Accurate, Large Minibatch SGD (Goyal et al.)
Optional · paper guide 1 of 2 for DLCV 01A · Training Deep Vision Models in Practice. Citation numbers refer to the reference list at the bottom of this page.
Navigate: this page and this track: Navigate: this page and this track
Challenge
Training on many GPUs at once requires a large total batch, because each GPU needs enough work to be efficient. But larger batches had been reported to hurt accuracy. Goyal et al. asked whether this loss was an optimization problem that could be fixed [9].

Figure source: [9], arXiv:1706.02677v2, PDF page 8. Original PDF.
Method
Two rules: scale the learning rate linearly with the batch size, and ramp it up gradually over the first epochs. The scaling rule follows from a simple comparison. SGD steps with small batches and rate move the weights by times the sum of small-batch mean gradients, each taken at a slightly different point; one step with a -times larger batch moves them by times the mean of the same gradients taken at the starting point. If the gradients barely change over the steps, the two updates match when . The paper names the two situations where that assumption fails: early in training, when the network changes quickly, which the warmup addresses, and beyond some batch size, where accuracy degrades.
Results
ResNet-50 on ImageNet with a batch of 8,192 images across 256 GPUs, trained in one hour, matched the accuracy of the batch-256 baseline [9].
Reading the evidence
The key table compares warmup strategies at batch 8k with the baseline, each as mean and standard deviation of the top-1 validation error over 5 trials: baseline 23.60% ± 0.12, no warmup 24.84% ± 0.37, constant warmup 25.88% ± 0.56, gradual warmup 23.74% ± 0.09 [9]. Read it in three steps. First, the spread of each row: differences smaller than about two standard deviations are not meaningful. Second, the surprise: a constant low rate for warmup is worse than no warmup, because the switch to the full rate causes a spike from which training never fully recovers. Third, the authors attribute the no-warmup gap to higher training error, not to worse generalization, which is what makes it an optimization problem.
Practical pitfalls
Scaling the cross-entropy loss is not equivalent to scaling the learning rate when weight decay is present; momentum needs a correction when the learning rate changes if the implementation folds the rate into the momentum buffer; BN statistics are computed per GPU over 32 images, and this per-worker batch is part of the definition of the loss; the zero- initialization of the last BN in each residual block helps, especially at large batches.
Problems
The experiments are ImageNet classification with SGD and momentum, plus transfer of the pretrained models to detection and instance segmentation; the rule is an approximation whose limit (about 8k images in their setting) depends on the task and model, and it says nothing direct about adaptive optimizers.
Conclusion
It is the source of two defaults in this chapter, linear scaling with warmup and zero- residual initialization, and an example of reporting variance over several trials.
References
- T. He, Z. Zhang, H. Zhang, Z. Zhang, J. Xie and M. Li, “Bag of tricks for image classification with convolutional neural networks,” in Proc. IEEE/CVF CVPR, 2019. doi
- A. Karpathy, “A recipe for training neural networks,” blog post, 2019. link
- scikit-learn developers, “load_digits,” scikit-learn documentation, version 1.9.1. link
- M. Buda, A. Maki and M. A. Mazurowski, “A systematic study of the class imbalance problem in convolutional neural networks,” Neural Networks, vol. 106, 2018. doi
- T.-Y. Lin, P. Goyal, R. Girshick, K. He and P. Dollár, “Focal loss for dense object detection,” in Proc. IEEE ICCV, 2017. doi
- A. Buslaev, V. I. Iglovikov, E. Khvedchenya, A. Parinov, M. Druzhinin and A. A. Kalinin, “Albumentations: Fast and flexible image augmentations,” Information, vol. 11, no. 2, 2020. doi
- C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” Journal of Big Data, vol. 6, 2019. doi
- K. He, X. Zhang, S. Ren and J. Sun, “Delving deep into rectifiers: Surpassing human-level performance on ImageNet classification,” in Proc. IEEE ICCV, 2015. doi
- P. Goyal, P. Dollár, R. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, A. Tulloch, Y. Jia and K. He, “Accurate, large minibatch SGD: Training ImageNet in 1 hour,” arXiv:1706.02677 (technical report), 2017. arXiv
- S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proc. ICML, PMLR 37, 2015. link
- Y. Wu and K. He, “Group normalization,” in Proc. ECCV, 2018. doi
- Y. Bengio, “Practical recommendations for gradient-based training of deep architectures,” in Neural Networks: Tricks of the Trade, 2nd ed., LNCS 7700, Springer, 2012. doi
- L. N. Smith, “Cyclical learning rates for training neural networks,” in Proc. IEEE WACV, 2017. doi
- L. N. Smith, “A disciplined approach to neural network hyper-parameters: Part 1 — learning rate, batch size, momentum, and weight decay,” US Naval Research Laboratory Technical Report 5510-026, arXiv:1803.09820, 2018. arXiv
- I. Loshchilov and F. Hutter, “SGDR: Stochastic gradient descent with warm restarts,” in Proc. ICLR, 2017. arXiv
