arXiv subjects

Yufei Ding

Publications and source records attributed to Yufei Ding.

At least 73 records · Page 4Linked to original sources

arXiv2024

Surf-Deformer: Mitigating Dynamic Defects on Surface Code via Adaptive Deformation

In this paper, we introduce Surf-Deformer, a code deformation framework that seamlessly integrates adaptive defect mitigation functionality into the current surface code workflow. It crafts several basic deformation instructions based on fundamental gauge transformations, which can be combined to explore a larger design space than previous methods. This enables more optimized deformation processes tailored to specific defect situations, restoring the QEC capability of deformed codes more efficiently with minimal qubit resources. Additionally, we design an adaptive code layout that accommodates our defect mitigation strategy while ensuring efficient execution of logical operations. Our evaluation shows that Surf-Deformer outperforms previous methods by significantly reducing the end-to-end failure rate of various quantum programs by 35x to 70x, while requiring only about 50% of the qubit resources compared to the previous method to achieve the same level of failure rate. Ablation studies show that Surf-Deformer surpasses previous defect removal methods in preserving QEC capability and facilitates surface code communication by achieving nearly optimal throughput.

quant-ph↗

arXiv2024

Improving GPU Multi-Tenancy Through Dynamic Multi-Instance GPU Reconfiguration

Continuous learning (CL) has emerged as one of the most popular deep learning paradigms deployed in modern cloud GPUs. Specifically, CL has the capability to continuously update the model parameters (through model retraining) and use the updated model (if available) to serve overtime arriving inference requests. It is generally beneficial to co-locate the retraining and inference together to enable timely model updates and avoid model transfer overheads. This brings the need for GPU sharing among retraining and inferences. Meanwhile, multiple CL workloads can share the modern GPUs in the cloud, leading to multi-tenancy execution. In this paper, we observe that prior GPU-sharing techniques are not optimized for multi-tenancy CL workloads. Specifically, they do not coherently consider the accuracy of the retraining model and the inference service level objective (SLO) attainment. Moreover, they cannot accommodate the overtime dynamics (e.g., inference arrival intensity) in CL execution. In this paper, we propose MIGRator, a novel GPU reconfiguration runtime that dynamically performs GPU reconfiguration for multi-tenancy CL workloads. MIGRator is based on the recent NVIDIA multi-instance GPU (MIG) to mitigate resource contention and formulates the reconfiguration optimization into Integer Linear Programming (ILP) to dynamically identify, reconfigure, and allocate the GPU instances. MIGRator leverages the "Goodput" metric in the ILP objective function to consider both inference SLO attainment and model accuracy in the reconfiguration exploration. We evaluate MIGRator using representative multi-tenancy CL workloads. The results show our approach outperforms the state-of-the-art GPU sharing techniques (i.e., Ekya, Astraea, and PARIS) by 17\%, 21\%, and 20\%, respectively.

cs.DC↗

arXiv2023

Quantum Memory: A Missing Piece in Quantum Computing Units

Memory is an indispensable component in classical computing systems. While the development of quantum computing is still in its early stages, current quantum processing units mainly function as quantum registers. Consequently, the actual role of quantum memory in future advanced quantum computing architectures remains unclear. With the rapid scaling of qubits, it is opportune to explore the potential and feasibility of quantum memory across different substrate device technologies and application scenarios. In this paper, we provide a full design stack view of quantum memory. We start from the elementary component of a quantum memory device, quantum memory cells. We provide an abstraction to a quantum memory cell and define metrics to measure the performance of physical platforms. Combined with addressing functionality, we then review two types of quantum memory devices: random access quantum memory (RAQM) and quantum random access memory (QRAM). Building on top of these devices, quantum memory units in the computing architecture, including building a quantum memory unit, quantum cache, quantum buffer, and using QRAM for the quantum input-output module, are discussed. We further propose the programming model for the quantum memory units and discuss their possible applications. By presenting this work, we aim to attract more researchers from both the Quantum Information Science (QIS) and classical memory communities to enter this emerging and exciting area.

