Tiny-ImageNet Reproduction

Hello and thank you for the great work! I am trying to reproduce your results for Tiny-ImageNet, would it possible to share your training hyperparameters (ie, learning rate, optimizer, annealing at which epoch)? I have also noticed that you chose to train for 300 epochs, whereas the original paper (Manifold Mixup) mentions training for 2000, is this because of the number of data generated from the augmentation? Thank you very much!