liguang190223 / once-for-all

[ICLR 2020] Once for All: Train One Network and Specialize it for Efficient Deployment

Once for All: Train One Network and Specialize it for Efficient Deployment [arXiv] [Slides] [Video]

@inproceedings{
  cai2020once,
  title={Once for All: Train One Network and Specialize it for Efficient Deployment},
  author={Han Cai and Chuang Gan and Tianzhe Wang and Zhekai Zhang and Song Han},
  booktitle={International Conference on Learning Representations},
  year={2020},
  url={https://arxiv.org/pdf/1908.09791.pdf}
}

News

• Fisrt place in the 4th Low-Power Computer Vision Challenge, both classification and detection track.
• First place in the 3rd Low-Power Computer Vision Challenge, DSP track at ICCV’19 using the Once-for-all Network.

Train once, specialize for many deployment scenarios

80% top1 ImageNet accuracy under mobile setting

Consistently outperforms MobileNetV3 on Diverse hardware platforms

How to use / evaluate OFA Specialized Networks

Use

""" OFA Specialized Networks.
Example: net, image_size = ofa_specialized('flops@[email protected]_finetune@75', pretrained=True)
""" 
from model_zoo import ofa_specialized
net, image_size = ofa_specialized(net_id, pretrained=True)

If the above scripts failed to download, you download it manually from Google Drive and put them under $HOME/.torch/ofa_specialized/.

Evaluate

python eval_specialized_net.py --path 'Your path to imagent' --net flops@[email protected]_finetune@75

OFA based on FLOPs

• flops@[email protected]_finetune@75
• flops@[email protected]_finetune@75
• flops@[email protected]_finetune@75

OFA for Mobile Phones

LG G8 LG-G8_lat@[email protected]_finetune@25 LG-G8_lat@[email protected]_finetune@25 LG-G8_lat@[email protected]_finetune@25 LG-G8_lat@[email protected]_finetune@25	Samsung Note8 note8_lat@[email protected]_finetune@25 note8_lat@[email protected]_finetune@25 note8_lat@[email protected]_finetune@25 note8_lat@[email protected]_finetune@25
Google Pixel1 pixel1_lat@[email protected]_finetune@75 pixel1_lat@[email protected]_finetune@75 pixel1_lat@[email protected]_finetune@75 pixel1_lat@[email protected]_finetune@75 pixel1_lat@[email protected]_finetune@25 pixel1_lat@[email protected]_finetune@25 pixel1_lat@[email protected]_finetune@25	Samsung Note10 note10_lat@[email protected]_finetune@75 note10_lat@[email protected]_finetune@75 note10_lat@[email protected]_finetune@75 note10_lat@[email protected]_finetune@75 note10_lat@[email protected]_finetune@25 note10_lat@[email protected]_finetune@25 note10_lat@[email protected]_finetune@25 note10_lat@[email protected]_finetune@25
Google Pixel2 pixel2_lat@[email protected]_finetune@25 pixel2_lat@[email protected]_finetune@25 pixel2_lat@[email protected]_finetune@25 pixel2_lat@[email protected]_finetune@25	Samsung S7 Edge s7edge_lat@[email protected]_finetune@25 s7edge_lat@[email protected]_finetune@25 s7edge_lat@[email protected]_finetune@25 s7edge_lat@[email protected]_finetune@25

OFA for Desktop (CPUs and GPUs)

1080ti GPU (Batch Size 64) 1080ti_gpu64@[email protected]_finetune@25 1080ti_gpu64@[email protected]_finetune@25 1080ti_gpu64@[email protected]_finetune@25 1080ti_gpu64@[email protected]_finetune@25	V100 GPU (Batch Size 64) v100_gpu64@[email protected]_finetune@25 v100_gpu64@[email protected]_finetune@25 v100_gpu64@[email protected]_finetune@25 v100_gpu64@[email protected]_finetune@25
Jetson TX2 GPU (Batch Size 16) tx2_gpu16@[email protected]_finetune@25 tx2_gpu16@[email protected]_finetune@25 tx2_gpu16@[email protected]_finetune@25 tx2_gpu16@[email protected]_finetune@25	Intel Xeon CPU with MKL-DNN (Batch Size 1) cpu_lat@[email protected]_finetune@25 cpu_lat@[email protected]_finetune@25 cpu_lat@[email protected]_finetune@25 cpu_lat@[email protected]_finetune@25

How to use / evaluate OFA Networks

Use

""" OFA Networks.
    Example: ofa_network = ofa_net('ofa_mbv3_d234_e346_k357_w1.0', pretrained=True)
""" 
from model_zoo import ofa_net
ofa_network = ofa_net(net_id, pretrained=True)
    
# Randomly sample sub-networks from OFA network
ofa_network.sample_active_subnet()
random_subnet = ofa_network.get_active_subnet(preserve_weight=True)
    
# Manually set the sub-network
ofa_network.set_active_subnet(ks=7, e=6, d=4)
manual_subnet = ofa_network.get_active_subnet(preserve_weight=True)

If the above scripts failed to download, you download it manually from Google Drive and put them under $HOME/.torch/ofa_nets/.

Evaluate

python eval_ofa_net.py --path 'Your path to imagenet' --net ofa_mbv3_d234_e346_k357_w1.0

How to train OFA Networks

mpirun -np 32 -H <server1_ip>:8,<server2_ip>:8,<server3_ip>:8,<server4_ip>:8 \
    -bind-to none -map-by slot \
    -x NCCL_DEBUG=INFO -x LD_LIBRARY_PATH -x PATH \
    python train_ofa_net.py

or

horovodrun -np 32 -H <server1_ip>:8,<server2_ip>:8,<server3_ip>:8,<server4_ip>:8 \
    python train_ofa_net.py

Introduction Video

Requirement

• Python 3.6
• Pytorch 1.0.0
• ImageNet Dataset
• Horovod

Related work on automated and efficient deep learning:

ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware (ICLR’19)

AutoML for Architecting Efficient and Specialized Neural Networks (IEEE Micro)

AMC: AutoML for Model Compression and Acceleration on Mobile Devices (ECCV’18)

HAQ: Hardware-Aware Automated Quantization (CVPR’19, oral)