nn.cross Entropy error

[anamce]

If anybody encountered this issue when run swin transformer please let me know I need your guidance

line 3086, in cross_entropy
return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing)
RuntimeError: only batches of spatial targets supported (3D tensors) but got targets of size: : [2, 512, 512, 3]