Ningxin Zheng

ByteDance Seed, Jiaze Chen, Tiantian Fan et al. · bytedance

We introduce Seed1.5-Thinking, capable of reasoning through thinking before responding, resulting in improved performance on a wide range of benchmarks. Seed1.5-Thinking achieves 86.7 on AIME 2024, 55.0 on Codeforces and 77.3 on GPQA, demonstrating excellent reasoning abilities in STEM and coding. Beyond reasoning tasks, the method demonstrates notable generalization across diverse domains. For instance, it surpasses DeepSeek R1 by 8% in win rate on non-reasoning tasks, indicating its broader applicability. Compared to other state-of-the-art reasoning models, Seed1.5-Thinking is a Mixture-of-Experts (MoE) model with a relatively small size, featuring 20B activated and 200B total parameters. As part of our effort to assess generalized reasoning, we develop two internal benchmarks, BeyondAIME and Codeforces, both of which will be publicly released to support future research. Model trial link: https://www.volcengine.com/experience/ark.

Ningxin Zheng, Huiqiang Jiang, Quanlu Zhang et al. · microsoft-research

Dynamic sparsity, where the sparsity patterns are unknown until runtime, poses a significant challenge to deep learning. The state-of-the-art sparsity-aware deep learning solutions are restricted to pre-defined, static sparsity patterns due to significant overheads associated with preprocessing. Efficient execution of dynamic sparse computation often faces the misalignment between the GPU-friendly tile configuration for efficient execution and the sparsity-aware tile shape that minimizes coverage wastes (non-zero values in tensor). In this paper, we propose PIT, a deep-learning compiler for dynamic sparsity. PIT proposes a novel tiling mechanism that leverages Permutation Invariant Transformation (PIT), a mathematically proven property, to transform multiple sparsely located micro-tiles into a GPU-efficient dense tile without changing the computation results, thus achieving both high GPU utilization and low coverage waste. Given a model, PIT first finds feasible PIT rules for all its operators and generates efficient GPU kernels accordingly. At runtime, with the novel SRead and SWrite primitives, PIT rules can be executed extremely fast to support dynamic sparsity in an online manner. Extensive evaluation on diverse models shows that PIT can accelerate dynamic sparsity computation by up to 5.9x (average 2.43x) over state-of-the-art compilers.

Li Lyna Zhang, Xudong Wang, Jiahang Xu et al. · microsoft-research, pku

The combination of Neural Architecture Search (NAS) and quantization has proven successful in automatically designing low-FLOPs INT8 quantized neural networks (QNN). However, directly applying NAS to design accurate QNN models that achieve low latency on real-world devices leads to inferior performance. In this work, we find that the poor INT8 latency is due to the quantization-unfriendly issue: the operator and configuration (e.g., channel width) choices in prior art search spaces lead to diverse quantization efficiency and can slow down the INT8 inference speed. To address this challenge, we propose SpaceEvo, an automatic method for designing a dedicated, quantization-friendly search space for each target hardware. The key idea of SpaceEvo is to automatically search hardware-preferred operators and configurations to construct the search space, guided by a metric called Q-T score to quantify how quantization-friendly a candidate search space is. We further train a quantized-for-all supernet over our discovered search space, enabling the searched models to be directly deployed without extra retraining or quantization. Our discovered models establish new SOTA INT8 quantized accuracy under various latency constraints, achieving up to 10.1% accuracy improvement on ImageNet than prior art CNNs under the same latency. Extensive experiments on diverse edge devices demonstrate that SpaceEvo consistently outperforms existing manually-designed search spaces with up to 2.5x faster speed while achieving the same accuracy.

Jun Xiao, Xinyang Jiang, Ningxin Zheng et al.

Deep learning-based models have achieved remarkable performance in video super-resolution (VSR) in recent years, but most of these models are less applicable to online video applications. These methods solely consider the distortion quality and ignore crucial requirements for online applications, e.g., low latency and low model complexity. In this paper, we focus on online video transmission, in which VSR algorithms are required to generate high-resolution video sequences frame by frame in real time. To address such challenges, we propose an extremely low-latency VSR algorithm based on a novel kernel knowledge transfer method, named convolutional kernel bypass graft (CKBG). First, we design a lightweight network structure that does not require future frames as inputs and saves extra time costs for caching these frames. Then, our proposed CKBG method enhances this lightweight base model by bypassing the original network with ``kernel grafts'', which are extra convolutional kernels containing the prior knowledge of external pretrained image SR models. In the testing phase, we further accelerate the grafted multi-branch network by converting it into a simple single-path structure. Experiment results show that our proposed method can process online video sequences up to 110 FPS, with very low model complexity and competitive SR performance.

