Are there any differences between nn.activations and nn.functional.activations?
Such as class torch.nn.ReLU and torch.nn.functional.relu,the two activation functions seem to be the same, can I use one of them to replace another?
PyTorch activations use the functional API under the hood. So you can safely use one or the other. However, the nn.module additionally implements _repr_
which lets you easily pretty print a summary of your network.