mask_decoder mismatch with raw SAM mask_decoder

[wade0604]

src = torch.repeat_interleave(image_embeddings, tokens.shape[0], dim=0)

Why did you comment out this line in your mask decoder

[Cjl-MedSeg]

The result of doing this is to expand the dimension of image_embeddings to match the size of tokens. We have already performed a similar operation during training, so commenting this line here will not affect the normal functioning of the model.

[wade0604]

这是来自QQ邮箱的假期自动回复邮件。   您好，我最近正在休假中，无法亲自回复您的邮件。我将在假期结束后，尽快给您回复。