Sparse Checkpointing for Fast and Reliable MoE Training
This provides a scalable fault-tolerance solution for training large MoE models, addressing a domain-specific bottleneck in distributed AI systems.
The paper tackles the problem of high checkpointing overheads in Mixture-of-Experts (MoE) model training, which degrade efficiency due to frequent failures, and presents MoEvement, a system that reduces checkpointing overhead by up to 4× and recovery overhead by up to 31×, achieving up to 8× overall training speedup.
As large language models scale, training them requires thousands of GPUs over extended durations--making frequent failures an inevitable reality. While checkpointing remains the primary fault-tolerance mechanism, existing methods fall short when applied to Mixture-of-Experts (MoE) models. Due to their substantially larger training state, MoE models exacerbate checkpointing overheads, often causing costly stalls or prolonged recovery that severely degrade training efficiency. We present MoEvement, a distributed, in-memory checkpointing system tailored for MoE models. MoEvement is built on three key ideas: (1) sparse checkpointing, which incrementally snapshots subsets of experts across iterations to reduce overhead; (2) a sparse-to-dense checkpoint conversion mechanism that incrementally reconstructs consistent dense checkpoints from sparse snapshots; and (3) upstream logging of activations and gradients at pipeline-stage boundaries, enabling localized recovery without re-executing unaffected workers. Evaluations across diverse MoE models with up to 64 experts show that MoEvement reduces checkpointing overhead by up to $4\times$ and recovery overhead by up to $31\times$ compared to state-of-the-art approaches, sustaining ETTR $\ge 0.94$ even under frequent failures (MTBF as low as 10 minutes) and delivering up to $8\times$ overall training speedup, all without compromising synchronous training semantics. Overall, MoEvement offers a robust and scalable fault-tolerance solution for the next generation of sparsely activated models.