RCV2023 Challenges: Benchmarking Model Training and Inference for Resource-Constrained Deep Learning