Incorrect error message when using nn.functional.convNd() and nn.functional.conv_transposeNd()

[vishwakftw]

• OS: Ubuntu 16.04
• PyTorch version: 0.3.0.post4
• How you installed PyTorch (conda, pip, source): pip
• Python version: 3.5.2
• CUDA/cuDNN version: CUDA 8

The below script produces an incorrect error message (torch.nn.functional.convNd)

>>> input = torch.autograd.Variable(torch.randn(10, 3, 28, 28, 28))
>>> weight = torch.autograd.Variable(torch.randn(20, 3, 4, 7))
>>> torch.nn.functional.conv3d(input, weight)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 126, in conv3d
    return f(input, weight, bias)
RuntimeError: Expected 5-dimensional input for 5-dimensional weight [10, 3, 4, 7], but got input of size [10, 3, 28, 28, 28] instead

>>> torch.nn.functional.conv2d(input[..., -1], weight[..., -1])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 90, in conv2d
    return f(input, weight, bias)
RuntimeError: Expected 4-dimensional input for 4-dimensional weight [10, 3, 4], but got input of size [10, 3, 28, 28] instead

>>> torch.nn.functional.conv1d(input[..., -1, -1], weight[..., -1, -1])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 54, in conv1d
    return f(input, weight, bias)
RuntimeError: Expected 3-dimensional input for 3-dimensional weight [10, 3], but got input of size [10, 3, 28] instead

The below script produces an incorrect error message (torch.nn.functional.conv_transposeNd)

>>> input = torch.autograd.Variable(torch.randn(10, 3, 28, 28, 28))
>>> weight = torch.autograd.Variable(torch.randn(3, 10, 2, 12))
>>> torch.nn.functional.conv_transpose3d(input, weight)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 224, in conv_transpose3d
    return f(input, weight, bias)
RuntimeError: Expected 5-dimensional input for 5-dimensional weight [3, 10, 2, 12], but got input of size [10, 3, 28, 28, 28] instead

>>> torch.nn.functional.conv_transpose2d(input[..., -1], weight[..., -1])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 192, in conv_transpose2d
    return f(input, weight, bias)
RuntimeError: Expected 4-dimensional input for 4-dimensional weight [3, 10, 2], but got input of size [10, 3, 28, 28] instead

>>> torch.nn.functional.conv_transpose1d(input[..., -1, -1], weight[..., -1, -1])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/torch/nn/functional.py", line 159, in conv_transpose1d
    return f(input, weight, bias)
RuntimeError: Expected 3-dimensional input for 3-dimensional weight [3, 10], but got input of size [10, 3, 28] instead