spar-project/Llama-3.2-3B-Instruct-attention-layers

TEXT GENERATIONConcurrency Cost:1Model Size:3.2BQuant:BF16Ctx Length:32kPublished:Mar 25, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The spar-project/Llama-3.2-3B-Instruct-attention-layers is a 3.2 billion parameter instruction-tuned Llama model developed by spar-project. This model was finetuned from unsloth/Llama-3.2-3B-Instruct and optimized for faster training using Unsloth and Huggingface's TRL library. It is designed for general instruction-following tasks, leveraging its efficient training methodology.

Loading preview...

Overview

spar-project/Llama-3.2-3B-Instruct-attention-layers is a 3.2 billion parameter instruction-tuned language model. It is a derivative of the unsloth/Llama-3.2-3B-Instruct model, developed by spar-project.

Key Characteristics

  • Efficient Training: This model was finetuned with a focus on speed, achieving 2x faster training times by utilizing the Unsloth library in conjunction with Huggingface's TRL library.
  • Instruction-Tuned: Designed to follow instructions effectively, making it suitable for a variety of conversational and task-oriented applications.
  • Llama Architecture: Based on the Llama model family, providing a robust foundation for language understanding and generation.

Use Cases

This model is well-suited for applications requiring a compact yet capable instruction-following LLM, particularly where training efficiency is a priority. Its optimized training process makes it an interesting choice for developers looking to deploy Llama-based models with reduced resource consumption during finetuning.