arXiv subjects

Haoran Wei

Publications and source records attributed to Haoran Wei.

At least 55 records · Page 3Linked to original sources

arXiv2024

DreamLLM: Synergistic Multimodal Comprehension and Creation

This paper presents DreamLLM, a learning framework that first achieves versatile Multimodal Large Language Models (MLLMs) empowered with frequently overlooked synergy between multimodal comprehension and creation. DreamLLM operates on two fundamental principles. The first focuses on the generative modeling of both language and image posteriors by direct sampling in the raw multimodal space. This approach circumvents the limitations and information loss inherent to external feature extractors like CLIP, and a more thorough multimodal understanding is obtained. Second, DreamLLM fosters the generation of raw, interleaved documents, modeling both text and image contents, along with unstructured layouts. This allows DreamLLM to learn all conditional, marginal, and joint multimodal distributions effectively. As a result, DreamLLM is the first MLLM capable of generating free-form interleaved content. Comprehensive experiments highlight DreamLLM's superior performance as a zero-shot multimodal generalist, reaping from the enhanced learning synergy. Project page: https://dreamllm.github.io.

cs.CV↗

arXiv2024

Anomalous thermal transport and high thermoelectric performance of Cu-based vanadate CuVO3

Thermoelectric (TE) conversion technology, capable of transforming heat into electricity, is critical for sustainable energy solutions. Many promising TE materials contain rare or toxic elements, so the development of cost-effective and eco-friendly high-performance TE materials is highly urgent. Herein, we explore the thermal transport and TE properties of transition metal vanadate CuVO3 by using first-principles calculation. On the basis of unified theory of heat conduction, we uncover the hierarchical thermal transport feature in CuVO3, where wave-like tunneling makes a significant contribution to the lattice thermal conductivity (\k{appa}l) and result in the anomalously weak temperature dependence of \k{appa}l. This is primarily attributable to the complex phononic band structure caused by the heterogeneity of Cu-O and V-O bonds. Simultaneously, we report a high power factor of 5.45 mW K-2 m-1 realized in hole-doped CuVO3, which arises from a high electrical conductivity and a large Seebeck coefficient enabled by the multiple valleys and large electronic density of states near the valence band edge. Impressively, the low \k{appa}l and the high power factor make p-typed CuVO3 have ZT of up to 1.39, with the excellent average ZT above 1.0 from 300 to 600 K, which is superior to most reported Cu-based TE materials. Our findings suggest that CuVO3 compound is promising candidate for energy conversion applications in innovative TE devices.

physics.comp-ph↗

arXiv2024

Half-Metallic Ferromagnetic Weyl Fermions Related to Dynamic Correlations in the Zinc-blende Compound VAs

The realization of 100\% polarized topological Weyl fermions in half-metallic ferromagnets is of particular importance for fundamental research and spintronic applications. Here, we theoretically investigate the electronic and topological properties of the zinc-blende compound VAs, which was deemed as a half-metallic ferromagnet related to dynamic correlations. Based on the combination of density functional theory and dynamical mean field theory, we uncover that the half-metallic ferromagnet VAs exhibit attractive Weyl semimetallic behaviors with twelve pairs of Weyl points, which are very close to the Fermi level. Meanwhile, we also investigate the magnetization-dependent topological properties; the results show that the change of magnetization directions only slightly affects the positions of Weyl points, which is attributed to the weak spin-orbital coupling effects. The topological surface states of VAs projected on semi-infinite (001) and (111) surfaces are investigated. The Fermi arcs of all Weyl points are clearly visible on the projected Fermi surfaces. Our findings suggest that VAs is a fully spin-polarized Weyl semimetal with many-body correlated effects for spintronic applications.

