The multi-line LiDAR is widely used in autonomous vehicles, so point cloud-based 3D detectors are essential for autonomous driving. Extracting rich multi-scale features is crucial for point cloud-based 3D detectors in autonomous driving due to significant differences in the size of different types of objects. However, because of the real-time requirements, large-size convolution kernels are rarely used to extract large-scale features in the backbone. Current 3D detectors commonly use feature pyramid networks to obtain large-scale features; however, some objects containing fewer point clouds are further lost during down-sampling, resulting in degraded performance. Since pillar-based schemes require much less computation than voxel-based schemes, they are more suitable for constructing real-time 3D detectors. Hence, we propose the *, a pillar-based scheme. We redesigned the feature encoding, the backbone, and the neck of the 3D detector. We propose the Voxel2Pillar feature encoding, which uses a sparse convolution constructor to construct pillars with richer point cloud features, especially height features. The Voxel2Pillar adds more learnable parameters to the feature encoding, enabling the initial pillars to have higher performance ability. We extract multi-scale and large-scale features in the proposed fully sparse backbone, which does not utilize large-size convolutional kernels; the backbone consists of the proposed multi-scale feature extraction module. The neck consists of the proposed sparse ConvNeXt, whose simple structure significantly improves the performance. We validate the effectiveness of the proposed * on the Waymo Open Dataset, and the object detection accuracy for vehicles, pedestrians, and cyclists is improved. We also verify the effectiveness of each proposed module in detail through ablation studies.
相關內容
Despite outstanding processes in many tasks, Large Language Models (LLMs) still lack accuracy when dealing with highly technical domains. Especially, telecommunications (telco) is a particularly challenging domain due the large amount of lexical, semantic and conceptual peculiarities. Yet, this domain holds many valuable use cases, directly linked to industrial needs. Hence, this paper studies how LLMs can be adapted to the telco domain. It reports our effort to (i) collect a massive corpus of domain-specific data (800M tokens, 80K instructions), (ii) perform adaptation using various methodologies, and (iii) benchmark them against larger generalist models in downstream tasks that require extensive knowledge of telecommunications. Our experiments on Llama-2-7b show that domain-adapted models can challenge the large generalist models. They also suggest that adaptation can be restricted to a unique instruction-tuning step, dicarding the need for any fine-tuning on raw texts beforehand.
This paper thoroughly surveys machine learning (ML) algorithms acceleration in hardware accelerators, focusing on Field-Programmable Gate Arrays (FPGAs). It reviews 287 out of 1138 papers from the past six years, sourced from four top FPGA conferences. Such selection underscores the increasing integration of ML and FPGA technologies and their mutual importance in technological advancement. Research clearly emphasises inference acceleration (81\%) compared to training acceleration (13\%). Additionally, the findings reveals that CNN dominates current FPGA acceleration research while emerging models like GNN show obvious growth trends. The categorization of the FPGA research papers reveals a wide range of topics, demonstrating the growing relevance of ML in FPGA research. This comprehensive analysis provides valuable insights into the current trends and future directions of FPGA research in the context of ML applications.
We investigate the proof complexity of systems based on positive branching programs, i.e. non-deterministic branching programs (NBPs) where, for any 0-transition between two nodes, there is also a 1-transition. Positive NBPs compute monotone Boolean functions, just like negation-free circuits or formulas, but constitute a positive version of (non-uniform) NL, rather than P or NC1, respectively. The proof complexity of NBPs was investigated in previous work by Buss, Das and Knop, using extension variables to represent the dag-structure, over a language of (non-deterministic) decision trees, yielding the system eLNDT. Our system eLNDT+ is obtained by restricting their systems to a positive syntax, similarly to how the 'monotone sequent calculus' MLK is obtained from the usual sequent calculus LK by restricting to negation-free formulas. Our main result is that eLNDT+ polynomially simulates eLNDT over positive sequents. Our proof method is inspired by a similar result for MLK by Atserias, Galesi and Pudl\'ak, that was recently improved to a bona fide polynomial simulation via works of Je\v{r}\'abek and Buss, Kabanets, Kolokolova and Kouck\'y. Along the way we formalise several properties of counting functions within eLNDT+ by polynomial-size proofs and, as a case study, give explicit polynomial-size poofs of the propositional pigeonhole principle.
To deploy LLMs on resource-contained platforms such as mobile robotics and wearables, non-transformers LLMs have achieved major breakthroughs. Recently, a novel RNN-based LLM family, Repentance Weighted Key Value (RWKV) models have shown promising results in text generation on resource-constrained devices thanks to their computational efficiency. However, these models remain too large to be deployed on embedded devices due to their high parameter count. In this paper, we propose an efficient suite of compression techniques, tailored to the RWKV architecture. These techniques include low-rank approximation, sparsity predictors, and clustering head, designed to align with the model size. Our methods compress the RWKV models by 4.95--3.8x with only 2.95pp loss in accuracy.
