I am trying to replicate a TF code in PyTorch. I see that in linear layer the initialization is done as:
initialization=‘he’
However I am not able to find a PyTorch equivalent of that. All I get is xavier as:
nn.init.xavier_…(layer.weight)
I am trying to replicate a TF code in PyTorch. I see that in linear layer the initialization is done as:
initialization=‘he’
However I am not able to find a PyTorch equivalent of that. All I get is xavier as:
nn.init.xavier_…(layer.weight)
PyTorch equivalent for he initialization
is
nn.init.kaiming_…(layer.weight)
https://pytorch.org/docs/stable/nn.init.html#torch.nn.init.kaiming_uniform_