Quantize Once, Train Fast: Allreduce-Compatible Compression with Provable Guarantees

• URL: http://arxiv.org/abs/2305.18627v2
• Date: Tue, 29 Jul 2025 12:28:13 GMT
• Title: Quantize Once, Train Fast: Allreduce-Compatible Compression with Provable Guarantees
• Authors: Jihao Xin, Marco Canini, Peter Richtárik, Samuel Horváth,
• Abstract summary: We introduce Global-QSGD, an All-reduce gradient-compatible quantization method.<n>We show that it accelerates distributed training by up to 3.51% over baseline quantization methods.
• Score: 53.950234267704
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Distributed training enables large-scale deep learning, but suffers from high communication overhead, especially as models and datasets grow. Gradient compression, particularly quantization, is a promising approach to mitigate this bottleneck. However, existing quantization schemes are often incompatible with Allreduce, the dominant communication primitive in distributed deep learning, and many prior solutions rely on heuristics without theoretical guarantees. We introduce Global-QSGD, an Allreduce-compatible gradient quantization method that leverages global norm scaling to reduce communication overhead while preserving accuracy. Global-QSGD is backed by rigorous theoretical analysis, extending standard unbiased compressor frameworks to establish formal convergence guarantees. Additionally, we develop a performance model to evaluate its impact across different hardware configurations. Extensive experiments on NVLink, PCIe, and large-scale cloud environments show that Global-QSGD accelerates distributed training by up to 3.51% over baseline quantization methods, making it a practical and efficient solution for large-scale deep learning workloads.

