The parameter about clip_adapter.clip_model.visual.cxt_decoder.

[shuo1874]

Hello, I noticed that the pre-trained model SCAN.pth you provided contains parameters related to clip_adapter.clip_model.visual.cxt_decoder. However, when I trained the model myself, this part was not present. As a result, I am unable to replace this part when using CS_CLIP.pt.
Besides, the same as parameter clip_adapter.clip_model.visual.learnable_weight .
Could you please advise what are these and what the issue might be? Thank you.