AurIA-G3-v1 Overview
AurIA-G3-v1 is a 4.3 billion parameter language model developed by wallacebf, specifically fine-tuned and optimized for efficient deployment. This model leverages the Gemma architecture and has been converted to the GGUF format using Unsloth, which also facilitated its training at twice the speed.
Key Characteristics
- Architecture: Based on the Gemma model family.
- Parameter Count: 4.3 billion parameters.
- Format: Provided in GGUF format for compatibility with tools like
llama.cpp. - Optimization: Fine-tuned and converted using Unsloth for enhanced performance and efficiency.
- Context Length: Supports a context length of 32768 tokens.
- GGUF Compatibility: Includes adjustments to the BOS token behavior to ensure seamless integration with GGUF-based inference engines.
Available Versions
The model is distributed with two primary GGUF files:
gemma-3-4b-it-abliterated.Q8_0.gguf: A quantized version suitable for text-only applications.gemma-3-4b-it-abliterated.F16-mmproj.gguf: A version that includes multimodal projection, designed for multimodal use cases.
Usage
AurIA-G3-v1 can be used with llama.cpp tools. For text-only applications, llama-cli is recommended, while llama-mtmd-cli is suitable for multimodal tasks.