README.md

---
license: apache-2.0
tags:
- pruned
- python
- optimized
base_model: Qwen/Qwen3-4B
---

# Qwen3-4B-python-medium-prune

This model is a **medium** pruned version of [Qwen/Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B), specialized for **PYTHON** tasks.

## Pruning Details

- **Base Model**: Qwen/Qwen3-4B
- **Specialization**: Python
- **Prune Mode**: Medium
- **Method**: Activation-based weight pruning

## Performance Comparison

| Category | Original | Pruned |
|----------|----------|--------|
| Python | 0.0% | 13.3% |
| HTML | 6.7% | 26.7% |
| Trivia | 86.7% | 86.7% |
| Math | 40.0% | 46.7% |
| Reasoning | 60.0% | 66.7% |

![Comparison Graph](comparison_graph.png)

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/Qwen3-4B-python-medium-prune-prune")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/Qwen3-4B-python-medium-prune-prune")
```

## License

This model inherits the license from the base model.