- L. N. Smith and N. Topin, “Super-convergence: Very fast training of neural networks using large learning rates,” in Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications, Proc. SPIE 11006, 2019. doi
- PyTorch contributors, “torch.optim.lr_scheduler.OneCycleLR,” PyTorch 2.15 documentation, 2026. link
- K. He, X. Zhang, S. Ren and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE CVPR, 2016. doi
- J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li and L. Fei-Fei, “ImageNet: A large-scale hierarchical image database,” in Proc. IEEE CVPR, 2009. doi
- D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in Proc. ICLR, 2015. arXiv
- I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” in Proc. ICLR, 2019. arXiv
- PyTorch contributors, “torch.optim.Adam,” PyTorch 2.15 documentation, 2026. link
- N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” Journal of Machine Learning Research, vol. 15, 2014. link
- G. Huang, Y. Sun, Z. Liu, D. Sedra and K. Q. Weinberger, “Deep networks with stochastic depth,” in Proc. ECCV, 2016. doi
- L. Prechelt, “Early stopping — but when?,” in Neural Networks: Tricks of the Trade, 2nd ed., LNCS 7700, Springer, 2012. doi
- V. Godbole, G. E. Dahl, J. Gilmer, C. J. Shallue and Z. Nado, “Deep learning tuning playbook,” Google Research, GitHub repository, 2023. link
- B. T. Polyak and A. B. Juditsky, “Acceleration of stochastic approximation by averaging,” SIAM Journal on Control and Optimization, vol. 30, no. 4, 1992. doi
- P. Izmailov, D. Podoprikhin, T. Garipov, D. Vetrov and A. G. Wilson, “Averaging weights leads to wider optima and better generalization,” in Proc. UAI, 2018. arXiv
- D. Morales-Brotons, T. Vogels and H. Hendrikx, “Exponential moving average of weights in deep learning: Dynamics and benefits,” Transactions on Machine Learning Research, 2024. arXiv
- P. Micikevicius, S. Narang, J. Alben, G. Diamos, E. Elsen, D. Garcia, B. Ginsburg, M. Houston, O. Kuchaiev, G. Venkatesh and H. Wu, “Mixed precision training,” in Proc. ICLR, 2018. arXiv
- PyTorch contributors, “Automatic mixed precision package — torch.amp,” PyTorch 2.15 documentation, 2026. link
- J. Yosinski, J. Clune, Y. Bengio and H. Lipson, “How transferable are features in deep neural networks?,” in Advances in Neural Information Processing Systems 27, 2014. link
- E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang and W. Chen, “LoRA: Low-rank adaptation of large language models,” in Proc. ICLR, 2022. arXiv
- J. Howard and S. Ruder, “Universal language model fine-tuning for text classification,” in Proc. ACL, 2018. doi
- A. Kumar, A. Raghunathan, R. Jones, T. Ma and P. Liang, “Fine-tuning can distort pretrained features and underperform out-of-distribution,” in Proc. ICLR, 2022. arXiv
- C. Zhang, S. Bengio, M. Hardt, B. Recht and O. Vinyals, “Understanding deep learning requires rethinking generalization,” in Proc. ICLR, 2017. arXiv
- R. Geirhos, J.-H. Jacobsen, C. Michaelis, R. Zemel, W. Brendel, M. Bethge and F. A. Wichmann, “Shortcut learning in deep neural networks,” Nature Machine Intelligence, vol. 2, 2020. doi
- PyTorch contributors, “Reproducibility,” PyTorch 2.15 documentation, 2026. link
- Y. Xin, J. Yang, S. Luo, Y. Du, Q. Qin, K. Cen, Y. He, Z. Zhang, B. Fu, X. Yang, G. Zhai, M.-H. Yang and X. Liu, “Parameter-efficient fine-tuning for pre-trained vision models: A survey and benchmark,” arXiv:2402.02242, 2024. arXiv
- G. E. Dahl, F. Schneider, Z. Nado, N. Agarwal, C. S. Sastry, P. Hennig, S. Medapati, R. Eschenhagen, P. Kasimbeg, D. Suo et al., “Benchmarking neural network training algorithms,” arXiv:2306.07179, 2023. arXiv
- P. Kasimbeg, F. Schneider, R. Eschenhagen, J. Bae, C. S. Sastry, M. Saroufim, B. Feng, L. Wright, E. Z. Yang, Z. Nado, S. Medapati, P. Hennig, M. Rabbat and G. E. Dahl, “Accelerating neural network training: An analysis of the AlgoPerf competition,” in Proc. ICLR, 2025. arXiv
- A. Defazio, X. A. Yang, H. Mehta, K. Mishchenko, A. Khaled and A. Cutkosky, “The road less scheduled,” in Advances in Neural Information Processing Systems 37, 2024. link
- S. Hu, Y. Tu, X. Han, C. He, G. Cui, X. Long, Z. Zheng, Y. Fang, Y. Huang, W. Zhao et al., “MiniCPM: Unveiling the potential of small language models with scalable training strategies,” in Proc. Conference on Language Modeling (COLM), 2024. arXiv
- A. Hägele, E. Bakouch, A. Kosson, L. Ben Allal, L. Von Werra and M. Jaggi, “Scaling laws and compute-optimal training beyond fixed training durations,” in Advances in Neural Information Processing Systems 37, 2024. link
- M. Wortsman, G. Ilharco, J. W. Kim, M. Li, S. Kornblith, R. Roelofs, R. Gontijo Lopes, H. Hajishirzi, A. Farhadi, H. Namkoong and L. Schmidt, “Robust fine-tuning of zero-shot models,” in Proc. IEEE/CVF CVPR, 2022. doi
- M. Wortsman, G. Ilharco, S. Y. Gadre, R. Roelofs, R. Gontijo-Lopes, A. S. Morcos, H. Namkoong, A. Farhadi, Y. Carmon, S. Kornblith and L. Schmidt, “Model soups: Averaging weights of multiple fine-tuned models improves accuracy without increasing inference time,” in Proc. ICML, PMLR 162, 2022. link
- PyTorch contributors, “Automatic mixed precision examples,” PyTorch 2.15 documentation, 2026. link
- PyTorch contributors, “DistributedDataParallel,” PyTorch 2.15.0 source documentation, 2026. Official source.