Retrieval finds a small number of relevant candidates from a large corpus for information retrieval and recommendation applications. A key component of retrieval is to model (user, item) similarity, which is commonly represented as the dot product of two learned embeddings. This formulation permits efficient inference, commonly known as Maximum Inner Product Search (MIPS). Despite its popularity, dot products cannot capture complex user-item interactions, which are multifaceted and likely high rank. We hence examine non-dot-product retrieval settings on accelerators, and propose mixture of logits (MoL), which models (user, item) similarity as an adaptive composition of elementary similarity functions. This new formulation is expressive, capable of modeling high rank (user, item) interactions, and further generalizes to the long tail. When combined with a hierarchical retrieval strategy, h-indexer, we are able to scale up MoL to 100M corpus on a single GPU with latency comparable to MIPS baselines. On public datasets, our approach leads to uplifts of up to 77.3% in hit rate (HR). Experiments on a large recommendation surface at Meta showed strong metric gains and reduced popularity bias, validating the proposed approach's performance and improved generalization.

We investigate the impact of aggressive low-precision representations of weights and activations in two families of large LSTM-based architectures for Automatic Speech Recognition (ASR): hybrid Deep Bidirectional LSTM -Hidden Markov Models (DBLSTM-HMMs) and Recurrent Neural Network -Transducers (RNN-Ts).Using a 4-bit integer representation, a naïve quantization approach applied to the LSTM portion of these models results in significant Word Error Rate (WER) degradation.On the other hand, we show that minimal accuracy loss is achievable with an appropriate choice of quantizers and initializations.In particular, we customize quantization schemes depending on the local properties of the network, improving recognition performance while limiting computational time.We demonstrate our solution on the Switchboard (SWB) and CallHome (CH) test sets of the NIST Hub5-2000 evaluation.DBLSTM-HMMs trained with 300 or 2000 hours of SWB data achieves <0.5% and <1% average WER degradation, respectively.On the more challenging RNN-T models, our quantization strategy limits degradation in 4-bit inference to 1.3%.

The growing prevalence and computational demands of Artificial Intelligence (AI) workloads has led to widespread use of hardware accelerators in their execution. Scaling the performance of AI accelerators across generations is pivotal to their success in commercial deployments. The intrinsic error-resilient nature of AI workloads present a unique opportunity for performance/energy improvement through precision scaling. Motivated by the recent algorithmic advances in precision scaling for inference and training, we designed RaPiD1, a 4-core AI accelerator chip supporting a spectrum of precisions, namely, 16 and 8-bit floating-point and 4 and 2-bit fixed-point. The 36mm2RaPiD chip fabricated in 7nm EUV technology delivers a peak 3.5 TFLOPS/W in HFP8 mode and 16.5 TOPS/W in INT4 mode at nominal voltage. Using a performance model calibrated to within 1% of the measurement results, we evaluated DNN inference using 4-bit fixed-point representation for a 4-core 1 RaPiD chip system and DNN training using 8-bit floating point representation for a 768 TFLOPs AI system comprising 4 32-core RaPiD chips. Our results show INT4 inference for batch size of 1 achieves 3 - 13.5 (average 7) TOPS/W and FP8 training for a mini-batch of 512 achieves a sustained 102 - 588 (average 203) TFLOPS across a wide range of applications.

Large-scale distributed training of Deep Neural Networks (DNNs) on state-of-the-art platforms are expected to be severely communication constrained. To overcome this limitation, numerous gradient compression techniques have been proposed and have demonstrated high compression ratios. However, most existing compression methods do not scale well to large scale distributed systems (due to gradient build-up) and / or lack evaluations in large datasets. To mitigate these issues, we propose a new compression technique, Scalable Sparsified Gradient Compression (ScaleComp), that (i) leverages similarity in the gradient distribution amongst learners to provide a commutative compressor and keep communication cost constant to worker number and (ii) includes low-pass filter in local gradient accumulations to mitigate the impacts of large batch size training and significantly improve scalability. Using theoretical analysis, we show that ScaleComp provides favorable convergence guarantees and is compatible with gradient all-reduce techniques. Furthermore, we experimentally demonstrate that ScaleComp has small overheads, directly reduces gradient traffic and provides high compression rates (70-150X) and excellent scalability (up to 64-80 learners and 10X larger batch sizes over normal training) across a wide range of applications (image, language, and speech) without significant accuracy loss.

