Open pawopawo opened 4 years ago
Yes, the code is easy to extend. You can create another file for shuffleNet the the folder https://github.com/zhanghang1989/RegNet-Search-PyTorch/tree/master/arch
Thank you for your reply,
Do different blocks (shufflenet or ResNeSt) and different FLOPs require manual adjustment of the range of these hyper-parameters?
The default ranges are quite large. It's better to adjust them according to the FLOPs to improve the sampling efficiency.
Thanks for your excellent work! Are the search criteria in regnet suitable for ShuffleNetV2? If I want to search a shuffleNet model with FLOPs 60M, can I design a search space based on this code?