We present Pippo, a generative model capable of producing 1K resolution dense turnaround videos of a person from a single casually clicked photo. Pippo is a multi-view diffusion transformer and does not require any additional inputs — e.g., a fitted parametric model or camera parameters of the input image. We pre-train Pippo on 3B human images without captions, and conduct multi-view mid-training and post-training on studio captured humans. During mid-training, to quickly absorb the studio dataset, we denoise several (upto 48) views at low-resolution, and encode target cameras coarsely using a shallow MLP. During post-training, we denoise fewer views at high-resolution and use pixel-aligned controls (e.g., Spatial anchor and Plucker rays) to enable 3D consistent generations. At inference, we propose an attention biasing technique that allows Pippo to simultaneously generate greater than 5× as many views as seen during training. Finally, we also introduce an improved metric to evaluate 3D consistency of multi-view generations, and show that Pippo outperforms existing works on multi-view human generation from a single image.

As training billion-scale transformers becomes increasingly common, employing multiple distributed GPUs along with parallel training methods has become a standard practice. However, existing transformer designs suffer from significant communication overhead, especially in Tensor Parallelism (TP), where each block’s MHA–MLP connection requires an all-reduce communication. Through our investigation, we show that the MHA-MLP connections can be bypassed for efficiency, while the attention output of the first layer can serve as an alternative signal for the bypassed connection. Motivated by the observations, we propose FAL (First Attentions Last), an efficient transformer architecture that redirects the first MHA output to the MLP inputs of the following layers, eliminating the per-block MHA-MLP connections. This removes the all-reduce communication and enables parallel execution of MHA and MLP on a single GPU. We also introduce FAL+, which adds the normalized first attention output to the MHA outputs of the following layers to augment the MLP input for the model quality. Our evaluation shows that FAL reduces multi-GPU training time by up to 44%, improves single-GPU throughput by up to 1.18×, and achieves better perplexity compared to the baseline GPT. FAL+ achieves even lower perplexity without increasing the training time than the baseline. Codes are available at: https://casl-ku.github.io/FAL/

Creating photorealistic avatars of existing people currently requires extensive person-specific data capture, which is usually only accessible to the VFX industry and not the general public. Our work aims to address this drawback by relying only on a short mobile phone capture to obtain a drivable 3D head avatar that matches a person's likeness faithfully. In contrast to existing approaches, our architecture avoids the complex task of directly modeling the entire manifold of human appearance, aiming instead to generate an avatar model that can be specialized to novel identities using only small amounts of data. The model dispenses with low-dimensional latent spaces that are commonly employed for hallucinating novel identities, and instead, uses a conditional representation that can extract person-specific information at multiple scales from a high resolution registered neutral phone scan. We achieve high quality results through the use of a novel universal avatar prior that has been trained on high resolution multi-view video captures of facial performances of hundreds of human subjects. By fine-tuning the model using inverse rendering we achieve increased realism and personalize its range of motion. The output of our approach is not only a high-fidelity 3D head avatar that matches the person's facial shape and appearance, but one that can also be driven using a jointly discovered shared global expression space with disentangled controls for gaze direction. Via a series of experiments we demonstrate that our avatars are faithful representations of the subject's likeness. Compared to other state-of-the-art methods for lightweight avatar creation, our approach exhibits superior visual quality and animateability.

Numerous existing works have shown that, key to the efficiency of distributed machine learning (ML) is proper system and algorithm co-design: system design should be tailored to the unique mathematical properties of ML algorithms, and algorithms can be re-designed to better exploit the system architecture. While existing research has made attempts along this direction, many algorithmic and system properties that are characteristic of ML problems remain to be explored. Through an exploration of system-algorithm co-design, we build a new decentralized system Orpheus to support distributed training of a general class of ML models whose parameters are represented with large matrices. Training such models at scale is challenging: transmitting and checkpointing large matrices incur substantial network traffic and disk IO, which aggravates the inconsistency among parameter replicas. To cope with these challenges, Orpheus jointly exploits system and algorithm designs which (1) reduce the size and number of network messages for efficient communication, 2) incrementally checkpoint vectors for light-weight and fine-grained fault tolerance without blocking computation, 3) improve the consistency among parameter copies via periodic centralized synchronization and parameter-replicas rotation. As a result of these co-designs, communication and fault tolerance costs are linear to both matrix dimension and number of machines in the network, as opposed to being quadratic in existing systems. And the improved parameter consistency accelerates algorithmic convergence. Empirically, we show our system outperforms several existing baseline systems on training several representative large-scale ML models.

