Temporal action detection (TAD) aims to identify and localize action instances in untrimmed videos, which is essential for various video understanding tasks. However, recent improvements in model performance, driven by larger feature extractors and datasets, have led to increased computational demands. This presents a challenge for applications like autonomous driving and robotics, which rely on limited computational resources. While existing channel pruning methods can compress these models, reducing the number of channels often hinders the parallelization efficiency of GPU, due to the inefficient multiplication between small matrices. Instead of pruning channels, we propose a Progressive Block Drop method that reduces model depth while retaining layer width. In this way, we still use large matrices for computation but reduce the number of multiplications. Our approach iteratively removes redundant blocks in two steps: first, we drop blocks with minimal impact on model performance; and second, we employ a parameter-efficient cross-depth alignment technique, fine-tuning the pruned model to restore model accuracy. Our method achieves a 25% reduction in computational overhead on two TAD benchmarks (THUMOS14 and ActivityNet-1.3) to achieve lossless compression. More critically, we empirically show that our method is orthogonal to channel pruning methods and can be combined with it to yield further efficiency gains.
We introduce UniCon, a novel architecture designed to enhance control and efficiency in training adapters for large-scale diffusion models. Unlike existing methods that rely on bidirectional interaction between the diffusion model and control adapter, UniCon implements a unidirectional flow from the diffusion network to the adapter, allowing the adapter alone to generate the final output. UniCon reduces computational demands by eliminating the need for the diffusion model to compute and store gradients during adapter training. Our results indicate that UniCon reduces GPU memory usage by one-third and increases training speed by 2.3 times, while maintaining the same adapter parameter size. Additionally, without requiring extra computational resources, UniCon enables the training of adapters with double the parameter volume of existing ControlNets. In a series of image conditional generation tasks, UniCon has demonstrated precise responsiveness to control inputs and exceptional generation capabilities.
Document layout analysis is a critical preprocessing step in document intelligence, enabling the detection and localization of structural elements such as titles, text blocks, tables, and formulas. Despite its importance, existing layout detection models face significant challenges in generalizing across diverse document types, handling complex layouts, and achieving real-time performance for large-scale data processing. To address these limitations, we present PP-DocLayout, which achieves high precision and efficiency in recognizing 23 types of layout regions across diverse document formats. To meet different needs, we offer three models of varying scales. PP-DocLayout-L is a high-precision model based on the RT-DETR-L detector, achieving 90.4% [email omitted] and an end-to-end inference time of 13.4 ms per page on a T4 GPU. PP-DocLayout-M is a balanced model, offering 75.2% [email omitted] with an inference time of 12.7 ms per page on a T4 GPU. PP-DocLayout-S is a high-efficiency model designed for resource-constrained environments and real-time applications, with an inference time of 8.1 ms per page on a T4 GPU and 14.5 ms on a CPU. This work not only advances the state of the art in document layout analysis but also provides a robust solution for constructing high-quality training data, enabling advancements in document intelligence and multimodal AI systems. Code and models are available at https://github.com/PaddlePaddle/PaddleX .
With the advancements in long-context inference capabilities of large language models (LLMs), the KV cache has become one of the foundational components. However, its substantial GPU memory consumption makes KV cache compression a key technique for enabling efficient LLM inference in industrial scenarios. While recent studies have focused on optimizing the memory occupied by the KV cache, they overlook two critical factors: preserving semantic coherence and considering task-specific characteristic during compression. To address these limitations, we propose a novel task-adaptive KV cache window selection method, WindowKV. WindowKV dynamically selects local semantic windows consisting of consecutive tokens, according to task-specific characteristics, ensuring the retained KV cache captures continuous, essential context. Additionally, we introduce an intra-group layer KV cache indices sharing strategy to reduce computational overhead, achieving a balance between performance and efficiency. We rigorously evaluate WindowKV on the LongBench benchmark, and the results demonstrate that it maintains a performance comparable to full KV cache retention while using only 12% of the original KV cache, significantly reducing memory requirements. Furthermore, our method also achieves state-of-the-art results in the Needle-in-a-Haystack evaluation, highlighting its effectiveness and robustness.