Related papers

• QuantVSR: Low-Bit Post-Training Quantization for Real-World Video Super-Resolution [53.13952833016505]
  We propose a low-bit quantization model for real-world video super-resolution (VSR)<n>We use a calibration dataset to measure both spatial and temporal complexity for each layer.<n>We refine the FP and low-bit branches to achieve simultaneous optimization.
  arXiv  Detail & Related papers  (2025-08-06T14:35:59Z)
• Supervised Optimism Correction: Be Confident When LLMs Are Sure [91.7459076316849]
  We establish a novel theoretical connection between supervised fine-tuning and offline reinforcement learning. We show that the widely used beam search method suffers from unacceptable over-optimism. We propose Supervised Optimism Correction, which introduces a simple yet effective auxiliary loss for token-level $Q$-value estimations.
  arXiv  Detail & Related papers  (2025-04-10T07:50:03Z)
• QT-DoG: Quantization-aware Training for Domain Generalization [58.439816306817306]
  We propose Quantization-aware Training for Domain Generalization (QT-DoG) QT-DoG exploits quantization as an implicit regularizer by inducing noise in model weights. We demonstrate that QT-DoG generalizes across various datasets, architectures, and quantization algorithms.
  arXiv  Detail & Related papers  (2024-10-08T13:21:48Z)
• Mask-Encoded Sparsification: Mitigating Biased Gradients in Communication-Efficient Split Learning [15.78336840511033]
  This paper introduces a novel framework designed to achieve a high compression ratio in Split Learning (SL) scenarios. Our investigations demonstrate that compressing feature maps within SL leads to biased gradients that can negatively impact the convergence rates. We employ a narrow bit-width encoded mask to compensate for the sparsification error without increasing the order of time complexity.
  arXiv  Detail & Related papers  (2024-08-25T09:30:34Z)
• Differential error feedback for communication-efficient decentralized learning [48.924131251745266]
  We propose a new decentralized communication-efficient learning approach that blends differential quantization with error feedback. We show that the resulting communication-efficient strategy is stable both in terms of mean-square error and average bit rate. The results establish that, in the small step-size regime and with a finite number of bits, it is possible to attain the performance achievable in the absence of compression.
  arXiv  Detail & Related papers  (2024-06-26T15:11:26Z)
• Communication-Efficient Distributed Learning with Local Immediate Error Compensation [95.6828475028581]
  We propose the Local Immediate Error Compensated SGD (LIEC-SGD) optimization algorithm. LIEC-SGD is superior to previous works in either the convergence rate or the communication cost.
  arXiv  Detail & Related papers  (2024-02-19T05:59:09Z)
• EControl: Fast Distributed Optimization with Compression and Error Control [8.624830915051021]
  We propose EControl, a novel mechanism that can regulate the strength of the feedback signal. We show that EControl mitigates the naive implementation of our method and support our findings.
  arXiv  Detail & Related papers  (2023-11-06T10:00:13Z)
• Asynchronous Federated Learning with Incentive Mechanism Based on Contract Theory [5.502596101979607]
  We propose a novel asynchronous FL framework that integrates an incentive mechanism based on contract theory. Our framework exhibits a 1.35% accuracy improvement over the ideal Local SGD under attacks.
  arXiv  Detail & Related papers  (2023-10-10T09:17:17Z)
• Learning Accurate Performance Predictors for Ultrafast Automated Model Compression [86.22294249097203]
  We propose an ultrafast automated model compression framework called SeerNet for flexible network deployment. Our method achieves competitive accuracy-complexity trade-offs with significant reduction of the search cost.
  arXiv  Detail & Related papers  (2023-04-13T10:52:49Z)
• Quantized Distributed Training of Large Models with Convergence Guarantees [34.054462975511996]
  We present QSDP, a variant of FSDP which supports both quant and weight gradientization with theoretical guarantees. We show that QSDP preserves accuracy, while completely removing the communication of FSDP, providing the speed-to-endups of up to 2.2x.
  arXiv  Detail & Related papers  (2023-02-05T14:20:55Z)
• On the Convergence of Heterogeneous Federated Learning with Arbitrary Adaptive Online Model Pruning [15.300983585090794]
  We present a unifying framework for heterogeneous FL algorithms with em arbitrary adaptive online model pruning. In particular, we prove that under certain sufficient conditions, these algorithms converge to a stationary point of standard FL for general smooth cost functions. We illuminate two key factors impacting convergence: pruning-induced noise and minimum coverage index.
  arXiv  Detail & Related papers  (2022-01-27T20:43:38Z)
• Optimizing the Communication-Accuracy Trade-off in Federated Learning with Rate-Distortion Theory [1.5771347525430772]
  A significant bottleneck in federated learning is the network communication cost of sending model updates from client devices to the central server. Our method encodes quantized updates with an appropriate universal code, taking into account their empirical distribution. Because quantization introduces error, we select quantization levels by optimizing for the desired trade-off in average total gradient and distortion.
  arXiv  Detail & Related papers  (2022-01-07T20:17:33Z)
• Optimal Rate Adaption in Federated Learning with Compressed Communications [28.16239232265479]
  Federated Learning incurs high communication overhead, which can be greatly alleviated by compression for model updates. tradeoff between compression and model accuracy in the networked environment remains unclear. We present a framework to maximize the final model accuracy by strategically adjusting the compression each iteration.
  arXiv  Detail & Related papers  (2021-12-13T14:26:15Z)
• Permutation Compressors for Provably Faster Distributed Nonconvex Optimization [68.8204255655161]
  We show that the MARINA method of Gorbunov et al (2021) can be considered as a state-of-the-art method in terms of theoretical communication complexity. Theory of MARINA to support the theory of potentially em correlated compressors, extends to the method beyond the classical independent compressors setting.
  arXiv  Detail & Related papers  (2021-10-07T09:38:15Z)
• Generalizable Mixed-Precision Quantization via Attribution Rank Preservation [90.26603048354575]
  We propose a generalizable mixed-precision quantization (GMPQ) method for efficient inference. Our method obtains competitive accuracy-complexity trade-off compared with the state-of-the-art mixed-precision networks.
  arXiv  Detail & Related papers  (2021-08-05T16:41:57Z)
• Smoothness-Aware Quantization Techniques [0.2578242050187029]
  We show that block quantization with $n$ blocks outperforms single block quantization. We also show that our smoothness-aware quantization strategies outperform existing quantization schemes.
  arXiv  Detail & Related papers  (2021-06-07T11:30:05Z)
• Federated Learning with Unreliable Clients: Performance Analysis and Mechanism Design [76.29738151117583]
  Federated Learning (FL) has become a promising tool for training effective machine learning models among distributed clients. However, low quality models could be uploaded to the aggregator server by unreliable clients, leading to a degradation or even a collapse of training. We model these unreliable behaviors of clients and propose a defensive mechanism to mitigate such a security risk.
  arXiv  Detail & Related papers  (2021-05-10T08:02:27Z)
• ScaleCom: Scalable Sparsified Gradient Compression for Communication-Efficient Distributed Training [74.43625662170284]
  Large-scale distributed training of Deep Neural Networks (DNNs) on state-of-the-art platforms is expected to be severely communication constrained. We propose a new compression technique that leverages similarity in the gradient distribution amongst learners to provide significantly improved scalability. We experimentally demonstrate that ScaleCom has small overheads, directly reduces gradient traffic and provides high compression rates (65-400X) and excellent scalability (up to 64 learners and 8-12X larger batch sizes over standard training) without significant accuracy loss.
  arXiv  Detail & Related papers  (2021-04-21T02:22:10Z)
• An Efficient Statistical-based Gradient Compression Technique for Distributed Training Systems [77.88178159830905]
  Sparsity-Inducing Distribution-based Compression (SIDCo) is a threshold-based sparsification scheme that enjoys similar threshold estimation quality to deep gradient compression (DGC) Our evaluation shows SIDCo speeds up training by up to 41:7%, 7:6%, and 1:9% compared to the no-compression baseline, Topk, and DGC compressors, respectively.
  arXiv  Detail & Related papers  (2021-01-26T13:06:00Z)
• Structured Sparsification with Joint Optimization of Group Convolution and Channel Shuffle [117.95823660228537]
  We propose a novel structured sparsification method for efficient network compression. The proposed method automatically induces structured sparsity on the convolutional weights. We also address the problem of inter-group communication with a learnable channel shuffle mechanism.
  arXiv  Detail & Related papers  (2020-02-19T12:03:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.