Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms.

No abstract

Commodity heterogeneous systems (e.g., integrated CPUs and GPUs), now support a unified, shared memory address space for all components. Because the latency of global communication in a heterogeneous system can be prohibitively high, heterogeneous systems (unlike homogeneous CPU systems) provide synchronization mechanisms that only guarantee ordering among a subset of threads, which we call a scope. Unfortunately, the consequences and semantics of these scoped operations are not yet well understood. Without a formal and approachable model to reason about the behavior of these operations, we risk an array of portability and performance issues.In this paper, we embrace scoped synchronization with a new class of memory consistency models that add scoped synchronization to data-race-free models like those of C++ and Java. Called sequential consistency for heterogeneousrace-free (SC for HRF), the new models guarantee SC for programs with "sufficient" synchronization (no data races) of "sufficient" scope. We discuss two such models. The first, HRF-direct, works well for programs with highly regular parallelism. The second, HRF-indirect, builds on HRFdirect by allowing synchronization using different scopes in some cases involving transitive communication. We quantitatively show that HRF-indirect encourages forward-looking programs with irregular parallelism by showing up to a 10% performance increase in a task runtime for GPUs.

Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we aim to develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to selfattention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameterlimited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutional models.

We present GSPMD, an automatic, compiler-based parallelization system for common machine learning computation graphs. It allows users to write programs in the same way as for a single device, then give hints through a few annotations on how to distribute tensors, based on which GSPMD will parallelize the computation. Its representation of partitioning is simple yet general, allowing it to express different or mixed paradigms of parallelism on a wide variety of models.GSPMD infers the partitioning for every operator in the graph based on limited user annotations, making it convenient to scale up existing single-device programs. It solves several technical challenges for production usage, such as static shape constraints, uneven partitioning, exchange of halo data, and nested operator partitioning. These techniques allow GSPMD to achieve 50% to 62% compute utilization on 128 to 2048 Cloud TPUv3 cores for models with up to one trillion parameters.GSPMD produces a single program for all devices, which adjusts its behavior based on a run-time partition ID, and uses collective operators for cross-device communication. This property allows the system itself to be scalable: the compilation time stays constant with increasing number of devices.