Self-supervised adaptation (SSA) improves foundation model transfer to medical domains but is computationally prohibitive. Although parameter efficient fine-tuning methods such as LoRA have been explored for supervised adaptation, their effectiveness for SSA remains unknown. In this work, we introduce efficient self-supervised adaptation (ESSA), a framework that applies parameter-efficient fine-tuning techniques to SSA with the aim of reducing computational cost and improving adaptation performance. Among the methods tested, Attention Projection Layer Adaptation (APLA) sets a new state-of-the-art, consistently surpassing full-parameter SSA and supervised fine-tuning across diverse medical tasks, while reducing GPU memory by up to 40.1% and increasing training throughput by 25.2%, all while maintaining inference efficiency.
Despite advanced token compression techniques, existing multimodal large language models (MLLMs) still struggle with hour-long video understanding. In this work, we propose Video-XL-Pro, an efficient method for extremely long video understanding, built upon Reconstructive Compression of Tokens (ReCoT), a learnable module that leverages self-supervised learning to generate comprehensive and compact video tokens. ReCoT introduces two key components: (i) Dynamic Token Synthesizer (DTS): DTS generates pseudo-video tokens from static image tokens by learning intra-token relationships, which are then used in masked video modeling. (ii) Semantic-Guided Masking (SGM): SGM adaptively masks redundant visual tokens to facilitate more effective reconstructive learning. To improve training efficiency in MLLMs fine-tuning, we introduce a video-specific dataset pruning strategy and design a simple yet Query-aware Selector that enables the model to precisely locate query-relevant video tokens. With only 3B parameters, Video-XL-Pro outperforms most 7B models trained on larger datasets across multiple long video understanding benchmarks. Moreover, it can process over 8K frames on a single A100 GPU while maintaining high-quality performance.
Recognizing arbitrary or previously unseen categories is essential for comprehensive real-world 3D scene understanding. Currently, all existing methods rely on 2D or textual modalities during training, or together at inference. This highlights a clear absence of a model capable of processing 3D data alone for learning semantics end-to-end, along with the necessary data to train such a model. Meanwhile, 3D Gaussian Splatting (3DGS) has emerged as the de facto standard for 3D scene representation across various vision tasks. However, effectively integrating semantic reasoning into 3DGS in a generalizable fashion remains an open challenge. To address these limitations we introduce SceneSplat, to our knowledge the first large-scale 3D indoor scene understanding approach that operates natively on 3DGS. Furthermore, we propose a self-supervised learning scheme that unlocks rich 3D feature learning from unlabeled scenes. In order to power the proposed methods, we introduce SceneSplat-7K, the first large-scale 3DGS dataset for indoor scenes, comprising of 6868 scenes derived from 7 established datasets like ScanNet, Matterport3D, etc. Generating SceneSplat-7K required computational resources equivalent to 119 GPU-days on an L4 GPU, enabling standardized benchmarking for 3DGS-based reasoning for indoor scenes. Our exhaustive experiments on SceneSplat-7K demonstrate the significant benefit of the proposed methods over the established baselines.
Understanding the geometric and semantic properties of the scene is crucial in autonomous navigation and particularly challenging in the case of Unmanned Aerial Vehicle (UAV) navigation. Such information may be by obtained by estimating depth and semantic segmentation maps of the surrounding environment and for their practical use in autonomous navigation, the procedure must be performed as close to real-time as possible. In this paper, we leverage monocular cameras on aerial robots to predict depth and semantic maps in low-altitude unstructured environments. We propose a joint deep-learning architecture that can perform the two tasks accurately and rapidly, and validate its effectiveness on MidAir and Aeroscapes benchmark datasets. Our joint-architecture proves to be competitive or superior to the other single and joint architecture methods while performing its task fast predicting 20.2 FPS on a single NVIDIA quadro p5000 GPU and it has a low memory footprint. All codes for training and prediction can be found on this link: https://github.com/Malga-Vision/Co-SemDepth
Large language models (LLMs) deliver impressive performance but require large amounts of energy. In this work, we present a MatMul-free LLM architecture adapted for Intel's neuromorphic processor, Loihi 2. Our approach leverages Loihi 2's support for low-precision, event-driven computation and stateful processing. Our hardware-aware quantized model on GPU demonstrates that a 370M parameter MatMul-free model can be quantized with no accuracy loss. Based on preliminary results, we report up to 3x higher throughput with 2x less energy, compared to transformer-based LLMs on an edge GPU, with significantly better scaling. Further hardware optimizations will increase throughput and decrease energy consumption. These results show the potential of neuromorphic hardware for efficient inference and pave the way for efficient reasoning models capable of generating complex, long-form text rapidly and cost-effectively.
