Model Card for Model ID
This model is a pruned version of the Llama-3.2-3b model, with a parameter reduction of 20% in the MLP Layers. The pruning process aims to enhance computational efficiency while maintaining acceptable performance across specific tasks. This model is not intended to be used directly, but rather to be fine-tuned for specific tasks where it can achieve equal or superior performance compared to fine-tuning the base model for the same task.
Model Details
- Model Type: Pruned version of LLaMA-3.2 using structured pruning
- Original Model: meta-llama/Llama-3.2-1B
- Pruning Method: Structured pruning of MLP layers using importance scores based on absolute maximum weights
- Size Reduction: 13.1% (from 3.21B to 2.79B parameters)
- Architecture: Same as original LLaMA but with reduced MLP layer sizes
- Language(s): Same as original model
- License: Same as original model
- Developed by: Pere Martra
These models are part of the study "Exploring GLU Expansion Ratios: Structured Pruning in Llama-3.2 Models". They explore structured pruning in GLU-based architectures using Llama-3.2 (1B and 3B variants). The pruning experiments target optimal expansion ratios to balance performance, computational efficiency, and environmental sustainability. The models were evaluated across multiple benchmarks, including BoolQ, ARC-Easy, and MUSR, and demonstrate significant efficiency gains while maintaining robust task performance.
Performance on Standard Benchmarks
| Benchmark | Original Model | Pruned Model | Relative Change |
|---|---|---|---|
| ARC-Easy | 65.19% | 58.54% | -10.2% |
| BoolQ | 64.16% | 39.97% | -37.7% |
| LAMBADA-OpenAI | 62.20% | 54.94% | -11.7% |
| LAMBADA-Standard | 53.46% | 49.25% | -7.9% |
Key Findings
- The pruned model shows a moderate degradation on reasoning tasks (ARC-Easy) but maintains reasonable performance relative to its size reduction.
- Performance on binary classification tasks (BoolQ) is more significantly impacted, indicating limitations for such use cases.
- For language completion tasks (LAMBADA), the model experiences mild to moderate degradation but remains usable for less demanding applications.
Limitations
- Reduced performance on tasks requiring complex reasoning or classification: Tasks such as BoolQ see significant drops in accuracy.
- Impacts on long-range comprehension: While less severe than BoolQ, tasks like LAMBADA show noticeable degradation.
- Limited utility for high-accuracy applications: The pruned model is less suitable for scenarios demanding peak performance in understanding or generating complex language.
Implementation Details
- Pruning Notebook: Detailed implementation and methodology
- GitHub Repository: LLM Course
- Article explaining pruning methodology: How to Prune LLaMA 3.2 and Similar Large Language Models
Pruning Method
- Technique: Structured pruning targeting MLP layers
- Pruning Ratio: 20% of neurons removed from MLP layers
- Selection Criteria: Importance scoring based on absolute maximum weights
- Architecture Specifics: Maintained GLU structure during pruning
Hardware Requirements
- Reduced memory footprint compared to original model
- Can run on hardware with ~15% less memory than original
Acknowledgments
- Thanks to Mariusz Kurman for creating llama-pruning, a library that extends and improve this pruning methodology.
- This model was created following the pruning method described in the paper: The Width Pruning Dichotomy in Llama-3.2
@misc{martra2025fragileknowledgerobustinstructionfollowing,
title={Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2},
author={Pere Martra},
year={2025},
eprint={2512.22671},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2512.22671},
}