README.md

metadata

license: apache-2.0
tags:
  - pruned
  - writing
  - optimized
  - wanda
  - activation-pruning
base_model: Qwen/Qwen3-0.6B
pipeline_tag: text-generation

Qwen3-0.6B-writing-aggressive

🎯 WRITING-optimized | 📦 Aggressive pruning | ⚡ 12% weights pruned

This model is a aggressively pruned version of Qwen/Qwen3-0.6B, 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: 12% weights pruned
• Use Case: Maximum compression for edge deployment

📊 Performance Comparison

Category	Original	Pruned	Change
Python	30.0%	0.0%	↓ 30.0%
Html	0.0%	0.0%	→
Trivia	90.0%	86.7%	↓ 3.3%
Math	96.7%	76.7%	↓ 20.0%
Reasoning	36.7%	36.7%	→
Medical	83.3%	70.0%	↓ 13.3%
Linux	93.3%	90.0%	↓ 3.3%
Writing	53.3%	40.0% ⭐	↓ 13.3%

Average: 60.4% → 50.0% (-10.4%)

Writing Retention: 75.0% of original performance

🚀 Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/Qwen3-0.6B-writing-aggressive")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/Qwen3-0.6B-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/Qwen3-0.6B
Specialization	Writing
Prune Mode	Aggressive
Pruning Method	Activation-based weight pruning (Wanda)
Weight Reduction	12% weights pruned

🔗 Related Models

This model is part of the Qwen3-0.6B 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/Qwen3-0.6B.

---

Generated by ZANNPS [Zeto Automatic Neural Network Pruning System]