Extreme Tensoring for Low-Memory Preconditioning
This addresses memory efficiency for training billion-parameter models, offering a significant reduction in hardware constraints.
The paper tackles the problem of high memory consumption in training large models by proposing extreme tensoring for low-memory preconditioning, reducing optimizer memory overhead by three orders of magnitude without performance degradation in a large-scale NLP model.
State-of-the-art models are now trained with billions of parameters, reaching hardware limits in terms of memory consumption. This has created a recent demand for memory-efficient optimizers. To this end, we investigate the limits and performance tradeoffs of memory-efficient adaptively preconditioned gradient methods. We propose extreme tensoring for high-dimensional stochastic optimization, showing that an optimizer needs very little memory to benefit from adaptive preconditioning. Our technique applies to arbitrary models (not necessarily with tensor-shaped parameters), and is accompanied by regret and convergence guarantees, which shed light on the tradeoffs between preconditioner quality and expressivity. On a large-scale NLP model, we reduce the optimizer memory overhead by three orders of magnitude, without degrading performance.