Guanghao Yin, Zefan Qu, Xinyang Jiang et al.

Online video streaming has fundamental limitations on the transmission bandwidth and computational capacity and super-resolution is a promising potential solution. However, applying existing video super-resolution methods to online streaming is non-trivial. Existing video codecs and streaming protocols (\eg, WebRTC) dynamically change the video quality both spatially and temporally, which leads to diverse and dynamic degradations. Furthermore, online streaming has a strict requirement for latency that most existing methods are less applicable. As a result, this paper focuses on the rarely exploited problem setting of online streaming video super resolution. To facilitate the research on this problem, a new benchmark dataset named LDV-WebRTC is constructed based on a real-world online streaming system. Leveraging the new benchmark dataset, we proposed a novel method specifically for online video streaming, which contains a convolution and Look-Up Table (LUT) hybrid model to achieve better performance-latency trade-off. To tackle the changing degradations, we propose a mixture-of-expert-LUT module, where a set of LUT specialized in different degradations are built and adaptively combined to handle different degradations. Experiments show our method achieves 720P video SR around 100 FPS, while significantly outperforms existing LUT-based methods and offers competitive performance compared to efficient CNN-based methods.

Hanshi Sun, Li-Wen Chang, Wenlei Bao et al.

With the widespread deployment of long-context large language models (LLMs), there has been a growing demand for efficient support of high-throughput inference. However, as the key-value (KV) cache expands with the sequence length, the increasing memory footprint and the need to access it for each token generation both result in low throughput when serving long-context LLMs. While various dynamic sparse attention methods have been proposed to speed up inference while maintaining generation quality, they either fail to sufficiently reduce GPU memory consumption or introduce significant decoding latency by offloading the KV cache to the CPU. We present ShadowKV, a high-throughput long-context LLM inference system that stores the low-rank key cache and offloads the value cache to reduce the memory footprint for larger batch sizes and longer sequences. To minimize decoding latency, ShadowKV employs an accurate KV selection strategy that reconstructs minimal sparse KV pairs on-the-fly. By evaluating ShadowKV on a broad range of benchmarks, including RULER, LongBench, and Needle In A Haystack, and models like Llama-3.1-8B, Llama-3-8B-1M, GLM-4-9B-1M, Yi-9B-200K, Phi-3-Mini-128K, and Qwen2-7B-128K, we demonstrate that it can support up to 6$\times$ larger batch sizes and boost throughput by up to 3.04$\times$ on an A100 GPU without sacrificing accuracy, even surpassing the performance achievable with infinite batch size under the assumption of infinite GPU memory. The code is available at https://github.com/bytedance/ShadowKV.

Chunyu Xue, Yangrui Chen, Jianyu Jiang et al.

As the foundational component of versatile AI applications, training an multimodal large language model (MLLM) relies on multimodal datasets with dynamic modality mixture proportions and sample length distributions. However, existing MLLM systems remain inefficient under dynamic workloads, due to statically coupled decisions of resource allocation and model parallelization between encoders and the LLM backbone. This paper presents MegaScale-Omni, an industrial-grade MLLM training system tailored for dynamic workload adaption and hyper-scale deployment. MegaScale-Omni is built upon the training scheme of encoder-LLM multiplexing with three key innovations: (1) Decoupled parallelism strategies with long-short sequence parallelism for encoders to process variable-length samples, and full-fledged 5D parallelism for the LLM backbone, both organized under a communication-efficient parallelization layout. (2) Unified encoder-LLM representations for flexible, extensible colocation, and a new paradigm of encoder-LLM joint pipeline with workload resilience. (3) Workload balancing techniques via decentralized grouped reordering in data loaders and adaptive resharding from encoder to LLM ranks. MegaScale-Omni is deployed as the foundation of our in-house large-scale MLLM training tasks with thousands of GPUs. Our experimental results demonstrate $1.27\times$-$7.57\times$ throughput improvement under production-grade dynamic workloads, as compared to four state-of-the-art systems.

