metadata
license: apache-2.0
tags:
- pruned
- reasoning
- optimized
- wanda
- activation-pruning
base_model: HuggingFaceTB/SmolLM3-3B
pipeline_tag: text-generation
SmolLM3-3B-reasoning-medium
🎯 REASONING-optimized | 📦 Medium pruning | ⚡ 20% weights pruned
This model is a moderately pruned version of HuggingFaceTB/SmolLM3-3B, specialized for REASONING tasks using activation-aware weight pruning (Wanda-style).
✨ Key Features
- Specialization: Optimized for Reasoning tasks
- Pruning Method: Wanda-style (|W| × |activation|) importance scoring
- Size Reduction: 20% weights pruned
- Use Case: Balanced trade-off between size and accuracy
📊 Performance Comparison
| Category | Original | Pruned | Change |
|---|---|---|---|
| Python | 80.0% | 40.0% | ↓ 40.0% |
| Html | 0.0% | 0.0% | → |
| Trivia | 100.0% | 80.0% | ↓ 20.0% |
| Math | 100.0% | 100.0% | → |
| Reasoning | N/A | N/A | |
| Medical | 100.0% | 100.0% | → |
| Linux | 100.0% | 100.0% | → |
| Writing | 93.3% | 73.3% | ↓ 20.0% |
Average: 81.9% → 70.5% (-11.4%)
🚀 Quick Start
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("CompactAI/SmolLM3-3B-reasoning-medium")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/SmolLM3-3B-reasoning-medium")
# 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 | Reasoning |
| Prune Mode | Medium |
| Pruning Method | Activation-based weight pruning (Wanda) |
| Weight Reduction | 20% 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]