Addressing the Memory Bottleneck in AI Model Training
This addresses memory limitations for scientists and researchers developing large, state-of-the-art AI models, though it is incremental as it applies existing methods to new hardware configurations.
The paper tackled the memory bottleneck in AI model training by demonstrating the training of a deep neural network with a ~1 TB memory footprint on a single-node server using Intel-optimized TensorFlow and 2nd Generation Intel Xeon Scalable Processors, enabling large-scale model development.
Using medical imaging as case-study, we demonstrate how Intel-optimized TensorFlow on an x86-based server equipped with 2nd Generation Intel Xeon Scalable Processors with large system memory allows for the training of memory-intensive AI/deep-learning models in a scale-up server configuration. We believe our work represents the first training of a deep neural network having large memory footprint (~ 1 TB) on a single-node server. We recommend this configuration to scientists and researchers who wish to develop large, state-of-the-art AI models but are currently limited by memory.