Zhichen Zeng, Chi-Chih Chang, Jiayi Wang et al.

Mixture-of-experts (MoE) architectures enable trillion-parameter LLMs with sparsely activated experts. Expert parallelism (EP) is a widely adopted MoE training strategy, but it suffers from severe all-to-all communication bottlenecks, which is exaggerated by the limited inter-node network bandwidth as the growing model size requires distributing experts across GPU nodes. Prior work focused on overlapping these all-to-all communications with feed-forward network (FFN) and self-attention computations, which often leaves residual network-bound stalls due to inherent imbalance in attention and FFN layers' computation-communication ratios. We present DisagMoE, a disaggregated MoE training system that jointly optimizes model placement and scheduling for maximal efficiency. DisagMoE separates attention and FFN layers into disjoint GPU groups, introduces a multi-stage pipeline with uni-directional, many-to-many communications, and employs a computation-communication roofline model to balance GPU and network bandwidth allocation among the attention and FFN groups. DisagMoE is implemented on Megatron-LM, and evaluation shows that DisagMoE improves training efficiency across multiple MoE models with up to 1.8x speedup on 16-node 8xH800 clusters.

Xinyu Liu, Houwen Peng, Ningxin Zheng et al.

Vision transformers have shown great success due to their high model capabilities. However, their remarkable performance is accompanied by heavy computation costs, which makes them unsuitable for real-time applications. In this paper, we propose a family of high-speed vision transformers named EfficientViT. We find that the speed of existing transformer models is commonly bounded by memory inefficient operations, especially the tensor reshaping and element-wise functions in MHSA. Therefore, we design a new building block with a sandwich layout, i.e., using a single memory-bound MHSA between efficient FFN layers, which improves memory efficiency while enhancing channel communication. Moreover, we discover that the attention maps share high similarities across heads, leading to computational redundancy. To address this, we present a cascaded group attention module feeding attention heads with different splits of the full feature, which not only saves computation cost but also improves attention diversity. Comprehensive experiments demonstrate EfficientViT outperforms existing efficient models, striking a good trade-off between speed and accuracy. For instance, our EfficientViT-M5 surpasses MobileNetV3-Large by 1.9% in accuracy, while getting 40.4% and 45.2% higher throughput on Nvidia V100 GPU and Intel Xeon CPU, respectively. Compared to the recent efficient model MobileViT-XXS, EfficientViT-M2 achieves 1.8% superior accuracy, while running 5.8x/3.7x faster on the GPU/CPU, and 7.4x faster when converted to ONNX format. Code and models are available at https://github.com/microsoft/Cream/tree/main/EfficientViT.

Size Zheng, Xuegui Zheng, Hanshi Sun et al.

The scaling of large language models (LLMs) is currently bottlenecked by the rigidity of distributed programming. While high-performance libraries like CuBLAS and NCCL provide optimized primitives, they lack the flexibility required for rapidly evolving model architectures. Conversely, existing tensor compilers fail to address the complex memory hierarchy of distributed clusters effectively. To bridge this gap, we propose DITRON, a scalable tile-level compiler that democratizes high-performance distributed kernel development. DITRON introduces a novel hierarchical programming abstraction spanning Core, Device, and Task levels to map tensor programs efficiently onto heterogeneous distributed hardware. This abstraction allows DITRON to support diverse parallelism strategies while abstracting away the complexity of inter-node and intra-node communication. Evaluated across large-scale clusters, DITRON achieves performance parity with or exceeding expert-tuned CUDA libraries, delivering speedups of $6\%-30\%$ on isolated kernels and $5\%-30\%$ on end-to-end inference in vLLM. Furthermore, DITRON demonstrates strong portability, achieving significant speedups on both NVIDIA and AMD platforms. \ours{} has been deployed at the enterprise level for both training and inference. It achieves an MFU improvement of over 10\% in training tasks, saving approximately 500,000 GPU hours of training cost per month. For inference tasks, it delivers an end-to-end gain of over 20\% and has been applied to cloud service inference and edge inference scenarios.

