Either too little or too many positional arguments when using nn.Sequential

1

Hi,

I am new to PyTorch, so please excuse my silly question.

I define a nn.Sequential in init of my Encoder object like this:

self.list_of_blocks = [EncoderBlock(n_features, n_heads, n_hidden, dropout) for _ in range(n_blocks)]
self.blocks = nn.Sequential(*self.list_of_blocks)

The forward of EncoderBlock looks like this

def forward(self, x, mask):

Then, in the forward() of my Encoder, I try to do:

z0 = self.blocks(z0, mask)

However, I get

TypeError: forward() takes 2 positional arguments but 3 were given

When I try:

z0 = self.blocks(z0)

I get (understandably):

TypeError: forward() takes 2 positional arguments but only 1 was given

When I do not use nn.Sequential and just execute one EncoderBlock after another, it works:

for i in range(self.n_blocks):
     z0 = self.list_of_blocks[i](z0, mask)

What am I doing wrong and how do I use nn.Sequential correctly in this case?

Thank you!

2

nn.Sequential is used for very simple models and, if I’m not mistaken, doesn’t accept multiple inputs.
You could try this approach or write a custom model.

2 Likes