I understand that FLOPS means floating-point operations per second, and throughput is the number of inputs (for example, images) per second. If a model has higher FLOPS, it means it performs faster.
However, in the article Container: Context Aggregation Network, they show that:
The container has higher FLOPS and less throughput, while the container-light has lower FLOPS and higher throughput.
What is the reason for that?