In this paper, we investigate how to push the performance limits of serving Deep Neural Network (DNN) models on CPU-based servers. Specifically, we observe that while intra-operator parallelism across multiple threads is an effective way to reduce inference latency, it provides diminishing returns. Our primary insight is that instead of running a single instance of a model with all available threads on a server, running multiple instances each with smaller batch sizes and fewer threads for intra-op parallelism can provide lower inference latency. However, the right configuration is hard to determine manually since it is workload- (DNN model and batch size used by the serving system) and deployment-dependent (number of CPU cores on server). We present Packrat, a new serving system for online inference that given a model and batch size (𝐵) algorithmically picks the optimal number of instances (𝑖), the number of threads each should be allocated (𝑡), and the batch sizes each should operate on (𝑏) that minimizes latency. Packrat is built as an extension to TorchServe and supports online reconfigurations to avoid serving downtime. Averaged across a range of batch sizes, Packrat improves inference latency by 1.43× to 1.83× on a range of commonly used DNNs.

New ML models are often derived from existing ones (e.g., through fine-tuning, quantization or distillation), forming an ecosystem where models are related to each other and can share structure or even parameter values. Managing such a large and evolving ecosystem of model derivatives is challenging. For instance, the overhead of storing all such models is high, and models may inherit bugs from related models, complicating error attribution and debugging. In this paper, we propose a model versioning and management system called MGit that makes it easier to store, test, update, and collaborate on related models. MGit introduces a lineage graph that records the relationships between models, optimizations to efficiently store model parameters, and abstractions over this lineage graph that facilitate model testing, updating and collaboration. We find that MGit works well in practice: MGit is able to reduce model storage footprint by up to 7$\times$. Additionally, in a user study with 20 ML practitioners, users complete a model updating task 3$\times$ faster on average with MGit.

While GPUs are responsible for training the vast majority of state-of-the-art deep learning models, the implications of their architecture are often overlooked when designing new deep learning (DL) models. As a consequence, modifying a DL model to be more amenable to the target hardware can significantly improve the runtime performance of DL training and inference. In this paper, we provide a set of guidelines for users to maximize the runtime performance of their transformer models. These guidelines have been created by carefully considering the impact of various model hyperparameters controlling model shape on the efficiency of the underlying computation kernels executed on the GPU. We find the throughput of models with “efficient” model shapes is up to 39% higher while preserving accuracy compared to models with a similar number of parameters but with unoptimized shapes.

The stunning qualitative improvement of text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (HEIM). Whereas previous evaluations focus mostly on image-text alignment and image quality, we identify 12 aspects, including text-image alignment, image quality, aesthetics, originality, reasoning, knowledge, bias, toxicity, fairness, robustness, multilinguality, and efficiency. We curate 62 scenarios encompassing these aspects and evaluate 26 state-of-the-art text-to-image models on this benchmark. Our results reveal that no single model excels in all aspects, with different models demonstrating different strengths. We release the generated images and human evaluation results for full transparency at https://crfm.stanford.edu/heim/latest and the code at https://github.com/stanford-crfm/helm, which is integrated with the HELM codebase

Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important, but requires an efficiency metric that is comparable across models from different providers. Unfortunately, raw runtimes measured through black-box APIs do not satisfy this property: model providers can implement software and hardware optimizations orthogonal to the model, and shared infrastructure introduces performance contention. We propose a new metric for inference efficiency called _idealized runtime_, that puts models on equal footing as though they were served on uniform hardware and software without performance contention, and a cost model to efficiently estimate this metric for autoregressive Transformer models. We also propose variants of the idealized runtime that incorporate the number and type of accelerators needed to serve the model. Using these metrics, we compare ten LLMs developed in 2022 to provide the first analysis of inference efficiency-capability tradeoffs; we make several observations from this analysis, including the fact that the superior inference runtime performance of certain APIs is often a byproduct of optimizations within the API rather than the underlying model. Our code is open sourced at https://github.com/stanford-crfm/helm-efficiency.

