Can't reach reported accuracy on MiniImagenet dataset of baseline++ with ResNet18

[bl0]

Hello, Thanks for your great work. I have run your code with the following command for several times, but the final accuracy is all about 73.99, which is slower than 75.68 reported in the paper. Am I missing something? Or I have to tune some hyperparameters?

for file in train save_features test; do
    python ./${file}.py --dataset miniImagenet --model ResNet18 --method baseline++ --train_aug --n_shot 5
done

Thanks very much.

[liuzechun]

Dear Author, Thanks a lot for your work! I also got the same issue with ResNet10 baseline++ on miniImageNet. I run: python ./train.py --dataset miniImagenet --model ResNet10 --method baseline++ --train_aug But got 74.10% +- 0.62%, instead of 75.90±0.61. Is there anything I can adjust? Thanks!

[XuChunqiao]

Hello, Thanks for your great work. I have the same problem . I run: python ./train.py --dataset miniImagenet --model ResNet18 --method baseline++ --train_aug python ./save_features.py --dataset miniImagenet --model ResNet18 --method baseline++ --train_aug python ./test.py --dataset miniImagenet --model ResNet18 --method baseline++ --train_aug Final accuracy is 74.38% +- 0.64% instead of 75.68% +- 0.64%.Did I miss something? Thank you very much.

[janetwise]

Maybe try to increase batch_size in the training. I used 128 as suggested in another issue and got 5%+ higher accuracy. You can try 32, 64... (default 16) based on your gpu mem.