cond-mat.mtrl-sci↗

arXiv2024

MegaScale: Scaling Large Language Model Training to More Than 10,000 GPUs

We present the design, implementation and engineering experience in building and deploying MegaScale, a production system for training large language models (LLMs) at the scale of more than 10,000 GPUs. Training LLMs at this scale brings unprecedented challenges to training efficiency and stability. We take a full-stack approach that co-designs the algorithmic and system components across model block and optimizer design, computation and communication overlapping, operator optimization, data pipeline, and network performance tuning. Maintaining high efficiency throughout the training process (i.e., stability) is an important consideration in production given the long extent of LLM training jobs. Many hard stability issues only emerge at large scale, and in-depth observability is the key to address them. We develop a set of diagnosis tools to monitor system components and events deep in the stack, identify root causes, and derive effective techniques to achieve fault tolerance and mitigate stragglers. MegaScale achieves 55.2% Model FLOPs Utilization (MFU) when training a 175B LLM model on 12,288 GPUs, improving the MFU by 1.34x compared to Megatron-LM. We share our operational experience in identifying and fixing failures and stragglers. We hope by articulating the problems and sharing our experience from a systems perspective, this work can inspire future LLM systems research.

cs.LG↗

arXiv2024

Small Language Model Meets with Reinforced Vision Vocabulary

Playing Large Vision Language Models (LVLMs) in 2023 is trendy among the AI community. However, the relatively large number of parameters (more than 7B) of popular LVLMs makes it difficult to train and deploy on consumer GPUs, discouraging many researchers with limited resources. Imagine how cool it would be to experience all the features of current LVLMs on an old GTX1080ti (our only game card). Accordingly, we present Vary-toy in this report, a small-size Vary along with Qwen-1.8B as the base ``large'' language model. In Vary-toy, we introduce an improved vision vocabulary, allowing the model to not only possess all features of Vary but also gather more generality. Specifically, we replace negative samples of natural images with positive sample data driven by object detection in the procedure of generating vision vocabulary, more sufficiently utilizing the capacity of the vocabulary network and enabling it to efficiently encode visual information corresponding to natural objects. For experiments, Vary-toy can achieve 65.6% ANLS on DocVQA, 59.1% accuracy on ChartQA, 88.1% accuracy on RefCOCO, and 29% on MMVet. The code will be publicly available on the homepage.

cs.CV↗

arXiv2023

Single pair of charge-two high-fold fermions with type-II van Hove singularities on the surface of ultralight chiral crystals

The realization of single-pair chiral fermions in Weyl systems remains challenging in topology physics, especially for the systems with higher chiral charges $C$. In this work, based on the symmetry analysis, low-energy effective model, and first-principles calculations, we identify the single-pair high-fold fermions in chiral cubic lattices. We first derive the minimal lattice model that exhibits a single pair of Weyl points with the opposite chiral charges of $C$ = $\pm{2}$. Furthermore, we show the ultralight chiral crystal P4$_3$32-type LiC$_2$ and its mirror enantiomer as high-quality candidate materials, which exhibit large energy windows to surmount the interruption of irrelevant bands. Since two enantiomers are connected by the mirror symmetry, we observe the opposite chiral charges $C$ and the reversal of the Fermi arc velocities, showing the correspondence of chirality in the momentum space and the real space. In addition, we also reveal type-II van Hove singularities on the helicoid surfaces, which may induce chirality-locked charge density waves on the crystal surface. Our work not only provides a promising platform for controlling the sign of topological charge through the structural chirality but also facilitates the exploration of electronic correlations on the surface of ultralight chiral crystals.

cond-mat.mtrl-sci↗

arXiv2023

Vary: Scaling up the Vision Vocabulary for Large Vision-Language Models

Modern Large Vision-Language Models (LVLMs) enjoy the same vision vocabulary -- CLIP, which can cover most common vision tasks. However, for some special vision task that needs dense and fine-grained vision perception, e.g., document-level OCR or chart understanding, especially in non-English scenarios, the CLIP-style vocabulary may encounter low efficiency in tokenizing the vision knowledge and even suffer out-of-vocabulary problem. Accordingly, we propose Vary, an efficient and effective method to scale up the vision vocabulary of LVLMs. The procedures of Vary are naturally divided into two folds: the generation and integration of a new vision vocabulary. In the first phase, we devise a vocabulary network along with a tiny decoder-only transformer to produce the desired vocabulary via autoregression. In the next, we scale up the vanilla vision vocabulary by merging the new one with the original one (CLIP), enabling the LVLMs can quickly garner new features. Compared to the popular BLIP-2, MiniGPT4, and LLaVA, Vary can maintain its vanilla capabilities while enjoying more excellent fine-grained perception and understanding ability. Specifically, Vary is competent in new document parsing features (OCR or markdown conversion) while achieving 78.2% ANLS in DocVQA and 36.2% in MMVet. Our code will be publicly available on the homepage.

