An instruction-tuned text embedding model based on Qwen2 with 1.5 billion parameters for retrieval tasks.