FlashAttention on a Napkin: A Diagrammatic Approach to Deep Learning IO-Awareness

• URL: http://arxiv.org/abs/2412.03317v2
• Date: Sun, 19 Jan 2025 08:31:39 GMT
• Title: FlashAttention on a Napkin: A Diagrammatic Approach to Deep Learning IO-Awareness
• Authors: Vincent Abbott, Gioele Zardini,
• Abstract summary: Methods like FlashAttention have achieved a x6 performance improvement over native PyTorch by avoiding unnecessary data transfers.<n>This paper extends Neural Circuit Diagrams for deep learning models to consider resource usage and the distribution of tasks across a GPU hierarchy.<n>We develop a methodology for representing intermediate-level pseudocode with diagrams, allowing hardware-aware algorithms to be derived step-by-step.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Optimizing deep learning algorithms currently requires slow, manual derivation, potentially leaving much performance untapped. Methods like FlashAttention have achieved a x6 performance improvement over native PyTorch by avoiding unnecessary data transfers, but required three iterations over three years to be developed. Automated compiled methods have consistently lagged behind. This paper extends Neural Circuit Diagrams for deep learning models to consider resource usage and the distribution of tasks across a GPU hierarchy. We show how diagrams can use simple relabellings to derive high-level streaming and tiling optimization strategies along with performance models. We show how this high-level performance model allows the effects of quantization and multi-level GPU hierarchies to be readily considered. We develop a methodology for representing intermediate-level pseudocode with diagrams, allowing hardware-aware algorithms to be derived step-by-step. Finally, we show how our methodology can be used to better understand existing techniques like FlashAttention. This work uses a theoretical framework to link assumptions about GPU behaviour to claims about performance. We aim to lay the groundwork for a scientific approach to GPU optimization where experiments can address clear hypotheses rather than post-hoc rationalizations.

