Henry Hu

Dong Xu, Han Meng, Xinyu Chen et al.

Large language models (LLMs) training or inference across multiple nodes introduces significant pressure on GPU memory and interconnect bandwidth. The Compute Express Link (CXL) shared memory pool offers a scalable solution by enabling memory sharing across nodes, reducing over-provisioning and improving resource utilization. We propose \name, a collective communication library, leveraging the CXL shared memory pool to support cross-node GPU operations without relying on traditional RDMA-based networking. Our design addresses the challenges on synchronization, data interleaving, and communication parallelization faced by using the CXL shared memory pool for collective communications. Evaluating on multiple nodes with a TITAN-II CXL switch and six Micron CZ120 memory cards, we show that \name achieves highly efficient collective operations across hosts, demonstrating CXL's potential for scalable, memory-centric GPU communication. Our evaluation demonstrates that \name achieves average performance improvements of 1.34$\times$ for AllGather, 1.84$\times$ for Broadcast, 1.94$\times$ for Gather, and 1.04$\times$ for Scatter, compared to the original RDMA-based implementation over 200 Gbps InfiniBand. \textcolor{dong}{In addition, the evaluation with a case of LLM training shows 1.11$\times$ speedup compared with the InfiniBand while saving production cost by $2.75\times$ in hardware.}

Xue Li, Xiaonan Song, Henry Hu

Real-world deployment of Vision-Language Models (VLMs) is hindered by high computational demands, as existing architectures inefficiently process all tokens uniformly. We introduce Adaptive Token Pruning (ATP), a dynamic inference mechanism that retains only the most informative tokens based on contextual relevance. ATP operates at the vision-language interface, assigning a hybrid importance score combining ViT CLS attention (intra-modal saliency) and CLIP text-image similarity (inter-modal relevance) to keep top-K tokens for the LLM. Unlike static compression, ATP adapts to each input without modifying the backbone. Proposed as a lightweight gating module, ATP is compatible with popular backbones like BLIP-2, LLaVA, and Flamingo. Preliminary evaluations across VQAv2, GQA, and COCO indicate that ATP reduces inference FLOPs by around 40% and achieves roughly 1.5x speedups in end-to-end latency with negligible accuracy loss (less than 1%). Qualitative analyses suggest ATP preserves visual grounding and enhances interpretability. Beyond efficiency, we investigate robustness under corruptions; observations suggest adaptive pruning suppresses spurious correlations, improving stability. These findings imply that resource-constrained inference and model reliability are not competing objectives. Finally, we discuss ATP's role in efficient multimodal edge computing pipelines.