cs.CV↗

arXiv2023

Autoencoder with Group-based Decoder and Multi-task Optimization for Anomalous Sound Detection

In industry, machine anomalous sound detection (ASD) is in great demand. However, collecting enough abnormal samples is difficult due to the high cost, which boosts the rapid development of unsupervised ASD algorithms. Autoencoder (AE) based methods have been widely used for unsupervised ASD, but suffer from problems including 'shortcut', poor anti-noise ability and sub-optimal quality of features. To address these challenges, we propose a new AE-based framework termed AEGM. Specifically, we first insert an auxiliary classifier into AE to enhance ASD in a multi-task learning manner. Then, we design a group-based decoder structure, accompanied by an adaptive loss function, to endow the model with domain-specific knowledge. Results on the DCASE 2021 Task 2 development set show that our methods achieve a relative improvement of 13.11% and 15.20% respectively in average AUC over the official AE and MobileNetV2 across test sets of seven machines.

cs.SD↗

arXiv2023

DWRSeg: Rethinking Efficient Acquisition of Multi-scale Contextual Information for Real-time Semantic Segmentation

Many current works directly adopt multi-rate depth-wise dilated convolutions to capture multi-scale contextual information simultaneously from one input feature map, thus improving the feature extraction efficiency for real-time semantic segmentation. However, this design may lead to difficult access to multi-scale contextual information because of the unreasonable structure and hyperparameters. To lower the difficulty of drawing multi-scale contextual information, we propose a highly efficient multi-scale feature extraction method, which decomposes the original single-step method into two steps, Region Residualization-Semantic Residualization. In this method, the multi-rate depth-wise dilated convolutions take a simpler role in feature extraction: performing simple semantic-based morphological filtering with one desired receptive field in the second step based on each concise feature map of region form provided by the first step, to improve their efficiency. Moreover, the dilation rates and the capacity of dilated convolutions for each network stage are elaborated to fully utilize all the feature maps of region form that can be achieved.Accordingly, we design a novel Dilation-wise Residual (DWR) module and a Simple Inverted Residual (SIR) module for the high and low level network, respectively, and form a powerful DWR Segmentation (DWRSeg) network. Extensive experiments on the Cityscapes and CamVid datasets demonstrate the effectiveness of our method by achieving a state-of-the-art trade-off between accuracy and inference speed, in addition to being lighter weight. Without pretraining or resorting to any training trick, we achieve an mIoU of 72.7% on the Cityscapes test set at a speed of 319.5 FPS on one NVIDIA GeForce GTX 1080 Ti card, which exceeds the latest methods of a speed of 69.5 FPS and 0.8% mIoU. The code and trained models are publicly available.

cs.CV↗

arXiv2023

ChatSpot: Bootstrapping Multimodal LLMs via Precise Referring Instruction Tuning

Human-AI interactivity is a critical aspect that reflects the usability of multimodal large language models (MLLMs). However, existing end-to-end MLLMs only allow users to interact with them through language instructions, leading to the limitation of the interactive accuracy and efficiency. In this study, we present precise referring instructions that utilize diverse reference representations such as points and boxes as referring prompts to refer to the special region. This enables MLLMs to focus on the region of interest and achieve finer-grained interaction. Based on precise referring instruction, we propose ChatSpot, a unified end-to-end multimodal large language model that supports diverse forms of interactivity including mouse clicks, drag-and-drop, and drawing boxes, which provides a more flexible and seamless interactive experience. We also construct a multi-grained vision-language instruction-following dataset based on existing datasets and GPT-4 generating. Furthermore, we design a series of evaluation tasks to assess the effectiveness of region recognition and interaction. Experimental results showcase ChatSpot's promising performance.

