ORANSight Gemma-2B Instruct: An O-RAN Expert Assistant

NextGLab/ORANSight_Gemma_2_2B_Instruct is the inaugural model in the ORANSight family, developed by NextG lab@ NC State. This instruction-tuned model, built on the Gemma architecture, is specifically designed to function as an expert assistant in the domain of Open Radio Access Networks (O-RAN).

Key Capabilities

• O-RAN Expertise: Fine-tuned to understand and explain complex concepts related to O-RAN, such as the E2 interface.
• Context Window: Features an 8192-token context window, allowing for processing of substantial O-RAN related queries and information.
• Framework: Fine-tuned using the Unsloth framework.

Good For

• O-RAN Specific Q&A: Ideal for developers and researchers seeking explanations or information on various aspects of Open Radio Access Networks.
• Educational Tools: Can be integrated into tools or platforms requiring an AI assistant with specialized knowledge in O-RAN.

This model leverages foundational work, as acknowledged by the forthcoming paper "Oran-bench-13k: An open source benchmark for assessing llms in open radio access networks" by Gajjar and Shah (2024).