen2.5-3B-Instruct-math-light |
|
|
|
|
|
> 🎯 **MATH-optimized** | 📦 **Light** pruning | ⚡ **3% weights pruned** |
|
|
|
|
|
This model is a **lightly pruned** version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), specialized for **MATH** tasks using activation-aware weight pruning (Wanda-style). |
|
|
|
|
|
## ✨ Key Features |
|
|
|
|
|
- **Specialization**: Optimized for Math tasks |
|
|
- **Pruning Method**: Wanda-style (|W| × |activation|) importance scoring |
|
|
- **Size Reduction**: 3% weights pruned |
|
|
- **Use Case**: Good balance of accuracy and size for production use |
|
|
|
|
|
## 📊 Performance Comparison |
|
|
|
|
|
| Category | Original | Pruned | Change | |
|
|
|----------|----------|--------|--------| |
|
|
| Python | 100.0% | 100.0% | → | |
|
|
| Html | 6.7% | 6.7% | → | |
|
|
| Trivia | 100.0% | 100.0% | → | |
|
|
| **Math** | 60.0% | 60.0% ⭐ | → | |
|
|
| Reasoning | N/A | N/A | | |
|
|
| Medical | 100.0% | 100.0% | → | |
|
|
| Linux | 100.0% | 100.0% | → | |
|
|
| Writing | 73.3% | 73.3% | → | |
|
|
|
|
|
**Average**: 77.1% → 77.1% (+0.0%) |
|
|
|
|
|
**Math Retention**: 100.0% of original performance |
|
|
|
|
|
 |
|
|
|
|
|
## 🚀 Quick Start |
|
|
|
|
|
```python |
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-math-light") |
|
|
tokenizer = AutoTokenizer.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-math-light") |
|
|
|
|
|
# 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 | Math | |
|
|
| Prune Mode | Light | |
|
|
| Pruning Method | Activation-based weight pruning (Wanda) | |
|
|
| Weight Reduction | 3% weights pruned | |
|
|
|
|
|
## 🔗 Related Models |
|
|
|
|
|
This model is part of the **Qwen2.5-3B-Instruct** pruned model collection. Other variants: |
|
|
- Extra-light (minimal pruning) |
|
|
- Light |
|
|
- Medium-light |
|
|
- Medium |
|
|
- Medium-heavy |
|
|
- Heavy |
|
|
- Extra-heavy (maximum compression) |
|
|
|
|
|
## 📜 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]* |
|
|
|
