Jun-Liang Lin

Jun-Liang Lin, Kamesh Madduri, Mahmut Taylan Kandemir

Graph foundation models have demonstrated remarkable adaptability across diverse downstream tasks through large-scale pretraining on graphs. However, existing implementations of the backbone model, graph transformers, are typically limited to single-GPU systems, leading to long training times or out-of-memory issues on large graphs. Moreover, parallelizing graph transformer training over the full graph is challenging, as efficiency depends heavily on both the graph structure and system characteristics, such as bandwidth and memory capacity. In this work, we introduce a distributed training framework for graph transformers, which automatically selects and optimizes parallelization strategies based on the graph structure and hardware configuration. With our implementation of distributed sparse operations, we accelerate sparse graph attention by up to 3.8x and reduce memory consumption by 78% compared to state-of-the-art frameworks. On large graph benchmarks, our proposed framework achieves up to 6x speedup with system scaling up to 8 GPUs. These results demonstrate that the proposed framework improves the scalability of graph transformers, bringing them closer to serving as practical graph foundation models.

Jun-Liang Lin, Sheng-De Wang

The inference of Neural Networks is usually restricted by the resources (e.g., computing power, memory, bandwidth) on edge devices. In addition to improving the hardware design and deploying efficient models, it is possible to aggregate the computing power of many devices to enable the machine learning models. In this paper, we proposed a novel method of exploiting model parallelism to separate a neural network for distributed inferences. To achieve a better balance between communication latency, computation latency, and performance, we adopt neural architecture search (NAS) to search for the best transmission policy and reduce the amount of communication. The best model we found decreases by 86.6% of the amount of data transmission compared to the baseline and does not impact performance much. Under proper specifications of devices and configurations of models, our experiments show that the inference of large neural networks on edge clusters can be distributed and accelerated, which provides a new solution for the deployment of intelligent applications in the internet of things (IoT).