airllm and Recurrent-LLM

These are complements: AirLLM enables efficient inference of large models on limited hardware, while Recurrent-LLM provides a specialized architecture for long-form text generation, and using them together would allow resource-constrained systems to generate extended content through recurrent token generation.