In this paper, we propose a number of novel techniques and numerical representation formats that enable, for the very first time, the precision of training systems to be aggressively scaled from 8-bits to 4-bits. To enable this advance, we explore a novel adaptive Gradient Scaling technique (Gradscale) that addresses the challenges of insufficient range and resolution in quantized gradients as well as explores the impact of quantization errors observed during model training. We theoretically analyze the role of bias in gradient quantization and propose solutions that mitigate the impact of this bias on model convergence. Finally, we examine our techniques on a spectrum of deep learning models in computer vision, speech, and NLP. In combination with previously proposed solutions for 4-bit quantization of weight and activation tensors, 4-bit training shows a non-significant loss in accuracy across application domains while enabling significant hardware acceleration (> 7X over state-of-the-art FP16 systems).

Advances in deep neural networks (DNNs) and the availability of massive real-world data have enabled superhuman levels of accuracy on many AI tasks and ushered the explosive growth of AI workloads across the spectrum of computing devices. However, their superior accuracy comes at a high computational cost, which necessitates approaches beyond traditional computing paradigms to improve their operational efficiency. Leveraging the application-level insight of error resilience, we demonstrate how approximate computing (AxC) can significantly boost the efficiency of AI platforms and play a pivotal role in the broader adoption of AI-based applications and services. To this end, we present RaPiD, a multi-tera operations per second (TOPS) AI hardware accelerator core (fabricated at 14-nm technology) that we built from the ground-up using AxC techniques across the stack including algorithms, architecture, programmability, and hardware. We highlight the workload-guided systematic explorations of AxC techniques for AI, including custom number representations, quantization/pruning methodologies, mixed-precision architecture design, instruction sets, and compiler technologies with quality programmability, employed in the RaPiD accelerator.

The resilience of Deep Learning (DL) training and inference workloads to low-precision computations, coupled with the demand for power-and area-efficient hardware accelerators for these workloads, has led to the emergence of 16-bit floating point formats as the precision of choice for DL hardware accelerators. This paper describes our optimized 16-bit format that has 6 exponent bits and 9 fraction bits, derived from a study of the range of values encountered in DL applications. We demonstrate that our format preserves the accuracy of DL networks, and we compare its ease-of-use for DL against IEEE-754 half-precision (5 exponent bits and 10 fraction bits) and bfloat16 (8 exponent bits and 7 fraction bits). Further, our format eliminated sub-normals and simplifies rounding modes and handling of corner cases. This streamlines floating-point unit logic and enables realization of a compact power-efficient computation engine.

Reducing the numerical precision of data and computation is extremely effective in accelerating deep learning training workloads. Towards this end, 8-bit floating point representations (FP8) were recently proposed for DNN training. However, its applicability was demonstrated on a few selected models only and significant degradation is observed when popular networks such as MobileNet and Transformer are trained using FP8. This degradation is due to the inherent precision requirement difference in the forward and backward passes of DNN training. Using theoretical insights, we propose a hybrid FP8 (HFP8) format and DNN end-to-end distributed training procedure. We demonstrate, using HFP8, the successful training of deep learning models across a whole spectrum of applications including Image Classification, Object Detection, Language and Speech without accuracy degradation. Finally, we demonstrate that, by using the new 8 bit format, we can directly quantize a pre-trained model down to 8-bits without losing accuracy by simply fine-tuning batch normalization statistics. These novel techniques enable a new generations of 8-bit hardware that are robust for building and deploying neural network models.

Ferroelectric Field Effect Transistor (FeFET) is a promising nonvolatile device which provides high integration density, fast programming speed, and excellent CMOS compatibility. In general, the non-volatility of FeFET is impacted by its physical structure and there is a trade-off between data retention time and device endurance. To improve the cell endurance, for example, the ferroelectric layer of FeFET needs to be programmed to a low polarization level, leading to a short retention time. In ferroelectric DRAM (FeDRAM) design, degradation in FeFET retention time and write-read disturbance requires the FeDRAM cells to b e periodically refreshed in order to prevent data loss. In this work, we propose a novel adaptive refreshing and read voltage control scheme to minimize the energy overheads associated with FeDRAM refreshing while still achieve high cell access reliability.