Shulai Zhang, Ningxin Zheng, Haibin Lin et al.

Mixture-of-experts (MoE) has been extensively employed to scale large language models to trillion-plus parameters while maintaining a fixed computational cost. The development of large MoE models in the distributed scenario encounters the problem of large communication overhead. The inter-device communication of a MoE layer can occupy 47% time of the entire model execution with popular models and frameworks. Therefore, existing methods suggest the communication in a MoE layer to be pipelined with the computation for overlapping. However, these coarse grained overlapping schemes introduce a notable impairment of computational efficiency and the latency concealing is sub-optimal. To this end, we present COMET, an optimized MoE system with fine-grained communication-computation overlapping. Leveraging data dependency analysis and task rescheduling, COMET achieves precise fine-grained overlapping of communication and computation. Through adaptive workload assignment, COMET effectively eliminates fine-grained communication bottlenecks and enhances its adaptability across various scenarios. Our evaluation shows that COMET accelerates the execution of a single MoE layer by $1.96\times$ and for end-to-end execution, COMET delivers a $1.71\times$ speedup on average. COMET has been adopted in the production environment of clusters with ten-thousand-scale of GPUs, achieving savings of millions of GPU hours.

Shulai Zhang, Ao Xu, Quan Chen et al.

Embodied AI systems operate in dynamic environments, requiring seamless integration of perception and generation modules to process high-frequency input and output demands. Traditional sequential computation patterns, while effective in ensuring accuracy, face significant limitations in achieving the necessary "thinking" frequency for real-world applications. In this work, we present Auras, an algorithm-system co-designed inference framework to optimize the inference frequency of embodied AI agents. Auras disaggregates the perception and generation and provides controlled pipeline parallelism for them to achieve high and stable throughput. Faced with the data staleness problem that appears when the parallelism is increased, Auras establishes a public context for perception and generation to share, thereby promising the accuracy of embodied agents. Experimental results show that Auras improves throughput by 2.54x on average while achieving 102.7% of the original accuracy, demonstrating its efficacy in overcoming the constraints of sequential computation and providing high throughput.

Li-Wen Chang, Wenlei Bao, Qi Hou et al.

Large deep learning models have demonstrated strong ability to solve many tasks across a wide range of applications. Those large models typically require training and inference to be distributed. Tensor parallelism is a common technique partitioning computation of an operation or layer across devices to overcome the memory capacity limitation of a single processor, and/or to accelerate computation to meet a certain latency requirement. However, this kind of parallelism introduces additional communication that might contribute a significant portion of overall runtime. Thus limits scalability of this technique within a group of devices with high speed interconnects, such as GPUs with NVLinks in a node. This paper proposes a novel method, Flux, to significantly hide communication latencies with dependent computations for GPUs. Flux over-decomposes communication and computation operations into much finer-grained operations and further fuses them into a larger kernel to effectively hide communication without compromising kernel efficiency. Flux can potentially overlap up to 96% of communication given a fused kernel. Overall, it can achieve up to 1.24x speedups for training over Megatron-LM on a cluster of 128 GPUs with various GPU generations and interconnects, and up to 1.66x and 1.30x speedups for prefill and decoding inference over vLLM on a cluster with 8 GPUs with various GPU generations and interconnects.

Bo Li, Xinyang Jiang, Donglin Bai et al.

The energy consumption of deep learning models is increasing at a breathtaking rate, which raises concerns due to potential negative effects on carbon neutrality in the context of global warming and climate change. With the progress of efficient deep learning techniques, e.g., model compression, researchers can obtain efficient models with fewer parameters and smaller latency. However, most of the existing efficient deep learning methods do not explicitly consider energy consumption as a key performance indicator. Furthermore, existing methods mostly focus on the inference costs of the resulting efficient models, but neglect the notable energy consumption throughout the entire life cycle of the algorithm. In this paper, we present the first large-scale energy consumption benchmark for efficient computer vision models, where a new metric is proposed to explicitly evaluate the full-cycle energy consumption under different model usage intensity. The benchmark can provide insights for low carbon emission when selecting efficient deep learning algorithms in different model usage scenarios.