ysj9909 / SHViT

[CVPR 2024] SHViT: Single-Head Vision Transformer with Memory Efficient Macro Design
Other
24 stars 0 forks source link

Classification task training issue #3

Open JackLiao0824 opened 1 week ago

JackLiao0824 commented 1 week ago

Hi, I'm trying to train your SHVIT model from S1 to S4 on the ImageNet-1k classification task, unfortunately, all my attempts have failed.The highest accuracy achieved was 0.15%.Did you make any modifications to the training code? thanks.

ysj9909 commented 1 week ago

Based on today's review, there are no issues with the code. Please check if there are any problems with the dataset you used or if you have modified the code.