quant-ph↗

arXiv2023

Enabling Full-Stack Quantum Computing with Changeable Error-Corrected Qubits

Executing quantum applications with quantum error correction (QEC) faces the gate non-universality problem imposed by the Eastin-Knill theorem. As one resource-time-efficient solution, code switching changes the encoding of logical qubits to implement universal logical gates. Unfortunately, it is still unclear how to perform full-stack fault-tolerant quantum computing (FTQC) based on the changeable logical qubit. Specifically, three critical problems remain unsolved: a) how to implement the dynamic logical qubit on hardware; b) how to determine the appropriate timing for logical qubit varying; c) how to improve the overall system performance for programs of different features. To overcome those design problems, We propose CECQ, to explore the large design space for FTQC based on changeable logical qubits. Experiments on various quantum programs demonstrate the effectiveness of CECQ.

quant-ph↗

arXiv2023

A Geometrical Approach to Evaluate the Adversarial Robustness of Deep Neural Networks

Deep Neural Networks (DNNs) are widely used for computer vision tasks. However, it has been shown that deep models are vulnerable to adversarial attacks, i.e., their performances drop when imperceptible perturbations are made to the original inputs, which may further degrade the following visual tasks or introduce new problems such as data and privacy security. Hence, metrics for evaluating the robustness of deep models against adversarial attacks are desired. However, previous metrics are mainly proposed for evaluating the adversarial robustness of shallow networks on the small-scale datasets. Although the Cross Lipschitz Extreme Value for nEtwork Robustness (CLEVER) metric has been proposed for large-scale datasets (e.g., the ImageNet dataset), it is computationally expensive and its performance relies on a tractable number of samples. In this paper, we propose the Adversarial Converging Time Score (ACTS), an attack-dependent metric that quantifies the adversarial robustness of a DNN on a specific input. Our key observation is that local neighborhoods on a DNN's output surface would have different shapes given different inputs. Hence, given different inputs, it requires different time for converging to an adversarial sample. Based on this geometry meaning, ACTS measures the converging time as an adversarial robustness metric. We validate the effectiveness and generalization of the proposed ACTS metric against different adversarial attacks on the large-scale ImageNet dataset using state-of-the-art deep networks. Extensive experiments show that our ACTS metric is an efficient and effective adversarial metric over the previous CLEVER metric.

cs.CV↗

arXiv2023

Multi-mode Cavity Centric Architectures for Quantum Simulation

Near-term quantum computing technologies grapple with huge complexity overheads, hindering their ability to induce algorithms, necessitating engineering and scientific innovations. One class of problems of interest is Quantum Simulation, whereby quantum systems are simulated using a quantum computer. However, current devices are yet to surpass classical tensor network techniques. For problems of interest, where classical simulation techniques fail, large degrees of entanglement are required. Another challenge of implementing quantum simulation problems is that qubits sit idle whilst alternating simulation terms are implemented, exposing the system to decoherence. In the near term, 2D planar superconducting lattices of circuit-QED elements such as the transmon continue to draw substantial attention, but they are hindered by their nearest neighbor topology and relatively short lifespan, two problems that are problematic for quantum simulation. One technology of particular interest is the multi-mode superconducting resonator capable of storing multiple qubits in one device. We observe that these cavities have a natural virtual topology that aligns particularly well with quantum simulation problems, and exhibit much longer lifespans in comparison to other planar superconducting hardware. In this paper we present MUCIC, we discuss the simple integration of these devices into the current landscape and their implications to quantum simulation, motivated by their alignment to the quantum simulation problem, and potential as a quantum memory candidate. We report the development of MUCICs transpiler, leading to reductions of up to 82% in quantum simulation circuit depths. Additionally, our investigation demonstrates improvements of up to 19.4% in converged results from Variational Quantum Algorithms.

