Issues
- 2
paper says adapters are added every M blocks(M=1) but checkpoints seems to add every two
#7 opened by Maxlinn - 2
How is the dimension of pos_embed handled when the image size is increased to 420x420 in the pre-training phase?
#6 opened by yutaidong - 1
Pretrained models
#4 opened by yuntaodu - 1
Could you please provide the pictures and videos used by benchmark for convenience?
#3 opened by ycsun1972 - 3
Prompt during training and inference
#2 opened by vateye - 7