This paper targets the challenge of real-time LiDAR re-simulation in dynamic driving scenarios. Recent approaches utilize neural radiance fields combined with the physical modeling of LiDAR sensors to achieve high-fidelity re-simulation results. Unfortunately, these methods face limitations due to high computational demands in large-scale scenes and cannot perform real-time LiDAR rendering. To overcome these constraints, we propose LiDAR-RT, a novel framework that supports real-time, physically accurate LiDAR re-simulation for driving scenes. Our primary contribution is the development of an efficient and effective rendering pipeline, which integrates Gaussian primitives and hardware-accelerated ray tracing technology. Specifically, we model the physical properties of LiDAR sensors using Gaussian primitives with learnable parameters and incorporate scene graphs to handle scene dynamics. Building upon this scene representation, our framework first constructs a bounding volume hierarchy (BVH), then casts rays for each pixel and generates novel LiDAR views through a differentiable rendering algorithm. Importantly, our framework supports realistic rendering with flexible scene editing operations and various sensor configurations. Extensive experiments across multiple public benchmarks demonstrate that our method outperforms state-of-the-art methods in terms of rendering quality and efficiency. Our project page is at //zju3dv.github.io/lidar-rt.
High order strong stability preserving (SSP) time discretizations ensure the nonlinear non-inner-product strong stability properties of spatial discretizations suited for the stable simulation of hyperbolic PDEs. Over the past decade multiderivative time-stepping have been used for the time-evolution hyperbolic PDEs, so that the strong stability properties of these methods have become increasingly relevant. In this work we review sufficient conditions for a two-derivative multistage method to preserve the strong stability properties of spatial discretizations in a forward Euler and different conditions on the second derivative. In particular we present the SSP theory for explicit and implicit two-derivative Runge--Kutta schemes, and discuss a special condition on the second derivative under which these implicit methods may be unconditionally SSP. This condition is then used in the context of implicit-explicit (IMEX) multi-derivative Runge--Kutta schemes, where the time-step restriction is independent of the stiff term. Finally, we present the SSP theory for implicit-explicit (IMEX) multi-derivative general linear methods, and some novel second and third order methods where the time-step restriction is independent of the stiff term.
A flexible specification approach for verifying total correctness of fine-grained concurrent modules
A well-established approach to proving progress properties such as deadlock-freedom and termination is to associate obligations with threads. For example, in most existing work the proof rule for lock acquisition prescribes a standard usage protocol by burdening the acquiring thread with an obligation to release the lock. The fact that the obligation creation is hardcoded into the acquire operation, however, rules out non-standard clients e.g. where the release happens in a different thread. We overcome this limitation by instead having the blocking operations take the obligation creation operations required for the specific client scenario as arguments. We dub this simple instance of higher-order programming with auxiliary code Sassy. To illustrate Sassy, we extend HeapLang, a simple, higher-order, concurrent programming language with erasable code and state. The resulting language gets stuck if no progress is made. Consequently, we can apply standard safety separation logic to compositionally reason about termination in a fine-grained concurrent setting. We validated Sassy by developing (non-foundational) machine-checked proofs of representative locks -- an unfair Spinlock (competitive succession), a fair Ticketlock (direct handoff succession) and the hierarchically constructed Cohortlock that is starvation-free if the underlying locks are starvation-free -- against our specifications using an encoding of the approach in the VeriFast program verifier for C and Java.
Language models are increasingly being incorporated as components in larger AI systems for various purposes, from prompt optimization to automatic evaluation. In this work, we analyze the construct validity of four recent, commonly used methods for measuring text-to-image consistency - CLIPScore, TIFA, VPEval, and DSG - which rely on language models and/or VQA models as components. We define construct validity for text-image consistency metrics as a set of desiderata that text-image consistency metrics should have, and find that no tested metric satisfies all of them. We find that metrics lack sufficient sensitivity to language and visual properties. Next, we find that TIFA, VPEval and DSG contribute novel information above and beyond CLIPScore, but also that they correlate highly with each other. We also ablate different aspects of the text-image consistency metrics and find that not all model components are strictly necessary, also a symptom of insufficient sensitivity to visual information. Finally, we show that all three VQA-based metrics likely rely on familiar text shortcuts (such as yes-bias in QA) that call their aptitude as quantitative evaluations of model performance into question.
In this study, we address the central issue of statistical inference for Markov jump processes using discrete time observations. The primary problem at hand is to accurately estimate the infinitesimal generator of a Markov jump process, a critical task in various applications. To tackle this problem, we begin by reviewing established methods for generating sample paths from a Markov jump process conditioned to endpoints, known as Markov bridges. Additionally, we introduce a novel algorithm grounded in the concept of time-reversal, which serves as our main contribution. Our proposed method is then employed to estimate the infinitesimal generator of a Markov jump process. To achieve this, we use a combination of Markov Chain Monte Carlo techniques and the Monte Carlo Expectation-Maximization algorithm. The results obtained from our approach demonstrate its effectiveness in providing accurate parameter estimates. To assess the efficacy of our proposed method, we conduct a comprehensive comparative analysis with existing techniques (Bisection, Uniformization, Direct, Rejection, and Modified Rejection), taking into consideration both speed and accuracy. Notably, our method stands out as the fastest among the alternatives while maintaining high levels of precision.
We present ResMLP, an architecture built entirely upon multi-layer perceptrons for image classification. It is a simple residual network that alternates (i) a linear layer in which image patches interact, independently and identically across channels, and (ii) a two-layer feed-forward network in which channels interact independently per patch. When trained with a modern training strategy using heavy data-augmentation and optionally distillation, it attains surprisingly good accuracy/complexity trade-offs on ImageNet. We will share our code based on the Timm library and pre-trained models.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191