LiDARs provide accurate geometric measurements, making them valuable for ego-motion estimation and reconstruction tasks. Although its success, managing an accurate and lightweight representation of the environment still poses challenges. Both classic and NeRF-based solutions have to trade off accuracy over memory and processing times. In this work, we build on recent advancements in Gaussian Splatting methods to develop a novel LiDAR odometry and mapping pipeline that exclusively relies on Gaussian primitives for its scene representation. Leveraging spherical projection, we drive the refinement of the primitives uniquely from LiDAR measurements. Experiments show that our approach matches the current registration performance, while achieving SOTA results for mapping tasks with minimal GPU requirements. This efficiency makes it a strong candidate for further exploration and potential adoption in real-time robotics estimation tasks.
The recent exponential growth of Large Language Models (LLMs) has relied on GPU-based systems. However, CPUs are emerging as a flexible and lower-cost alternative, especially when targeting inference and reasoning workloads. RISC-V is rapidly gaining traction in this area, given its open and vendor-neutral ISA. However, the RISC-V hardware for LLM workloads and the corresponding software ecosystem are not fully mature and streamlined, given the requirement of domain-specific tuning. This paper aims at filling this gap, focusing on optimizing LLM inference on the Sophon SG2042, the first commercially available many-core RISC-V CPU with vector processing capabilities. On two recent state-of-the-art LLMs optimized for reasoning, DeepSeek R1 Distill Llama 8B and DeepSeek R1 Distill QWEN 14B, we achieve 4.32/2.29 token/s for token generation and 6.54/3.68 token/s for prompt processing, with a speed up of up 2.9x/3.0x compared to our baseline.
The growing adoption of long-context Large Language Models (LLMs) has introduced significant memory and computational challenges in autoregressive decoding due to the expanding Key-Value (KV) cache. KV cache quantization has emerged as a promising solution, with prior work showing that 4-bit or even 2-bit quantization can maintain model accuracy while reducing memory costs. However, despite these benefits, preliminary implementations for the low-bit KV cache struggle to deliver the expected speedup due to quantization and dequantization overheads and the lack of Tensor Cores utilization. In this work, we propose BitDecoding, a GPU-optimized framework that unlocks Tensor Cores for efficient decoding with low-bit KV cache. Efficiently leveraging Tensor Cores for low-bit KV cache is challenging due to the dynamic nature of KV cache generation at each decoding step. BitDecoding addresses these challenges with a Tensor Cores-Centric BitFusion Scheme that ensures data layout compatibility to enable high utilization of Tensor Cores. Additionally, BitDecoding incorporates a warp-efficient parallel decoding kernel and a fine-grained asynchronous pipeline, minimizing dequantization overhead and improving computational efficiency. Experiments show that BitDecoding achieves up to 7.5x speedup on RTX 4090, 4.8x on A100, and 8.9x on H100, compared to FP16 FlashDecoding-v2. It also outperforms the state-of-the-art low-bit KV cache implementation (QServe) by up to 4.3x. On LLaMA-3.1-8B with a 128K sequence length, BitDecoding reduces single-batch decoding latency by 3x, demonstrating its effectiveness in long-context generation scenarios. The code is available at https://github.com/DD-DuDa/BitDecoding.