Many state-of-the-art ML results have been obtained by scaling up the number of parameters in existing models. However, parameters and activations for such large models often do not fit in the memory of a single accelerator device; this means that it is necessary to distribute training of large models over multiple accelerators. In this work, we propose PipeDream-2BW, a system that supports memory-efficient pipeline parallelism. PipeDream-2BW uses a novel pipelining and weight gradient coalescing strategy, combined with the double buffering of weights, to ensure high throughput, low memory footprint, and weight update semantics similar to data parallelism. In addition, PipeDream-2BW automatically partitions the model over the available hardware resources, while respecting hardware constraints such as memory capacities of accelerators and interconnect topologies. PipeDream-2BW can accelerate the training of large GPT and BERT language models by up to 20x with similar final model accuracy.

The rapid increase in sizes of state-of-the-art DNN models, and consequently the increase in the compute and memory requirements of model training, has led to the development of many execution schemes such as data parallelism, pipeline model parallelism, tensor (intra-layer) model parallelism, and various memory-saving optimizations. However, no prior work has tackled the highly complex problem of optimally partitioning the DNN computation graph across many accelerators while combining all these parallelism modes and optimizations.In this work, we introduce Piper, an efficient optimization algorithm for this problem that is based on a two-level dynamic programming approach. Our two-level approach is driven by the insight that being given tensor-parallelization techniques for individual layers (e.g., Megatron-LM's splits for transformer layers) significantly reduces the search space and makes the global problem tractable, compared to considering tensor-parallel configurations for the entire DNN operator graph.

Large language models have led to state-of-the-art accuracies across several tasks. However, training these models efficiently is challenging because: a) GPU memory capacity is limited, making it impossible to fit large models on even a multi-GPU server, and b) the number of compute operations required can result in unrealistically long training times. Consequently, new methods of model parallelism such as tensor and pipeline parallelism have been proposed. Unfortunately, naive usage of these methods leads to scaling issues at thousands of GPUs. In this paper, we show how tensor, pipeline, and data parallelism can be composed to scale to thousands of GPUs. We propose a novel interleaved pipelining schedule that can improve throughput by 10+% with memory footprint comparable to existing approaches. Our approach allows us to perform training iterations on a model with 1 trillion parameters at 502 petaFLOP/s on 3072 GPUs (per-GPU throughput of 52% of theoretical peak).

Resource allocation problems in many computer systems can be formulated as mathematical optimization problems. However, finding exact solutions to these problems using off-the-shelf solvers is often intractable for large problem sizes with tight SLAs, leading system designers to rely on cheap, heuristic algorithms. We observe, however, that many allocation problems are granular: they consist of a large number of clients and resources, each client requests a small fraction of the total number of resources, and clients can interchangeably use different resources. For these problems, we propose an alternative approach that reuses the original optimization problem formulation and leads to better allocations than domain-specific heuristics. Our technique, Partitioned Optimization Problems (POP), randomly splits the problem into smaller problems (with a subset of the clients and resources in the system) and coalesces the resulting sub-allocations into a global allocation for all clients. We provide theoretical and empirical evidence as to why random partitioning works well. In our experiments, POP achieves allocations within 1.5% of the optimal with orders-of-magnitude improvements in runtime compared to existing systems for cluster scheduling, traffic engineering, and load balancing.

Systems for ML inference are widely deployed today, but they typically optimize ML inference workloads using techniques designed for conventional data serving workloads and miss critical opportunities to leverage the statistical nature of ML. In this demo, we present Willump, an optimizer for ML inference that introduces statistically-motivated optimizations targeting ML applications whose performance bottleneck is feature computation. Willump automatically cascades feature computation for classification queries: Willump classifies most data inputs using only high-value, low-cost features selected by a cost model, improving query performance by up to 5 x without statistically significant accuracy loss. In this demo, we use interactive and easily-downloadable Jupyter notebooks to show VLDB attendees which applications Willump can speed up, how to use Willump, and how Willump produces such large performance gains.

