Abstract:Multimodal Large Language Models (MLLMs) have shown remarkable versatility in understanding diverse multimodal data and tasks. However, these capabilities come with an increased model scale. While post-training pruning reduces model size in unimodal models, its application to MLLMs often yields limited success. Our analysis discovers that conventional methods fail to account for the unique token attributes across layers and modalities inherent to MLLMs. Inspired by this observation, we propose TAMP, a simple yet effective pruning framework tailored for MLLMs, featuring two key components: (1) Diversity-Aware Sparsity, which adjusts sparsity ratio per layer based on diversities among multimodal output tokens, preserving more parameters in high-diversity layers; and (2) Adaptive Multimodal Input Activation, which identifies representative multimodal input tokens using attention scores to guide unstructured weight pruning. We validate our method on two state-of-the-art MLLMs: LLaVA-NeXT, designed for vision-language tasks, and VideoLLaMA2, capable of processing audio, visual, and language modalities. Empirical experiments across various multimodal evaluation benchmarks demonstrate that each component of our approach substantially outperforms existing pruning techniques.
Abstract:Large-scale graphs with billions of edges are ubiquitous in many industries, science, and engineering fields such as recommendation systems, social graph analysis, knowledge base, material science, and biology. Graph neural networks (GNN), an emerging class of machine learning models, are increasingly adopted to learn on these graphs due to their superior performance in various graph analytics tasks. Mini-batch training is commonly adopted to train on large graphs, and data parallelism is the standard approach to scale mini-batch training to multiple GPUs. In this paper, we argue that several fundamental performance bottlenecks of GNN training systems have to do with inherent limitations of the data parallel approach. We then propose split parallelism, a novel parallel mini-batch training paradigm. We implement split parallelism in a novel system called gsplit and show that it outperforms state-of-the-art systems such as DGL, Quiver, and PaGraph.
Abstract:Representation learning is a fundamental task in machine learning. It consists of learning the features of data items automatically, typically using a deep neural network (DNN), instead of selecting hand-engineered features that typically have worse performance. Graph data requires specific algorithms for representation learning such as DeepWalk, node2vec, and GraphSAGE. These algorithms first sample the input graph and then train a DNN based on the samples. It is common to use GPUs for training, but graph sampling on GPUs is challenging. Sampling is an embarrassingly parallel task since each sample can be generated independently. However, the irregularity of graphs makes it hard to use GPU resources effectively. Existing graph processing, mining, and representation learning systems do not effectively parallelize sampling and this negatively impacts the end-to-end performance of representation learning. In this paper, we present NextDoor, the first system specifically designed to perform graph sampling on GPUs. NextDoor introduces a high-level API based on a novel paradigm for parallel graph sampling called transit-parallelism. We implement several graph sampling applications, and show that NextDoor runs them orders of magnitude faster than existing systems