TAH-QUANT: Effective Activation Quantization in Pipeline Parallelism over Slow Network
This addresses network bandwidth limitations for distributed LLM training, though it appears incremental as an improvement over existing compression methods.
The paper tackles the network communication bottleneck in decentralized pipeline-parallel training of large language models by introducing TAH-Quant, an activation quantization framework that achieves 3-4 bit quantization with up to 4.3× end-to-end speedup without compromising convergence.
Decentralized training of large language models offers the opportunity to pool computational resources across geographically distributed participants but faces significant network communication bottlenecks, particularly in pipeline-parallel settings. While pipeline parallelism partitions model layers across devices to handle large-scale models, it necessitates frequent communication of intermediate activations, creating challenges when network bandwidth is limited. Existing activation compression methods, such as AQ-SGD, mitigate quantization-induced errors through error compensation but impose prohibitive memory overhead by requiring storage of previous activations. To address these issues, we introduce TAH-Quant (Tile-wise Adaptive Hadamard Quantization), a novel activation quantization framework designed specifically for pipeline parallelism. Our approach integrates fine-grained tile-wise quantization for precise control, entropy-guided token-level adaptive bit allocation for optimal bit usage, and a Hadamard-based transform with pivot element swapping to effectively suppress quantization outliers. We further provide a theoretical analysis, proving that pipeline parallel training equipped with TAH-Quant maintains a convergence rate of $\mathcal{O}(1/\sqrt{T})$, matching that of vanilla stochastic gradient descent. Extensive experiments on diverse LLM tasks demonstrate that TAH-Quant achieves aggressive activation quantization (3-4 bits) ratio, which provides up to 4.3$\times$ end-to-end speedup without compromising training convergence, matches state-of-the-art methods, incurs no extra memory overhead, and generalizes well across different training scenarios.