ROAM: memory-efficient large DNN training via optimized operator ordering and memory layout
This addresses memory efficiency for researchers and practitioners training large models, offering a novel optimization approach that is not incremental but provides substantial improvements.
The paper tackles the problem of high memory requirements in training large deep neural networks by proposing ROAM, a method that optimizes operator ordering and memory layout at the computation graph level, resulting in a memory reduction of up to 35.7% and a speedup of 53.7x compared to existing methods.
As deep learning models continue to increase in size, the memory requirements for training have surged. While high-level techniques like offloading, recomputation, and compression can alleviate memory pressure, they also introduce overheads. However, a memory-efficient execution plan that includes a reasonable operator execution order and tensor memory layout can significantly increase the models' memory efficiency and reduce overheads from high-level techniques. In this paper, we propose ROAM which operates on computation graph level to derive memory-efficient execution plan with optimized operator order and tensor memory layout for models. We first propose sophisticated theories that carefully consider model structure and training memory load to support optimization for large complex graphs that have not been well supported in the past. An efficient tree-based algorithm is further proposed to search task divisions automatically, along with delivering high performance and effectiveness to solve the problem. Experiments show that ROAM achieves a substantial memory reduction of 35.7%, 13.3%, and 27.2% compared to Pytorch and two state-of-the-art methods and offers a remarkable 53.7x speedup. The evaluation conducted on the expansive GPT2-XL further validates ROAM's scalability.