Dissecting FLOPs along input dimensions for GreenAI cost estimations
This work provides a more accurate cost estimation method for GreenAI practitioners, though it is incremental as it refines an existing metric rather than introducing a new paradigm.
The authors tackled the problem that FLOPs do not correlate well with energy consumption on parallel hardware like GPUs, proposing a refined formula called α-FLOPs for convolutional layers to better estimate computational costs. This correction addresses discrepancies across layers and aligns more closely with real-world performance.
The term GreenAI refers to a novel approach to Deep Learning, that is more aware of the ecological impact and the computational efficiency of its methods. The promoters of GreenAI suggested the use of Floating Point Operations (FLOPs) as a measure of the computational cost of Neural Networks; however, that measure does not correlate well with the energy consumption of hardware equipped with massively parallel processing units like GPUs or TPUs. In this article, we propose a simple refinement of the formula used to compute floating point operations for convolutional layers, called α-FLOPs, explaining and correcting the traditional discrepancy with respect to different layers, and closer to reality. The notion of α-FLOPs relies on the crucial insight that, in case of inputs with multiple dimensions, there is no reason to believe that the speedup offered by parallelism will be uniform along all different axes.