cs.CL↗

arXiv2023

PolyLM: An Open Source Polyglot Large Language Model

Large language models (LLMs) demonstrate remarkable ability to comprehend, reason, and generate following nature language instructions. However, the development of LLMs has been primarily focused on high-resource languages, such as English, thereby limiting their applicability and research in other languages. Consequently, we present PolyLM, a multilingual LLM trained on 640 billion (B) tokens, avaliable in two model sizes: 1.7B and 13B. To enhance its multilingual capabilities, we 1) integrate bilingual data into training data; and 2) adopt a curriculum learning strategy that increases the proportion of non-English data from 30% in the first stage to 60% in the final stage during pre-training. Further, we propose a multilingual self-instruct method which automatically generates 132.7K diverse multilingual instructions for model fine-tuning. To assess the model's performance, we collect several existing multilingual tasks, including multilingual understanding, question answering, generation, and translation. Extensive experiments show that PolyLM surpasses other open-source models such as LLaMA and BLOOM on multilingual tasks while maintaining comparable performance in English. Our models, alone with the instruction data and multilingual benchmark, are available at: \url{https://modelscope.cn/models/damo/nlp_polylm_13b_text_generation}.

cs.CL↗

arXiv2023

Phonetic-assisted Multi-Target Units Modeling for Improving Conformer-Transducer ASR system

Exploiting effective target modeling units is very important and has always been a concern in end-to-end automatic speech recognition (ASR). In this work, we propose a phonetic-assisted multi target units (PMU) modeling approach, to enhance the Conformer-Transducer ASR system in a progressive representation learning manner. Specifically, PMU first uses the pronunciation-assisted subword modeling (PASM) and byte pair encoding (BPE) to produce phonetic-induced and text-induced target units separately; Then, three new frameworks are investigated to enhance the acoustic encoder, including a basic PMU, a paraCTC and a pcaCTC, they integrate the PASM and BPE units at different levels for CTC and transducer multi-task training. Experiments on both LibriSpeech and accented ASR tasks show that, the proposed PMU significantly outperforms the conventional BPE, it reduces the WER of LibriSpeech clean, other, and six accented ASR testsets by relative 12.7%, 6.0% and 7.7%, respectively.

eess.AS↗

arXiv2023

Multi-pass Training and Cross-information Fusion for Low-resource End-to-end Accented Speech Recognition

Low-resource accented speech recognition is one of the important challenges faced by current ASR technology in practical applications. In this study, we propose a Conformer-based architecture, called Aformer, to leverage both the acoustic information from large non-accented and limited accented training data. Specifically, a general encoder and an accent encoder are designed in the Aformer to extract complementary acoustic information. Moreover, we propose to train the Aformer in a multi-pass manner, and investigate three cross-information fusion methods to effectively combine the information from both general and accent encoders. All experiments are conducted on both the accented English and Mandarin ASR tasks. Results show that our proposed methods outperform the strong Conformer baseline by relative 10.2% to 24.5% word/character error rate reduction on six in-domain and out-of-domain accented test sets.

cs.SD↗

arXiv2023

1st Place Solution for PVUW Challenge 2023: Video Panoptic Segmentation

Video panoptic segmentation is a challenging task that serves as the cornerstone of numerous downstream applications, including video editing and autonomous driving. We believe that the decoupling strategy proposed by DVIS enables more effective utilization of temporal information for both "thing" and "stuff" objects. In this report, we successfully validated the effectiveness of the decoupling strategy in video panoptic segmentation. Finally, our method achieved a VPQ score of 51.4 and 53.7 in the development and test phases, respectively, and ultimately ranked 1st in the VPS track of the 2nd PVUW Challenge. The code is available at https://github.com/zhang-tao-whu/DVIS

cs.CV↗

arXiv2023

Collaborative Multi-BS Power Management for Dense Radio Access Network using Deep Reinforcement Learning

