We study a novel and important communication pattern in large-scale model-parallel deep learning (DL), which we call cross-mesh resharding. This pattern emerges when the two paradigms of model parallelism - intra-operator and inter-operator parallelism - are combined to support large models on large clusters. In cross-mesh resharding, a sharded tensor needs to be sent from a source device mesh to a destination device mesh, on which the tensor may be distributed with the same or different layouts. We formalize this as a many-to-many multicast communication problem, and show that existing approaches either are sub-optimal or do not generalize to different network topologies or tensor layouts, which result from different model architectures and parallelism strategies. We then propose two contributions to address cross-mesh resharding: an efficient broadcast-based communication system, and an "overlapping-friendly" pipeline schedule. On microbenchmarks, our overall system outperforms existing ones by up to 10x across various tensor and mesh layouts. On end-to-end training of two large models, GPT-3 and U-Transformer, we improve throughput by 10% and 50%, respectively.

在各种设备上部署深度学习模型已成为一个重要的话题。硬件专业化的浪潮为多维张量计算带来了一套多样化的加速度原始图。这些新的加速原始基原料以及新兴的机器学习模型带来了巨大的工程挑战。在本文中，我们提出了Tensorir，这是一种编译器抽象，用于通过这些张量计算原始素优化程序。Tensorir概括了现有机器学习编译器中使用的循环巢表示，以将张量计算作为一流的公民。最后，我们在抽象之上构建了一个端到端框架，以自动优化给定的张量计算原始图的深度学习模型。实验结果表明，Tensorir编译会自动使用给定硬件后端的张量计算原始图，并提供与跨平台的最新手工精制系统竞争性能的性能。

科学家越来越依靠Python工具使用丰富的，类似于Numpy的表达式执行可扩展的分布式内存阵列操作。但是，这些工具中的许多工具都依赖于针对抽象任务图进行了优化的动态调度程序，这些调度图通常遇到内存和网络带宽相关的瓶颈，这是由于亚最佳数据和操作员的放置决策。在消息传递接口（MPI）（例如Scalapack和Slate）上构建的工具具有更好的缩放属性，但是这些解决方案需要使用专门的知识。在这项工作中，我们提出了NUMS，这是一个数组编程库，可在基于任务的分布式系统上优化类似Numpy的表达式。这是通过称为负载模拟层次调度（LSHS）的新型调度程序来实现的。 LSHS是一种本地搜索方法，可通过最大程度地减少分布式系统中任何给定节点上的最大内存和网络加载来优化操作员放置。再加上用于负载平衡数据布局的启发式，我们的方法能够在某些常见的数值操作上达到通信下限，我们的经验研究表明，LSHS通过减少2倍的降低2倍来增强RAR上的性能，需要减少4倍的内存， ，在逻辑回归问题上减少10倍的执行时间。在Terabyte尺度数据上，NUMS在DGEMM上实现了竞争性能，与Dask ML和Spark的Mllib相比，在键盘分解的密钥操作中，DASK高达20倍的速度以及logistic回归的2倍加速。

训练大型神经网络（NN）模型需要广泛的记忆资源，而激活压缩训练（ACT）是减少训练记忆足迹的一种有前途的方法。本文介绍了GACT，这是一个ACT框架，旨在支持具有有限域知识的通用NN体系结构的广泛机器学习任务。通过分析ACT近似梯度的线性化版本，我们证明了GACT的收敛性，而没有有关操作员类型或模型体系结构的先验知识。为了使训练保持稳定，我们提出了一种算法，该算法通过估计运行时对梯度的影响来决定每个张量的压缩比。我们将GACT实施为Pytorch库，很容易适用于任何NN体系结构。GACT将卷积NN，变压器和图形NNS的激活记忆降低到8.1倍，从而使4.2倍至24.7倍的训练能够较大，而精度损失可忽略不计。

ALPA通过生成统一数据，操作员和管道并行性的执行计划来自动对大型深度学习（DL）模型的模型平行训练。现有的模型并行训练系统要求用户手动创建并行化计划，或者自动从有限的模型并行性配置中生成一个计划。它们不足以在分布式计算设备上扩展复杂的DL模型。 ALPA通过将并行性视为两个层次级别来分配大型DL模型的训练：操作员和操作员并行性。基于它，ALPA构建了一个新的分层空间，用于大规模的模型并行执行计划。 ALPA设计了许多汇编，以在每个并行性级别自动得出有效的并行执行计划。 ALPA实现了有效的运行时，以在分布式计算设备上协调两级并行执行。我们的评估表明，ALPA生成的并行化计划，即使在其设计的型号上，也可以匹配或超过手动模型并联训练系统。与专业系统不同，ALPA还推广到具有异质体系结构和模型的模型，而没有手动设计的计划。 ALPA的源代码可在https://github.com/alpa-projects/alpa上公开获得

There is an increasing need to bring machine learning to a wide diversity of hardware devices. Current frameworks rely on vendor-specific operator libraries and optimize for a narrow range of server-class GPUs. Deploying workloads to new platforms -such as mobile phones, embedded devices, and accelerators (e.g., FPGAs, ASICs) -requires significant manual effort. We propose TVM, a compiler that exposes graph-level and operator-level optimizations to provide performance portability to deep learning workloads across diverse hardware back-ends. TVM solves optimization challenges specific to deep learning, such as high-level operator fusion, mapping to arbitrary hardware primitives, and memory latency hiding. It also automates optimization of low-level programs to hardware characteristics by employing a novel, learning-based cost modeling method for rapid exploration of code optimizations. Experimental results show that TVM delivers performance across hardware back-ends that are competitive with state-ofthe-art, hand-tuned libraries for low-power CPU, mobile GPU, and server-class GPUs. We also demonstrate TVM's ability to target new accelerator back-ends, such as the FPGA-based generic deep learning accelerator.The system is open sourced and in production use inside several major companies.

