Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
samithuang 966b180bc1 | 1 year ago | |
---|---|---|
.. | ||
README.md | 1 year ago | |
shufflenet_v2_0.5_ascend.yaml | 1 year ago | |
shufflenet_v2_1.0_ascend.yaml | 1 year ago | |
shufflenet_v2_1.5_ascend.yaml | 1 year ago | |
shufflenet_v2_2.0_ascend.yaml | 1 year ago |
ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
A key point was raised in ShuffleNetV2, where previous lightweight networks were guided by computing an indirect measure of network complexity, namely FLOPs. The speed of lightweight networks is described by calculating the amount of floating point operations. But the speed of operation was never considered directly. The running speed in mobile devices needs to consider not only FLOPs, but also other factors such as memory accesscost and platform characterics.
Therefore, based on these two principles, ShuffleNetV2 proposes four effective network design principles.
Figure 1. Architecture Design in ShuffleNetV2 [1]
Our reproduced model performance on ImageNet-1K is reported as follows.
Model | Context | Top-1 (%) | Top-5 (%) | Params (M) | Recipe | Download |
---|---|---|---|---|---|---|
shufflenet_v2_x0_5 | D910x8-G | 60.68 | 82.44 | 1.37 | yaml | weights |
shufflenet_v2_x1_0 | D910x8-G | 69.51 | 88.67 | 2.29 | yaml | weights |
shufflenet_v2_x1_5 | D910x8-G | 72.59 | 90.79 | 3.53 | yaml | weights |
shufflenet_v2_x2_0 | D910x8-G | 75.14 | 92.13 | 7.44 | yaml | weights |
Please refer to the installation instruction in MindCV.
Please download the ImageNet-1K dataset for model training and validation.
It is easy to reproduce the reported results with the pre-defined training recipe. For distributed training on multiple Ascend 910 devices, please run
# distrubted training on multiple GPU/Ascend devices
mpirun -n 8 python train.py --config configs/shufflenet_v2/shufflenet_v2_0.5_ascend.yaml --data_dir /path/to/imagenet
Similarly, you can train the model on multiple GPU devices with the above mpirun
command.
For detailed illustration of all hyper-parameters, please refer to config.py.
Note: As the global batch size (batch_size x num_devices) is an important hyper-parameter, it is recommended to keep the global batch size unchanged for reproduction or adjust the learning rate linearly to a new global batch size.
If you want to train or finetune the model on a smaller dataset without distributed training, please run:
# standalone training on a CPU/GPU/Ascend device
python train.py --config configs/shufflenet_v2/shufflenet_v2_0.5_ascend.yaml --data_dir /path/to/dataset --distribute False
To validate the accuracy of the trained model, you can use validate.py
and parse the checkpoint path with --ckpt_path
.
python validate.py -c configs/shufflenet_v2/shufflenet_v2_0.5_ascend.yaml --data_dir /path/to/imagenet --ckpt_path /path/to/ckpt
To deploy online inference services with the trained model efficiently, please refer to the deployment tutorial.
[1] Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 116-131.
No Description
Python Markdown Text other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》