SageAttention2 Technical Report: Accurate 4 Bit Attention for Plug-and-play Inference Acceleration

• URL: http://arxiv.org/abs/2411.10958v1
• Date: Sun, 17 Nov 2024 04:35:49 GMT
• Title: SageAttention2 Technical Report: Accurate 4 Bit Attention for Plug-and-play Inference Acceleration
• Authors: Jintao Zhang, Haofeng Huang, Pengle Zhang, Jia Wei, Jun Zhu, Jianfei Chen,
• Abstract summary: We propose SageAttention2, which utilizes significantly faster 4-bit matrix multiplication (Matmul) alongside precision-enhancing techniques. We analyze the quantization accuracy across timesteps and layers, then propose an adaptive quantization method to ensure the end-to-end metrics. Experiments confirm that our approach incurs negligible end-to-end metrics loss across diverse models.
• Score: 22.551095978580147
• License:
• Abstract: Although quantization for linear layers has been widely used, its application to accelerate the attention process remains limited. SageAttention utilizes 8-bit matrix multiplication, 16-bit matrix multiplication with 16-bit accumulator, and precision-enhancing methods, implementing an accurate and 2x speedup kernel compared to FlashAttention2. To further enhance the efficiency of attention computation while maintaining precision, we propose SageAttention2, which utilizes significantly faster 4-bit matrix multiplication (Matmul) alongside additional precision-enhancing techniques. First, we propose to quantize matrixes $(Q, K)$ to INT4 in a warp-level granularity and quantize matrixes $(\widetilde P, V)$ to FP8. Second, we propose a method to smooth $Q$ and $V$, enhancing the accuracy of attention with INT4 $QK$ and FP8 $PV$. Third, we analyze the quantization accuracy across timesteps and layers, then propose an adaptive quantization method to ensure the end-to-end metrics over various models. The operations per second (OPS) of SageAttention2 surpass FlashAttention2 and xformers by about 3x and 5x on RTX4090, respectively. Comprehensive experiments confirm that our approach incurs negligible end-to-end metrics loss across diverse models, including those for large language processing, image generation, and video generation. The codes are available at https://github.com/thu-ml/SageAttention.

Related papers

• SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models [58.5019443418822]
  Diffusion models have been proven highly effective at generating high-quality images. As these models grow larger, they require significantly more memory and suffer from higher latency. In this work, we aim to accelerate diffusion models by quantizing their weights and activations to 4 bits.
  arXiv  Detail & Related papers  (2024-11-07T18:59:58Z)
• COMET: Towards Partical W4A4KV4 LLMs Serving [37.30529940231099]
  Quantization is a compression technology to reduce the overhead of serving large language models (LLMs) on terminal devices and in cloud data centers. We propose a novel mixed-precision quantization algorithm (FMPQ) that compresses most activations into 4-bit with negligible accuracy loss. We integrate the optimized W4Ax kernel into our inference framework, COMET, and provide efficient management to support popular LLMs.
  arXiv  Detail & Related papers  (2024-10-16T02:16:53Z)
• FlatQuant: Flatness Matters for LLM Quantization [58.28221892035609]
  We propose FlatQuant, a new post-training quantization approach to enhance flatness of weights and activations. Our approach identifies optimal affine transformations tailored to each linear layer, calibrated in hours via a lightweight objective runtime. For inference latency, FlatQuant reduces the slowdown induced by pre-quantization transformation from 0.26x of QuaRot to merely $textbf0.07x$, bringing up to $textbf2.3x$ speedup for prefill and $textbf1.7x$ speedup for decoding.
  arXiv  Detail & Related papers  (2024-10-12T08:10:28Z)
• EXAQ: Exponent Aware Quantization For LLMs Acceleration [15.610222058802005]
  We propose an analytical approach to determine the optimal clipping value for the input to the softmax function. This method accelerates the calculations of both $ex$ and $sum(ex)$ with minimal to no accuracy degradation. This ultra-low bit quantization allows, for the first time, an acceleration of approximately 4x in the accumulation phase.
  arXiv  Detail & Related papers  (2024-10-04T06:54:30Z)
• SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration [22.551095978580147]
  We propose SageAttention, a highly efficient and accurate quantization method for attention. Our approach incurs almost no end-to-end metrics loss across diverse models.
  arXiv  Detail & Related papers  (2024-10-03T10:25:23Z)
• Rotated Runtime Smooth: Training-Free Activation Smoother for accurate INT4 inference [54.2589824716527]
  Large language models incur substantial computation and memory movement costs due to their large scale. Existing approaches separate outliers and normal values into two matrices or migrate outliers from activations to weights, suffering from high latency or accuracy degradation. We propose Rotated Smooth (RRS), a plug-and-play activation smoother for quantization, consisting of Smooth and Rotation operation. The proposed method outperforms the state-of-the-art method in the LLaMA and Qwen families and improves WikiText-2 perplexity from 57.33 to 6.66 for INT4 inference.
  arXiv  Detail & Related papers  (2024-09-30T14:59:22Z)
• QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving [52.31791050376249]
  Quantization can accelerate large language model (LLM) inference. Existing INT4 quantization methods suffer from significant runtime overhead when dequantizing weights or partial sums. We introduce QoQ, a W4A8KV4 quantization algorithm with 4-bit weight, 8-bit activation, and 4-bit KV cache. QServe improves the maximum achievable serving of Llama-3-8B by 1.2x on A100, 1.4x on L40S; and Qwen-721.5B by 2.4x on A100, 3.5x on L40S.
  arXiv  Detail & Related papers  (2024-05-07T17:59:30Z)
• QUIK: Towards End-to-End 4-Bit Inference on Generative Large Language Models [57.04178959678024]
  We show that the majority of inference computations for large generative models can be performed with both weights and activations being cast to 4 bits. We achieve this via a hybrid quantization strategy called QUIK, which compresses most of the weights and activations to 4-bit. We provide GPU kernels matching the QUIK format with highly-efficient layer-wise runtimes, which lead to practical end-to-end throughput improvements of up to 3.4x.
  arXiv  Detail & Related papers  (2023-10-13T17:15:05Z)
• Dual Grained Quantization: Efficient Fine-Grained Quantization for LLM [6.85331857224501]
  Large Language Models (LLMs) pose significant hardware challenges related to memory requirements and computational ability. There are two mainstream quantization schemes for LLMs: coarse-grained ($textite.g.,$ channel-wise) quantization and fine-grained ($textite.g.,$ group-wise) quantization. We introduce Dual Grained Quantization (DGQ), a novel A8W4 quantization for LLM that maintains superior performance while ensuring fast inference speed.
  arXiv  Detail & Related papers  (2023-10-07T14:50:28Z)
• Training Transformers with 4-bit Integers [21.861232105539933]
  Quantizing activation, weight, and gradient to 4-bit is promising to accelerate neural network training. Existing 4-bit training methods require custom numerical formats which are not supported by contemporary hardware. In this work, we propose a training method for transformers with all matrix multiplications implemented with the INT4 arithmetic.
  arXiv  Detail & Related papers  (2023-06-21T02:45:01Z)
• 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)

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.