Related papers

• Numerical Pruning for Efficient Autoregressive Models [87.56342118369123]
  This paper focuses on compressing decoder-only transformer-based autoregressive models through structural weight pruning. Specifically, we propose a training-free pruning method that calculates a numerical score with Newton's method for the Attention and modules, respectively. To verify the effectiveness of our method, we provide both theoretical support and extensive experiments.
  arXiv  Detail & Related papers  (2024-12-17T01:09:23Z)
• Efficient LLM Inference with I/O-Aware Partial KV Cache Recomputation [7.204881999658682]
  Inference for Large Language Models (LLMs) is computationally demanding. To reduce the cost of auto-regressive decoding, Key-Value ( KV) caching is used to store intermediate activations. The memory required for KV caching grows rapidly, often exceeding the capacity of GPU memory. A cost-effective alternative is to offload KV cache to CPU memory, which alleviates GPU memory pressure but shifts the bottleneck to the limited bandwidth of the PCIe connection between the CPU and GPU.
  arXiv  Detail & Related papers  (2024-11-26T04:03:14Z)
• Implementation and Analysis of GPU Algorithms for Vecchia Approximation [0.8057006406834466]
  Vecchia Approximation is widely used to reduce the computational complexity and can be calculated with embarrassingly parallel algorithms. While multi-core software has been developed for Vecchia Approximation, software designed to run on graphics processing units ( GPU) is lacking. We show that our new method outperforms the other two and then present it in the GpGpU R package.
  arXiv  Detail & Related papers  (2024-07-03T01:24:44Z)
• Deep Symbolic Optimization for Combinatorial Optimization: Accelerating Node Selection by Discovering Potential Heuristics [10.22111332588471]
  We propose a novel deep symbolic optimization learning framework that combines their advantages. Dso4NS guides the search for mathematical expressions within the high-dimensional discrete symbolic space and then incorporates the highest-performing mathematical expressions into a solver. Experiments demonstrate the effectiveness of Dso4NS in learning high-quality expressions, outperforming existing approaches on a CPU machine.
  arXiv  Detail & Related papers  (2024-06-14T06:02:14Z)
• Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity [12.663030430488922]
  We propose Flash-LLM for enabling low-cost and highly-efficient large generative model inference on high-performance Cores. At SpMM kernel level, Flash-LLM significantly outperforms the state-of-the-art library, i.e., Sputnik and SparTA by an average of 2.9x and 1.5x, respectively.
  arXiv  Detail & Related papers  (2023-09-19T03:20:02Z)
• INR-Arch: A Dataflow Architecture and Compiler for Arbitrary-Order Gradient Computations in Implicit Neural Representation Processing [66.00729477511219]
  Given a function represented as a computation graph, traditional architectures face challenges in efficiently computing its nth-order gradient. We introduce INR-Arch, a framework that transforms the computation graph of an nth-order gradient into a hardware-optimized dataflow architecture. We present results that demonstrate 1.8-4.8x and 1.5-3.6x speedup compared to CPU and GPU baselines respectively.
  arXiv  Detail & Related papers  (2023-08-11T04:24:39Z)
• SqueezeLLM: Dense-and-Sparse Quantization [80.32162537942138]
  Main bottleneck for generative inference with LLMs is memory bandwidth, rather than compute, for single batch inference. We introduce SqueezeLLM, a post-training quantization framework that enables lossless compression to ultra-low precisions of up to 3-bit. Our framework incorporates two novel ideas: (i) sensitivity-based non-uniform quantization, which searches for the optimal bit precision assignment based on second-order information; and (ii) the Dense-and-Sparse decomposition that stores outliers and sensitive weight values in an efficient sparse format.
  arXiv  Detail & Related papers  (2023-06-13T08:57:54Z)
• Adaptable Butterfly Accelerator for Attention-based NNs via Hardware and Algorithm Co-design [66.39546326221176]
  Attention-based neural networks have become pervasive in many AI tasks. The use of the attention mechanism and feed-forward network (FFN) demands excessive computational and memory resources. This paper proposes a hardware-friendly variant that adopts a unified butterfly sparsity pattern to approximate both the attention mechanism and the FFNs.
  arXiv  Detail & Related papers  (2022-09-20T09:28:26Z)
• 8-bit Optimizers via Block-wise Quantization [57.25800395197516]
  Statefuls maintain statistics over time, e.g., the exponentially smoothed sum (SGD with momentum) or squared sum (Adam) of past values. This state can be used to accelerate optimization compared to plain gradient descent but uses memory that might otherwise be allocated to model parameters. In this paper, we develop first gradients that use 8-bit statistics while maintaining the performance levels of using 32-bit gradient states.
  arXiv  Detail & Related papers  (2021-10-06T15:43:20Z)
• Using Graph Neural Networks to model the performance of Deep Neural Networks [2.1151356984322307]
  We develop a novel performance model that adopts a graph representation. Experimental evaluation shows a 7:75x and 12x reduction in prediction error compared to the Halide and TVM models, respectively.
  arXiv  Detail & Related papers  (2021-08-27T20:20:17Z)
• Top-KAST: Top-K Always Sparse Training [50.05611544535801]
  We propose Top-KAST, a method that preserves constant sparsity throughout training. We show that it performs comparably to or better than previous works when training models on the established ImageNet benchmark. In addition to our ImageNet results, we also demonstrate our approach in the domain of language modeling.
  arXiv  Detail & Related papers  (2021-06-07T11:13:05Z)
• Faster Meta Update Strategy for Noise-Robust Deep Learning [62.08964100618873]
  We introduce a novel Faster Meta Update Strategy (FaMUS) to replace the most expensive step in the meta gradient with a faster layer-wise approximation. We show our method is able to save two-thirds of the training time while still maintaining the comparable or achieving even better generalization performance.
  arXiv  Detail & Related papers  (2021-04-30T16:19:07Z)
• Fast Few-Shot Classification by Few-Iteration Meta-Learning [173.32497326674775]
  We introduce a fast optimization-based meta-learning method for few-shot classification. Our strategy enables important aspects of the base learner objective to be learned during meta-training. We perform a comprehensive experimental analysis, demonstrating the speed and effectiveness of our approach.
  arXiv  Detail & Related papers  (2020-10-01T15:59:31Z)
• ConfuciuX: Autonomous Hardware Resource Assignment for DNN Accelerators using Reinforcement Learning [5.251940442946459]
  We propose an autonomous strategy called ConfuciuX to find optimized HW resource assignments for a given model and dataflow style. It converges to the optimized hardware configuration 4.7 to 24 times faster than alternate techniques.
  arXiv  Detail & Related papers  (2020-09-04T04:59:26Z)
• Kernel methods through the roof: handling billions of points efficiently [94.31450736250918]
  Kernel methods provide an elegant and principled approach to nonparametric learning, but so far could hardly be used in large scale problems. Recent advances have shown the benefits of a number of algorithmic ideas, for example combining optimization, numerical linear algebra and random projections. Here, we push these efforts further to develop and test a solver that takes full advantage of GPU hardware.
  arXiv  Detail & Related papers  (2020-06-18T08:16:25Z)
• Heterogeneous CPU+GPU Stochastic Gradient Descent Algorithms [1.3249453757295084]
  We study training algorithms for deep learning on heterogeneous CPU+GPU architectures. Our two-fold objective -- maximize convergence rate and resource utilization simultaneously -- makes the problem challenging. We show that the implementation of these algorithms achieves both faster convergence and higher resource utilization than on several real datasets.
  arXiv  Detail & Related papers  (2020-04-19T05:21:20Z)
• Optimizing Streaming Parallelism on Heterogeneous Many-Core Architectures: A Machine Learning Based Approach [16.702537371391053]
  This article presents an automatic approach to derive a good solution for hardware resource partition and task granularity for task-based parallel applications on heterogeneous many-core architectures. Our approach employs a performance model to estimate the resulting performance of the target application under a given resource partition and task granularity configuration. Compared to the single-stream version, our approach achieves a 1.6x and 1.1x speedup on the XeonPhi and the GPU platform, respectively.
  arXiv  Detail & Related papers  (2020-03-05T21:18:21Z)
• Image Matching across Wide Baselines: From Paper to Practice [80.9424750998559]
  We introduce a comprehensive benchmark for local features and robust estimation algorithms. Our pipeline's modular structure allows easy integration, configuration, and combination of different methods. We show that with proper settings, classical solutions may still outperform the perceived state of the art.
  arXiv  Detail & Related papers  (2020-03-03T15:20:57Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.