Ouro-2.6B

**⚠️ IMPORTANT: This model is intended for research purposes only. It is provided as-is without warranties for production use.** **Ouro-2.6B** is a 2.6 billion parameter Looped Language Model (LoopLM) that achieves exceptional parameter efficiency through iterative shared-weight computation. > **Note**: vLLM does not currently support the adaptive exit feature due to its inference optimization characteristics. When using vLLM, the model will always execute the full number of `total_ut_steps`.

译文尚未就绪，显示来源原文