quant-ph↗

arXiv2023

QASMTrans: A QASM based Quantum Transpiler Framework for NISQ Devices

The success of a quantum algorithm hinges on the ability to orchestrate a successful application induction. Detrimental overheads in mapping general quantum circuits to physically implementable routines can be the deciding factor between a successful and erroneous circuit induction. In QASMTrans, we focus on the problem of rapid circuit transpilation. Transpilation plays a crucial role in converting high-level, machine-agnostic circuits into machine-specific circuits constrained by physical topology and supported gate sets. The efficiency of transpilation continues to be a substantial bottleneck, especially when dealing with larger circuits requiring high degrees of inter-qubit interaction. QASMTrans is a high-performance C++ quantum transpiler framework that demonstrates up to 369X speedups compared to the commonly used Qiskit transpiler. We observe speedups on large dense circuits such as uccsd_n24 and qft_n320 which require O(10^6) gates. QASMTrans successfully transpiles the aforementioned circuits in 69s and 31s, whilst Qiskit exceeded an hour of transpilation time. With QASMTrans providing transpiled circuits in a fraction of the time of prior transpilers, potential design space exploration, and heuristic-based transpiler design becomes substantially more tractable. QASMTrans is released at http://github.com/pnnl/qasmtrans.

quant-ph↗

arXiv2023

MGG: Accelerating Graph Neural Networks with Fine-grained intra-kernel Communication-Computation Pipelining on Multi-GPU Platforms

The increasing size of input graphs for graph neural networks (GNNs) highlights the demand for using multi-GPU platforms. However, existing multi-GPU GNN systems optimize the computation and communication individually based on the conventional practice of scaling dense DNNs. For irregularly sparse and fine-grained GNN workloads, such solutions miss the opportunity to jointly schedule/optimize the computation and communication operations for high-performance delivery. To this end, we propose MGG, a novel system design to accelerate full-graph GNNs on multi-GPU platforms. The core of MGG is its novel dynamic software pipeline to facilitate fine-grained computation-communication overlapping within a GPU kernel. Specifically, MGG introduces GNN-tailored pipeline construction and GPU-aware pipeline mapping to facilitate workload balancing and operation overlapping. MGG also incorporates an intelligent runtime design with analytical modeling and optimization heuristics to dynamically improve the execution performance. Extensive evaluation reveals that MGG outperforms state-of-the-art full-graph GNN systems across various settings: on average 4.41X, 4.81X, and 10.83X faster than DGL, MGG-UVM, and ROC, respectively.

cs.DC↗

arXiv2023

OneQ: A Compilation Framework for Photonic One-Way Quantum Computation

In this paper, we propose OneQ, the first optimizing compilation framework for one-way quantum computation towards realistic photonic quantum architectures. Unlike previous compilation efforts for solid-state qubit technologies, our innovative framework addresses a unique set of challenges in photonic quantum computing. Specifically, this includes the dynamic generation of qubits over time, the need to perform all computation through measurements instead of relying on 1-qubit and 2-qubit gates, and the fact that photons are instantaneously destroyed after measurements. As pioneers in this field, we demonstrate the vast optimization potential of photonic one-way quantum computing, showcasing the remarkable ability of OneQ to reduce computing resource requirements by orders of magnitude.

quant-ph↗

arXiv2023

Biologically Inspired Dynamic Thresholds for Spiking Neural Networks

The dynamic membrane potential threshold, as one of the essential properties of a biological neuron, is a spontaneous regulation mechanism that maintains neuronal homeostasis, i.e., the constant overall spiking firing rate of a neuron. As such, the neuron firing rate is regulated by a dynamic spiking threshold, which has been extensively studied in biology. Existing work in the machine learning community does not employ bioinspired spiking threshold schemes. This work aims at bridging this gap by introducing a novel bioinspired dynamic energy-temporal threshold (BDETT) scheme for spiking neural networks (SNNs). The proposed BDETT scheme mirrors two bioplausible observations: a dynamic threshold has 1) a positive correlation with the average membrane potential and 2) a negative correlation with the preceding rate of depolarization. We validate the effectiveness of the proposed BDETT on robot obstacle avoidance and continuous control tasks under both normal conditions and various degraded conditions, including noisy observations, weights, and dynamic environments. We find that the BDETT outperforms existing static and heuristic threshold approaches by significant margins in all tested conditions, and we confirm that the proposed bioinspired dynamic threshold scheme offers homeostasis to SNNs in complex real-world tasks.

