My weight initialization method is shown in 2. However, the weight value remains unchanged after initialization. Is there a proper weight initialization method that can be trained?

1

2.My weight initialization method
w=torch.tensor([[ 1.3562],[ 1.3095], [-0.6110],[ 1.6676], [-0.3462],[-1.8372],[-1.0733],[ 1.5269],[-0.9958],
[-0.7507]], requires_grad=True)
net.hidden1.weight=torch.nn.Parameter(w)

2

1.my network

network
network1399×376 57.3 KB