|
|
@ -180,7 +180,7 @@ class RandomMixing(nn.Module):
|
|
|
|
data=torch.softmax(torch.rand(num_tokens, num_tokens), dim=-1),
|
|
|
|
data=torch.softmax(torch.rand(num_tokens, num_tokens), dim=-1),
|
|
|
|
requires_grad=False)
|
|
|
|
requires_grad=False)
|
|
|
|
'''
|
|
|
|
'''
|
|
|
|
self.random_matrix = torch.softmax(torch.rand(num_tokens, num_tokens)
|
|
|
|
self.random_matrix = torch.softmax(torch.rand(num_tokens, num_tokens))
|
|
|
|
def forward(self, x):
|
|
|
|
def forward(self, x):
|
|
|
|
B, H, W, C = x.shape
|
|
|
|
B, H, W, C = x.shape
|
|
|
|
x = x.reshape(B, H*W, C)
|
|
|
|
x = x.reshape(B, H*W, C)
|
|
|
|