Corpus ID: 201666112

Once for All: Train One Network and Specialize it for Efficient Deployment

@article{Cai2020OnceFA,
  title={Once for All: Train One Network and Specialize it for Efficient Deployment},
  author={Han Cai and Chuang Gan and S. Han},
  journal={ArXiv},
  year={2020},
  volume={abs/1908.09791}
}
  • Han Cai, Chuang Gan, S. Han
  • Published 2020
  • Mathematics, Computer Science
  • ArXiv
  • We address the challenging problem of efficient inference across many devices and resource constraints, especially on edge devices. Conventional approaches either manually design or use neural architecture search (NAS) to find a specialized neural network and train it from scratch for each case, which is computationally prohibitive (causing $CO_2$ emission as much as 5 cars' lifetime) thus unscalable. In this work, we propose to train a once-for-all (OFA) network that supports diverse… CONTINUE READING
    PONAS: Progressive One-shot Neural Architecture Search for Very Efficient Deployment
    1
    Automating Deep Neural Network Model Selection for Edge Inference
    Discovering Multi-Hardware Mobile Models via Architecture Search
    Domain-Aware Dynamic Networks
    1
    X X Table 1 : Comparisons of architecture search approaches for efficient models ( ProxylessNAS
    • 2019
    Scaling Up Deep Neural Network Optimization for Edge Inference
    DA-NAS: Data Adapted Pruning for Efficient Neural Architecture Search
    1
    APQ: Joint Search for Network Architecture, Pruning and Quantization Policy
    6
    Neural Architecture Transfer
    2
    MoGA: Searching Beyond Mobilenetv3
    9

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 51 REFERENCES
    FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search
    268
    MnasNet: Platform-Aware Neural Architecture Search for Mobile
    556
    Single Path One-Shot Neural Architecture Search with Uniform Sampling
    128
    ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware
    471
    Efficient Architecture Search by Network Transformation
    248
    Learning both Weights and Connections for Efficient Neural Network
    2370
    Slimmable Neural Networks
    75