DNN training is extremely time-consuming, necessitating efficient multi-accelerator parallelization. Current approaches to parallelizing training primarily use intra-batch parallelization, where a single iteration of training is split over the available workers, but suffer from diminishing returns at higher worker counts. We present PipeDream, a system that adds inter-batch pipelining to intra-batch parallelism to further improve parallel training throughput, helping to better overlap computation with communication and reduce the amount of communication when possible. Unlike traditional pipelining, DNN training is bi-directional, where a forward pass through the computation graph is followed by a backward pass that uses state and intermediate data computed during the forward pass. Naïve pipelining can thus result in mismatches in state versions used in the forward and backward passes, or excessive pipeline flushes and lower hardware efficiency. To address these challenges, PipeDream versions model parameters for numerically correct gradient computations, and schedules forward and backward passes of different minibatches concurrently on different workers with minimal pipeline stalls. PipeDream also automatically partitions DNN layers among workers to balance work and minimize communication. Extensive experimentation with a range of DNN tasks, models, and hardware configurations shows that PipeDream trains models to high accuracy up to 5.3X faster than commonly used intra-batch parallelism techniques.

Modern analytics applications use a diverse mix of libraries and functions. Unfortunately, there is no optimization across these libraries, resulting in performance penalties as high as an order of magnitude in many applications. To address this problem, we proposed Weld, a common runtime for existing data analytics libraries that performs key physical optimizations such as pipelining under existing, imperative library APIs. In this work, we further develop the Weld vision by designing an automatic adaptive optimizer for Weld applications, and evaluating its impact on realistic data science workloads. Our optimizer eliminates multiple forms of overhead that arise when composing imperative libraries like Pandas and NumPy, and uses lightweight measurements to make data-dependent decisions at run-time in ad-hoc workloads where no statistics are available, with sub-second overhead. We also evaluate which optimizations have the largest impact in practice and whether Weld can be integrated into libraries incrementally. Our results are promising: using our optimizer, Weld accelerates data science workloads by up to 23X on one thread and 80X on eight threads, and its adaptive optimizations provide up to a 3.75X speedup over rule-based optimization. Moreover, Weld provides benefits if even just 4--5 operators in a library are ported to use it. Our results show that common runtime designs like Weld may be a viable approach to accelerate analytics.

As data volumes continue to rise, manual inspection is becoming increasingly untenable. In response, we present MacroBase, a data analytics engine that prioritizes end-user attention in high-volume fast data streams. MacroBase enables efficient, accurate, and modular analyses that highlight and aggregate important and unusual behavior, acting as a search engine for fast data. MacroBase is able to deliver order-of-magnitude speedups over alternatives by optimizing the combination of explanation (i.e., feature selection) and classification tasks and by leveraging a new reservoir sampler and heavy-hitters sketch specialized for fast data streams. As a result, MacroBase delivers accurate results at speeds of up to 2M events per second per query on a single core. The system has delivered meaningful results in production, including at a telematics company monitoring hundreds of thousands of vehicles.

As data volumes continue to rise, manual inspection is becoming increasingly untenable. In response, we present MacroBase, a data analytics engine that prioritizes end-user attention in high-volume fast data streams. MacroBase enables efficient, accurate, and modular analyses that highlight and aggregate important and unusual behavior, acting as a search engine for fast data. MacroBase is able to deliver order-of-magnitude speedups over alternatives by optimizing the combination of explanation and classification tasks and by leveraging a new reservoir sampler and heavy-hitters sketch specialized for fast data streams. As a result, MacroBase delivers accurate results at speeds of up to 2M events per second per query on a single core. The system has delivered meaningful results in production, including at a telematics company monitoring hundreds of thousands of vehicles.