The mean dice is very low after testing

[wanrat40]

I run train and test code following the guide, I use R50+ViT-B_16 as pretrained model and processing with Synapse data, I fixed n_skip=3, img_size=224, max_epochs=150, vit_patches_size=16 num_classes=9 but I fixed batch_size = 12 and base_lr=0.001 to prevent a memory leak. After testing I got the best mean_dice just 0.0163 as the picture below, I don't know what I was wrong. Can somebody help me?

[1583364217]

有很多空白的标签，他这个代码计算dice有点问题，所以导致dice很低

[thundercult]

I also encountered a similar problem. After I changed the dataset, the test dice was basically 0. Have you solved this problem？