In-Place Activated BatchNorm for Memory-Optimized Training of DNNs
This work addresses memory optimization for training deep neural networks, which is a critical bottleneck for researchers and practitioners with limited hardware resources, though it is incremental in combining existing techniques into a novel layer.
The paper tackles the problem of high memory usage during deep neural network training by introducing In-Place Activated BatchNorm, which reduces memory footprint by up to 50% with only a minor increase in computation time, while achieving state-of-the-art results on semantic segmentation tasks.
In this work we present In-Place Activated Batch Normalization (InPlace-ABN) - a novel approach to drastically reduce the training memory footprint of modern deep neural networks in a computationally efficient way. Our solution substitutes the conventionally used succession of BatchNorm + Activation layers with a single plugin layer, hence avoiding invasive framework surgery while providing straightforward applicability for existing deep learning frameworks. We obtain memory savings of up to 50% by dropping intermediate results and by recovering required information during the backward pass through the inversion of stored forward results, with only minor increase (0.8-2%) in computation time. Also, we demonstrate how frequently used checkpointing approaches can be made computationally as efficient as InPlace-ABN. In our experiments on image classification, we demonstrate on-par results on ImageNet-1k with state-of-the-art approaches. On the memory-demanding task of semantic segmentation, we report results for COCO-Stuff, Cityscapes and Mapillary Vistas, obtaining new state-of-the-art results on the latter without additional training data but in a single-scale and -model scenario. Code can be found at https://github.com/mapillary/inplace_abn .