An Intel FPGA AI NIC implementation that offloads All-Reduce in distributed training - View it on GitHub