why out_channels=4 * self.hidden_dim?

[cuixianheng]

why out_channels=4 * self.hidden_dim?

[Berndinio]

Because they split it afterwards in

cc_i, cc_f, cc_o, cc_g = torch.split(combined_conv, self.hidden_dim, dim=1)

f, g, o and g all have the same convolution operation (but with independent weights).
If you have a hidden.channel + x.channel = 4 input

[ndrplz]

Thanks @Berndinio

Also this answer might help