Masked image modeling (MIM) performs strongly in pre-training large vision Transformers (ViTs). However, small models that are critical for real-world applications cannot or only marginally benefit from this pre-training approach. In this paper, we explore distillation techniques to transfer the success of large MIM-based pre-trained models to smaller ones. We systematically study different options in the distillation framework, including distilling targets, losses, input, network regularization, sequential distillation, etc, revealing that: 1) Distilling token relations is more effective than CLS token- and feature-based distillation; 2) An intermediate layer of the teacher network as target perform better than that using the last layer when the depth of the student mismatches that of the teacher; 3) Weak regularization is preferred; etc. With these findings, we achieve significant fine-tuning accuracy improvements over the scratch MIM pre-training on ImageNet-1K classification, using all the ViT-Tiny, ViT-Small, and ViT-base models, with +4.2%/+2.4%/+1.4% gains, respectively. Our TinyMIM model of base size achieves 52.2 mIoU in AE20K semantic segmentation, which is +4.1 higher than the MAE baseline. Our TinyMIM model of tiny size achieves 79.6% top-1 accuracy on ImageNet-1K image classification, which sets a new record for small vision models of the same size and computation budget. This strong performance suggests an alternative way for developing small vision Transformer models, that is, by exploring better training methods rather than introducing inductive biases into architectures as in most previous works. Code is available at https://github.com/OliverRensu/TinyMIM.

Dataset distillation has emerged as a prominent technique to improve data efficiency when training machine learning models. It encapsulates the knowledge from a large dataset into a smaller synthetic dataset. A model trained on this smaller distilled dataset can attain comparable performance to a model trained on the original training dataset. However, the existing dataset distillation techniques mainly aim at achieving the best trade-off between resource usage efficiency and model utility. The security risks stemming from them have not been explored. This study performs the first backdoor attack against the models trained on the data distilled by dataset distillation models in the image domain. Concretely, we inject triggers into the synthetic data during the distillation procedure rather than during the model training stage, where all previous attacks are performed. We propose two types of backdoor attacks, namely NAIVEATTACK and DOORPING. NAIVEATTACK simply adds triggers to the raw data at the initial distillation phase, while DOORPING iteratively updates the triggers during the entire distillation procedure. We conduct extensive evaluations on multiple datasets, architectures, and dataset distillation techniques. Empirical evaluation shows that NAIVEATTACK achieves decent attack success rate (ASR) scores in some cases, while DOORPING reaches higher ASR scores (close to 1.0) in all cases. Furthermore, we conduct a comprehensive ablation study to analyze the factors that may affect the attack performance. Finally, we evaluate multiple defense mechanisms against our backdoor attacks and show that our attacks can practically circumvent these defense mechanisms.

Benefiting from the intrinsic supervision information exploitation capability, contrastive learning has achieved promising performance in the field of deep graph clustering recently. However, we observe that two drawbacks of the positive and negative sample construction mechanisms limit the performance of existing algorithms from further improvement. 1) The quality of positive samples heavily depends on the carefully designed data augmentations, while inappropriate data augmentations would easily lead to the semantic drift and indiscriminative positive samples. 2) The constructed negative samples are not reliable for ignoring important clustering information. To solve these problems, we propose a Cluster-guided Contrastive deep Graph Clustering network (CCGC) by mining the intrinsic supervision information in the high-confidence clustering results. Specifically, instead of conducting complex node or edge perturbation, we construct two views of the graph by designing special Siamese encoders whose weights are not shared between the sibling sub-networks. Then, guided by the high-confidence clustering information, we carefully select and construct the positive samples from the same high-confidence cluster in two views. Moreover, to construct semantic meaningful negative sample pairs, we regard the centers of different high-confidence clusters as negative samples, thus improving the discriminative capability and reliability of the constructed sample pairs. Lastly, we design an objective function to pull close the samples from the same cluster while pushing away those from other clusters by maximizing and minimizing the cross-view cosine similarity between positive and negative samples. Extensive experimental results on six datasets demonstrate the effectiveness of CCGC compared with the existing state-of-the-art algorithms.

As one of the prevalent methods to achieve automation systems, Imitation Learning (IL) presents a promising performance in a wide range of domains. However, despite the considerable improvement in policy performance, the corresponding research on the explainability of IL models is still limited. Inspired by the recent approaches in explainable artificial intelligence methods, we proposed a model-agnostic explaining framework for IL models called R2RISE. R2RISE aims to explain the overall policy performance with respect to the frames in demonstrations. It iteratively retrains the black-box IL model from the randomized masked demonstrations and uses the conventional evaluation outcome environment returns as the coefficient to build an importance map. We also conducted experiments to investigate three major questions concerning frames' importance equality, the effectiveness of the importance map, and connections between importance maps from different IL models. The result shows that R2RISE successfully distinguishes important frames from the demonstrations.