Syntax error in Matlab MatConvNet vl_nnconv function implemented as Pytorch nn.conv1d function

Question

In Matlab, I have the following code:

x = vl_nnconv(x, 'size', [3 3 3 nfilters(i)], 'pad', 1);

When I do the same thing in Python (Pytorch), I get a syntax error:

for i in range(1, nlayers-1):
    if i == 1:
        x = nn.conv1d(x,[3 3 3 nfilters(i)], pad())

with the error marked at the [3 3 3 nfilters(i)] area. I'm not sure what is the syntax error.

Answer 1

The proper formatting for the functional API F.conv1d (where F is torch.nn.functional ) is:

F.conv1d(x, weight, bias=None, stride=1, padding=0)

Assuming you have weight (and optionally bias ) as tensors stored somewhere.