cs.RO↗

arXiv2023

TC-GNN: Bridging Sparse GNN Computation and Dense Tensor Cores on GPUs

Recently, graph neural networks (GNNs), as the backbone of graph-based machine learning, demonstrate great success in various domains (e.g., e-commerce). However, the performance of GNNs is usually unsatisfactory due to the highly sparse and irregular graph-based operations. To this end, we propose TC-GNN, the first GNN acceleration framework based on GPU Tensor Core Units (TCUs). The core idea is to reconcile the "Sparse" GNN computation with the high-performance "Dense" TCUs. Specifically, we conduct an in-depth analysis of the sparse operations in mainstream GNN computing frameworks. We introduce a novel sparse graph translation technique to facilitate TCU processing of the sparse GNN workload. We implement an effective CUDA core and TCU collaboration design to fully utilize GPU resources. We integrate TC-GNN with the PyTorch framework for high programmability. Rigorous experiments show an average of 1.70X speedup over the state-of-the-art DGL framework across various models and datasets.

cs.LG↗

arXiv2023

ALCOP: Automatic Load-Compute Pipelining in Deep Learning Compiler for AI-GPUs

Pipelining between data loading and computation is a critical tensor program optimization for GPUs. In order to unleash the high performance of latest GPUs, we must perform a synergetic optimization of multi-stage pipelining across the multi-level buffer hierarchy of GPU. Existing frameworks rely on hand-written libraries such as cuBLAS to perform pipelining optimization, which is inextensible to new operators and un-composable with prior tensor compiler optimizations. This paper presents ALCOP, the first framework that is compiler-native and fully supports multi-stage multi-level pipelining. ALCOP overcomes three critical obstacles in generating code for pipelining: detection of pipelining-applicable buffers, program transformation for multi-level multi-stage pipelining, and efficient schedule parameter search by incorporating static analysis. Experiments show that ALCOP can generate programs with 1.23x speedup on average (up to 1.73x) over vanilla TVM. On end-to-end models, ALCOP can improve upon TVM by up to 1.18x, and XLA by up to 1.64x. Besides, our performance model significantly improves the efficiency of the schedule tuning process and can find schedules with 99% of the performance given by exhaustive search while costing 40x fewer trials.

cs.DC↗

arXiv2023

On Adversarial Robustness of Point Cloud Semantic Segmentation

Recent research efforts on 3D point cloud semantic segmentation (PCSS) have achieved outstanding performance by adopting neural networks. However, the robustness of these complex models have not been systematically analyzed. Given that PCSS has been applied in many safety-critical applications like autonomous driving, it is important to fill this knowledge gap, especially, how these models are affected under adversarial samples. As such, we present a comparative study of PCSS robustness. First, we formally define the attacker's objective under performance degradation and object hiding. Then, we develop new attack by whether to bound the norm. We evaluate different attack options on two datasets and three PCSS models. We found all the models are vulnerable and attacking point color is more effective. With this study, we call the attention of the research community to develop new approaches to harden PCSS models.

cs.CV↗

arXiv2023

Quantum Bayesian Error Mitigation Employing Poisson Modelling over the Hamming Spectrum for Quantum Error Mitigation

The field of quantum computing has experienced a rapid expansion in recent years, with ongoing exploration of new technologies, a decrease in error rates, and a growth in the number of qubits available in quantum processors. However, near-term quantum algorithms are still unable to be induced without compounding consequential levels of noise, leading to non-trivial erroneous results. Quantum Error Correction and Mitigation are rapidly advancing areas of research in the quantum computing landscape, with a goal of reducing errors. IBM has recently emphasized that Quantum Error Mitigation is the key to unlocking the full potential of quantum computing. A recent work, namely HAMMER, demonstrated the existence of a latent structure regarding post-circuit induction errors when mapping to the Hamming spectrum. However, they assumed that errors occur solely in local clusters, whereas we observe that at higher average Hamming distances this structure falls away. Our study demonstrates that the correlated structure is not just limited to local patterns, but it also encompasses certain non-local clustering patterns that can be accurately characterized through a Poisson distribution model. This model takes into account the input circuit, the current state of the device, including calibration statistics, and the qubit topology. Using this quantum error characterizing model, we developed an iterative algorithm over the generated Bayesian network state-graph for post-induction error mitigation. Our Q-Beep approach delivers state-of-the-art results, thanks to its problem-aware modeling of the error distribution's underlying structure and the implementation of an Bayesian network state-graph. This has resulted in an increase of up to 234.6% in circuit execution accuracy on BV circuits and an average improvement of 71.0% in the quality of QAOA solutions when tested on 16 IBMQ quantum processors.