Modern Large Language Model serving system batches multiple requests to achieve high throughput, while batching attention operations is challenging, rendering memory bandwidth a critical bottleneck. The community relies on high-end GPUs with multiple high-bandwidth memory channels. Unfortunately, HBM's high bandwidth often comes at the expense of limited memory capacity, which reduces core utilization and increases costs. Recent advancements enabling longer contexts for LLMs have substantially increased the key-value cache size, further intensifying the pressures on memory capacity. The literature has explored KV cache quantization techniques, which commonly use low bitwidth for most values, selectively using higher bitwidth for outlier values. While this approach helps achieve high accuracy and low bitwidth simultaneously, it comes with the limitation that cost for online outlier detection is excessively high, negating the advantages. We propose Oaken, an acceleration solution that achieves high accuracy and high performance simultaneously through co-designing algorithm and hardware. To effectively find a sweet spot in the accuracy-performance trade-off space of KV cache quantization, Oaken employs an online-offline hybrid approach, setting outlier thresholds offline, which are then used to determine the quantization scale online. To translate the proposed algorithmic technique into tangible performance gains, Oaken also comes with custom quantization engines and memory management units that can be integrated with any LLM accelerators. We built an Oaken accelerator on top of an LLM accelerator, LPU, and conducted a comprehensive evaluation. Our experiments show that for a batch size of 256, Oaken achieves up to 1.58x throughput improvement over NVIDIA A100 GPU, incurring a minimal accuracy loss of only 0.54\% on average, compared to state-of-the-art KV cache quantization techniques.
The ability of machine learning (ML) classification models to resist small, targeted input perturbations - known as adversarial attacks - is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6% when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets. We present a novel black-box attack using Bayesian optimization to determine external workloads that bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach, to learn floating point operation orderings that lead to misclassifications, making the assumption that any reduction or permutation ordering is possible. This LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, we investigate parallel reduction ordering across different GPU architectures for a reduction under three conditions: (1) executing external background workloads, (2) utilizing multi-GPU virtualization, and (3) applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions. The results and methods developed here can help to include machine-level considerations into adversarial robustness assessments.
As large language models (LLMs) have shown great success in many tasks, they are used in various applications. While a lot of works have focused on the efficiency of single-LLM application (e.g., offloading, request scheduling, parallelism strategy selection), multi-LLM applications receive less attention, particularly in offline inference scenarios. In this work, we aim to improve the offline end-to-end inference efficiency of multi-LLM applications in the single-node multi-GPU environment. The problem involves two key decisions: (1) determining which LLMs to run concurrently each time (we may not run all the models at the same time), and (2) selecting a parallelism strategy to use for each LLM. This problem is NP-hard. Naive solutions may not work well because the running time for a model to complete a set of requests depends on the request workload and the selected parallelism strategy, and they lack an accurate model of the running time. As the LLM output lengths are unknown before running, to estimate the model running time, we propose a sampling-then-simulation method which first estimates the output lengths by sampling from an empirical cumulative function we obtained from a large dataset in advance, and then simulates the LLM inference process accordingly. Based on the simulation, we estimate the per-iteration latencys to get the total latency. A greedy method is proposed to optimize the scheduling of the LLMs in the application across the GPUs. We then propose a framework SamuLLM which contains two phases: planning, which calls the greedy method for an application and running, which runs the application and dynamically adjust the model scheduling based on the runtime information. Experiments on 3 applications and a mixed application show that SamuLLM can achieve 1.0-2.4$\times$ end-to-end speedups compared to the competitors.
3D intelligence leverages rich 3D features and stands as a promising frontier in AI, with 3D rendering fundamental to many downstream applications. 3D Gaussian Splatting (3DGS), an emerging high-quality 3D rendering method, requires significant computation, making real-time execution on existing GPU-equipped edge devices infeasible. Previous efforts to accelerate 3DGS rely on dedicated accelerators that require substantial integration overhead and hardware costs. This work proposes an acceleration strategy that leverages the similarities between the 3DGS pipeline and the highly optimized conventional graphics pipeline in modern GPUs. Instead of developing a dedicated accelerator, we enhance existing GPU rasterizer hardware to efficiently support 3DGS operations. Our results demonstrate a 23$\times$ increase in processing speed and a 24$\times$ reduction in energy consumption, with improvements yielding 6$\times$ faster end-to-end runtime for the original 3DGS algorithm and 4$\times$ for the latest efficiency-improved pipeline, achieving 24 FPS and 46 FPS respectively. These enhancements incur only a minimal area overhead of 0.2\% relative to the entire SoC chip area, underscoring the practicality and efficiency of our approach for enabling 3DGS rendering on resource-constrained platforms.
