Is torch.nn.Sequential able to take a list as the input?

Hi, guys,
I am working on reimplementing the DeepLabV3+ these days,
and I am thinking about whether the torch.nn.Sequential can take a list of modules as the input?
I think this would be an awesome feature!

Any answer or idea will be appreciated!

We had discussions in the past about passing multiple inputs to nn.Sequential e.g. in this PR and here, but in the end the changes were rejected.

You could of course use the posted custom sequential containers for your model, if that makes its definition easier.

1 Like