Dynamic Batch Processing with FlexiDecode Scheduler for Efficient LLM Inference in IIoT | TUP Journals & Magazine | IEEE Xplore