Open honghess opened 7 months ago
from ss2d:
x_dbl = torch.einsum("b k d l, k c d -> b k c l", xs, self.x_proj_weight)
RuntimeError: einsum(): subscript d has size 64 for operand 1 which does not broadcast with previously seen size 32 这里的通道数xs为d,但是 self.x_proj_weight为2d啊,怎么可能对的上,我的输入是32通道
from ss2d:
RuntimeError: einsum(): subscript d has size 64 for operand 1 which does not broadcast with previously seen size 32 这里的通道数xs为d,但是 self.x_proj_weight为2d啊,怎么可能对的上,我的输入是32通道