Multiplication of two tensors

Question

I'm trying to add a new block in yolov7. I want to multiply the output of this block with the previous layer coming from other layer.

I used mul (previous and output), but how can I modify in order to precise this two layers in YAML file.

Answer 1

1. You should add this lines in yolo.py

elif m is Multiply:
    c2 = ch[f[0]]

2. Add a class Multiply in common.py

class Multiply(nn.Module):
    def __init__(self):
        super(Multiply, self).__init__()

    def forward(self, x):
        return torch.mul(x[0], x[1])

3. Add in Yaml.file now

for example: [[-1, 6], 1, Multiply, []]  # cat backbone P4