Machine learning (ML) algorithms are commonly applied to big data, using distributed systems that partition the data across machines and allow each machine to read and update all ML model parameters --- a strategy known as data parallelism. An alternative and complimentary strategy, model parallelism, partitions the model parameters for non-shared parallel access and updates, and may periodically repartition the parameters to facilitate communication. Model parallelism is motivated by two challenges that data-parallelism does not usually address: (1) parameters may be dependent, thus naive concurrent updates can introduce errors that slow convergence or even cause algorithm failure; (2) model parameters converge at different rates, thus a small subset of parameters can bottleneck ML algorithm completion. We propose scheduled model parallelism (SchMP), a programming approach that improves ML algorithm convergence speed by efficiently scheduling parameter updates, taking into account parameter dependencies and uneven convergence. To support SchMP at scale, we develop a distributed framework STRADS which optimizes the throughput of SchMP programs, and benchmark four common ML applications written as SchMP programs: LDA topic modeling, matrix factorization, sparse least-squares (Lasso) regression and sparse logistic regression. By improving ML progress per iteration through SchMP programming whilst improving iteration throughput through STRADS we show that SchMP programs running on STRADS outperform non-model-parallel ML implementations: for example, SchMP LDA and SchMP Lasso respectively achieve 10x and 5x faster convergence than recent, well-established baselines.

How can one build a distributed framework that allows efficient deployment of a wide spectrum of modern advanced machine learning (ML) programs for industrial-scale problems using Big Models (100s of billions of parameters) on Big Data (terabytes or petabytes)- Contemporary parallelization strategies employ fine-grained operations and scheduling beyond the classic bulk-synchronous processing paradigm popularized by MapReduce, or even specialized operators relying on graphical representations of ML programs. The variety of approaches tends to pull systems and algorithms design in different directions, and it remains difficult to find a universal platform applicable to a wide range of different ML programs at scale. We propose a general-purpose framework that systematically addresses data- and model-parallel challenges in large-scale ML, by leveraging several fundamental properties underlying ML programs that make them different from conventional operation-centric programs: error tolerance, dynamic structure, and nonuniform convergence; all stem from the optimization-centric nature shared in ML programs' mathematical definitions, and the iterative-convergent behavior of their algorithmic solutions. These properties present unique opportunities for an integrative system design, built on bounded-latency network synchronization and dynamic load-balancing scheduling, which is efficient, programmable, and enjoys provable correctness guarantees. We demonstrate how such a design in light of ML-first principles leads to significant performance improvements versus well-known implementations of several ML programs, allowing them to run in much less time and at considerably larger model sizes, on modestly-sized computer clusters.

What is a systematic way to efficiently apply a wide spectrum of advanced ML programs to industrial scale problems, using Big Models (up to 100 s of billions of parameters) on Big Data (up to terabytes or petabytes)? Modern parallelization strategies employ fine-grained operations and scheduling beyond the classic bulk-synchronous processing paradigm popularized by MapReduce, or even specialized graph-based execution that relies on graph representations of ML programs. The variety of approaches tends to pull systems and algorithms design in different directions, and it remains difficult to find a universal platform applicable to a wide range of ML programs at scale. We propose a general-purpose framework, Petuum, that systematically addresses data- and model-parallel challenges in large-scale ML, by observing that many ML programs are fundamentally optimization-centric and admit error-tolerant, iterative-convergent algorithmic solutions. This presents unique opportunities for an integrative system design, such as bounded-error network synchronization and dynamic scheduling based on ML program structure. We demonstrate the efficacy of these system designs versus well-known implementations of modern ML algorithms, showing that Petuum allows ML programs to run in much less time and at considerably larger model sizes, even on modestly-sized compute clusters.

We propose a parameter server system for distributed ML, which follows a Stale Synchronous Parallel (SSP) model of computation that maximizes the time computational workers spend doing useful work on ML algorithms, while still providing correctness guarantees. The parameter server provides an easy-to-use shared interface for read/write access to an ML model's values (parameters and variables), and the SSP model allows distributed workers to read older, stale versions of these values from a local cache, instead of waiting to get them from a central storage. This significantly increases the proportion of time workers spend computing, as opposed to waiting. Furthermore, the SSP model ensures ML algorithm correctness by limiting the maximum age of the stale values. We provide a proof of correctness under SSP, as well as empirical results demonstrating that the SSP model achieves faster algorithm convergence on several different ML problems, compared to fully-synchronous and asynchronous schemes.

In this paper, the stability analysis and stabilization problem for a discrete-time Markovian jump nonlinear systems (MJLNS) with time-varying delays are investigated. The time-delay is considered to be time-varying and has a upper bound. The transition probabilities of the mode jumps are considered to be completely known. Sufficient conditions for stochastic stability of the Markovian jump fuzzy systems (MJFS) are derived via the linear matrix inequality (LMI) formulation, and the design of the stabilizing controller is further given. A numerical example is used to illustrate the developed theory.

NAND flash-based storage is widely used in embedded systems due to its numerous benefits: low cost, high density, small form factor and so on. However, NAND flash-based storage is still suffering from serious performance degradation for random or small size write access. This degradation mainly comes from the physical constraints of NAND flash: erase-before-program and different unit size of erase and program operations. To overcome these constraints, we propose to use PRAM (Phase-change RAM) which supports advanced features: fast byte access capability and no requirement for erase-before-program.