The emergent dynamics of complex gene regulatory networks govern various cellular processes. However, understanding these dynamics is challenging due to the difficulty of parameterizing the computational models for these networks, especially as the network size increases. Here, we introduce a simulation library, Gene Regulatory Interaction Network Simulator (GRiNS), to address these challenges. GRiNS integrates popular parameter-agnostic simulation frameworks, RACIPE and Boolean Ising formalism, into a single Python library capable of leveraging GPU acceleration for efficient and scalable simulations. GRiNS extends the ordinary differential equations (ODE) based RACIPE framework with a more modular design, allowing users to choose parameters, initial conditions, and time-series outputs for greater customisability and accuracy in simulations. For large networks, where ODE-based simulation formalisms do not scale well, GRiNS implements Boolean Ising formalism, providing a simplified, coarse-grained alternative, significantly reducing the computational cost while capturing key dynamical behaviours of large regulatory networks. The documentation and installation instructions for GRiNS can be found at https://moltenecdysone09.github.io/GRiNS/.
As large language models (LLMs) scale, model compression is crucial for edge deployment and accessibility. Weight-only quantization reduces model size but suffers from performance degradation at lower bit widths. Moreover, standard finetuning is incompatible with quantized models, and alternative methods often fall short of full finetuning. In this paper, we propose ClusComp, a simple yet effective compression paradigm that clusters weight matrices into codebooks and finetunes them block-by-block. ClusComp (1) achieves superior performance in 2-4 bit quantization, (2) pushes compression to 1-bit while outperforming ultra-low-bit methods with minimal finetuning, and (3) enables efficient finetuning, even surpassing existing quantization-based approaches and rivaling full FP16 finetuning. Notably, ClusComp supports compression and finetuning of 70B LLMs on a single A6000-48GB GPU.
Transformer-based large language models (LLMs) have already achieved remarkable results on long-text tasks, but the limited GPU memory (VRAM) resources struggle to accommodate the linearly growing demand for key-value (KV) cache as the sequence length increases, which has become a bottleneck for the application of LLMs on long sequences. Existing KV cache compression methods include eviction, merging, or quantization of the KV cache to reduce its size. However, compression results in irreversible information forgetting, potentially affecting the accuracy of subsequent decoding. In this paper, we propose SpeCache, which takes full advantage of the large and easily expandable CPU memory to offload the complete KV cache, and dynamically fetches KV pairs back in each decoding step based on their importance measured by low-bit KV cache copy in VRAM. To avoid inference latency caused by CPU-GPU communication, SpeCache speculatively predicts the KV pairs that the next token might attend to, allowing us to prefetch them before the next decoding step which enables parallelization of prefetching and computation. Experiments on LongBench and Needle-in-a-Haystack benchmarks verify that SpeCache effectively reduces VRAM usage while avoiding information forgetting for long sequences without re-training, even with a 10x high KV cache compression ratio.
The advancement of remote sensing technology has improved the spatial resolution of satellite imagery, facilitating more detailed visual representations for diverse interpretations. However, existing methods exhibit limited generalization capabilities across varied applications. While some contemporary foundation models demonstrate potential, they are hindered by insufficient cross-task adaptability and primarily process low-resolution imagery of restricted sizes, thus failing to fully exploit high-resolution data or leverage comprehensive large-scene semantics. Crucially, remote sensing imagery differs fundamentally from natural images, as key foreground targets (eg., maritime objects, artificial structures) often occupy minimal spatial proportions (~1%) and exhibit sparse distributions. Efficiently modeling cross-task generalizable knowledge from lengthy 2D tokens (~100,000) poses a significant challenge yet remains critical for remote sensing image understanding. Motivated by the selective attention mechanisms inherent to the human visual system, we propose DynamicVis, a dynamic visual perception foundation model for remote sensing imagery. The framework integrates a novel dynamic region perception backbone based on the selective state space model, which strategically balances localized detail extraction with global contextual integration, enabling computationally efficient encoding of large-scale data while maintaining architectural scalability. To enhance cross-task knowledge transferring, we introduce a multi-instance learning paradigm utilizing meta-embedding representations, trained on million-scale region-level annotations. Evaluations across nine downstream tasks demonstrate the model's versatility. DynamicVis achieves multi-level feature modeling with exceptional efficiency, processing (2048x2048) pixels with 97 ms latency (6% of ViT's) and 833 MB GPU memory (3% of ViT's).