conv = nn.Conv2d(c1, 1, 1, bias=False).requires_grad_(False)

[williamlzw]

conv = nn.Conv2d(c1, 1, 1, bias=False).requires_grad_(False)
self.conv.weight.data[:] = nn.Parameter(x.view(1, c1, 1, 1))

https://github.com/ultralytics/ultralytics/blob/main/ultralytics/nn/modules.py
line73
requires_grad_(False) ? not support in torchsharp.
conv.weight.data[:] = nn.Parameter(x.view(1, c1, 1, 1))

[NiklasGustafsson]

Thanks for reporting the issue. It's been implemented incorrectly (easy fix):

            public Tensor requires_grad_(bool requires_grad = true)
            {
                this.requires_grad = true;
                return this;
            }

In the meantime, use with_requires_grad() or set the requires_grad property directly.