RuntimeError: Mismatch in shape: grad_output[0] has a shape of torch.Size([1, 2404, 4]) and output[0] has a shape of torch.Size([1, 1963, 4]).

[OswaldoBornemann]

I have a 3D point cloud, which the number of each sample is vary. And i met the mismatch in shape problem during the back propagation, while i set the batch size to 2.

RuntimeError: Mismatch in shape: grad_output[0] has a shape of torch.Size([1, 2404, 4]) and output[0] has a shape of torch.Size([1, 1963, 4]).

I guess the problem came from torchdiffeq.