Open chengtao-lv opened 2 years ago
And I must add that the inner and outer learning rate in your paper about maml are both 0.01,but in your code are 0.001 and 0.005.
Hi, which pre-trained backbone did you use?
hey! resnet10,do I download or use the wrong backbone?
thanks for your reply, appreciate it!
I used the Resnet10 pre-trained using Simpleshot. Please download here: https://drive.google.com/file/d/1wWFQibiLL6yJWNXsHIxioucp0DnxOelP/view?usp=sharing
是啊,哥们。我就是从这里下载的,导入也没有啥问题,就是达不到你的效果.....方便加一个wx吗?如果不介意的话可以将wx号发我邮箱,897674362@qq.com。
under your instructions, i run python main.py --dataset miniImagenet --method metatrain --train_aug --test maml5_resnet and finally get the result of 66.5%,far from the result in your paper about 70%, is the Hyperparametric problems? ps: I did not use the novel.hdf5 file,because I don't think this file is getting anything useful,does file make the situation?
hello,I want to reproduce results 76.37% from the paper (MAML + ifsl with dataset miniImagenet and model ResNet10 ). But when I run an command (python main.py --dataset miniImagenet --model na --method metatrain --train_aug --test maml5_ifsl_resnet ),get a wrong answer whose acc=32. Did you reproduce the result successfully? Could you tell me how to get the 76.37% from the paper?If my command was wrong , what is the correct command? thank you!
under your instructions, i run python main.py --dataset miniImagenet --method metatrain --train_aug --test maml5_resnet and finally get the result of 66.5%,far from the result in your paper about 70%, is the Hyperparametric problems? ps: I did not use the novel.hdf5 file,because I don't think this file is getting anything useful,does file make the situation?