README.md

---
license: apache-2.0
tags:
- pruned
- python
- optimized
- wanda
base_model: LiquidAI/LFM2.5-1.2B-Instruct
pipeline_tag: text-generation
---

# LFM2.5-1.2B-Instruct-python-safe

> **PYTHON-optimized** | **Safe** pruning | **30% weights pruned**

This model is a **conservatively pruned** version of [LiquidAI/LFM2.5-1.2B-Instruct](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct).



> **Pruning Alert:** The benchmarks show virtually NO quality drop! This isn't a bug -- it is a feature. The Wanda pruning algorithm is so effective at identifying unimportant weights that it can remove a large percentage of parameters without affecting performance. Think of it like pruning dead leaves from a tree -- the tree does not miss them because they were not doing anything anyway!



## Performance Comparison

| Category | Original | Pruned | Change |
|----------|----------|--------|--------|
| **Python** | 0.0% | 0.0% ⭐ | → |
| Html | 0.0% | 0.0% | → |
| Trivia | 90.0% | 90.0% | → |
| Math | 55.0% | 55.0% | → |
| Reasoning | 40.0% | 40.0% | → |
| Medical | 80.0% | 80.0% | → |
| Linux | 45.0% | 45.0% | → |
| Writing | 20.0% | 20.0% | → |

**Average**: 41.2% -> 41.2% (+0.0%)



![Comparison Graph](comparison_graph.png)

## Quick Start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/LFM2.5-1.2B-Instruct-python-safe")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/LFM2.5-1.2B-Instruct-python-safe")

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 | [LiquidAI/LFM2.5-1.2B-Instruct](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct) |
| Specialization | Python |
| Prune Mode | Safe |
| Weight Reduction | 30% weights pruned |

## License

This model inherits the license from the base model.