Hello, I have read the paper and this is a great work on vision transformer compression!
However, when I want to reproduce the code following the command on github. I always get 0 accuracy after joint training. Could you provide the pre-trained model checkpoint so I can easily evaluate it on ImageNet?
I simply followed the command on github for setting the environment and run the stage 1 uvc training command. I installed apex library without global option. I used 2 NVIDIA GeForce RTX 2080 GPU to run it.
Hello, I have read the paper and this is a great work on vision transformer compression!
However, when I want to reproduce the code following the command on github. I always get 0 accuracy after joint training. Could you provide the pre-trained model checkpoint so I can easily evaluate it on ImageNet?
I simply followed the command on github for setting the environment and run the stage 1 uvc training command. I installed apex library without global option. I used 2 NVIDIA GeForce RTX 2080 GPU to run it.
Best regards,