Errors with adapter

[351246241]

Thank you for your awesome work. But I run the test.py found a error: RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 1024 but got size 768 for tensor number 1 in the list. It seems the shape of results of clip encoder and vae encoder are wrong. How can I fix it?

" def forward(self, clip, vae): # clip (1 257 1024) vae = self.pool(vae) # 1 4 80 64 --> 1 4 40 32 vae = rearrange(vae, 'b c h w -> b c (h w)') # 1 4 40 32 --> 1 4 1280 vae = self.vae2clip(vae) # 1 4 768 # Concatenate them is difficult concat = torch.cat((clip, vae), 1) "