1x1 convolution.

Separate parameters for the main optimizer and the auxiliary optimizer. Return two optimizers

Helper function to find a named buffer. Returns a `torch.Tensor` or `None` Args: module (nn.Module): the root module query (s

Forward pass of Window Multi-head Self-attention module. Args: x: input tensor with shape of [b h w c]; attn_mask:

generating the mask of SW-MSA Args: shift: shift parameters in CyclicShift. Returns: attn_mask: should be

Update the registered buffers in a module according to the tensors sized in a state_dict. (There's no way in torch to directly load a buff

SwinTransformer Block

Helper function to find a named module. Returns a `nn.Module` or `None` Args: module (nn.Module): the root module query (str)

Return a new model instance from `state_dict`.