README.md

metadata

license: mit
base_model:
  - google/gemma-3-1b-it
pipeline_tag: text-generation
tags:
  - Google
  - Gemini
  - Gemma-3
  - LLm

Gemma-3-1b-it Q4_0 Quantized Model

This is a Q4_0 quantized version of the google/gemma-3-1b-it model, converted to GGUF format and optimized for efficient inference. It was created using llama.cpp tools in Google Colab.

Model Details

• Base Model: google/gemma-3-1b-it
• Quantization: Q4_0 (4-bit quantization)
• Format: GGUF
• Size: ~1–1.5 GB
• Converted Using: llama.cpp (commit from April 2025)
• License: Inherits the license from google/gemma-3-1b-it

Usage

To use this model with llama.cpp:

./llama-cli -m gemma-3-1b-it-Q4_0.gguf --prompt "Hello, world!" --no-interactive

How It Was Created

1. Downloaded google/gemma-3-1b-it from Hugging Face.
2. Converted to GGUF using convert_hf_to_gguf.py.
3. Quantized to Q4_0 using llama-quantize from llama.cpp.
4. Tested in Google Colab with llama-cli.

Limitations

• Quantization may reduce accuracy compared to the original model.
• Requires llama.cpp or compatible software for inference.

Acknowledgments

• Based on the work of bartowski for GGUF quantization.
• Uses llama.cpp by Georgi Gerganov.