How do I create a dataloder with list of tensors in different shape

data
1

let say I have two 3d tensors. each tensor contain an entire batch (batch size = 64) and I want to stack them together in a data-loader object so that when I iterate over the data-loader with Batch size = 64 will first get the first object and the section…
( I don’t need the shuffle so shuffle=False)
code e.g:

a_torch.shape
Out[97]: torch.Size([64, 347, 5])
b_torch.shape
Out[98]: torch.Size([64, 151, 5])
t_iter = DataLoader([a_torch, b_torch], batch_size=64, shuffle=False)
for data in t_iter:
    print(type(data))
    print(data.shape)

Out[98]:torch.Tensor
Out[99]:torch.Size([64, 347, 5])
Out[100]:torch.Tensor
Out[101]:torch.Size([64, 151, 5])

I tried to set my own collate_fn but it did not work.

Thank you