SmolLM3-3B-math-light

🎯 MATH-optimized | 📦 Light pruning | ⚡ 3% weights pruned

This model is a lightly pruned version of HuggingFaceTB/SmolLM3-3B, 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	80.0%	100.0%	↑ 20.0%
Html	0.0%	0.0%	→
Trivia	100.0%	100.0%	→
Math	100.0%	100.0% ⭐	→
Reasoning	N/A	N/A
Medical	100.0%	100.0%	→
Linux	100.0%	100.0%	→
Writing	93.3%	80.0%	↓ 13.3%

Average: 81.9% → 82.9% (+1.0%)

Math Retention: 100.0% of original performance

🚀 Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/SmolLM3-3B-math-light")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/SmolLM3-3B-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	HuggingFaceTB/SmolLM3-3B
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 SmolLM3-3B 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 HuggingFaceTB/SmolLM3-3B.

---

Generated by ZANNPS [Zeto Automatic Neural Network Pruning System]