Network energy efficiency is a main pillar in the design and operation of wireless communication systems. In this paper, we investigate a dense radio access network (dense-RAN) capable of radiated power management at the base station (BS). Aiming to improve the long-term network energy efficiency, an optimization problem is formulated by collaboratively managing multi-BSs radiated power levels with constraints on the users traffic volume and achievable rate. Considering stochastic traffic arrivals at the users and time-varying network interference, we first formulate the problem as a Markov decision process (MDP) and then develop a novel deep reinforcement learning (DRL) framework based on the cloud-RAN operation scheme. To tackle the trade-off between complexity and performance, the overall optimization of multi-BSs energy efficiency with the multiplicative complexity constraint is modeled to achieve nearoptimal performance by using a deep Q-network (DQN). In DQN,each BS first maximizes its individual energy efficiency, and then cooperates with other BSs to maximize the overall multiBSs energy efficiency. Simulation results demonstrate that the proposed algorithm can converge faster and enjoy a network energy efficiency improvement by 5% and 10% compared with the benchmarks of the Q-learning and sleep schemes, respectively.

eess.SP↗

arXiv2022

Dimensionality-confined superconductivity within SrNbO3-SrTiO3 heterostructures

Interfaces between transition-metal oxides are able to host two-dimensional electron gases (2DEGs) and exhibit exotic quantum phenomena. Here we report the observation of superconductivity below 230 mK for the heterostructure composed of SrNbO3 (SNO) and SrTiO3 (STO). Different from some other counterparts with two insulators, the metallic SNO provides a novel mechanism to form a quasi 2DEG by charge transfer from bulk towards interface under strain. The superconductivity, residing within the strained SNO layer near the interface, is contributed by an electron system with record-low carrier density. Notably, although embedded in a normal metallic layer with a carrier density 4 to 5 orders higher, the electron system is still uniquely well-protected to retain high mobility and lies deep in extreme quantum regime.

cond-mat.supr-con↗

arXiv2022

Competency-Aware Neural Machine Translation: Can Machine Translation Know its Own Translation Quality?

Neural machine translation (NMT) is often criticized for failures that happen without awareness. The lack of competency awareness makes NMT untrustworthy. This is in sharp contrast to human translators who give feedback or conduct further investigations whenever they are in doubt about predictions. To fill this gap, we propose a novel competency-aware NMT by extending conventional NMT with a self-estimator, offering abilities to translate a source sentence and estimate its competency. The self-estimator encodes the information of the decoding procedure and then examines whether it can reconstruct the original semantics of the source sentence. Experimental results on four translation tasks demonstrate that the proposed method not only carries out translation tasks intact but also delivers outstanding performance on quality estimation. Without depending on any reference or annotated data typically required by state-of-the-art metric and quality estimation methods, our model yields an even higher correlation with human quality judgments than a variety of aforementioned methods, such as BLEURT, COMET, and BERTScore. Quantitative and qualitative analyses show better robustness of competency awareness in our model.

cs.CL↗

arXiv2022

Non-Homogeneous Haze Removal via Artificial Scene Prior and Bidimensional Graph Reasoning

Due to the lack of natural scene and haze prior information, it is greatly challenging to completely remove the haze from a single image without distorting its visual content. Fortunately, the real-world haze usually presents non-homogeneous distribution, which provides us with many valuable clues in partial well-preserved regions. In this paper, we propose a Non-Homogeneous Haze Removal Network (NHRN) via artificial scene prior and bidimensional graph reasoning. Firstly, we employ the gamma correction iteratively to simulate artificial multiple shots under different exposure conditions, whose haze degrees are different and enrich the underlying scene prior. Secondly, beyond utilizing the local neighboring relationship, we build a bidimensional graph reasoning module to conduct non-local filtering in the spatial and channel dimensions of feature maps, which models their long-range dependency and propagates the natural scene prior between the well-preserved nodes and the nodes contaminated by haze. To the best of our knowledge, this is the first exploration to remove non-homogeneous haze via the graph reasoning based framework. We evaluate our method on different benchmark datasets. The results demonstrate that our method achieves superior performance over many state-of-the-art algorithms for both the single image dehazing and hazy image understanding tasks. The source code of the proposed NHRN is available on https://github.com/whrws/NHRNet.

cs.CV↗