Gemma 3-1B — Quick Reference

Project Dependencies

To achieve efficient fine-tuning on consumer hardware, this project relies on a specific set of optimized libraries. The dependencies are chosen to balance performance (speed/memory) with ease of use (Hugging Face integration).

Key Libraries

• unsloth: Accelerates training and reduces VRAM usage
• torch (PyTorch): Core deep learning framework
• transformers: Hugging Face model loading/manipulation
• peft: Enables LoRA parameter-efficient fine-tuning
• trl: SFTTrainer for supervised fine-tuning
• bitsandbytes: 4/8-bit quantization
• accelerate: Hardware optimization
• datasets: Data loading and manipulation
• tensorboard, matplotlib: Visualization and reporting

Installation

Recommended: use a virtual environment, then install dependencies via pip:

pip install -r requirements.txt

requirements.txt

torch
unsloth
trl
peft
accelerate
bitsandbytes
datasets
transformers
tensorboard
matplotlib

🔗 Related Documents

• index
• REFERENCES

---

Last Updated: 2026-04-29
Version: 1.0
Status: Complete