quant-ph↗

arXiv2022

AutoComm: A Framework for Enabling Efficient Communication in Distributed Quantum Programs

Distributed quantum computing (DQC) is a promising approach to extending the computational power of near-term quantum devices. However, the non-local quantum communication between quantum devices is much more expensive and error-prone than the local quantum communication within each quantum device. Previous work on the DQC communication optimization focus on optimizing the communication protocol for each individual non-local gate and then adopt quantum compilation designs which are designed for local multi-qubit gates (such as controlled-x or CX gates) in a single quantum computer. The communication patterns in distributed quantum programs are not yet well studied, leading to a far-from-optimal communication cost. In this paper, we identify burst communication, a specific qubit-node communication pattern that widely exists in many distributed programs and can be leveraged to guide communication overhead optimization. We then propose AutoComm, an automatic compiler framework to first extract the burst communication patterns from the input programs, and then optimize the communication steps of burst communication discovered. Experimental results show that our proposed AutoComm can reduce the communication resource consumption and the program latency by 75.6% and 71.4% on average, respectively.

quant-ph↗

arXiv2022

Faith: An Efficient Framework for Transformer Verification on GPUs

Transformer verification draws increasing attention in machine learning research and industry. It formally verifies the robustness of transformers against adversarial attacks such as exchanging words in a sentence with synonyms. However, the performance of transformer verification is still not satisfactory due to bound-centric computation which is significantly different from standard neural networks. In this paper, we propose Faith, an efficient framework for transformer verification on GPUs. We first propose a semantic-aware computation graph transformation to identify semantic information such as bound computation in transformer verification. We exploit such semantic information to enable efficient kernel fusion at the computation graph level. Second, we propose a verification-specialized kernel crafter to efficiently map transformer verification to modern GPUs. This crafter exploits a set of GPU hardware supports to accelerate verification specialized operations which are usually memory-intensive. Third, we propose an expert-guided autotuning to incorporate expert knowledge on GPU backends to facilitate large search space exploration. Extensive evaluations show that Faith achieves $2.1\times$ to $3.4\times$ ($2.6\times$ on average) speedup over state-of-the-art frameworks.

cs.LG↗

arXiv2022

CollComm: Enabling Efficient Collective Quantum Communication Based on EPR buffering

The noisy and lengthy nature of quantum communication hinders the development of distributed quantum computing. The inefficient design of existing compilers for distributed quantum computing worsens the situation. Previous compilation frameworks couple communication hardware with the implementation of expensive remote gates. However, we discover that the efficiency of quantum communication, especially collective communication, can be significantly boosted by decoupling communication resources from remote operations, that is, the communication hardware would be used only for preparing remote entanglement, and the computational hardware, the component used to store program information, would be used for conducting remote gates. Based on the observation, we develop a compiler framework to optimize the collective communication happening in distributed quantum programs. In this framework, we decouple the communication preparation process in communication hardware from the remote gates conducted in computational hardware by buffering EPR pairs generated by communication hardware in qubits of the computational hardware. Experimental results show that the proposed framework can halve the communication cost of various distributed quantum programs, compared to state-of-the-art compilers for distributed quantum computing.

quant-ph↗

arXiv2022

LightSeq2: Accelerated Training for Transformer-based Models on GPUs

Transformer-based neural models are used in many AI applications. Training these models is expensive, as it takes huge GPU resources and long duration. It is challenging because typical data like sentences have variable lengths, and Transformer's computation patterns are more complex than convolutional neural networks. Existing systems either only focus on model inference or optimization for only BERT-like encoder models. In this paper, we present LightSeq2, a system to accelerate training for a general family of Transformer models on GPUs. We propose a series of GPU optimization techniques tailored to the specific computation flow and memory access patterns of Transformer models. LightSeq2 supports many model architectures, including BERT (encoder-only), GPT (decoder-only), Transformer (encoder-decoder), and vision Transformer. Our experiments for a variety of models and benchmarks show that LightSeq2 is consistently faster (1.4-3.5x) than previous systems on different GPUs. In particular, it gains 308% training speedup compared with existing systems on a large public machine translation benchmark (WMT14 English-German).

cs.CL↗