Reversible Vision Transformers
This addresses the problem of high GPU memory requirements for training large vision models, making them more accessible for hardware-limited settings, though it is an incremental improvement on existing architectures.
The paper tackles the memory inefficiency of scaling Vision Transformers by introducing Reversible Vision Transformers, which reduce memory footprint by up to 15.5x while maintaining similar accuracy and parameters, and increase throughput by up to 2.3x for deeper models.
We present Reversible Vision Transformers, a memory efficient architecture design for visual recognition. By decoupling the GPU memory requirement from the depth of the model, Reversible Vision Transformers enable scaling up architectures with efficient memory usage. We adapt two popular models, namely Vision Transformer and Multiscale Vision Transformers, to reversible variants and benchmark extensively across both model sizes and tasks of image classification, object detection and video classification. Reversible Vision Transformers achieve a reduced memory footprint of up to 15.5x at roughly identical model complexity, parameters and accuracy, demonstrating the promise of reversible vision transformers as an efficient backbone for hardware resource limited training regimes. Finally, we find that the additional computational burden of recomputing activations is more than overcome for deeper models, where throughput can increase up to 2.3x over their non-reversible counterparts. Full code and trained models are available at https://github.com/facebookresearch/slowfast. A simpler, easy to understand and modify version is also available at https://github.com/karttikeya/minREV