Overview

radan01/galaxy-qa-merged is a 1.5 billion parameter instruction-tuned language model based on the Qwen2.5 architecture. Developed by radan01, this model was specifically finetuned using the Unsloth library and Huggingface's TRL library, which facilitated a 2x acceleration in its training process. It operates under an Apache-2.0 license.

Key Capabilities

• Efficient Training: Leverages Unsloth for significantly faster finetuning.
• Instruction-Tuned: Optimized for following instructions and generating relevant responses.
• Question Answering: Designed to excel in question-answering scenarios.
• Context Length: Supports a substantial context window of 32768 tokens, allowing for processing longer inputs.

Good For

• Applications requiring a compact yet capable model for question-answering.
• Scenarios where efficient inference and deployment of a 1.5B parameter model are beneficial.
• Developers looking for a Qwen2.5-based model with accelerated training origins.