search

floodsung / LearningToCompare_FSL

PyTorch code for CVPR 2018 paper: Learning to Compare: Relation Network for Few-Shot Learning (Few-Shot Learning part)
MIT License
1.04k stars 267 forks source link

fix .data[0] to .item(), fix in test episode error variable should in… #21

Closed Duan-JM closed 5 years ago

Duan-JM commented 5 years ago
  1. Pytorch's new version, do not support loss.data[0] form, we need to change it into loss.item().
  2. The origin code will generate error for predict_labels and test_labels are not in GPU. So I fixed it.
  3. After unzip the Omniglot Dataset, there still exist .DStore in dir Alphabet_of_the_Magi.

The code is clean and elegant, reading your code is my honor.

Vincent

Duan-JM commented 5 years ago

Some Issue in the train mini-imagenet