TinyLlama-1.1B-Chat-v1.0-python-safe

🎯 PYTHON-optimized | 📦 Safe pruning | ⚡ 1% weights pruned

This model is a conservatively pruned version of TinyLlama/TinyLlama-1.1B-Chat-v1.0, specialized for PYTHON tasks using activation-aware weight pruning (Wanda-style).

✨ Key Features

• Specialization: Optimized for Python tasks
• Pruning Method: Wanda-style (|W| × |activation|) importance scoring
• Size Reduction: 1% weights pruned
• Use Case: High accuracy retention, ideal for production use

📊 Performance Comparison

Category	Original	Pruned	Change
Python	0.0%	0.0% ⭐	→
Html	0.0%	0.0%	→
Trivia	6.7%	6.7%	→
Math	0.0%	0.0%	→
Reasoning	13.3%	13.3%	→
Medical	0.0%	0.0%	→
Linux	0.0%	0.0%	→
Writing	0.0%	0.0%	→

Average: 2.5% → 2.5% (+0.0%)

🚀 Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/TinyLlama-1.1B-Chat-v1.0-python-safe")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/TinyLlama-1.1B-Chat-v1.0-python-safe")

# 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	TinyLlama/TinyLlama-1.1B-Chat-v1.0
Specialization	Python
Prune Mode	Safe
Pruning Method	Activation-based weight pruning (Wanda)
Weight Reduction	1% weights pruned

🔗 Related Models

This model is part of the TinyLlama-1.1B-Chat-v1.0 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 TinyLlama/TinyLlama-1.1B-Chat-v1.0.

---

Generated by ZANNPS [Zeto Automatic Neural Network Pruning System]