CODEWITHSUNDEEP
pythonpytorchdetection
w_h_d
w_h_d

Reputation: 43

How to understand this block of python code?

Found a piece of python code that can run perfectly,but I couldn't understand how it works. Would appreciate it if you could explain the lining part for me. I totally don't know what it does

class BoxHead(nn.Module):#pending
    def __init__(self, lengths, num_classes):
        super(BoxHead, self).__init__()
#-------------------------------------------------------
        self.cls_score = nn.Sequential(*tuple([
            module for i in range(len(lengths) - 1)
            for module in (nn.Linear(lengths[i], lengths[i + 1]), nn.ReLU())]
            + [nn.Linear(lengths[-1], num_classes)]))
#-----------------------------------------------------------------------

        self.bbox_pred = nn.Sequential(*tuple([
            module for i in range(len(lengths) - 1)
            for module in (nn.Linear(lengths[i], lengths[i + 1]), nn.ReLU())]
            + [nn.Linear(lengths[-1], 4)]))

Upvotes: 2

Views: 72

Answers (2)

23Xor
23Xor

Reputation: 44

After reading, I assert lengths as a list of numbers.

This is the input of nn.Sequential for cls_score:

# lengths = [num1, num2, num3, ..., numN-1, numN]
# numN-1 just represents its meaning, and is not valid in python syntax
[
    nn.Linear(num1, num2),
    nn.ReLU(),
    nn.Linear(num2, num3),
    nn.ReLU(),
    nn.Linear(num3, num4),
    nn.ReLU(),
    ...,
    nn.Linear(numN-1, numN), 
    nn.ReLU(),
    nn.Linear(numN, num_classes)
]

Input of nn.Sequential for bbox_pred is alike, but its last item is nn.Linear(numN, 4)

Upvotes: 1

Jubick
Jubick

Reputation: 448

nn.Sequential accepts *args which means you can pass any number of positional arguments which supposed to be PyTorch layers. In this case It creates len(lengts) - 1 blocks of nn.Linear followed by nn.ReLU each and also one linear layer at last.

Upvotes: 2

Related Questions